Correct bypassing of power supplies
Sorry if that has been discussed a lot of times - I could not find any specific posts about this.
I am building an IC-based(LM3886) dual channel amplifier. The main power supply will be on a separate board(2x4700 uF).
The bypass caps on amp's board would be 2x470 uf.
The problem is about ground return paths of the bypass capacitors.
What happens if all grounds(signal, load, bypass) are connected in a star formation on an amp board and then return to main supply?
I suspect this way of connecting may be prone to picking up power supply/load noise to amp's input if large currents are flowing between main and local bypass caps?
If so , what is the proper solution to this?
You are referring to decoupling of the supplies.
Try that for a search.
I'm fairly certain that you will not want the input section's ground references and returns to share a conductor with the load ground or the power decoupling/bypass ground, on the way to the star ground.
And I think that you will probably want the decoupling capacitors' grounds to connect directly to the load ground. So you'll want to place the decoupling caps as close to the power pins as possible and connect their grounds together as directly as possible and then bring the load ground to THEM.
Theoretically, you can get better high-frequency performance, transient response, and stability, as well as lower distortion at all frequencies, if you replace each decoupling cap with multiple smaller caps in parallel, since that could reduce the caps' and their associated conductors'/leads' total inductance and resistance to less that that of one cap, lowering the power supply impedance seen by the chip's power pins at frequencies of interest for those purposes.
One of the potential problems with implementing that is that if parallel caps share conductors, then the inductances won't necessarily be reduced as much as they could have been. So it would probably be best to pick a point near both the positive power supply pins and the negative power supply pin and take the load ground there, and then solder one end of each cap lead to an appropriate power pin and the other end to the load ground point.
If soldering directly to chip pins, remember to clamp long nose pliers to the pin, between the chip and where you're going to solder, to keep the heat out of the chip (or use some equivalent or better method).
Personally, I'd probably just fit as many caps as possible, in parallel, for each power pin. 5000 uF per rail would not be too much, certainly. But it will only help a lot if it can be done with multiple smaller caps in parallel. Maybe multiple 1000 uF caps would work well.
For one of the next LM3886 boards that I make I'm considering standing it on edge, so I can fit caps on both sides, more easily.
If you want to go for overkill (and who doesn't?), you could run multiple parallel power supply traces for each rail and for ground, and use a different pair of power & gnd rails for each smoothing cap (or each set of smoothing caps), since they also benefit similarly from paralleling with separate conductors, and take each rail separately to the amp board and join them only at the power pins and the load ground point, and use separate decoupling caps between each pair, there, too. That should give the lowest psu impedance as seen by the power pins, unless you can use a multilayer PCB.
Thanks for replies.
I have read a lot of posts/threads regarding to supply capacitors/bypassing. Many contradicting opinions , ideas etc.
So I decided to do a common ground point on the amp's PCB by using star grounding, which would return to PSU board with as short leads as possible.
I am going to make the board using the laser printer toner transfer so ground plane is not an option.
Do you see any problems with the layout and stability?
The star point should be in the power supply with seperate ground
returns to it, usually just the supply return and a signal return.
Seems churlish to compromise the signal return for one connection.
What you are planning can not be called star grounding.
"Conflicting opinions" is not a reasonable rationale, because opinion is not a factor in circuit analysis. If you cannot do the analysis yourself but still want to do the right thing, that's very good, but then why plan based on the opinions of others who also can't do the analysis?
If anything, the existence of some doubt or uncertainty about it should mean that you just take every ground back to the PSU separately, instead of wondering which ones might be OK to combine.
By the way, using the toner transfer method does not mean that you can't have a ground plane. See my page here: Easy PCB (Printed Circuit Board) Fabrication, Using Laser Printer Toner Transfer, with a Household Clothes Iron and Glossy Inkjet Photo Paper; DIY at Home; Better AND Cheaper than Press-n-Peel ( PnP / P-n-P )! Making, Cheap , Economical , fastest fas .
OK, I also looked at your PCB, but only briefly so far. The decoupling capacitors are not nearly close-enough to the power pins. Also, SHARING them between two chips might not be the best idea, and has also already made the two chips' circuits asymmetrical, so they won't quite respond identically, which is probably not what you want. A purist would probably also want the two chips' circuits and PSU connections to be completely separate, anyway, for as far upstream as is possible, even if separate PSUs were not used for the two channels. Do you really want the output signal ground return currents from one channel modulating the input signal ground reference point and power rail voltages of the other channel? It's bad enough that they would be modulating their own.
Keep in mind that all ground return currents will induce voltages across the inductance and resistance of the ground return conductor, which will appear back at the non-PSU ends of every ground return path that shares their ground return conductor. If an amplifier input's ground reference voltage is bouncing around like that, the bouncing voltage is arithmetically summed with the input signal voltage! You shouldn't want that. (Not only will it add distortion, it is also another feedback path from output to input, which might cause bigger problems.)
|All times are GMT. The time now is 10:54 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio