Hey guys,
Long time reader, first time poster. I am just starting to play with PCB layout software (I have the demo version of Eagle) and I was looking for feedback (/bad pun) on my first attempt at designing one.
Here it is:
www.hayenga.com/mark/ckt.gif
It's just a simple active filter, each of the three sections implements two second order lowpass filters, input on far left and output on far right.
Questions:
1) The opamps are tied to +Vcc (top layer, uppermost thick red line) and -Vcc (thick red line @ bottom). On the bottom layer right underneath -Vcc I have an 'analog ground' point. Let's say the opamp rails are +/-10V. My filter design software says I need to tie certain elements surrounding the opamp (some of the R's and C's) to ground. Is this absolute ground, or signal ground? Is this the right way to do it? DoI have to tie it to something that's absolute 0V or can I just tie it to a supply rail?
2) How do I mark the physical dimensions of the board in Eagle? (So the PCB mfgr knows what size the board actually is).
3) Resistor/Capacitor sizes. Eagle has a ton of R/C types (like an 0204 resistor, an 0207 resistor, etc). How do I know what size to use? Do I buy parts first and figure out what they are and then design the board? I didn't notice any indications of 0204 or 0207 on mouser's metal film resistor page.
4) Pads. I assume for DIP, non-surface mount resistors, etc that the pads are plated through holes? ie I can connect to them from both the bottom and/or top layers? (Surface mount I surmise is only accessible from the top layer).
5) Finally, does this look good or is it crappy? Do my traces need to be wider, more compact, more space to solder, etc? Anything technically or stylistically wrong?
Thanks for any help, I'm a long time speaker builder and senior in computer engineering but I know jack about PCB design.
Long time reader, first time poster. I am just starting to play with PCB layout software (I have the demo version of Eagle) and I was looking for feedback (/bad pun) on my first attempt at designing one.
Here it is:
www.hayenga.com/mark/ckt.gif
It's just a simple active filter, each of the three sections implements two second order lowpass filters, input on far left and output on far right.
Questions:
1) The opamps are tied to +Vcc (top layer, uppermost thick red line) and -Vcc (thick red line @ bottom). On the bottom layer right underneath -Vcc I have an 'analog ground' point. Let's say the opamp rails are +/-10V. My filter design software says I need to tie certain elements surrounding the opamp (some of the R's and C's) to ground. Is this absolute ground, or signal ground? Is this the right way to do it? DoI have to tie it to something that's absolute 0V or can I just tie it to a supply rail?
2) How do I mark the physical dimensions of the board in Eagle? (So the PCB mfgr knows what size the board actually is).
3) Resistor/Capacitor sizes. Eagle has a ton of R/C types (like an 0204 resistor, an 0207 resistor, etc). How do I know what size to use? Do I buy parts first and figure out what they are and then design the board? I didn't notice any indications of 0204 or 0207 on mouser's metal film resistor page.
4) Pads. I assume for DIP, non-surface mount resistors, etc that the pads are plated through holes? ie I can connect to them from both the bottom and/or top layers? (Surface mount I surmise is only accessible from the top layer).
5) Finally, does this look good or is it crappy? Do my traces need to be wider, more compact, more space to solder, etc? Anything technically or stylistically wrong?
Thanks for any help, I'm a long time speaker builder and senior in computer engineering but I know jack about PCB design.
I use PCAD, but all PCB packages have common features. I'll tell you what I do...
1) Grounding is a "special" subject. Ultimately all ground connections should go back to a single point in the box you enclose the circuit in. How you handle that with a PCB is subject to some debate. Some would create a ground on the PCB for ALL the grounded components thereon, then run a wire from the ground on the board (ah, but at which point do you attach the wire?) to the single point chassis ground. Others would connect some ground points to wires and run them directly to the chassis single point ground, so you might end up with multiple ground wires from your board to the single point chassis ground. In any case, the power supply ground should go directly to the chassis ground point, and your circuit ground(s) should connect to the supply ground at the chassis ground point.
2) In PCAD and Protel, you make a layer called BOARD and then draw the outline of your board on that layer. Don't forget to include that layer in your Gerber output for the board house.
3) Component sizes are always a sticky point. For whatever reason, PCAD libraries don't include pads for standard surface mount component sizes. You actually have to create them yourself- quite annoying, but then how do they know what you preferences are?- there are many possible valid pad dimensions for an 0805 part. Generally, you pick the parts you're going to use, then create the libraries around them. Good board layout requires intimate knowledge of the circuit's electrical AND mechanical characteristics.
