THAT 1646 noise in PSU

audiomixer said:

I think we all missed that point.

Click to expand...

I don't think we missed it really; it's just that it is so alien that we just can't believe our eyes.
It's just another example of how myths permeate and result in unnecessarily complicated arrangements. Grounding starts with a plain and simple understanding that what comes in must go out; implementing sophisticated schemes without understanding is a recipe for disaster.
As Monte wrote earlier: "Having that many grounds is a good recipe for having no ground at all." How true!

Thank you for the clarification!~
Best,
Bruno2000

abbey road d enfer said:

bruno2000 said:

It seems a bit odd that the major console mfgs (SSL, Neve) only put in 100nF ceramics next to 5532/4 chips.  Comments?

Click to expand...

There are always(?) electrolytics in addition, generally 47-220uF.
IMO Ricardo's recommandation is not the only possible scheme.
There are two aspects that govern the choices: stability and minimization of distortion due to improper current circulation.
Stability is relatively easy to achieve with the typical pair of 100nF ceramic close to the opamp. minimization of distortion takes significant capacitance, but also requires careful analysis of current circulation, in particular how the current that is injected into the load comes back to the decoupling caps.
Ricardo's arrangement offers excellent local bypass instead of centralized bypass, but very often most of the loads are rather light, except for the output stage (that would be typical of many mixers). In that case, a centralized set of caps is perfectly adequate, provided the return path is correctly routed.

Click to expand...

GND et GND1 were connected to the same point on the PSU, but were not connected together on the PCB

ricardo said:

Duu.uh!  Do you mean GND & GND1 WEREN'T CONNECTED in your initial tests? 
I hang my head in shame for not noticing this  :-[

Click to expand...

Yes they were but only at the psu side

Have you tried 2 x Electrolytics AT each 553x and also the 1646?

I said earlier the EXACT physical position where your GND & GND1 join is important.

Can you install the Electrolytics and provide 2  schematics. 

• One with your original Grounding scheme
• The other with the Grounding scheme that stops the oscillation.

Both should show how your scope is connected too.

When we sort these basic points out, we can look at the stuff in my #6 post

ricardo said:

Have you tried 2 x Electrolytics AT each 553x and also the 1646?

Click to expand...

Hello Ricardo, yep didn't change much, but I understand the meaning of this for lowering distorsion and improving transcient response (I guess we can't measure this with a simple sine-wave, and it's not really the point here.. ) , but however thank you for providing theese great infos!

ricardo said:

• One with your original Grounding scheme
• The other with the Grounding scheme that stops the oscillation.

Both should show how your scope is connected too.

Click to expand...

Here it is,
http://hpics.li/b5e4049
I put the probe's ground at the psu side (PSU_0) and I measure the rails anywhere on the PCB (same noise everywhere)

An interesting fact is that when I put the probe's ground at "PSU_0" and measure "GND" or "GND1" on the pcb I could also see the same oscillation on the scope.

saint gillis said:

An interesting fact is that when I put the probe's ground at "PSU_0" and measure "GND" or "GND1" on the pcb I could also see the same oscillation on the scope.

Click to expand...

That makes sense.
It's somewhat difficult to give a definite explanation why the system oscillates, because it involves unknown factors (internal stray capacitance in the chips, lead and PCB trace inductance...) but it was somewhat predictable. A stiff connection is needed between the reference ground (audio ground) and the "sewer"* ground.

One of Murphy's law in electronics says: "A circuit that is designed to oscillate won't ; a circuit that is designed not to oscillate will.

* I don't like this word because it implies it receives all the "garbage" and dumps it in a hole and it's lost and forgotten. It is not, Kirchoff ensures that all "dirty" currents are returned back to the circuit. The whole trick is to know where and how to return them in a way that does minimal damage.

saint gillis said:

ricardo said:

Have you tried 2 x Electrolytics AT each 553x and also the 1646?

Click to expand...

Hello Ricardo, yep didn't change much

Click to expand...

Put the 2 electrolytics back AT each 553x to your GND1

You are doing AT LEAST 3 things which by themselves could result in oscillation and WILL give worse THD even if there is no oscillation.  You need to get ALL 3 things right.

Here it is, http://hpics.li/b5e4049
This is the 2nd of those things.  In your "cure", the junction of your C15/16 is your "Electrican's Star Earth" (Sorry Abbey  ;D) which you have brought MUCH closer to where it should be.

It wants to be as close to important bits as possible.  Certainly NOT at the PSU 20cm away.  The only connection from the PSU_0 should be to this star point.

The 3rd thing is you are guilty of EXACTLY the sin I warn against in #6.  You need 1n or more from p3 of each 1646 to Vee AT the chip.

You really need to read and re-read Kingston's thread several times and try to understand what it means.

