Hello all,
I have just finished desigining a PCB for LM4781 amplifier. I have used the schematic in the datasheet (except for the resistor for mute). PCB gurus please have a look at the same and let me know if this is ok. I had to decrease the size of the tracks which goes in between two pins of the IC (some of these are power tracks). Any comments or suggestions are welcome.
Thanks in advance,
Venki.
I have just finished desigining a PCB for LM4781 amplifier. I have used the schematic in the datasheet (except for the resistor for mute). PCB gurus please have a look at the same and let me know if this is ok. I had to decrease the size of the tracks which goes in between two pins of the IC (some of these are power tracks). Any comments or suggestions are welcome.
Thanks in advance,
Venki.
Hi
Admirable effort.
Is there any reason why you're not using the National layout as a base?
Per se, I can only see one side of the layout. What does the other side look like?
AFAIK the power traces look very very thin. Though the 4781 isn't that power hungry, maybe thicker traces will be a little more desirable?
Is it normal practice to have a huge ground plane around the circuit area? I would assume a linear plane with a single ground point be more optimised for noise rejection. The way you have designed, the plane may increase noise.
Output traces are also very thin and you've attempted to lay out all tracks as straight parallel lines. I think more work is required. Do you want to use the datasheet layout as a base and progress from there? If you're doing scratch look at BrianGTs threads when they were designing PCBs for the 4780 and 3875. A wealth of information on layout. The discussion itself is very important in your case, more so than the final PCB.
4781 is also a very complex first project. I have some of those chips laying around, so will look forward to your completing the PCBs, and if you take orders, would love to get a couple of them from you if you plan to sell.
Admirable effort.
Is there any reason why you're not using the National layout as a base?
Per se, I can only see one side of the layout. What does the other side look like?
AFAIK the power traces look very very thin. Though the 4781 isn't that power hungry, maybe thicker traces will be a little more desirable?
Is it normal practice to have a huge ground plane around the circuit area? I would assume a linear plane with a single ground point be more optimised for noise rejection. The way you have designed, the plane may increase noise.
Output traces are also very thin and you've attempted to lay out all tracks as straight parallel lines. I think more work is required. Do you want to use the datasheet layout as a base and progress from there? If you're doing scratch look at BrianGTs threads when they were designing PCBs for the 4780 and 3875. A wealth of information on layout. The discussion itself is very important in your case, more so than the final PCB.
4781 is also a very complex first project. I have some of those chips laying around, so will look forward to your completing the PCBs, and if you take orders, would love to get a couple of them from you if you plan to sell.
Mmmm, I see you're going for a single sided layout.
I am not sure it is going to work well. But definitely do not have this surrounding ground plane, it will be very noisy.
If you have major routing problems consider jumpers as an alternative. Though not the best solution, will help some of the track load demands. I would say that beauty of component layout in this case should be secondary. Also the tracks do not need to run parallel to each other all the time.
Return load grounds to power ground and work out a single-point ground for best results. With the 4781 it's a little involved because it's three amps in one package.
I think this is the first time this chip has been brought up, so a component layout would help other members visualise what you are talking about.
Ideally I would still look at the datasheet layout and start from there. It's a little spread out and open, some traces could be a lot shorter, but it works well I guess. It's also got a good grounding scheme. You could start by removing the on-board connectors which take up about 30% of the board space. Then start moving the passives closer to the chip area. Try not to extend the layout behind the chip as your cooling options become limited.
I am not sure it is going to work well. But definitely do not have this surrounding ground plane, it will be very noisy.
If you have major routing problems consider jumpers as an alternative. Though not the best solution, will help some of the track load demands. I would say that beauty of component layout in this case should be secondary. Also the tracks do not need to run parallel to each other all the time.
Return load grounds to power ground and work out a single-point ground for best results. With the 4781 it's a little involved because it's three amps in one package.
