Searching \ for '[EE] Suggestions for saving board space on my desi' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: techref.massmind.org/techref/index.htm?key=suggestions+saving
Search entire site for: 'Suggestions for saving board space on my desi'.

Exact match. Not showing close matches.
PICList Thread
'[EE] Suggestions for saving board space on my desi'
2011\05\23@083815 by V G

picon face
Hi all,

I'm having three boards made for my university. The design is a SCADA
control system terminal that will sit inside a chamber and switch on and off
heaters/humidifiers/stuff like that. It has all the necessary protection in
place (MOV, zeners, fueses, optoisolators, etc), most of it on the board,
where it is needed. Unfortunately, due to all the protection, LEDs, and
stuff, the board is quite large (around 6x6 inches). It's also using only
through hole components.

Here's a roughly (messily) laid out, unrouted picture of the board:
http://postimage.org/image/11z7x17dw/full/

I would like to save as much space as possible. This will probably mean
switching all the components to surface mount. There's also four huge
MOSFETS near the top right which probably wont be laying down on the board
to save space, but are outlined as such in the picture. The huge DIP package
in the middle is a UBW32 board. Other ICs are mostly 74 series logic
buffers, etc. Most of the 6 pin DIPs are optoisolators.

Note: I wouldn't know how to route this beast by hand. The autorouter routes
this thing fine in two layers.

What suggestions would you guys have for saving space and designing for
better efficiency

2011\05\23@090936 by Michael Watterson

face picon face
On 23/05/2011 13:38, V G wrote:
> Note: I wouldn't know how to route this beast by hand. The autorouter routes
> this thing fine in two layers.
>
> What suggestions would you guys have for saving space and designing for
> better efficiency?

Get someone that *can* route it by hand. Or practice till you can.

The most important aspect is initial position and orientation of the parts

2011\05\23@093115 by PICdude

flavicon
face
I'm guessing the connectors will pretty much need to be through-hole,  but what's really bugging me here is all those DIP chips.  If you can  do SSOP's, that would be much better, but even SOIC's will help.  You  can put SMD components on the bottom side of the board, under the  chips perhaps to save space.

If you keep through-hole passives, you can mount the resistors on end,  rather than flat, to save more space.  That has made a big difference  for me in the past.

Yes, putting the MOSFETs vertical will save a lot of space, but I'm  guessing they're in those large packages for temperature reasons, so  move them to the edges of the board, so you can heatsink if required.

Cheers,
-Neil.



Quoting V G <spam_OUTx.solarwind.xTakeThisOuTspamgmail.com>:

{Quote hidden}

>

2011\05\23@102138 by Geo

picon face
V G wrote:
<snip>
> Note: I wouldn't know how to route this beast by hand. The autorouter  >routes this thing fine in two layers.
>
> What suggestions would you guys have for saving space and designing  >for better efficiency?

I know you said "roughly" laid out so the following is advice - not criticism...

Location, location, location...
You need to spend a *lot* of time on component placement.
This reduces track lengths and allows easier routing.
Make life easier for the router by placing components on a 0.1" grid wherever possible.

Are you sure U3 and U5 are the right way round? They look like RJ connectors to me.

Will any of the (ribbon?) connectors have "ears"? You would not be able to get the mating half in J6 if so (too close to J1).

Rotate the 4 leds 180 degrees and their resistors and you gain tracking space.

Treat the connectivity lines as pieces of elastic - you can then see that the 5 component in the bottom right-hand corner would /really/ like to be near the TO220s. There are 5 connection pulling them up - and only two down - so the tracking would get easier.

The components down the left hand side could also do with some movement (e.g R89 swap with R3). The aim is to reduce the track length to as near zero as possible.

So take about a day  - run the autorouter and see where it has problems that could be solved by re-arrangement. After the auto router has run, highlight some tracks and see if they have taken sensible or circuitous routes where some tweaking by hand can improve. Watch out for it using the space between the optocoupler pins if you need safety clearance.

luck,
George Smith


2011\05\23@114057 by PICdude

flavicon
face
Quoting Geo <.....buggiesmithKILLspamspam@spam@gmail.com>:

> ...
> So take about a day  - run the autorouter and see where it has problems
> that could be solved by re-arrangement. ...

After autorouting, I turn off all layers except vias, so I can clearly  see where the heavy clusters/difficult spots are, and re-arrange  components accordingly.

Cheers,
-Neil.

2011\05\23@141706 by YES NOPE9

flavicon
face
Why are you trying to save space ?
If you told us what kind of space you are trying to fit into.... we could offer more help.
Would a planar circuit work best ?  A "cube" ?  A series of small boards that can be connected together ?
Resistor packs will save space , diode packs , transistor packs
You can stand the UWB32 on it's side with a right angle socket.

Your original post contained much good information regarding your question..........
Excellent effort....
99guspuppet


2011\05\23@154357 by Denny Esterline

picon face
On Mon, May 23, 2011 at 5:38 AM, V G <x.solarwind.xspamKILLspamgmail.com> wrote:

{Quote hidden}

About a million things :-)
I would suggest that you start with the understanding that the "perfect"
design is impossible, then work to "functional".

First, I see a classic newbie mistake - no mounting holes.

You said the autorouter does it "fine", why is that not an acceptable
solution?

You want to "save as much space as possible" - why?  If it physically
doesn't fit in the mounting location, ok, that's a good reason. If you're
trying to reduce costs on the board, you need to re-evaluate your motives.
Saving $5 off the PCB on a total production run of three units, compared to
the dozen hours or more it takes to lay out a space optimized version may
not be your best choice.

Also consider the future - if you were able to shrink this down to a 2"
square using the tiniest parts available, how convenient would it be to
assemble or debug? A little extra space at this stage can make your life a
lot easier later.


-Denn

2011\05\23@171622 by V G

picon face
On Mon, May 23, 2011 at 3:43 PM, Denny Esterline <.....desterlineKILLspamspam.....gmail.com>wrote:

> About a million things :-)
> I would suggest that you start with the understanding that the "perfect"
> design is impossible, then work to "functional".
>

There would be lots of ways to make it functional, but I would like to save
a few square inches on the board, it would save quite a bit of money. Also,
I'm gonna be using these guys: http://www.pcbgeek.com/ and they have a  7x7"
restriction. I'd still like to use surface mount parts, mostly.


> First, I see a classic newbie mistake - no mounting holes.
>

My mistake. Thanks for pointing that out.


> You said the autorouter does it "fine", why is that not an acceptable
> solution?
>

It is! I just need to place the components properly. The board will be
operating at low speeds. The microcontroller already has it's own board
(UBW32), and everything else will be running at around 9600 bits/s or less


> You want to "save as much space as possible" - why?  If it physically
> doesn't fit in the mounting location, ok, that's a good reason. If you're
> trying to reduce costs on the board, you need to re-evaluate your motives..
> Saving $5 off the PCB on a total production run of three units, compared to
> the dozen hours or more it takes to lay out a space optimized version may
> not be your best choice.
>

Difference between 6x6 and 7x7 is 13 sq in. That can mean a $65 difference
if I use Dorkbot, for example. So squeezing everything down a couple of
inches on each dimension would save a lot of money.


> Also consider the future - if you were able to shrink this down to a 2"
> square using the tiniest parts available, how convenient would it be to
> assemble or debug? A little extra space at this stage can make your life a
> lot easier later.
>

That's true, but I have no intention of going smaller than I can comfortably
solder by hand. Someone else pointed out that 1206 size surface mount
components are a good choice, so I'll go with those

2011\05\23@173330 by Denny Esterline

picon face
On Mon, May 23, 2011 at 2:16 PM, V G <EraseMEx.solarwind.xspam_OUTspamTakeThisOuTgmail.com> wrote:

{Quote hidden}

Hmmm... don't know about Dorkbot. I tend to use PCBFabExpress for low volume
and proto work. Just poked on their website, 2 layer 6"x6", 5 day turn time,
three boards $35 ea. vs. 7"x7" $41.67 ea. So about a $20 difference. Not
enough to make me redesign, but your economics may figure differently.

For my money, 1206 isn't enough smaller than through hole to make it worth
doing. SOIC and 0806 are easy enough to deal with. 0604 and TSSOP is doable,
but at least for me, passes the point of being "fun". I've done smaller by
hand, but I don't enjoy it.

-Denn

2011\05\23@194302 by William \Chops\ Westfield

face picon face

On May 23, 2011, at 5:38 AM, V G wrote:

> What suggestions would you guys have for saving space and designing  
> for
> better efficiency?

Don't use a UBW32 ?  Seriously; that's going to be a major factor  limiting how much you can shrink the board.   Mmm.  Could be that the  new ChipKit boards change your equation...

There are a lot of components on there, and I don't think you'll be  able to shrink the board very much just by better routing and  placement.  Protection circuits are annoyingly large :-(

In general, your passives are further apart than they need to be.  A  set of signal-protecting diodes, resistors, and caps (as you have on  the left edge) can generally fit on a 0.1 inch grid unless there are  other reasons (power dissipation, isolation) for keeping them far apart.

Consider quad opto-isolators instead of one-per-package devices.

Consider resistor arrays (SIPs) where appropriate (current limiting  for LEDs, especially.)

I can't quite figure out what your TTL devices are, but it looks like  you have one 20pin DIP  (eg U8) driving four LEDs, which is pretty  awful, density-wise.  Use higher efficiency LEDs and the PIC32 can  probably drive the LED directly (4mA).  I don't know if you can get  optos for low drives or not  (Hmm.  How much CAN the pic32 drive on IO  pins?  I see a 25mA absolute max, but not a "recommended" value...)   You could stick in something like a TLC5940 LED driver and drive 16  leds from a 28pin chip (with internal current limiting, so you get rid  of the resistors as well.)

The other thing to think about is that $65 is NOT a lot of money.   About one day's worth of labor at minimum wage, or less than a  billable hour for a research project.  (of course, it's REAL money of  the sort that escapes the university instead of going around in  circles making everyone look good, but still...)   A TH design that  costs an extra hundred bucks that is robust and can be put together  easily may be preferred over a crowded SMT board that was cheaper but  more difficult to assemble)

BillW


BillW

2011\05\23@200824 by V G

picon face
On Mon, May 23, 2011 at 7:43 PM, William "Chops" Westfield
<@spam@westfwKILLspamspammac.com>wrote:

>  Don't use a UBW32 ?  Seriously; that's going to be a major factor
> limiting how much you can shrink the board.   Mmm.  Could be that the
> new ChipKit boards change your equation...
>

The UBW32 is pretty cool and I'm comfortable using it. The layout is nice
and I think it's smaller overall than the ChipKit boards.

>
> There are a lot of components on there, and I don't think you'll be
> able to shrink the board very much just by better routing and
> placement.  Protection circuits are annoyingly large :-(
>

You're right. I rearranged the board and switched some components to surface
mount. Check it out:

Unrouted: http://postimage.org/image/22skvifk4/full/
Routed: http://postimage.org/image/22sj7z32c/full/

I got the board down to 4.9x5.6 inches. Not bad and not overly crowded.


> In general, your passives are further apart than they need to be.  A
> set of signal-protecting diodes, resistors, and caps (as you have on
> the left edge) can generally fit on a 0.1 inch grid unless there are
> other reasons (power dissipation, isolation) for keeping them far apart.
>

Edited them. Switched to surface mount and pushed them a little closer
together.


> Consider quad opto-isolators instead of one-per-package devices.
>

I looked them up, but I found the price and simplicity of 4N25s attractive.
I've used them before and I'm comfortable designing with them.


> Consider resistor arrays (SIPs) where appropriate (current limiting
> for LEDs, especially.)
>

I've looked at the arrays as you've suggested. I didn't find that it would
save much space compared to the 1206 resistors. There are only 8 LEDs on
there anyway.

I can't quite figure out what your TTL devices are, but it looks like
> you have one 20pin DIP  (eg U8) driving four LEDs, which is pretty
> awful, density-wise.  Use higher efficiency LEDs and the PIC32 can
> probably drive the LED directly (4mA).  I don't know if you can get
> optos for low drives or not  (Hmm.  How much CAN the pic32 drive on IO
> pins?  I see a 25mA absolute max, but not a "recommended" value...)
> You could stick in something like a TLC5940 LED driver and drive 16
> leds from a 28pin chip (with internal current limiting, so you get rid
> of the resistors as well.)
>

The TTL devices are 74 series buffers. I'm keeping the pins of the UC as
unloaded as possible. The buffers drive the LEDs, and eventually the
MOSFETS, etc.

The other thing to think about is that $65 is NOT a lot of money.
> About one day's worth of labor at minimum wage, or less than a
> billable hour for a research project.  (of course, it's REAL money of
> the sort that escapes the university instead of going around in
> circles making everyone look good, but still...)   A TH design that
> costs an extra hundred bucks that is robust and can be put together
> easily may be preferred over a crowded SMT board that was cheaper but
> more difficult to assemble)
>

You're right. This IS for a university, but specifically, I'm working for a
professor researching in biology. Most of their funding goes toward lab
tools related to biology. I quoted her a price of a few hundred dollars for
everything, including time and labour, and it seems I'm still well within
that budget. The project was my suggestion to her to improve certain lab
procedures and so on.


Also, I'm going to be using PCBGeek for making these boards:
http://www.pcbgeek.com/. Their prices seem reasonable, and I get a few
boards out of it. I'm only making 3 units, but if I squeeze it down enough,
I might get a spare board

2011\05\23@220151 by Denny Esterline

picon face
>
>
> You're right. I rearranged the board and switched some components to
> surface
> mount. Check it out:
>
> Unrouted: http://postimage.org/image/22skvifk4/full/
> Routed: http://postimage.org/image/22sj7z32c/full/
>
> I got the board down to 4.9x5.6 inches. Not bad and not overly crowded.
>
>
>
Uhm... still no mounting holes...

I don't know what you're doing with the TO220 parts, but anything that needs
a package that big will likely not be happy with those skinny little
traces.

-Denn

2011\05\24@002030 by V G

picon face
On Mon, May 23, 2011 at 10:01 PM, Denny Esterline <KILLspamdesterlineKILLspamspamgmail.com>wrote:

> Uhm... still no mounting holes...
>

It's not finished yet! I'll be sure to put appropriate mounting holes for M3
hex standoffs.



> I don't know what you're doing with the TO220 parts, but anything that
> needs
> a package that big will likely not be happy with those skinny little
> traces.
>

Those are specifically BUZ71 N channel MOSFETs. The MOSFETS switch a 12V AC
relay. It doesn't actually take /that/ much power and the MOSFET won't even
get a little warm because the relay won't that /that/ much power. I just
found that that particular MOSFET was suitable for my needs.

And BUZ71 is only available in that package:
http://ca.mouser.com/Search/Refine.aspx?Keyword=buz71

Specifically, they're being used to switch this relay:
http://futurlec.com/Relays/PR12VDCSPST.shtml

- Coil Power: 1200 mW
- Nominal Voltage: 12Vdc
- Coil Resistance: 120ohm

Could you please suggest an alternative N channel MOSFET suitable for this
use? The gate is being switched at 3.3V

2011\05\24@003740 by V G

picon face
On Tue, May 24, 2011 at 12:20 AM, V G <RemoveMEx.solarwind.xTakeThisOuTspamgmail.com> wrote:

{Quote hidden}

Anyway, I agree, I don't think a TO220 MOSFET is required. The relay only
needs 100mA, so I'm going to switch to a SOT223 package

2011\05\24@023431 by Ruben Jönsson

flavicon
face
>
> You're right. I rearranged the board and switched some components to surface
> mount. Check it out:
>
> Unrouted: http://postimage.org/image/22skvifk4/full/
> Routed: http://postimage.org/image/22sj7z32c/full/
>
> I got the board down to 4.9x5.6 inches. Not bad and not overly crowded.
>
>
Try to put SMD components on the bottom side and keep TH components on the top side. This way you could put some TH components on top of some SMD components. Under U1 you could probably fit some of the SMD ICs. You can put the current limiting resistors for the LEDs under the LEDs and move them much closer to J2. SMD components can also be placed much closer to the connectors if they are placed on the bottom side. These are just some examples.

If you also try to keep the signal and power traces on the bottom side, you can perhaps use most of the top side as a ground plane. It will also make it easier to place decoupling capacitors where they work best.

The routing also takes quite a lot of board real estate. Just by looking at the current limiting resistors for the LEDs, it seems like you can save some of it by doing it by hand. Remember that the resistors can be put in series with either the anode or the cathode and this may make routing easier.

/Ruben




==============================
Ruben Jönsson
AB Liros Electronic
Box 9124, 200 39 Malmö, Sweden
TEL INT +46 40142078
FAX INT +46 40947388
TakeThisOuTrubenEraseMEspamspam_OUTpp.sbbs.se
==============================

2011\05\24@032259 by Michael Watterson

face picon face
On 24/05/2011 05:20, V G wrote:
> Could you please suggest an alternative N channel MOSFET suitable for this
> use? The gate is being switched at 3.3V.

loads
VN2222LM is only one.
260mA @ 25C
160mA @100C
you need 100mA. There are T092ish parts (with SM alternatives) that do 200mA
Note that VGS in data sheets may be to allow MAXIMUM current, or MINIMUM resistance at Max Temp, rather than "good enough"

ie. might be 8 Ohms at 3V drive, 2 Ohms at 10V drive, but  8 Ohms is 80mW at 100mA, and not enough to be a problem for relay. Since input impedance is high there may be more than 3V drive.

There are loads of parts that may be suitable BS170 is lower power but may only dissipate 20mW @ 500mA
Check out 2N7002 etc About 4 Ohms @ 3V drive. SOT and T092, 200mA

The BS170 looks a little better.

BS170 & 2N700x Very cheap.

2011\05\24@081009 by Olin Lathrop

face picon face
V G wrote:
> Routed: http://postimage.org/image/22sj7z32c/full/

That looks to be a two layer board.  Since it has lots of routing on both
layers, that means there won't be any ground plane.  That means the board
will be more susceptible to external noise, and will radiate likewise.
Since you don't have a ground plane, you need to be extra careful about the
ground current paths, make sure everything is locally well bypassed, and
hopefully hook up the ground with fat traces in a star configuration to the
extent you can manage.

On two layer boards of similar complexity, I try to keep all the traces on
the top layer and only use the bottom layer for short "jumpers" as needed.
That way the rest of the bottom layer can be a ground plane.  A good example
of this is the USBProg.  The top side drawing is at
http://www.embedinc.com/products/usbprog/eusb3_top.gif, and the bottom at
http://www.embedinc.com/products/usbprog/eusb3_bot.gif.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\24@084927 by Olin Lathrop

face picon face
V G wrote:
>> Uhm... still no mounting holes...
>
> It's not finished yet! I'll be sure to put appropriate mounting holes
> for M3 hex standoffs.

Adding holes is part of the placement process.  In other words, this is done
before routing, not after.  It was a appropriate comment since you showed a
routed board.

> - Coil Power: 1200 mW
> - Nominal Voltage: 12Vdc
> - Coil Resistance: 120ohm

If that's all you need, you can use much smaller packages.  All you need is
a transistor that can withstand 12V and carry 100mA.  Just about anything
can do that.  Don't forget the diode to catch the coil kickback when turning
off the relay else whatever transistor you choose will get fried.

Something like a IRLML2502 should do nicely, but it doesn't need to be a
FET.  Especially considering your extreme cost sensitivity, you could use a
MMBT4401, emitter to ground, base to PIC pin with resistor in series.  Let's
say you count on a minimum gain of 50.  That means you need a minimum of 2mA
base current.  To be pessimistic, let's say the B-E drop is 700mV.  That
leaves 2.6V accross the base resistor.  2.6V / 2mA = 1.3kOhms, so a 1kOhm
resistor sounds about right.  The combination of MMBT4401 and 1KOhm resistor
will cost a small fraction of a suitable FET.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\24@091808 by alan.b.pearce

face picon face
> Especially considering your extreme cost sensitivity, you could use a
> MMBT4401, emitter to ground, base to PIC pin with resistor in series.  Let's
> say you count on a minimum gain of 50.  That means you need a minimum of 2mA
> base current.  To be pessimistic, let's say the B-E drop is 700mV.  That
> leaves 2.6V accross the base resistor.  2.6V / 2mA = 1.3kOhms, so a 1kOhm
> resistor sounds about right.  The combination of MMBT4401 and 1KOhm resistor
> will cost a small fraction of a suitable FET.

If going for surface mount an MMBT2222A or (IIRC) PMBT2222A, both SOT-23 versions (from different manufacturers, hence different part numbers) of the ubiquitous 2N2222A NPN transistor would apply equally well. The resistor value would stay the same.
-- Scanned by iCritical.

2011\05\24@093823 by Bob Ammerman

flavicon
face
And don't forget the snubber diode!

-- Bob Ammerman
RAm System

2011\05\24@121520 by Josh Koffman

face picon face
On Tue, May 24, 2011 at 8:10 AM, Olin Lathrop <RemoveMEolin_piclistspamTakeThisOuTembedinc.com> wrote:
> On two layer boards of similar complexity, I try to keep all the traces on
> the top layer and only use the bottom layer for short "jumpers" as needed..
> That way the rest of the bottom layer can be a ground plane.  A good example
> of this is the USBProg.  The top side drawing is at
> http://www.embedinc.com/products/usbprog/eusb3_top.gif, and the bottom at
> http://www.embedinc.com/products/usbprog/eusb3_bot.gif.

Olin,

Is that directly output from Eagle as GIF? I like the addition of the scales!

Josh
-- A common mistake that people make when trying to design something
completely foolproof is to underestimate the ingenuity of complete
fools.
        -Douglas Adams

2011\05\24@124526 by Olin Lathrop

face picon face
Josh Koffman wrote:
>> http://www.embedinc.com/products/usbprog/eusb3_top.gif, and the
>> bottom at http://www.embedinc.com/products/usbprog/eusb3_bot.gif.
>
> Is that directly output from Eagle as GIF? I like the addition of the
> scales!