An interesting fact is that when I put the probe's ground at "PSU_0" and measure "GND" or "GND1" on the pcb I could also see the same oscillation on the scope.

Click to expand...

If there's oscillation, the yucky current MUST flow in the PSU rails and also in the track to the PSU.

Click to expand...

ricardo said:

This is the 2nd of those things.  In your "cure", the junction of your C15/16 is your "Electrican's Star Earth" (Sorry Abbey  ;D) which you have brought MUCH closer to where it should be.

It wants to be as close to important bits as possible.  Certainly NOT at the PSU 20cm away.  The only connection from the PSU_0 should be to this star point.

An interesting fact is that when I put the probe's ground at "PSU_0" and measure "GND" or "GND1" on the pcb I could also see the same oscillation on the scope.

Click to expand...

If there's oscillation, the yucky current MUST flow in the PSU rails and also in the track to the PSU.

Click to expand...

Hello Ricardo!
Yes thanks it makes sense.

I tried to bypass all the op amps rail to ground with a 47µ electrolytic and a 100n ceramic.
And I tried to put as you said a 1n cap between the 1646 pin3 and Vee.

I made THD measurements with RMAA (I don't know how accurate theese measurements are), but here's what I saw :

  - Quite the same curves with electrolytics rail to ground bypassing and with rail to rail bypassing using 100n caps for the nee553x
  - Quite the same curves with or without the 1n cap between  1646 pin3 and vee

  But when the ground paths are not linked at the pcb side causing oscillation, the 1n cap between 1646 pin3 and Vee stops the oscillation !

saint gillis said:

But when the ground paths are not linked at the pcb side causing oscillation, the 1n cap between 1646 pin3 and Vee stops the oscillation !

Click to expand...

Thanks for your test and results.

Can you post a pic of your RMAA results.  Don't forget to tell us the output levels & loads.

Each of the 3 things I mentioned could have solved your oscillation problem by themselves.  It would have depended on your EXACT layout .. especially for the different grounds.

My favourite spot for the Star Earth would be at the Output of your box.

But I would still do all the other stability stuff.  Trouble-shooting stability is VERY frustrating.  Often just connecting a scope makes it go away.  Even Great Guru Baxandall has stories to tell about this.

Thus the better designers do ALL the stuff that is good for stability .. whether they have a problem or not.

I wasn't quite sure 1n would be enough but your tests show its good.  Using 100n from p3 to Vee at the 1646 is 'certain' to work but it then puts the yucky stuff from Vee onto your nice clean GND. 

ricardo said:

Can you post a pic of your RMAA results.  Don't forget to tell us the output levels & loads.

Click to expand...

Hello Ricardo, sorry the last days 've been quite busy, here are the thd curves, output level on RMAA was something like -6db and I don't remember the manual of the tascam us-122mkII I use for my tests says it has a 2.2k ohms input impedance

http://hpics.li/3dd3b36

  "1" is with rail to ground decoupling with electrolytics and 1n cap across that1646, with the 2 ground paths linked at the PCB side
  "2" is with rail to ground decoupling with electrolytics but without 1n cap across that1646,  with the 2 ground paths linked at the PCB side
  "3" is rail to rail decoupling without electrolytics, without 1n cap, with the 2 ground paths linked at the PCB side
  "4" is rail to rail decoupling without electrolytics, without the 1n cap across that1646, with the 2 ground paths not linked at the PCB side
  "5" is rail to rail decoupling without electrolytics, with the 1n cap across that1646, with the 2 ground paths not linked at the PCB side
  "6"  is with rail to ground decoupling with electrolytics and without the 1n cap across that1646, with the 2 ground paths not linked at the PCB side

saint gillis said:

  "1" is with rail to ground decoupling with electrolytics and 1n cap across that1646, with the 2 ground paths linked at the PCB side
  "2" is with rail to ground decoupling with electrolytics but without 1n cap across that1646,  with the 2 ground paths linked at the PCB side
  "3" is rail to rail decoupling without electrolytics, without 1n cap, with the 2 ground paths linked at the PCB side
  "4" is rail to rail decoupling without electrolytics, without the 1n cap across that1646, with the 2 ground paths not linked at the PCB side
  "5" is rail to rail decoupling without electrolytics, with the 1n cap across that1646, with the 2 ground paths not linked at the PCB side
  "6"  is with rail to ground decoupling with electrolytics and without the 1n cap across that1646, with the 2 ground paths not linked at the PCB side

Click to expand...

Since pin 3 of the 1646 is the audio "ground", it just shows that it can't be left floating at RF frequencies; either a hard connection or via a suitable capacitor establishes a stiff enough HF path for restoring stability.
Both "wrong" cases are typical of impedant grounding.