4) Usually through-hole components in the libraries are plated through holes. There should be a way to check it in Eagle- probably select the component pattern and right click and then open the properties dialog. It works that way in both PCAD and Protel. Surface mount components can be put on either the top or bottom layer, or both, and unless there is a hole through the pad (not usually a good idea) they are only accessable from the layer on which they are placed. Holes in surface mount pads can cause soldering problems, so it is best to use a via slightly moved away from the pad if you have to go to another layer.
5) It looks very nice- symmetry always makes for an easy to troubleshoot and professional looking board.
a) Generally it is a good idea to neck down the trace width where it meets a pad so that solder won't flow down the length of the trace- the surface tension of the molten solder will keep it from going far with just a little narrowing (i.e. make pads a little wider than the traces). Solder mask will also prevent solder flow.
b) You could reduce trace lengths by eliminating "U"s- for example, far left, blue trace- no need to run below the cap and then back up. There are quite a few of these structures in your layout. Looking at the same IC, the blue trace to the third cap could be run between the pads of the fourth cap to shorten that trace considerably.
c) At pins 1 and 2, and 6 and 7, you could simply connect the two pads with a straight line.
d) It doesn't look like there are any supply bypass caps on the ICs. That is a real bad idea. You should have bypass caps at both supply pins at each IC.
e) more trace shortening could be easily accomplished by standing the resistors up.
f) depending on the circuit, you might have enough capacitive coupling between the capacitors to cause problems. You might try swapping the R and C positions in every other RC pair to put more space between the caps.
MR
1) Grounding is a "special" subject. Ultimately all ground connections should go back to a single point in the box you enclose the circuit in. How you handle that with a PCB is subject to some debate. Some would create a ground on the PCB for ALL the grounded components thereon, then run a wire from the ground on the board (ah, but at which point do you attach the wire?) to the single point chassis ground. Others would connect some ground points to wires and run them directly to the chassis single point ground, so you might end up with multiple ground wires from your board to the single point chassis ground. In any case, the power supply ground should go directly to the chassis ground point, and your circuit ground(s) should connect to the supply ground at the chassis ground point.
2) In PCAD and Protel, you make a layer called BOARD and then draw the outline of your board on that layer. Don't forget to include that layer in your Gerber output for the board house.
3) Component sizes are always a sticky point. For whatever reason, PCAD libraries don't include pads for standard surface mount component sizes. You actually have to create them yourself- quite annoying, but then how do they know what you preferences are?- there are many possible valid pad dimensions for an 0805 part. Generally, you pick the parts you're going to use, then create the libraries around them. Good board layout requires intimate knowledge of the circuit's electrical AND mechanical characteristics.
4) Usually through-hole components in the libraries are plated through holes. There should be a way to check it in Eagle- probably select the component pattern and right click and then open the properties dialog. It works that way in both PCAD and Protel. Surface mount components can be put on either the top or bottom layer, or both, and unless there is a hole through the pad (not usually a good idea) they are only accessable from the layer on which they are placed. Holes in surface mount pads can cause soldering problems, so it is best to use a via slightly moved away from the pad if you have to go to another layer.
5) It looks very nice- symmetry always makes for an easy to troubleshoot and professional looking board.
a) Generally it is a good idea to neck down the trace width where it meets a pad so that solder won't flow down the length of the trace- the surface tension of the molten solder will keep it from going far with just a little narrowing (i.e. make pads a little wider than the traces). Solder mask will also prevent solder flow.
b) You could reduce trace lengths by eliminating "U"s- for example, far left, blue trace- no need to run below the cap and then back up. There are quite a few of these structures in your layout. Looking at the same IC, the blue trace to the third cap could be run between the pads of the fourth cap to shorten that trace considerably.
c) At pins 1 and 2, and 6 and 7, you could simply connect the two pads with a straight line.
d) It doesn't look like there are any supply bypass caps on the ICs. That is a real bad idea. You should have bypass caps at both supply pins at each IC.
e) more trace shortening could be easily accomplished by standing the resistors up.
f) depending on the circuit, you might have enough capacitive coupling between the capacitors to cause problems. You might try swapping the R and C positions in every other RC pair to put more space between the caps.
MR
Some answers
Hi Mark,
In Eagle you make board dimensions with WIRE in DIMENSION layer - just make a rectangle with it.
Through-hole components have pads on both outside layers and SMD parts are by default put on top layer (component side) and you can get them on the bottom layer (solder side) with MIRROR command.