I have Eagle export the drawing as a TIF file, then convert to GIF with my
own tools.  I don't trust Eagle to pick the right choices in the GIF.  The
colors I use are all deliberately gray scale, so I set up the LUT in the GIF
to only the gray scale range.  I haven't actually tried it, but I expect
Eagle wouldn't do that.

The scales and other stuff are put there by my GRID_BRD ULP.  After that,
you run either the GRID_TOP or GRID_BOT script to have it show the top or
bottom of the board.  The ULP sets up everything so that selecting the view
is just a matter of setting the colors differently.

These ULP and scripts, and a bunch of other stuff, is available in my Eagle
Tools release at http://www.embedinc.com/pic/dload.htm.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\24@174200 by V G

picon face
On Tuesday, May 24, 2011, Olin Lathrop <olin_piclistEraseMEspam.....embedinc.com> wrote:
>> - Coil Power: 1200 mW
>> - Nominal Voltage: 12Vdc
>> - Coil Resistance: 120ohm
>
> If that's all you need, you can use much smaller packages.  All you need is
> a transistor that can withstand 12V and carry 100mA.  Just about anything
> can do that.  Don't forget the diode to catch the coil kickback when turning
> off the relay else whatever transistor you choose will get fried.
>
> Something like a IRLML2502 should do nicely, but it doesn't need to be a
> FET.  Especially considering your extreme cost sensitivity, you could use a
> MMBT4401, emitter to ground, base to PIC pin with resistor in series.  Let's
> say you count on a minimum gain of 50.  That means you need a minimum of 2mA
> base current.  To be pessimistic, let's say the B-E drop is 700mV.  That
> leaves 2.6V accross the base resistor.  2.6V / 2mA = 1.3kOhms, so a 1kOhm
> resistor sounds about right.  The combination of MMBT4401 and 1KOhm resistor
> will cost a small fraction of a suitable FET.

Would a VN2222 N channel MOSFET be suitable? Or an equivalent in surface mount?

I checked the datasheet and everything seems suitable, but there may
be something I missed.

Just checked your IRLML2502. Loving the current drain capability at
Vgs = 3.3V. I guess I'll go with that.

My only concern is that the SOT23 part will be a bit hard to solder by hand..

My cost sensitivity isn't really toward $1 dollar components. I don't
mind spending for those. I do, however, mind the $20 more per board
for a few sq inches. That's unjustifiably unacceptable. But I found a
good fab house so it shouldn't be that bad.

2011\05\24@180254 by Bob Blick

face
flavicon
face


On Tue, 24 May 2011 17:41 -0400, "V G" wrote:

> My only concern is that the SOT23 part will be a bit hard to solder by
> hand.

You'll have no problems with it. Only thing to watch for, if you etch
your own board, don't run a trace between the source and gate pads. But
if you have a resist layer you're good to go.

Bob

-- http://www.fastmail.fm - Choose from over 50 domains or use your own

2011\05\24@180319 by Olin Lathrop

face picon face
V G wrote:
> Would a VN2222 N channel MOSFET be suitable? Or an equivalent in
> surface mount?

I don't know what a VN2222 MOSFET is.  The number 2222 hints at some variant
of the very common 2N2222A NPN transistor.

The reason I mentioned MBT4401 is that's the NPN jellybean transistor I use..
It's just as cheap and plentiful as 2222A or 3906, but is rather more
robust.

> Just checked your IRLML2502. Loving the current drain capability at
> Vgs = 3.3V. I guess I'll go with that.

That's a very nice part if you don't need more than 20V capability, which
you don't in this case.  The IRLML0030 is close to the same thing with 30V
capability.  However, these are many times more expensive than something
like a MMBT4401, although in absolute terms its only a few 10s of cents.

Too often it seems people knee jerk to using a FET when a plain old bipolar
will do fine.

> My only concern is that the SOT23 part will be a bit hard to solder
> by hand.

SOT-23 is no big deal to do by hand.  The pins are 50 mil pitch if I
remember right.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\24@182912 by V G

picon face
On Tue, May 24, 2011 at 6:03 PM, Olin Lathrop <EraseMEolin_piclistspamembedinc.com>wrote:

> V G wrote:
> > Would a VN2222 N channel MOSFET be suitable? Or an equivalent in
> > surface mount?
>

It's the MOSFET analog to a 2N2222 BJT.


> Too often it seems people knee jerk to using a FET when a plain old bipolar
> will do fine.
>

Then in which case would you suggest a FET over a BJT

2011\05\24@200125 by Dwayne Reid

flavicon
face
At 03:41 PM 5/24/2011, V G wrote:

>My cost sensitivity isn't really toward $1 dollar components. I don't
>mind spending for those. I do, however, mind the $20 more per board
>for a few sq inches. That's unjustifiably unacceptable. But I found a
>good fab house so it shouldn't be that bad.

I have another board house for you to try if you want: <http://dorkbotpdx.org/wiki/pcb_order> .  This was mentioned previously on the PIClist sometime last year.

The guy who runs it is named Laen and he is really easy to work with.  I've ordered from him previously with good results.  You get 3 copies of your board for $5 per square inch (two layer).  His price for 4 layer is reasonable: you get 3 copies of your board for $10 per square inch.

Note that price is per square inch of your design.  The actual cost per board is that cost divided by 3.

He is also offering a 'beta' service for those who want larger quantities: 10 board minimum, 150 square inch minimum for $1 per square inch.  That could be 10 boards at 15 Sq In each, or 30 boards at 5 Sq In each, or whatever works best for you.  That cost is inclusive of setup fees and ship cost to you.

My previous orders took about 3 weeks from date of order to receiving them but check with him for your specific delivery time.

Hope this helps!

dwayne

-- Dwayne Reid   <RemoveMEdwaynerEraseMEspamEraseMEplanet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
(780) 489-3199 voice          (780) 487-6397 fax
http://www.trinity-electronics.com
Custom Electronics Design and Manufacturing

2011\05\24@203446 by V G

picon face
On Tue, May 24, 2011 at 8:01 PM, Dwayne Reid <RemoveMEdwaynerspam_OUTspamKILLspamplanet.eon.net> wrote:

>  I have another board house for you to try if you want:
> <http://dorkbotpdx.org/wiki/pcb_order> .  This was mentioned
> previously on the PIClist sometime last year.
>

Hi Dwayne,

I have already considered this service. This is everything I looked at:

List of fab services:
http://digital-diy.com/General-Electronics/pcb-fabrication-sources.html

Also this: http://www.4pcb.com/index.php?load=content&page_id=130

I determined that http://www.pcbgeek.com/ is the best deal. $60 + $30
shipping for as many boards that can fit in 200 sq inches, maximum 7"x7"
board size.

Ordering PCBs is REALLY expensive for low quantities. I'm never going to do
this again unless I'm doing a project for someone else and they're paying
for it (like in this case). I'll get myself a $100 laminator, some TRF, and
dirt cheap copper clad board and make my own. Did it before, and results
were excellent. Drilling takes less than a second per hole with a tungsten
carbide bit, and vias can be done by threading a thin wire through them and
soldering. Wouldn't need much more than a 2 layer board for most things

2011\05\24@205809 by V G

picon face
On Tue, May 24, 2011 at 8:50 AM, Olin Lathrop <RemoveMEolin_piclistTakeThisOuTspamspamembedinc.com>wrote:

> Something like a IRLML2502 should do nicely, but it doesn't need to be a
> FET.
>

Mouser doesn't stock that part :(

I'm going with this one instead:
ca.mouser.com/ProductDetail/Fairchild-Semiconductor/FDT86106LZ/?qs=sGAEpiMZZMtCrm2fS1SYQoLbm0WbMNYXAFZoR%252bkbktE%3d
Datasheet: http://www.fairchildsemi.com/ds/FD/FDT86106LZ.pdf

Very similar characteristics. SOT223 package, so a little bigger, much
easier to solder, price is very reasonable

2011\05\24@211938 by V G

picon face
Took many of your suggestions and updated the board:

Unrouted: http://postimage.org/image/1j9m4mphg/full/
Routed: http://postimage.org/image/1j9kh3czo/full/

MOSFETS and two ICs are now under UBW32 (on bottom layer) to save board
space. Mounting holes added.

Can anyone think of anything else

2011\05\24@224806 by Denny Esterline

picon face
>
>
>
> Ordering PCBs is REALLY expensive for low quantities. I'm never going to do
> this again unless I'm doing a project for someone else and they're paying
> for it (like in this case). I'll get myself a $100 laminator, some TRF, and
> dirt cheap copper clad board and make my own. Did it before, and results
> were excellent. Drilling takes less than a second per hole with a tungsten
> carbide bit, and vias can be done by threading a thin wire through them and
> soldering. Wouldn't need much more than a 2 layer board for most things.
> --
>  <http://www.piclist.com>
>
>
"Expensive" _really_ depends on why your doing it.

I started in electronics years ago as a hobby. Plastic breadboards were my
friend, everything was available with leads, disassembling old TVs was worth
it to me. I recall one day when I needed a 7400 quad nand gate. I spent
about four hours cutting one off an old ISA card and soldering on little
bits of wire for new leads.

One day breadboards aren't good enough anymore, you want something
"permanent", so I graduated to perfboard and stripboard and a dozen variants
of it (and the cool little PCB's that are layed out like a breadboard, can't
forget them!) I even did a couple where I just poked all the parts though a
piece of cardboard from a cereal box. Marvelous learning opportunity about
humidity...

Then I bought a laminator and started doing toner transfer, Oh glorious
day!! circuits were actually reliable and didn't fall apart on me as I
worked with them. I went though about 5 sq ft of PCB stock (gotta love
ebay!) learning what makes a PCB good, what layout actually works and what's
a total PITA to build or work with. And I learned about mounting holes the
hard way. A lesson I had to learn several times before it stuck, I'm sorry
to say. About this time I learned about surface mount - not really trying to
get smaller, I was sick and tired of drilling holes! What a pain!!!! Unless
you use carbide bits, the fiberglass is _really_ hard on them (I've had as
few as 10 holes on one bit - but they were cheap bits) and the carbide is
_really_ brittle. Unless you have a decent drill press, you _will_ snap them
off about every third hole. And then there's the etchant... Everybody starts
with Ferric Chloride. Vile, nasty stuff, stains everything, dark so you
can't see what's happening to your board, and a disposal headache. I moved
on to Cupric Chloride. (there's a good writeup in the archives I did a few
years ago about creating it from readily available sources if you _really_
want to go down that road)

Eventually I worked on some projects for other people and had "real" boards
made at their expense. (university is a great place for this...) Wow, what a
difference. It turns out that "green stuff" you see on circuit boards
actually has a purpose! Who knew! And now after spending a day laying out a
board, you just send off the files and a few dollars and you get boards in a
few days, awesome! I didn't have to spend an hour trying to get my printer
to feed a page out of some magazine, and trying to get the two layers to
line up on opposite sides of the board and another hour breathing etchant
fumes and then lacquer thinner to remove the toner, and two hours squinting
at the drill press. Wow, life is easy!!

Now I have a real job. And though I've never seen it on a job description,
your first responsibility is to make sure that you use your
time effectively. Just today I raked a box of odd bits off my desk and
tossed it in the trash. Components, hardware, everything, at least $20 worth
of usable tidbits - in the can. Not very long ago I would have cried at the
waste. "But I could'a used that!!", now I realize that it's $20 worth of
stuff and I'd spend 4 hours trying to sort it out. $20 does not buy 4 hours
of my time.

It's been a long road, and I seriously doubt that I'm the only one to have
traveled it. I've learned _immensely_ from the journey and I wouldn't change
it if I could. But I don't want to repeat it either :-)

-Denn

2011\05\24@235604 by V G

picon face
On Tue, May 24, 2011 at 10:48 PM, Denny Esterline <EraseMEdesterlinespamspamspamBeGonegmail.com>wrote:

>
> [..] and a few dollars and you get boards in a
> few days, awesome! I didn't have to spend an hour trying to get my printer
> to feed a page out of some magazine, and trying to get the two layers to
> line up on opposite sides of the board and another hour breathing etchant
> fumes and then lacquer thinner to remove the toner, and two hours squinting
> at the drill press. Wow, life is easy!!
>

Sounds like a fairy tale :)

Except for the "few dollars" part. More like $100 including shipping and
everything. Not to mention the wait time.

Also, ferric chloride doesn't emit too much nasty vapour. If you're worried
about that, use a dollar store lunch box and etch under the kitchen stove
"fume hood". Also, as someone else said, label backing paper is the way to
go. I agree. As for alignment, what I do is print both layers with markings
where I drill holes. Then line up the holes with a BBQ skewer or something.
Worked perfectly. Took about two seconds. For me the DIP carbide bits didn't
break at all, and I drilled at low speed with a Dremel by hand. It took less
than a second per hole. As for removing the toner, I used 70% isopropyl
alcohol under the kitchen fume hood and it worked like a charm. Steel wool
was also used. And TWO hours under the drill press? Seriously!? The only
holes I needed to drill were DIP through holes and smaller ones where the
vias went. The DIP holes were hollow in the center and the bit aligned
itself instantly. Very quick to drill. Drilling took like 5 minutes for a
decent sized development board. The via holes were a little tougher if I
wanted small vias, but still very easy. I do not agree with your hour
numbers. I don't understand why it took you so much time

2011\05\25@000200 by V G

picon face
On Tue, May 24, 2011 at 11:55 PM, V G <RemoveMEx.solarwind.xKILLspamspamgmail.com> wrote:

{Quote hidden}

Of course, if EE is your career, than the professional PCB route is
completely justified. If you have to make a board for a customer, and you're
getting paid for the entire process, then I agree, it's cool. All I'm saying
is, if I'm not getting paid for what I'm making, then I cant justify it to
myself to get a board made when I can do it myself in under 2 hours - start
to finish. Unless, of course, if I were to use very fine pitch components,
or something requiring 4 layers..

2011\05\25@024857 by Ruben Jönsson

flavicon
face
> Took many of your suggestions and updated the board:
>
> Unrouted: http://postimage.org/image/1j9m4mphg/full/
> Routed: http://postimage.org/image/1j9kh3czo/full/
>
> MOSFETS and two ICs are now under UBW32 (on bottom layer) to save board
> space. Mounting holes added.
>
> Can anyone think of anything else?

If you really want to save board space you could shave off quite a lot around the edges. The board edge doesn't have to go outside the boundary for the connectors (J1, J3 and J5).

If U5 and U3(?) are RJ45 connectors, the lower board edge can even go under the connector boundary. It might even be a good thing if you want these connectors mounted in a box and the connector edge flush even with the outside surface of the box.

The board can also be made more narrow. Put the LEDs closer together and move components closest to the left and right board edge longer in on the board (U2, J1, U11, U7...).

The board always (to me anyway) looks more dense in the cad than the finished board will do. When you look at the finished board you will see what I mean :-)

/Ruben
==============================
Ruben Jönsson
AB Liros Electronic
Box 9124, 200 39 Malmö, Sweden
TEL INT +46 40142078
FAX INT +46 40947388
spamBeGonerubenSTOPspamspamEraseMEpp.sbbs.se
==============================

2011\05\25@035152 by Geo

picon face
V G wrote:
<snip>
> Can anyone think of anything else?

You did not look at rotating the LEDs 180 degrees so that their anodes were closer to the drivers and the cathodes to the resistors - would have saved some ugly routing.
I would have made the power and ground tracks 50 thou.

George Smit

2011\05\25@042054 by Oli Glaser

flavicon
face
On 25/05/2011 07:48, Ruben Jönsson wrote:
> The board always (to me anyway) looks more dense in the cad than the finished
> board will do. When you look at the finished board you will see what I mean:-)

I find this to be true also.
If your CAD tool has a 3D view then that can be useful for giving you a better idea of the finished board, especially with things like connectors.
I would check the RJ45s will be flush with the board edge (the footprint suggest they are not) otherwise you will have problems disconnecting the cable from them (as you won't have access to the tab)
Always try and make sure all the connections and similar stuff will be accessible, taking into account the dimensions of whatever has to plug into them (e.g. vertical headers - will the plug interfere with adjacent components?)

2011\05\25@064012 by Gordon

flavicon
face
Hi VG,

On Tue, May 24, 2011 at 1:08 AM, V G <KILLspamx.solarwind.xspamBeGonespamgmail.com> wrote:

> Unrouted: http://postimage.org/image/22skvifk4/full/
> Routed: http://postimage.org/image/22sj7z32c/full/

I'd consider finding somewhere else to host images, postimage is
blocked at my workplace as I think others use it to put rude *blush*
photos on there.

You could upload to google docs and share a link ( I think ).

Ta,
Gordo

2011\05\25@070610 by Herbert Graf

picon face
On Wed, 2011-05-25 at 11:40 +0100, Gordon wrote:
> Hi VG,
>
> On Tue, May 24, 2011 at 1:08 AM, V G <EraseMEx.solarwind.xspamEraseMEgmail.com> wrote:
>
> > Unrouted: http://postimage.org/image/22skvifk4/full/
> > Routed: http://postimage.org/image/22sj7z32c/full/
>
> I'd consider finding somewhere else to host images, postimage is
> blocked at my workplace as I think others use it to put rude *blush*
> photos on there.

Most image locker type sites are blocked on most work networks.

I'd recommend posting an image to the list if at all possible.

I usually just skip over posts with image locker links since I know I
won't be able to view the image.

TTYL

2011\05\25@072716 by Michael Watterson

face picon face
On 25/05/2011 12:06, Herbert Graf wrote:
> On Wed, 2011-05-25 at 11:40 +0100, Gordon wrote:
>> Hi VG,
>>
>> On Tue, May 24, 2011 at 1:08 AM, V G<@spam@x.solarwind.x@spam@spamspam_OUTgmail.com>  wrote:
>>
>>> Unrouted: http://postimage.org/image/22skvifk4/full/
>>> Routed: http://postimage.org/image/22sj7z32c/full/
>> I'd consider finding somewhere else to host images, postimage is
>> blocked at my workplace as I think others use it to put rude *blush*
>> photos on there.
> Most image locker type sites are blocked on most work networks.
>
> I'd recommend posting an image to the list if at all possible.
>
> I usually just skip over posts with image locker links since I know I
> won't be able to view the image.
>
> TTYL
>
Or sign up somewhere like here and write article/blog and upload suitable images. Spam, smut and advert free.
http://www.techtir.ie  you still own the copyright (unlike many other sites).

The list has a very  low limit on upload size. On techtir if you are doing nice article your upload limit is increased.

2011\05\25@073525 by Michael Watterson

face picon face
On 25/05/2011 12:06, Herbert Graf wrote:
> On Wed, 2011-05-25 at 11:40 +0100, Gordon wrote:
>> >  Hi VG,
>> >  
>> >  On Tue, May 24, 2011 at 1:08 AM, V G<spamBeGonex.solarwind.xspamKILLspamgmail.com>  wrote:
>> >  
>>> >  >  Unrouted:http://postimage.org/image/22skvifk4/full/
>>> >  >  Routed:http://postimage.org/image/22sj7z32c/full/
>> >  
>> >  I'd consider finding somewhere else to host images, postimage is
>> >  blocked at my workplace as I think others use it to put rude*blush*
>> >  photos on there.
> Most image locker type sites are blocked on most work networks.

Demo page....

http://www.techtir.ie/demo/solar1

Free, no strings other than it has to be technical content and some text to explain what it's about and Techtir is non-exclusively "publishing" it, But all other rights remain with Author.

See http://www.techtir.ie/about

2011\05\25@080611 by Olin Lathrop

face picon face
V G wrote:
> Then in which case would you suggest a FET over a BJT?

For power applications, the two main things to consider are power
dissipation and how the part needs to be driven.  FETs look more like a
resistor when on, and bipolars more like a voltage source.  A FET with
really low Rdson can often have a lower voltage drop accross it, and
therefore lower power dissipation, at the design current.  This becomes less
of a advantage as the current goes up.

Drive issues are a consideration too.  FETs are driven by a voltage, but for
large FETs it needs to swing 12 to 15 volts between on and off, and that
quickly changing voltage will see some capacitance.  This is not something a
microcontroller can do directly.  This is why there are dedicated chips just
to drive FETs.  They take logic levels in and put out the gate voltage.
They are optimized for a short but high burst of current when switching to
overcome the effective gate capacitance of the FET.  High end FET drivers
are rated for over 1A during switching.  "Logic level" FETs are designed so
that they turn on pretty well with only 3.3V or 5V base drive.  The
IRLML2502 and IRLML0030 are in this catagory.  However, these FETs are
limited to relatively low voltages.  The two I mentioned above are rated for
20V and 30V, respectively.  These are desirable characteristics, so also
cost a premium.

Bipolars are driven with current, not voltage.  That means that a micro
running at 5V or 3.3V supply can drive a bipolar gate, up to some current
limit.  You can switch 100V from a micro easily with just a NPN transistor
and a base resistor.  Drawbacks here are that the base current needs to be
on continuously to keep the transistor on, and that the output current is
limited by the gain.  Let's say the 100V NPN can be counted on to have a
gain of 50.  That means with 10mA base drive is can only switch 500mA.  If
that's all you need then that's not bad.  If you need 5A, then it will take
extra circuitry to provide the larger base current.

Bipolar transistors are also generally cheaper than than FETs with similar
voltage and current capabilities.  Sometimes that's a important
consideration.  They also tend to be a little more robust against the kinds
of transients and the like things sometimes get exposed to despite best
intentions.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\25@080725 by Gerhard Fiedler

picon face
Ruben Jönsson wrote:

> The board always (to me anyway) looks more dense in the cad than the
> finished board will do. When you look at the finished board you will
> see what I mean :-)

It helps to always have a few actual boards nearby when routing,
comparing areas with similar density between the screen and reality.
Helps you keep the perspective.

Gerhard

2011\05\25@082646 by Olin Lathrop

face picon face
Michael Watterson wrote:
> Or sign up somewhere like here and write article/blog and upload
> suitable images. Spam, smut and advert free.
> http://www.techtir.ie  you still own the copyright (unlike many other sites).
>
> The list has a very  low limit on upload size. On techtir if you are
> doing nice article your upload limit is increased.