I think this is the first time this chip has been brought up, so a component layout would help other members visualise what you are talking about.
Ideally I would still look at the datasheet layout and start from there. It's a little spread out and open, some traces could be a lot shorter, but it works well I guess. It's also got a good grounding scheme. You could start by removing the on-board connectors which take up about 30% of the board space. Then start moving the passives closer to the chip area. Try not to extend the layout behind the chip as your cooling options become limited.
Hello sangram,
Thanks a lot for the comments, Actually I thought increasing the ground plane will reduce the noise (Also etching is quicker
). The PCB is single sided, for making double sided DIY is really hard for me. So I try to stick with single sided board (even if I end up using jumpers). I will increase the power tracks as I had also felt the same . I will post the updated layout once I complete it. I used the same layout as per the datasheet. I will post the layout once i am home. I start designing PCBs by drawing a rough sketch of the PCB on a paper with connections to each pins. Then i draw it using a PCB designing tool (currently i use ExpressPCB). I dont know if this is a wrong procedure. Also i know little about the rules to be followed while designing a PCB (means component placement). (This includes the reply to you previous post, i was about to submit when i got logged out ).
Thanks in advance,
Venki.
Thanks a lot for the comments, Actually I thought increasing the ground plane will reduce the noise (Also etching is quicker
). The PCB is single sided, for making double sided DIY is really hard for me. So I try to stick with single sided board (even if I end up using jumpers). I will increase the power tracks as I had also felt the same . I will post the updated layout once I complete it. I used the same layout as per the datasheet. I will post the layout once i am home. I start designing PCBs by drawing a rough sketch of the PCB on a paper with connections to each pins. Then i draw it using a PCB designing tool (currently i use ExpressPCB). I dont know if this is a wrong procedure. Also i know little about the rules to be followed while designing a PCB (means component placement). (This includes the reply to you previous post, i was about to submit when i got logged out ).Thanks in advance,
Venki.
No dear, I meant the PCB layout as per datasheet. There is one PCB layout on page 23. I did not mean the circuit diagram.
Basically you should look at that then convert it to single sided.
I you're etching the PCB yourself, then you have some way to go. If you're running small numbers, don't hesitate to look at post assembly. In the sense use thicker tracks and jumpers.
Basically you should read a book called bebop to the boolean boogie (seriously!!) and Douglas Self. They have good tips on laying out PCBs, specially how to run ground loops, sizing traces, how to route, etc.
I have never designed a PCB myself, but some of the rules I do know are:
If the trace is expected to carry large current, make it thick. Power and output specially`
Don't run power, signal and output close to each other or in parallel, to prevent crosstalk.
Use single grounding method.
Use planes carefully.
Keep signal and feedback traces as short as possible.
Where possible, keep components close to the chip to minimise trace length.
Decoupling components should as close as possible to chip pins. If possible directly on the pins.
Basically you should look at that then convert it to single sided.
I you're etching the PCB yourself, then you have some way to go. If you're running small numbers, don't hesitate to look at post assembly. In the sense use thicker tracks and jumpers.
Basically you should read a book called bebop to the boolean boogie (seriously!!) and Douglas Self. They have good tips on laying out PCBs, specially how to run ground loops, sizing traces, how to route, etc.
I have never designed a PCB myself, but some of the rules I do know are:
If the trace is expected to carry large current, make it thick. Power and output specially`
Don't run power, signal and output close to each other or in parallel, to prevent crosstalk.
Use single grounding method.
Use planes carefully.
Keep signal and feedback traces as short as possible.
Where possible, keep components close to the chip to minimise trace length.
Decoupling components should as close as possible to chip pins. If possible directly on the pins.
Hello sangram,
I have re-designed the PCB with thicker power lines, output lines. I ended up adding two jumpers and removing ground plane. I know the PCB may look very different from the one i earlier designed. Please have a look at this also. Comment are always welcome.
Thanks in advance,
Venki.