You can have some problems with pad sizes (pad diameters -or how much of the copper ring is left when you drill the hole), if you understood me
In libraries, most parts are defined with ZERO pad radius meaning that it's diameter is derived from drill diameter, and it's controlled with RESTRING value in Design Rule Check.
It would be a good thing to first buy the parts, measure lead distances and find appropriate parts from libraries.
BTW, have you downloaded help file in PDF format from their site, and read it?
You can get most of the answers if you point your newsreader to
news.cadsoft.de
you will find a few groups in english language like userchat.eng and ask some questions, but most have been answered already, so first download all articles - 1000's of them and do some searching and reading.
I am also a newbie so better not ask me some more complicated questions
Hi Mark,
In Eagle you make board dimensions with WIRE in DIMENSION layer - just make a rectangle with it.
Through-hole components have pads on both outside layers and SMD parts are by default put on top layer (component side) and you can get them on the bottom layer (solder side) with MIRROR command.
You can have some problems with pad sizes (pad diameters -or how much of the copper ring is left when you drill the hole), if you understood me
In libraries, most parts are defined with ZERO pad radius meaning that it's diameter is derived from drill diameter, and it's controlled with RESTRING value in Design Rule Check.It would be a good thing to first buy the parts, measure lead distances and find appropriate parts from libraries.
BTW, have you downloaded help file in PDF format from their site, and read it?
You can get most of the answers if you point your newsreader to
news.cadsoft.de
you will find a few groups in english language like userchat.eng and ask some questions, but most have been answered already, so first download all articles - 1000's of them and do some searching and reading.
I am also a newbie so better not ask me some more complicated questions
Hi MR,
Thank you for the very detailed post. I've tried to incorporate all of your suggestions into design #2, save for the narrowed traces at the pins (only because I was too lazy to do it at 1am last night). I still have one 'U' but I think that's necessitated by the layout. Traces are a good bit shorter and the layout is more compact. Did I make any mistakes this time around?
www.hayenga.com/mark/ckt2.gif
Also, about the capacitive coupling. I have no real experience in building anything more complicated than, say, a DAC interfaced with a microcomputer or the like. I take it the coupling is generated by a stray E field generated by the caps? Are any caps more prone to this than others? Thanks again for the excellent help. (BTW, what part of TI do you work in? I've got a brother who's been doing DSL modem stuff there).
Rendisha,
Thank you also for the help. I actually managed to have dimensions to my board this time around and will download the help file from their site. Might also try a few other programs, just to get a feel for what's out there.
Thanks again guys,
Thank you for the very detailed post. I've tried to incorporate all of your suggestions into design #2, save for the narrowed traces at the pins (only because I was too lazy to do it at 1am last night
). I still have one 'U' but I think that's necessitated by the layout. Traces are a good bit shorter and the layout is more compact. Did I make any mistakes this time around?www.hayenga.com/mark/ckt2.gif
Also, about the capacitive coupling. I have no real experience in building anything more complicated than, say, a DAC interfaced with a microcomputer or the like. I take it the coupling is generated by a stray E field generated by the caps? Are any caps more prone to this than others? Thanks again for the excellent help. (BTW, what part of TI do you work in? I've got a brother who's been doing DSL modem stuff there).
Rendisha,
Thank you also for the help. I actually managed to have dimensions to my board this time around
and will download the help file from their site. Might also try a few other programs, just to get a feel for what's out there.Thanks again guys,
Mark Hayenga said:Hi MR,
Thank you for the very detailed post. I've tried to incorporate all of your suggestions into design #2, save for the narrowed traces at the pins (only because I was too lazy to do it at 1am last night
It looks much better. Be sure you left enough space between traces and pads for your board maker to be able to manufacture the board. I think you could save a little more real estate by putting the two power traces at the bottom of the board on top of each other. The extra capacitance there wouldn't hurt a bit.
Make sure that you capacitors will fit in the space provided before you make the boards. It is a real sad thing to make a beautiful board layout and then not be able to fit the parts on it when you get the boards into your hot little hands.
It is a good idea to silk screen the component designators onto the board so that assembly and troubleshooting are made easier. If you have space even the component value is useful.
If you put two foil/film caps close together, there is capacitance between the two caps' outer foil/film layers (with air dielectric). The capacitance is small, but in high Z circuits, or high order filters, the stray capacitance can be enough to screw up the operation of the circuit.
In the old days some foil caps were marked to indicate which terminal connected to the outermost layer of foil. If one end of the cap were connected to ground, you would connect the outer foil end so that the cap was effectively shielded. Modern caps don't seem to have that marking. It doesn't make much difference when neither side of the caps connects to ground (ground as in ac ground- i.e. either 0V or other power supply connection).