Surely he has some web space for being a student at his university.  Just
copy the image file somewhere there and give us the URL.  There is no need
to create a link to it from any of the web pages.  I do a similar thing when
posting large files.  I even have a TEMP directory at the top of my WWW tree
just for dumping temporary files.  I've used that here many times and it
works fine.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\25@083944 by RussellMc

face picon face
> This is why there are dedicated chips just
> to drive FETs.  They take logic levels in and put out the gate voltage.
> They are optimized for a short but high burst of current when switching to
> overcome the effective gate capacitance of the FET.  High end FET drivers
> are rated for over 1A during switching.

A FET gate driver good enough for many applications where voltage is
OK but current is lacking can be made with 2 jelly-bean bipolars and
little  else. (Any 'glue' parts needed would be needed regardless of
driver used)

A better driver that also allows whatever voltage drive is required
can be made with 2 bipolars and a little extra glue and a better again
version with 3 transistors plus glue.
When you get to this sort of drive requirement a bipolar will probably
also need a driver if using a uP but a different circuit.

Circuits available if wanted - information access and time both
unavailable at this moment.

Simplest FET current driver.

PNP = P  Pe = emitter etc
NPN = N

Pc - V+
Pe - Ne
Pb - Nb
PIC drive to Pb + Nb
Drive output from Pe + Ne.
Rseries series R from drive output to FET gate to limit current.
Reverse Zener -  FET gate to ground to limit Miller effect gate
spiking. Not usually optional*.
Optional reverse biased Schottky FET gate to ground mounted as close
to FET as possible to kill parasitic gate ringing.

Commoned-emitter, commoned-base arrangement (rare) tends to stop shoot
through which appears likely at 1st glance.

BC327 / BC337 or SMD equivalent pair good for 1A gate drive provided
adequate Beta jellybeans used.
Using highest Beta BC337 (BC337-400) of 250 guaranteed needs ~4 mA
base drive for 1A FET drive.
Rseries adjusted to suit. Can be SC / 0R but provides super fast gate
drive and more EMI and switching losses.
Actual drive power low as gate current on;y drawn to charge add
discharge FET gate capacitance.
Power ~~~~~= 2 x 0.5 x C_gate_FET x Vgate^2 x frequency = low
(0.5 CV^2 with two edges per cycle).

This stunningly simple circuit works very well. There is drive voltage
loss due to emitter follower drive so only FETs with Vgs-drive well
inside Vcc can be driven.

Adding 1 more bipolar will allow any desired gate drive voltage.
Anon.



          Russell

2011\05\25@084622 by RussellMc

face picon face
Pictures (JPG work, a few other formats also)

Can be emailed to

  RussellMc+.....PICDTspam_OUTspamfotki.com

and will turn up at

            http://public.fotki.com/RussellMc/misc-1/piclistdesktops/

within a minute or two.

email subject line becomes photo title.

N images per email appear with same title.

LARGE images OK (up to 10MB, maybe more).

Sample (Wouter's desk)

   http://public.fotki.com/RussellMc/misc-1/piclistdesktops/img8514.html


Advise when/if deletion wanted.


               Russell

On 26 May 2011 00:27, Olin Lathrop <TakeThisOuTolin_piclist.....spamTakeThisOuTembedinc.com> wrote:
{Quote hidden}

>

2011\05\25@085113 by Michael Watterson

face picon face
On 25/05/2011 13:27, Olin Lathrop wrote:
> Surely he has some web space for being a student at his university.  Just
> copy the image file somewhere there and give us the URL.  There is no need
> to create a link to it from any of the web pages.  I do a similar thing when
> posting large files.  I even have a TEMP directory at the top of my WWW tree
> just for dumping temporary files.  I've used that here many times and it
> works fine.

Sometimes that's not publicly accessible, or can't be uploaded to outside or gets randomly deleted.

Certainly the local students here avoid the University server.

2011\05\25@085705 by Michael Watterson

face picon face
On 25/05/2011 13:39, RussellMc wrote:
> Adding 1 more bipolar will allow any desired gate drive voltage.

Sometimes you need a small buck inverter to get the Gate supply voltage rail. Esp off 3.3V or 5V when driving FETs for very low on, or part of a high power or high voltage buck.

LT1073 (1V to 5V supply) + coil, diode & cap. (or a single LM393 + low voltage FET with 2V to 5V supply) will give up to +50V easily

2011\05\25@092254 by Oli Glaser

flavicon
face
On 25/05/2011 13:56, Michael Watterson wrote:
> On 25/05/2011 13:39, RussellMc wrote:
>> >  Adding 1 more bipolar will allow any desired gate drive voltage.
> Sometimes you need a small buck inverter to get the Gate supply voltage
> rail. Esp off 3.3V or 5V when driving FETs for very low on, or part of a
> high power or high voltage buck.

Buck inverter? Don't you mean a boost regulator?

> LT1073 (1V to 5V supply) + coil, diode&  cap. (or a single LM393 + low
> voltage FET with 2V to 5V supply) will give up to +50V easily.

2011\05\25@101946 by Herbert Graf

picon face
On Wed, 2011-05-25 at 12:27 +0100, Michael Watterson wrote:
> Or sign up somewhere like here and write article/blog and upload
> suitable images. Spam, smut and advert free.
> http://www.techtir.ie  you still own the copyright (unlike many other sites).
>
> The list has a very  low limit on upload size. On techtir if you are
> doing nice article your upload limit is increased.

The list's limit is 40k. That is FAR more the is necessary for pretty
much all purposes here.

People just don't understand that a 2MB JPG of a schematic isn't the
right way of doing things. Use GIF/PNG for line art type stuff. As for
photos, 40k of jpg at a reasonable res is more then enough IMHO.

What kind of image do you think would require more then 40k?

TTYL

2011\05\25@105116 by Michael Watterson

face picon face
On 25/05/2011 14:22, Oli Glaser wrote:
> Buck inverter? Don't you mean a boost regulator?

Duh..
Yes of course


Not well today. Either terminal hayfever or a cold..
Sniff..

2011\05\25@105423 by Michael Watterson

face picon face
On 25/05/2011 15:20, Herbert Graf wrote:
> What kind of image do you think would require more then 40k?

ones that are more than 40k :-)

Is it a trick question?

2011\05\25@111244 by Oli Glaser

flavicon
face
On 25/05/2011 15:51, Michael Watterson wrote:
> On 25/05/2011 14:22, Oli Glaser wrote:
>> Buck inverter? Don't you mean a boost regulator?
> Duh..
> Yes of course
>
>
> Not well today. Either terminal hayfever or a cold..
> Sniff..

Oh dear, no fun...
Terminal hayfever sounds interesting - is that a really nasty version, or an allergy to certain viewing apparatus?
Either way, a plain old cold sounds preferable.. :-)

2011\05\25@133807 by Herbert Graf

picon face
On Wed, 2011-05-25 at 15:54 +0100, Michael Watterson wrote:
> On 25/05/2011 15:20, Herbert Graf wrote:
> > What kind of image do you think would require more then 40k?
>
> ones that are more than 40k :-)
>
> Is it a trick question?

No.

IMHO the majority of the kinds of images I've seen referenced in a
message to the piclist was either already <40k, or could have easily
served it's purpose having been reduced to <40k.

For the odd case where more the 40k is NEEDED then yes, other options
can be used, many of us probably won't see them though.

TTYL

2011\05\25@141506 by Olin Lathrop

face picon face
Michael Watterson wrote:
>> What kind of image do you think would require more then 40k?
>
> ones that are more than 40k :-)

Right, but the point is there is little excuse for needing that much.  If
it's a photograph, then it should be JPG.  You can show a lot of stuff with
JPG compression in 40kbytes.  If it's line art, then it should compress very
well with LZW or even runlength compression.  A GIF file would be a good
choice, especially if you limit the image to black and white or gray scale.
40kbytes is huge for a black and white line drawing properly compressed.

The problem is that all too often people use inappropriate image formats and
compression.  Note how often we see people sending Eagle screen shots as
JPEG files, for example.  Think of the 40kbyte limit as a moron filter.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\25@161241 by V G

picon face
On Wed, May 25, 2011 at 3:51 AM, Geo <TakeThisOuTbuggiesmithKILLspamspamspamgmail.com> wrote:

> You did not look at rotating the LEDs 180 degrees so that their anodes
> were closer to the drivers and the cathodes to the resistors - would
> have saved some ugly routing.
>

Done.


> I would have made the power and ground tracks 50 thou.
>

Couldn't afford to make them 0.050 inches. The default net class is 0.013"
width, 0.013" clearance. I made the GND net 0.020", 0.015". There's not much
current flowing through them anyway. The LEDs are the biggest current drains
on there. Also, the drains on the MOSFETs, so I'll make those a bit bigger
too.

Thanks for the suggestions

2011\05\25@161830 by V G

picon face
On Wed, May 25, 2011 at 4:20 AM, Oli Glaser <.....oli.glaserspamRemoveMEtalktalk.net> wrote:

>  I find this to be true also.
> If your CAD tool has a 3D view then that can be useful for giving you a
> better idea of the finished board, especially with things like connectors..
> I would check the RJ45s will be flush with the board edge (the footprint
> suggest they are not) otherwise you will have problems disconnecting the
> cable from them (as you won't have access to the tab)
> Always try and make sure all the connections and similar stuff will be
> accessible, taking into account the dimensions of whatever has to plug
> into them (e.g. vertical headers - will the plug interfere with adjacent
> components?)


Thanks for the suggestions. I moved the RJ45 connectors

2011\05\25@161935 by V G

picon face
On Wed, May 25, 2011 at 8:27 AM, Olin Lathrop <RemoveMEolin_piclistspamspamBeGoneembedinc.com>wrote:

>  Surely he has some web space for being a student at his university.
>

Yeah, but it's no good. It's slow, hard to access, and contains information,
which I would rather not disclose to the public

2011\05\25@162239 by V G

picon face
On Wed, May 25, 2011 at 2:15 PM, Olin Lathrop <spamBeGoneolin_piclist@spam@spamspam_OUTembedinc.com>wrote:

>  Right, but the point is there is little excuse for needing that much.  If
> it's a photograph, then it should be JPG.  You can show a lot of stuff with
> JPG compression in 40kbytes.  If it's line art, then it should compress
> very
> well with LZW or even runlength compression.  A GIF file would be a good
> choice, especially if you limit the image to black and white or gray scale.
> 40kbytes is huge for a black and white line drawing properly compressed.
>
> The problem is that all too often people use inappropriate image formats
> and
> compression.  Note how often we see people sending Eagle screen shots as
> JPEG files, for example.  Think of the 40kbyte limit as a moron filter.
>

I don't agree with this. 40k is an arbitrary limit not based on any actual
statistical or scientific research. I compressed my "line art" pictures with
PNG and GIF (I really hate GIF), and both times, it was slightly over 40k
(about 60-80k). Yes, I could make it black and white, but I would be losing
a lot of information, like layer colours and so on. It shouldn't take this
much time and effort to display an image. This is 2011

2011\05\25@162324 by Lyle Hazelwood

picon face
On Wed, May 25, 2011 at 2:15 PM, Olin Lathrop <TakeThisOuTolin_piclistspamspamembedinc.com> wrote:
> You can show a lot of stuff with
> JPG compression in 40kbytes.