I have re-designed the PCB with thicker power lines, output lines. I ended up adding two jumpers and removing ground plane. I know the PCB may look very different from the one i earlier designed. Please have a look at this also. Comment are always welcome.
Thanks in advance,
Venki.
Yes, a component layout would be helpful.
I still see you are using one long line for the ground and connecting various things at various points. You should avoid that. What will happen is along the length the voltage will drop, creating hum problems.
Is it possible to make the power traces bigger? You still have a lot of space between them. Maybe so much is not required.
Honestly though, it looks a lot better. You can solve the grounding problem by deciding on one single ground point and routing traces individually to it, rather than couple components along a single length of trace. Again I urge you to look at page 23 of the datasheet (which this has somewhat started to resemble) for a good example of star ground. It has one single point, connecte to all grounds individually.
I still see you are using one long line for the ground and connecting various things at various points. You should avoid that. What will happen is along the length the voltage will drop, creating hum problems.
Is it possible to make the power traces bigger? You still have a lot of space between them. Maybe so much is not required.
Honestly though, it looks a lot better. You can solve the grounding problem by deciding on one single ground point and routing traces individually to it, rather than couple components along a single length of trace. Again I urge you to look at page 23 of the datasheet (which this has somewhat started to resemble) for a good example of star ground. It has one single point, connecte to all grounds individually.
Hey thanks for all the comments,
I will post the full layout after the modifications. I have a doubt about the PCB layout shown in the datasheet. They have actually connected the signal ground to the circuit ground via a 2.7 ohm resistor, is there any significance for this resistor with respect to noise? Also in the PCB, the traces going to the IC ground is really thin. Actually I am confused why they have thick ground lines only to the output?
Thanks in advance,
Venki.
I will post the full layout after the modifications. I have a doubt about the PCB layout shown in the datasheet. They have actually connected the signal ground to the circuit ground via a 2.7 ohm resistor, is there any significance for this resistor with respect to noise? Also in the PCB, the traces going to the IC ground is really thin. Actually I am confused why they have thick ground lines only to the output?
Thanks in advance,
Venki.
this is to keep noise from entering the input, the the wires/traces in the rest of the ground ....down to the actual ground connection can form a path of less resistanse for noise on the grounds, which will then rather follow it than go up the thin trace/wire back to the input... I think the resistor is just taking the concept a step further... the output aso carries the largest current so thats why you need thick stuff there.
Yes, I think the trace is too long, it can be on the main rails next to the - cap, and avoid the long wire. Actually it's not too much worry. I'm using LM4780 without any addiitonal bypass caps and it works just great, no instability or oscillation. Actually I tried some 100nF and they messed up the sound. So now running without them.
Serious problem in your PCB is the fact that input bias resistor is missing which will result in smoked speakers in 2 seconds. I would remove the input caps/resistors as those can be mounted offboard and can also be optional depending on configuration of the upstream components, if you have problems in size/space/routing terms.
But the input bias resistor is not optional and needs to be used. Of course they can be mounted on the track side of the board using some jiggery-pokery but that's not ideal.
Have you built these things yet? Any chance of getting hold of a few of these PCBs from you? It would make a nice entry-level 5.1 system in a very tight space (1 ATX PSU case and a shared heatsink, with leadouts for in/outputs)
Serious problem in your PCB is the fact that input bias resistor is missing which will result in smoked speakers in 2 seconds. I would remove the input caps/resistors as those can be mounted offboard and can also be optional depending on configuration of the upstream components, if you have problems in size/space/routing terms.
But the input bias resistor is not optional and needs to be used. Of course they can be mounted on the track side of the board using some jiggery-pokery but that's not ideal.
Have you built these things yet? Any chance of getting hold of a few of these PCBs from you? It would make a nice entry-level 5.1 system in a very tight space (1 ATX PSU case and a shared heatsink, with leadouts for in/outputs)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.