Groundplane
Hi Mark
Why don't you add a groundplane to your design?
There is plenty of room for it, and it's alvays a good idear.
Try to avoid small pads, you can edit the Library and change the components used. This takes a long time the first time you do it, but soon you will have all your favorite components added to a LIB. (Be sure to make a backup!!)
I think your PCB looks fine, but would get the components and measure their pin spacing and size before making the final artwork.
Also if you are making non plated prototypes yourself first, it's diffecult to solder the caps when the tracks leading to them are on the top layer.... I always try to keep tracks for caps on the bottom layer. With resistors it's no problem.
Good luck with the project
\Jens
Hi Mark
Why don't you add a groundplane to your design?
There is plenty of room for it, and it's alvays a good idear.
Try to avoid small pads, you can edit the Library and change the components used. This takes a long time the first time you do it, but soon you will have all your favorite components added to a LIB. (Be sure to make a backup!!)
I think your PCB looks fine, but would get the components and measure their pin spacing and size before making the final artwork.
Also if you are making non plated prototypes yourself first, it's diffecult to solder the caps when the tracks leading to them are on the top layer.... I always try to keep tracks for caps on the bottom layer. With resistors it's no problem.
Good luck with the project
\Jens
Pad sizes
In Eagle, one can use the restring function to change pad sizes and styles for the whole board, without editing the libraries at all.
Re: groundplanes, they are not a universal panacea, if you can't figure why it's needed, then I suggest it isn't
I'd second the option to put +ve rail over -ve rail if possible, then run the power tracks inside the opamp pins. Both of these actions reduce loop areas, and inductance, which as you progress to higher speed devices will become critical.
Analyse the current flowing in the tracks, and ensure that current flow from one device is not modulating another signficantly (and significant can be very small difficult to measure amounts for good audio).
The suggestion to stand resistor up to shorten traces should be considered for a while to realise why that's slightly silly - you just add component lead lenght in place of trackwork!
There's definitely room for shortening traces though, just play around with component placement then re-route, use 45 deg angles rather than 90 deg, it looks nicer and helps reduce trace lengths amongst other things.
We could do with a schematic to help further - I'm concerned by your '0V' comments.
PSU decoupling is a very good idea, and essential in high-speed op-amps.
Some suggest a cap between 0V and each rail, for audio I prefer a cap between + and - as this reduces current flow in the critical 0V signal reference . This cap must be close to the pins, for high speed devices SMD can be essential to get the trace inductance low enough. One cap per op-amp, unless close together.
Tracks for caps on the bottom layer is excellent advice.
As the circuit is simple you could place the eagle schematic file here, and I'd be happy to have a quick go at it myself if that would help you understand some of the advice.
We all have to learn somewhere, and PCB layout is one of those skills you get better at the more you do it, it becomes a bit intuitive after a while.
Andy.
In Eagle, one can use the restring function to change pad sizes and styles for the whole board, without editing the libraries at all.
Re: groundplanes, they are not a universal panacea, if you can't figure why it's needed, then I suggest it isn't
I'd second the option to put +ve rail over -ve rail if possible, then run the power tracks inside the opamp pins. Both of these actions reduce loop areas, and inductance, which as you progress to higher speed devices will become critical.
Analyse the current flowing in the tracks, and ensure that current flow from one device is not modulating another signficantly (and significant can be very small difficult to measure amounts for good audio).
The suggestion to stand resistor up to shorten traces should be considered for a while to realise why that's slightly silly - you just add component lead lenght in place of trackwork!
There's definitely room for shortening traces though, just play around with component placement then re-route, use 45 deg angles rather than 90 deg, it looks nicer and helps reduce trace lengths amongst other things.
We could do with a schematic to help further - I'm concerned by your '0V' comments.
PSU decoupling is a very good idea, and essential in high-speed op-amps.
Some suggest a cap between 0V and each rail, for audio I prefer a cap between + and - as this reduces current flow in the critical 0V signal reference . This cap must be close to the pins, for high speed devices SMD can be essential to get the trace inductance low enough. One cap per op-amp, unless close together.
Tracks for caps on the bottom layer is excellent advice.
As the circuit is simple you could place the eagle schematic file here, and I'd be happy to have a quick go at it myself if that would help you understand some of the advice.
We all have to learn somewhere, and PCB layout is one of those skills you get better at the more you do it, it becomes a bit intuitive after a while.
Andy.
Thank you guys for all the excellent advice. I couldn't have imagined I'd get so many helpful replies when I posted. I really do appreciate it.
ALW,
I don't have the schematic for that exact circuit on hand, but it's just three of these circuits in series:
http://www.hayenga.com/mark/DSP/second_page.bmp
(edit: it's six of those in series as those are dual opamps, and there's two 2nd order low passes per opamp.)
The ckt diagram is at the top, ignore the rest (just some experimentation back when I was doing signal processing in matlab). I am also concerned about my comments re: 0V The signal coming from the preamp, I can assume that the 'hot' line swings both + and - referenced to the ground line? So the opamp needs both + and - 10V rails, and that the ground point in that opamp ckt needs to be at 0V, not just signal ground?
Thank you guys for all the help. With all the help I'm getting it will just be another couple weeks before I'll be making boards for TI C6000's
ALW,
I don't have the schematic for that exact circuit on hand, but it's just three of these circuits in series:
http://www.hayenga.com/mark/DSP/second_page.bmp
(edit: it's six of those in series as those are dual opamps, and there's two 2nd order low passes per opamp.)
The ckt diagram is at the top, ignore the rest (just some experimentation back when I was doing signal processing in matlab). I am also concerned about my comments re: 0V
The signal coming from the preamp, I can assume that the 'hot' line swings both + and - referenced to the ground line? So the opamp needs both + and - 10V rails, and that the ground point in that opamp ckt needs to be at 0V, not just signal ground? Thank you guys for all the help. With all the help I'm getting it will just be another couple weeks before I'll be making boards for TI C6000's
Mark,
One question - you did start from a schematic in Eagle didn't you?
One should always do this for several reasons, it gives you nice documentation for future use, and it makes the PCB layout far easier, since the connections (or netlist) are transferred to the layout, reducing the chance for mistakes and errors in connection.
Andy.
P.S. Rounded traces are rarely used at RF, usually one is working with stripline / microstrip transmission lines, where mitreing the corners is the accepted approach to negating the impedance change when a track changes direction.
Depending upon the speed of your signals it can become necesssary to think about such things with op-amps, if only from an EMC perspective.
With audio it's rarely a problem, if you bandwidth limit input signals (as you should).
One question - you did start from a schematic in Eagle didn't you?
One should always do this for several reasons, it gives you nice documentation for future use, and it makes the PCB layout far easier, since the connections (or netlist) are transferred to the layout, reducing the chance for mistakes and errors in connection.
Andy.
P.S. Rounded traces are rarely used at RF, usually one is working with stripline / microstrip transmission lines, where mitreing the corners is the accepted approach to negating the impedance change when a track changes direction.
Depending upon the speed of your signals it can become necesssary to think about such things with op-amps, if only from an EMC perspective.
With audio it's rarely a problem, if you bandwidth limit input signals (as you should).
On Part Sizes
The first time you use the part sizes the specifications are a little overwhelming. I did the following: Draw a circuit design of just capacitors, just resistors, just diodes, just DCDC blocks, just AMP connectors etc -- keep each as a separate design file in a separate folder within your CAD library, then print the layouts and keep them in a notebook.
After a while, you will get to standardize your purchases and know from memory that you bought 10nF/50v bypass caps from Mouser and they are KERK04X3R2_5 or that a LM317LZ is going to be...
I would not suggest starting out with 10 mil traces for your first home-made boards. 20 mil is fine. When you reduce the trace size other factors such as exposure time, development time, etch time become more critical. and all you need is one small bubble to screw up a path.
As someone else has suggested you might want to edit the pads on IC's and other devices -- but keep these modified devices in a separate folder, or modify the description when you save, i.e. don't over-write the original design.
The first time you use the part sizes the specifications are a little overwhelming. I did the following: Draw a circuit design of just capacitors, just resistors, just diodes, just DCDC blocks, just AMP connectors etc -- keep each as a separate design file in a separate folder within your CAD library, then print the layouts and keep them in a notebook.
After a while, you will get to standardize your purchases and know from memory that you bought 10nF/50v bypass caps from Mouser and they are KERK04X3R2_5 or that a LM317LZ is going to be...
I would not suggest starting out with 10 mil traces for your first home-made boards. 20 mil is fine. When you reduce the trace size other factors such as exposure time, development time, etch time become more critical. and all you need is one small bubble to screw up a path.
As someone else has suggested you might want to edit the pads on IC's and other devices -- but keep these modified devices in a separate folder, or modify the description when you save, i.e. don't over-write the original design.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.