So, you're saying a picture is worth 20480 words?
:)

Lyl

2011\05\25@162613 by V G

picon face
On Wed, May 25, 2011 at 8:06 AM, Olin Lathrop <olin_piclistEraseMEspamembedinc.com>wrote:

> V G wrote:
> > Then in which case would you suggest a FET over a BJT?
>
> For power applications, the two main things to consider are power
> dissipation and how the part needs to be driven.  FETs look more like a
> resistor when on, and bipolars more like a voltage source.


I've seen you say this before, but I don't understand this. Could you please
explain this? Aren't they both like base/gate controlled variable resistors?


> A FET with
> really low Rdson can often have a lower voltage drop accross it, and
> therefore lower power dissipation, at the design current.  This becomes
> less
> of a advantage as the current goes up.
>

Can you explain that last sentence? Of course, as the current goes up, the
FET will dissipate more power, but will it dissipate more than a competitive
BJT (is that what you're saying when you say "less of an advantage")?


> Drive issues are a consideration too.  FETs are driven by a voltage, but
> for
> large FETs it needs to swing 12 to 15 volts between on and off, and that
> quickly changing voltage will see some capacitance.


What effect will the capacitance have

2011\05\25@165123 by IVP

face picon face
> (I really hate GIF)

Why's that

2011\05\25@165509 by Forrest W Christian

flavicon
face
On 5/25/2011 2:25 PM, V G wrote:
> On Wed, May 25, 2011 at 8:06 AM, Olin Lathrop<RemoveMEolin_piclistEraseMEspamspam_OUTembedinc.com>wrote:
>
>> For power applications, the two main things to consider are power
>> dissipation and how the part needs to be driven.  FETs look more like a
>> resistor when on, and bipolars more like a voltage source.
>
> I've seen you say this before, but I don't understand this. Could you please
> explain this? Aren't they both like base/gate controlled variable resistors?

When fully on, a FET has a resistance - RdsON.

When fully on, a Bipolar has a voltage drop - Vce(sat)

Maybe better put - a FET is like a resistor, a Bipolar is like a diode.

-forres

2011\05\25@165524 by Dwayne Reid

flavicon
face
At 02:22 PM 5/25/2011, V G wrote:
> >
> > Note how often we see people sending Eagle screen shots as
> > JPEG files, for example.  Think of the 40kbyte limit as a moron filter.
> >
>
>I don't agree with this. 40k is an arbitrary limit not based on any actual
>statistical or scientific research. I compressed my "line art" pictures with
>PNG and GIF (I really hate GIF), and both times, it was slightly over 40k
>(about 60-80k). Yes, I could make it black and white, but I would be losing
>a lot of information, like layer colours and so on. It shouldn't take this
>much time and effort to display an image. This is 2011.

Yeah - it *IS* 2011.  That doesn't mean that you shouldn't try to minimize bandwidth whenever possible.

MIT pays for the bandwidth that this group uses.  Sending huge image files to thousands of people for no need just is not being a responsible person.

Most GIF schematics and drawings are anywhere from 3k to 10k in size.  That's entirely reasonable.

If you need help in generating GIF files, just holler on the list.  There are lots of people who do this all the time (Joe - IVP, for example).

dwayne

-- Dwayne Reid   <@spam@dwaynerRemoveMEspamEraseMEplanet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
(780) 489-3199 voice          (780) 487-6397 fax
http://www.trinity-electronics.com
Custom Electronics Design and Manufacturing

2011\05\25@170749 by IVP

face picon face
> If you need help in generating GIF files, just holler on the
> list.  There are lots of people who do this all the time (Joe - IVP,
> for example).
>
> dwayne

Thanks for noticing Dwayne

My general preferences are - jpg (palette-reduced if not critical) for
pictures, b/w gif for schmatics, 16-colour gif for boards/overlays/etc

Being a prennial dial-up user I'm always conscious of transfer times,
and remind my relatives (all too often) not to send pictures straight
out of the flipping camera ;-) but use the utility I gave them

AFAIK IrfanView is still fre

2011\05\25@171443 by Carey Fisher

face picon face
>
> What kind of image do you think would require more then 40k?
>
> A picture of J Lo

2011\05\25@172453 by V G

picon face
On Wed, May 25, 2011 at 4:55 PM, Forrest W Christian <EraseMEforrestcspam@spam@imach.com>wrote:

> When fully on, a FET has a resistance - RdsON.
>

So the gate controls the resistance?


> When fully on, a Bipolar has a voltage drop - Vce(sat)
>

So what does the base control exactly? If the current through the base
controls the current through the collector, and I = V/R, is it also
controlling the resistance? Or is it controlling the voltage drop?


> Maybe better put - a FET is like a resistor, a Bipolar is like a diode.

2011\05\25@173725 by V G

picon face
On Wed, May 25, 2011 at 5:07 PM, IVP <@spam@joecolquittspam_OUTspam.....clear.net.nz> wrote:

> Being a prennial dial-up user I'm always conscious of transfer times,
> and remind my relatives (all too often) not to send pictures straight
> out of the flipping camera ;-) but use the utility I gave them
>
> AFAIK IrfanView is still free
>
>
Why do you use dial up

2011\05\25@174306 by Oli Glaser

flavicon
face
On 25/05/2011 13:06, Olin Lathrop wrote:
> and bipolars more like a voltage source.

Do you mean current source here?

2011\05\25@175011 by Gary Crowell

picon face
I've gotten several boards from pcbgeek, and it's worked very well for me.

Gary

On Tue, May 24, 2011 at 6:34 PM, V G <spamBeGonex.solarwind.xEraseMEspamgmail.com> wrote:

{Quote hidden}

>

2011\05\25@175325 by IVP

face picon face
> Why do you use dial up?

Because that's all I need

And so-called broadband in NZ is generally an over-priced rip-off. I
have plenty of friends whose speed is often barely double dial-up, at
10x the price I pa

2011\05\25@182359 by V G

picon face
On Wed, May 25, 2011 at 5:50 PM, Gary Crowell <RemoveMEgaryacrowellsr@spam@spamspamBeGonegmail.com>wrote:

> I've gotten several boards from pcbgeek, and it's worked very well for me..
>
>
Which country do you live in? How long did it take from the time you sent in
the design to the time the PCBs arrived

2011\05\25@182643 by V G

picon face
On Wed, May 25, 2011 at 5:52 PM, IVP <.....joecolquitt@spam@spamEraseMEclear.net.nz> wrote:

> Because that's all I need
>
> And so-called broadband in NZ is generally an over-priced rip-off. I
> have plenty of friends whose speed is often barely double dial-up, at
> 10x the price I pay
>
>
Just checked NZ prices. Damn. That's pretty bad. I would suffocate in NZ

2011\05\25@183701 by Gerhard Fiedler

picon face
V G wrote:

> On Wed, May 25, 2011 at 4:55 PM, Forrest W Christian <.....forrestcRemoveMEspamimach.com>wrote:
>
>> When fully on, a FET has a resistance - RdsON.
>
> So the gate controls the resistance?

Basically, yes.

>> When fully on, a Bipolar has a voltage drop - Vce(sat)
>
> So what does the base control exactly? If the current through the
> base controls the current through the collector, and I = V/R, is it
> also controlling the resistance?
I is only V/R for a linear R, or within a linear region of a non-linear
R. The C-E path of a bipolar transistor that is used as a switch and is
"on" is not usually in a linear region, differently from a FET.
The Vce(sat) voltage increases with current, but not linearly, whereas
the drain-source voltage of a FET that's "on" increases linearly (in
first approximation) with the current. Hence the specification of an
RdsON for switching FETS and Vce(sat) for switching bipolars.
Gerhar

2011\05\25@184517 by Gary Crowell

picon face
I'm in the US.  I think each order has been right at 10 days total, except
for one occasion when I sent it right before a week-long Chinese holiday.  I
think there is a holiday calendar on the pcbcore web site.  I've also had
good luck with pcbcore (but their web site doesn't seem to be working for me
at the moment).

Gary

On Wed, May 25, 2011 at 4:23 PM, V G <.....x.solarwind.xSTOPspamspam@spam@gmail.com> wrote:

> On Wed, May 25, 2011 at 5:50 PM, Gary Crowell <garyacrowellsrEraseMEspam@spam@gmail.com
> >wrote:
>
> > I've gotten several boards from pcbgeek, and it's worked very well for
> me.
> >
> >
> Which country do you live in? How long did it take from the time you sent
> in
> the design to the time the PCBs arrived?
> -

2011\05\25@184955 by Gerhard Fiedler

picon face
V G wrote:

> On Wed, May 25, 2011 at 2:15 PM, Olin Lathrop wrote:
>
>> Right, but the point is there is little excuse for needing that much.
>>  If it's a photograph, then it should be JPG.  You can show a lot of
>> stuff with JPG compression in 40kbytes.  If it's line art, then it
>> should compress very well with LZW or even runlength compression.  A
>> GIF file would be a good choice, especially if you limit the image
>> to black and white or gray scale. 40kbytes is huge for a black and
>> white line drawing properly compressed.
>>
>> The problem is that all too often people use inappropriate image
>> formats and compression.  Note how often we see people sending Eagle
>> screen shots as JPEG files, for example.  Think of the 40kbyte limit
>> as a moron filter.
>
> I don't agree with this. 40k is an arbitrary limit not based on any
> actual statistical or scientific research. I compressed my "line art"
> pictures with PNG and GIF (I really hate GIF), and both times, it was
> slightly over 40k (about 60-80k).
Statistical/scientific or not: it's enough for most purposes, including
this one. Just for fun, I cropped your routed picture to the board (as
PNG, just like the original): 34.9 kB.
> This is 2011.

Yup, and the basics are still the same :)

Gerhar

2011\05\25@185835 by V G

picon face
On Wed, May 25, 2011 at 6:36 PM, Gerhard Fiedler <RemoveMElistsspamspamBeGoneconnectionbrazil.com
> wrote:

> I is only V/R for a linear R, or within a linear region of a non-linear
> R. The C-E path of a bipolar transistor that is used as a switch and is
> "on" is not usually in a linear region, differently from a FET.


>From what I understand (I could be wrong), I = V/R for any point, as in,
when a voltage V is applied across a resistance R, then the current flowing
through R is equal to I = V/R. Are you saying the internals of a BJT violate
ohm's "law"? I mean, the violations of the law have been discussed on this
list before, if I can remember correctly.

I'm just not understanding the part about the linear region

2011\05\25@190414 by V G

picon face
On Wed, May 25, 2011 at 6:45 PM, Gary Crowell <spamBeGonegaryacrowellsrKILLspamspam@spam@gmail.com>wrote:

> I'm in the US.  I think each order has been right at 10 days total, except
> for one occasion when I sent it right before a week-long Chinese holiday.
>  I
> think there is a holiday calendar on the pcbcore web site.  I've also had
> good luck with pcbcore (but their web site doesn't seem to be working for
> me
> at the moment).
>

Thanks!

The website seems to be working: http://www.isup.me/pcbcore.com. But even
these guys seem to be way more expensive than PCBGeek

2011\05\25@191844 by Oli Glaser

flavicon
face
On 25/05/2011 22:24, V G wrote:
> So what does the base control exactly? If the current through the base
> controls the current through the collector, and I = V/R, is it also
> controlling the resistance?

