Bypass Cap ground ?

[jeth]

Is there any advantage of going that extra mile to keep the ground path for currents from the power supply bypass caps (referring to opamp circuits with regular bypassing of +- supplies to ground) seperate as far as possible from the audio circuits ground path.

I have been working on (one of my first) pcb designs and it seemed to make sense to me to have a seperate 'branch' of the ground distribution (its not quite a plane or a network of traces) to carry ground curent back to each boards ground wire connection point, seperating them as much as possible from signal ground currents.
As usual after enthusiastically spending hours trying to work my idea into the design it occured to me It may be worth checking if there are any benefits....

 

[JohnRoberts]

I don't know if there's a simple answer to that. It helps to look at the entire circuit to form a specific grounding strategy.

You clearly need to be careful about commingling current from PS decoupling caps with signal currents to prevent contamination. Likewise you want to keep PS ground traces low impedance for decoupling to be effective.

Ideally the audio ground follows a star approach with no loops. Decoupling caps shunting power rails and power ground rail form a loop at high frequency. This loop may be above audio, but it's not a bad idea to keep that loop area small.

Several variables, it depends on the frequency spectra of current you are capturing. Resistive decopling at strategic circuit points can limit peal current that must be handled in decoupling capacitor paths.

JR

 

[Guest]

"Straight ground wire" may be very good if no current loops are possible, for example in a mic only pre; but in case of a gear that has ins/outs grounds from which may be mixed through cables and other gear it is better to have a "Star" ground from output of PS that has the minimal dynamic resistance.

 

[jdbakker]

It also depends on the internal topology of the amplifiers. Have a look at this application note: An IC Amplifier User’s Guide to Decoupling, Grounding, and Making Things Go Right for a Change.

JDB.

 

[jeth]

Thanks for that folks... will give that article a read.. at a glance looks like I might be on the right lines through pure guesswork.

 

[jeth]

Ok, as a newbie I have only ever dabbled in smaller circuits on stripboard. Everything's worked in the end, somethings even first time but noise levels have been a bit high, I think due to compromised and overcrowded stripboard layouts with their unreliable connections and closely spaced traces and probably not helped by never having got round to building anything complete enough to put in a proper case.
So, now I'm embarking on my first bigger project and my first PCB designs.
I've been reading around on layout and as usual finding lots of other information along the way to confuse me all the more. This confusion prompted my original question and unfortunately the confusion prevails. Perhaps someone would be kind enough to give me some comments/practical advice on this..

The board will be a 4 way crossover using sallen key filter sections with a buffered level pot on each output. The screenshot shows my power and ground layout which I want to get right before I proceed, problem being as a beginner it's easy to be waylaid by others differences of opinion and practice and not have the capacity to draw ones own conclusions.
As you can see, all opamp bypass caps have their own substantial ground trace with it's own wire back to star ground. The pot bottom connections and audio signal ground (which will provide ground ref. for the sallen key filters and the pot buffers) also have independant returns to star ground.

Questions, Am I really likely to experience problems with a HF loop through the lengthy connections of ps rails and the independant ground connections? If so would adding a LPF at say 20KHz to the HF band input be of benefit, or perhaps a simple small value cap in the feedback path of one of the amps to roll of HF gain?
I'm also really unsure if it's necessary for the pot to be referenced to the same ground point as the buffers input ref. resistor to ground, which is also functioning as a law faking resistor?

Answers or other comments more than welcome..

 

[rlaury]

Without knowing the details, I would maybe approach it like this:

Signal ground is a plane and connects to system ground back at
the power supply. You may also be ok tying the two together
at the PCB.

You could also reverse the system and make power gnd as
a plane and signal gnd as a star. Just don't mix them.

Power ground is done in a star fashion. Both legs connect together
at the PCB connector.

The POT ground should go back to the associated circuit they
originate from. The POT metal work can be connected to the chassis
ground.

Add an additional set of PC caps about every 2 or 3 IC packages
and a single filter set across the PS buss at the end point.

Good Luck
Hope this helps.

RonL

http://www.nashaudio.com

 

[Samuel Groner]

As long as we deal with low power circuits and regulated supplies (e.g. no power amps) I see little point of separate ground returns as the power supply currents are likely too small to inject detectable distortion residuals into audio ground (at least with dual layer boards where it is easy to maintain a low impedance ground plane).

In addition to this I may suggest the use of a single capacitor across the supply lines instead of a pair referenced to ground. For most audio opamps and applications this will give perfectly sufficient decoupling (though I'd check stability on breadboard for high-speed amps such as AD797 or AD811) while preventing any distorted ground currents flowing into audio ground.

Samuel

 

[JohnRoberts]

This is arguably over design for the needs of that modest application but Ron's layout is IMO much better. Since we are already over engineering this I would only consider changing the audio ground from a singlel brute force ground plane to, a more serpentine but specific local ground routing where the signal currents dumped into this ground have well defined paths. these local signal ground paths could then tie to the ground bus at one point each. Not a loop issue but more the (miniscule) IxR voltage errors.

Note: This is over engineering the problem and there may be little if any measurable difference between your layout and Ron's. One consideration is what loads will these opamps ever have to drive, as that load current travelling from PS and ultimately returning to PS should be accounted for. That may also dominate HF noise being dumped into supply lines and grounds.

Enjoy,

JR

 

[jeth]

Many thanks for the replies here.

Rons layout is a lot tidier, but unfortunately the design is restricted by the small range of parts I have available, lots of big caps have to go round those ICs. I have split the ics into two vertical columns to accomodate.

The rather minimal use of PS filter caps is due to the use of TL series ICs which I'm informed are not too fussy. Though they are the only chips available to me here I'm thinking of increasing the supply filtering so I can drop in other ICs along the line when funds allow.

I'm very grateful for a definitive answer re: pot grounds as this is something I have had conflicting advice on.

Apart from the pot connection and the fact that I have split the ICs into two columns the general layout is as Ron suggests, but with the signal ground on a single 2.54mm trace that runs below the ICs with branches off to each sub-circuits local ground. The PS bypass ground is not a full plane, as I felt this would complicte layout bearing in mind this will be a single sided board, instead it's kind of a partial plane that surrounds the main circuit.

I think I will provide an extra terminal so that I can jumper the PS and audio grouind together on the board and try it woth one connection back to star ground.

Thanks again for all input, feeling more confident of things now...