The base current does control the C-E current , but it acts more like a current source as opposed to a resistance (over a region of compliance)
To get an idea, you could run a test in LTSpice - bias an NPN transistor with say 1uA into the base (just use a current source), emitter to ground and vary a voltage from e.g. 1 to 10 volts on the collector.
Do the same with a (say 100 or 1K) resistor (from voltage to ground) and note the difference in current behaviour.
Use the Pulse setting for the voltage source and select a slow rise/fall time, then monitor the results with transient simulation.
You can press add trace and use the d() function to check the dynamic impedance - enter e.g. d(V(V+)/d(Ic(Q1)) assuming your voltage source is called V+ and NPN is Q1.

2011\05\25@192351 by William \Chops\ Westfield

face picon face
>>
>> You can show a lot of stuff with JPG compression in 40kbytes.

27kbytes ?  Or is the limit NOT on the asciified encoding of the  picture?
27kbytes seems to be pretty small, even for grayscale PNG or GIF files.

BillW

2011\05\26@025126 by Ruben Jönsson

flavicon
face
>
> Couldn't afford to make them 0.050 inches. The default net class is 0.013"
> width, 0.013" clearance. I made the GND net 0.020", 0.015". There's not much
> current flowing through them anyway. The LEDs are the biggest current drains
> on there. Also, the drains on the MOSFETs, so I'll make those a bit bigger
> too.
>
With a digital design (like much of this one) you also need to take into account the switching currents of transistors inside the digital chips (MCU, FPGA, 74 logic, drivers...). Especially for chips which are working with clock signals (changes state with the clock) such as an MCU. These chips have a lot of transistors switching every clock and it is when they are switching that they draw current. This means that you have a current spike for every clock transition.
If you have long and thin power/ground tracks for these chips, the tracks will look like a relative large resistance between the current supply and the chip. This resistance varies with frequency which means that at certain frequencies you will have a higher resistance and at other frequencies you will have lower resistance.

Whenever you have a resistance in a current path, you also have a voltage. The higher this voltage becomes, the more trouble it can cause. One of the trouble it can cause is to make the trace an unintentional RF transmitter. Another is to make a difference in the ground voltage (or any reference voltage) between two chips on the same board which could cause false transitions of logical inputs.

Remember that the switching frequencies (clock) can be quite high and since this mostly is a square wave it contains a lot of overtones with much higher frequencies and it is the higher frequencies that can cause troble here (remember that the resistance is varying with frequency and usually is much higher at higher frequencies).

In my experience, when a circuit transmitts RF unintentionally (emitting diturbances), it also works as an RF receiver (has low immunity against RF disturbances). The received RF (either conductively via connected wires or over the air) is then converted to voltages in the tracks on the board which can make the board malfunction. Such an RF transmitter can be a relay switching a load.
How do we handle these problems then? The first thing is to make a groundplane which ensures that you have very low resistance over the entire frequency range between the current sources and the chips and between the chips themselves. This way the ground level is the same for all chips on the board and all chips sees the digital signals connected between them at the level it was supposed to be (the same for the transmitter as for the receiver), which means that there will be no false transitions.

The other thing is to make sure that the ICs which need a lot of current when its internal transistors are switching, has this current available close to the power and ground pins, with a low resistance for the entire frequency range.. This is done by making the power tracks as wide as possible and by using decoupling capacitors. The decoupling capacitor acts as a tiny current reservoir, which the chip can draw current from in very short bursts when it needs it. If the decoupling capacitor is placed close to the power and ground pins and the tracks between them are wide enough, the high switching currents are limited to those short tracks. Since not all current can be drawn from the decoupling capacitors and since they need to be recharged between the current bursts, you also need wide tracks between the power supply and the decoupling capacitor.

Also remember that when a chip draws a lot of current on the power pins, the same amount of current is going through the ground pins. This is why we need to have the decoupling capacitor as close to the power and ground pins on the same chip as possible (a ground plane helps here). Otherwise you will have a long way for the high current to travel on the board. This long way is called a current loop which, the longer it is the more trouble it can casue. In other words, keep the current loops as short as possible.

In short (which did become much longer than I intended), this is why you need big fat tracks for the power supply traces on your board. It is also the reason you need a good decoupling for the digital chips.

This becomes more important when you make boards professionally since these have to comply with certain EMC (ElectroMagnetic Compatibility) rules. These rules says that your electronics may not emit RF energy over a certain level at certain frequency ranges (both conductively and over air) and it must also be immune to a certain level of  RF energy thrown at it at certain frequency ranges. This is really also important when making board unprofessionally. Many times you hear that something dosen't work and noise or RF disturbance is blamed (a CPU gets reset for example). Most of the time I think that it is more due to bad board design casuing too low immunity thresholds for some frequencies than a very high disturbing signal.

/Ruben

==============================
Ruben Jönsson
AB Liros Electronic
Box 9124, 200 39 Malmö, Sweden
TEL INT +46 40142078
FAX INT +46 40947388
rubenspam_OUTspam@spam@pp.sbbs.se
==============================

2011\05\26@030817 by Ruben Jönsson

flavicon
face
> > Drive issues are a consideration too.  FETs are driven by a voltage, but
> > for
> > large FETs it needs to swing 12 to 15 volts between on and off, and that
> > quickly changing voltage will see some capacitance.
>
>
> What effect will the capacitance have?

This matters when the FET is used in a switching application, where you want the FET to be either completely on or completely off as much of the time as possible. This is because when the FET is half on, much of the load voltage appears across the FET instead of over the load, which causes a lot of power dissipation in the FET.

So to prevent this you need to switch it on or off as fast as possible. This becomes harder to do if the gate driver (a PIC output pin for example) sees a lot of capacitance. You actually need to charge and discharge this capacitance each time the FET is switched on or of. Since the driver has limitations in how much current it can deliver (it has an internal resistance which is more than 0 ohms) the switching on and off takes some time and during this time the FET is somwhere inbetween fully on and fully of and it is dissipating power (its Drain Source resistance is changing from infinitely high to very low). The bigger the capacitance is, the more current is needed to switch the FET on or off in the same time.

This really only matters when the FET is coninously switched on and of at a high frequency such as in PWM applications or switching power supplies. When switching relays on and of at a low frequency, the gate capacitance is of less importance.

/Ruben



> -

2011\05\26@043933 by V G

picon face
2011/5/26 Ruben Jönsson <spamBeGoneruben@spam@spampp.sbbs.se>

{Quote hidden}

Wow, okay. Thicker power/ground tracks it is. What thickness/clearance do
you recommend

2011\05\26@044000 by V G

picon face
2011/5/26 Ruben Jönsson <RemoveMErubenEraseMEspamKILLspampp.sbbs.se>

{Quote hidden}

Very well written, explains a lot! Thanks

2011\05\26@045253 by Michael Watterson

face picon face
On 25/05/2011 22:42, Oli Glaser wrote:
> On 25/05/2011 13:06, Olin Lathrop wrote:
>> and bipolars more like a voltage source.
> Do you mean current source here?
>

In saturation it looks like a voltage source, i.e. like a 0.2V or 0.15 Diode drop.
Out of saturation (>0.2V and up to Vsupply) it's more like a constant current source programmed by the base current.
Ic = Ib x Hfe

A FET looks like a resistor when "fully on" as  the current is not flowing through any PN or NP junctions.

2011\05\26@045720 by Michael Watterson

face picon face
On 25/05/2011 22:37, V G wrote:
> On Wed, May 25, 2011 at 5:07 PM, IVP<spamBeGonejoecolquittspam_OUTspamRemoveMEclear.net.nz>  wrote:
>
>> >  Being a prennial dial-up user I'm always conscious of transfer times,
>> >  and remind my relatives (all too often) not to send pictures straight
>> >  out of the flipping camera;-)  but use the utility I gave them
>> >
>> >  AFAIK IrfanView is still free
>> >
>> >
> Why do you use dial up?

Maybe there is nothing else?

BTW "Mobile" is a fast dialup.

Lots of people can't get Broadband

2011\05\26@050516 by Michael Watterson

face picon face
On 26/05/2011 09:39, V G wrote:
> Wow, okay. Thicker power/ground tracks it is. What thickness/clearance do
> you recommend?

For ground, all the unused space

For power 0.1" if possible.

It's not just resistance but inductance and the fact a 1/4 wavelength is good transmitting aerial, hence decoupling close to chip.

if 10MHz signal has about 1nS rise time or better that could be 1GHz or higher harmonics. That's 15cm for a 1/4 wave.

The Power and ground of IC can have nano second spikes when turning on /off the FET on the relay as it's rapidly charging/discharging the FET gate capacitor. Some designs with long traces put 50 Ohm to 220 Ohm at CPU and at Gate end of trace to reduce rise time and also for the FET reduce likelyhood of self oscillation or nearby radio switching it.

2011\05\26@053429 by Oli Glaser

flavicon
face
On 26/05/2011 09:52, Michael Watterson wrote:
> On 25/05/2011 22:42, Oli Glaser wrote:
>> On 25/05/2011 13:06, Olin Lathrop wrote:
>>> and bipolars more like a voltage source.
>> Do you mean current source here?
>>
> In saturation it looks like a voltage source, i.e. like a 0.2V or 0.15
> Diode drop.
> Out of saturation (>0.2V and up to Vsupply) it's more like a constant
> current source programmed by the base current.
> Ic = Ib x Hfe
>
> A FET looks like a resistor when "fully on" as  the current is not
> flowing through any PN or NP junctions.
>

Oh right, of course - thanks. I missed the context the statement was made in.



2011\05\26@061820 by V G

picon face
On Thu, May 26, 2011 at 5:05 AM, Michael Watterson <.....mikespamRemoveMEradioway.org>wrote:

>  For ground, all the unused space
>
> For power 0.1" if possible.
>

You mean the /entire/ power net should be 0.1"?

If so, the board looks like this and violates several design rules:
http://postimage.org/image/2572oi46c/

I tried attaching, but could not get it to under 60k without it looking like
a mess. 40k is much too low a limit. Should be at least 100k or 200k

2011\05\26@062100 by V G

picon face
On Thu, May 26, 2011 at 6:18 AM, V G <x.solarwind.xspam@spam@gmail.com> wrote:

> On Thu, May 26, 2011 at 5:05 AM, Michael Watterson <EraseMEmikeRemoveMEspamSTOPspamradioway.org>wrote:
>
>>  For ground, all the unused space
>>
>> For power 0.1" if possible.
>>
>
> You mean the /entire/ power net should be 0.1"?
>
> If so, the board looks like this and violates several design rules:
> http://postimage.org/image/2572oi46c/
>
> I tried attaching, but could not get it to under 60k without it looking
> like a mess. 40k is much too low a limit. Should be at least 100k or 200k..
>

On the power and ground nets, I can go 0.03" trace, 0.02" clearance max,
before things start getting unroutable

2011\05\26@062947 by Geo

picon face
V G wrote:

> On the power and ground nets, I can go 0.03" trace, 0.02" clearance max,
> before things start getting unroutable.

You do need need any extra spacing - 8 thou is enough.
What we used to do (on mainly TTL boards) was to route the power rails first so they took up the obvious direct routes.
I don't know your CAD program but the power/ground thicker tracks should go just past an SMD pad then neck down to say 10 thou to the pad to prevent uneven soldering. With the CAD package we used this was often a manual operation.

George Smit

2011\05\26@063405 by Michael Watterson

face picon face
On 26/05/2011 11:18, V G wrote:
>> >    For ground, all the unused space
>> >
>> >  For power 0.1" if possible.
>> >
> You mean the/entire/  power net should be 0.1"?
no. That wouldn't be possible.

"if possible" You can adjust the width of any segment just with a couple of clicks.

You ALWAYS have to manually adjust.

Get some experience and learn to manually place (the most important bit) and then manually route

2011\05\26@065732 by V G

picon face
On Thu, May 26, 2011 at 6:29 AM, Geo <RemoveMEbuggiesmithKILLspamspamTakeThisOuTgmail.com> wrote:

>  You do need need any extra spacing - 8 thou is enough.
> What we used to do (on mainly TTL boards) was to route the power rails
> first so they took up the obvious direct routes.
> I don't know your CAD program but the power/ground thicker tracks should
> go just past an SMD pad then neck down to say 10 thou to the pad to
> prevent uneven soldering. With the CAD package we used this was often a
> manual operation.
>

So you're saying, instead of making the entire power net 0.05", branch off
of it to the pads with shorter traces?

And Diptrace seriously rocks in every way. I can specify net routing
priority

2011\05\26@065840 by V G

picon face
On Thu, May 26, 2011 at 6:33 AM, Michael Watterson <spamBeGonemikespam@spam@radioway.org>wrote:

>  > You mean the/entire/  power net should be 0.1"?
> no. That wouldn't be possible.
>
> "if possible" You can adjust the width of any segment just with a couple
> of clicks.
>
> You ALWAYS have to manually adjust.
>
> Get some experience and learn to manually place (the most important bit)
> and then manually route.
>
>
Makes sense. I'll keep adjusting and routing sequentially until all nets can
be routed successfully

2011\05\26@075558 by Geo

picon face
V G wrote:

> So you're saying, instead of making the entire power net 0.05", branch off
> of it to the pads with shorter traces?
>
Yes - just for a few thou.
Have a quick look at SECTION 6: TRACE ROUTING starting on page 32 of this pdf book.

Note - clicking the preview link downloads a 900k pdf file.
http://preview.tinyurl.com/3ftmo3m

George Smit

2011\05\26@081903 by Olin Lathrop

face picon face
V G wrote:
> The default net class is
> 0.013" width, 0.013" clearance.

That's quite large for ordinary signal traces without unusual constraint.  I
use 8/8 mil design rules because any board house in the world can do that
and still have a little margin.  Some nowadays allow 6/6 before they start
charging extra.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\26@082517 by Olin Lathrop

face picon face
V G wrote:
> I don't agree with this. 40k is an arbitrary limit not based on any
> actual statistical or scientific research.

I think it's more driven by how much bandwidth they are willing to let the
server use in distributing a single list message.  40kbytes x 2000
subscribers = 80Mbytes.  Some of those will be grouped, but it's still a lot
to ask someone else to send for free as a result of a single message.

> I compressed my "line art"
> pictures with PNG and GIF (I really hate GIF), and both times, it was
> slightly over 40k (about 60-80k).

The pictures you showed had a lot of extraneous fluff in them.  They looked
like nothing more than untrimmed screen shots, and certainly could have been
smaller with very little effort.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\26@083906 by Olin Lathrop

face picon face
V G wrote:
>> FETs look more
>> like a resistor when on, and bipolars more like a voltage source.
>
> I've seen you say this before, but I don't understand this. Could you
> please explain this? Aren't they both like base/gate controlled
> variable resistors?

No, that's the point.  A saturated bipolar will have a roughly constant
voltage accross C-E.  Of course that's not perfect, and under high current
it goes up some.  As a result, there are some operating points where a
bipolar can result in less voltage drop than a FET.  This is more likely
when the transistor has to withstand higher voltage (this forces a tradeoff
with Rdson in FETs) and at moderate to high currents.

{Quote hidden}

The FET when on looks roughly like a fixed resistor to first approximation.
That means the dissipation is proportional to the square of the current.  If
a BJT has a constant C-E voltage, then the dissipation is proportional to
the current.  As the current goes up, the FET's relative advantage
diminishes, until in some cases the BJT can have lower dissipation.

> What effect will the capacitance have?

It takes large current to switch quickly.  For example, let's see what it
takes to change a 1nF capacitor by 15V in 50ns.

 15V * 1nF / 50ns = 300mA

Take a look at the effective gate capacitance of some large power FETs and
you will see why high end FET drivers are rated for 1A or more switching
current.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\26@084353 by Olin Lathrop

face picon face
V G wrote:
>> When fully on, a Bipolar has a voltage drop - Vce(sat)
>
> So what does the base control exactly? If the current through the base
> controls the current through the collector, and I = V/R, is it also
> controlling the resistance? Or is it controlling the voltage drop?

A BJT is a current multiplier.  The base controls how much current the
collector *can* allow to flow.  If the external circuit does not supply that
much current, then the transistor will be "saturated" and look like a fixed
voltage drop to rough first order approximation.  In power switching
applications, a BJT is deliberately operated in saturation to minimize the
on-state voltage drop.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\26@084523 by Oli Glaser

flavicon
face
On 26/05/2011 13:25, Olin Lathrop wrote:
>> >  I compressed my "line art"
>> >  pictures with PNG and GIF (I really hate GIF), and both times, it was
>> >  slightly over 40k (about 60-80k).
> The pictures you showed had a lot of extraneous fluff in them.  They looked
> like nothing more than untrimmed screen shots, and certainly could have been
> smaller with very little effort.

This is true - try the tool Joe suggested, IrfanView. It's free and very good for trimming and converting pictures. Far better than e.g. Miscrosoft Picture Manager and faster/less bloated than Photoshop.

2011\05\26@084602 by Olin Lathrop

face picon face
Oli Glaser wrote:
>> and bipolars more like a voltage source.
>
> Do you mean current source here?

No, I was talking about the on-state of switching applications when the BJT
is saturated.  The do look like a controlled current source in the linear
operating mode, but that's not relevant when we're talking about switching
relay coils on and off.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2011\05\26@085628 by Oli Glaser

flavicon
face
On 26/05/2011 13:46, Olin Lathrop wrote:
> Oli Glaser wrote:
>>> and bipolars more like a voltage source.
>> Do you mean current source here?
> No, I was talking about the on-state of switching applications when the BJT
> is saturated.  The do look like a controlled current source in the linear
> operating mode, but that's not relevant when we're talking about switching
> relay coils on and off.
>

Yes, sorry for the confusion - I missed the switching part as I didn't read the rest of your post. Michael pointed it out just before too.

2011\05\26@093244 by Michael Watterson

face picon face
On 26/05/2011 13:39, Olin Lathrop wrote:
>> What effect will the capacitance have?
> It takes large current to switch quickly.  For example, let's see what it
> takes to change a 1nF capacitor by 15V in 50ns.
>
>    15V * 1nF / 50ns = 300mA
>
> Take a look at the effective gate capacitance of some large power FETs and
> you will see why high end FET drivers are rated for 1A or more switching
> current.
>
A 900V 15A FET may a have 0.8nF to 2nF gate capacitance. The lower capacitance ones cost a lot more!

Higher current rating is often achieved by paralleling FET structures or even entire FETs on the chip. So gate capacitance is roughly related to IdMax
I've tried design/building some RF amplifiers using cheap $1 SMPSU instead of $50 RF FETs. At 400W you are limited to about 4MHz. At 100W to about  50Mhz and at 20W or more  over 70MHz is possible from parallel IRF510 in push pull.
The trick is to drive at 12.5 Ohm or even 3 Ohm

50 Ohm 2:1 transformer gives 12.5 Ohm. You start needing significant power to drive it. Still for a 137KHz 1W ERP you may need a 2KW PA and cheap $1 SMPSU fets running direct off full wave bridge/Capacitor on the mains will do that with suitable isolated input and output transformers. Here with 230V mains

2011\05\26@115536 by RussellMc

face picon face
> > Take a look at the effective gate capacitance of some large power FETs and
> > you will see why high end FET drivers are rated for 1A or more switching
> > current.

The two or three x bipolar jellybean gate driver that I mentioned can
provide about 1A of drive with transistors worth cents each.
Getting really high switching rates will require some additional glue
and special transistors for the very highest frequencies.
For use into the low MHz range (most smps aplications) very ordinary
parts may be used. For HF and VHF applications all but the more
entgused and capable may be advised to resort to existing driver ICs.


             Russel

2011\05\26@121115 by William \Chops\ Westfield

face picon face
>> For ground, all the unused space
>> For power 0.1" if possible.

Keep in mind that the critical high speed parts of VG's design are  mostly on that UBW32 daughter-board, which has its own proven (more or  less) layout.  A bunch of LEDs and relay drivers doesn't need to be  too over-designed.

BillW

2011\05\26@141926 by Gerhard Fiedler

picon face
V G wrote:

> On Wed, May 25, 2011 at 6:36 PM, Gerhard Fiedler wrote:
>
>> I is only V/R for a linear R, or within a linear region of a
>> non-linear R. The C-E path of a bipolar transistor that is used as a
>> switch and is "on" is not usually in a linear region, differently
>> from a FET.
>
> From what I understand (I could be wrong), I = V/R for any point, as
> in, when a voltage V is applied across a resistance R, then the
> current flowing through R is equal to I = V/R.
Depending on what you consider "a resistance R", this could be correct
or not. It is correct for an ideal "Ohm" resistor, as it is the
definition of it. It is approximately correct for a real resistor as
long as you stay within its linear region. It is not correct for
anything else.

For example, take a normal 1/4 W metal film resistor of 1k. Apply
voltage across it, say 1V. You expect 1mA of current. That's what you
get -- approximately, as long as you don't look too close at noise, for
example (which is one divergence of real parts from Ohm's "law"). Then
increase the voltage. The current will follow as expected, but at some
point it will not anymore -- the resistor is heating up, and eventually
will burn, and the current will change drastically. That's where it came
out of its linear region. There are other effects besides exceeding the
max power spec, for example when exceeding the max voltage spec, and all
those max/min specs delimit the linear region of that resistor.

> Are you saying the internals of a BJT violate ohm's "law"? I mean, the
> violations of the law have been discussed on this list before, if I
> can remember correctly.
A bipolar transistor (and most other semiconductors) is no resistor, and
definitely no ideal "Ohm" resistor, so it can't really violate the "law"
-- it doesn't apply to it.
> I'm just not understanding the part about the linear region.

Under certain operating conditions, semiconductors can present a linear
dependency between current and voltage, and such regions may be called
"linear regions" (in this context). For example, the source-drain
connection of a MOSFET that's "on" has a region of drain-source voltage
where it behaves very close to a resistor.
Gerhar

More... (looser matching)
- Last day of these posts
- In 2011 , 2012 only
- Today
- New search...