differences of noise coupling between transformers (G9 B+ problem solved)

[Kingston]

Furthermore, I didn't know it was possible to receive/induce any of the post rectification 100hz noise magnetically.

Me too,.
but if I were you, before doing major works, I'd quickly test with a thick shield between the PSU transformer and the audio PCB, or move the transf outside and see what happens..
just to rule out the PSU transformer's electromagnetically induced hum.

I've moved around and shielded the transformers (audio and PSU), I've moved the case to another room. I can optimise the 50hz (and oddly, 150hz) hum this way.

100hz stays at those fixed points.

 

[Jonte Knif]

I can optimise the 50hz (and oddly, 150hz) hum this way.

Wall power is rarely "pure" and furthermore the magnetic radiation from the mains trannie is affected by the current drawn from it. So, DO expect to see harmonics in the hum.

Any how, could it still be remotely possible that the current pulses in the power supply on the PCB cause enough stray magnetics to be picked up by the trannies? The nice 100,200,400Hz series fits just too neatly to the power supply scenario. This is just speculation, but the board is somewhat crowded. You would need an external power supply to try.

-Jonte

 

[Kingston]

Hahaa! success at last!

I moved B+ PSU out of the main PCB, and onto a separate veroboard. A little evening chore.

That nicely curving 100hz, 200hz, and 400hz in those images, completely gone. ;D Maybe it was the rectifier, maybe general positioning of those components, but that B+ part really has NO place on that audio PCB. And by the way, no external PSU needed here, just a separate B+ PCB or some p2p equivalent.

All that's left is some minor 50hz and 150hz, but all of that under -95dB. I think I can get it to below -100dB, and then I can finally bolt on all those transformers into the chassis and start recording.

I had the exact same problem with the GSSL PCB. Maybe it would be time for a little update, since they are so popular projects.. My guess is there are quite a lot of people who needlessly fight with these exact same problems.

 

[Neeno]

I learned a lesson about noise issues from SSLTech.
50Hz hum, is normally, induced hum from the power trannie,
100Hz hum is something "happening" after the rectifier.

Dunno if this helps, just my 2 cents.

 

[1954U1]

Great find, congrats! 

An important question:
do you've desoldered and used in the veroboard the same components?
diodes, the 2 big c14 & c15, tl783?

Just to be sure it wasnt a bad cap, etc..

 

[Kingston]

Yeah, same components, and my veroboard is anything but clean. I happened to only have the kind of veroboard that has those "pairs of three" solder slabs, and since I only had these used cut-for-pcb parts around I had to get creative. looks plain nasty from under the board but there's no noise.

Wish I knew what exactly it was that this fixed. The really interesting thing here is how much more edcor reacted to that hum, and how those spikes seemed to "stand" on the transformer noise floor. I wonder why that was as well.

And now I know I can't trust even a 250 euro multi-meter. Dammit. As if it wasn't difficult enough already.

 

[Samuel Groner]

Maybe it was the rectifier.

A strong indication that it is the rectifier is the presence of noise along with hum/buzz.

Samuel

 

[Jonte Knif]

Nice!

Kind of strange in a way, one would assume that the heater supply would cause the problematic magnetic fields because of much higher currents, but here we are. I also wonder why you didn't get different hum figures when you disconnected the other channel tubes. That would reduce the currents and therefore the hum too. I'm puzzled.

The really interesting thing here is how much more edcor reacted to that hum, and how those spikes seemed to "stand" on the transformer noise floor. I wonder why that was as well.

Why? I mean, Lundahl has a proper mu shield, edcor not, right?

-Jonte

 

[nielsk]

On this I have had the best results using a bridge very close to the power transformer for the B+ and +12, ( the shorter the high current AC leads the better) and an extra 100uf or so cap on the bridge.... and solder a ground lead to your star from each output pot, that dropped the noise floor way down....

 

[Kingston]

Nice!

Kind of strange in a way, one would assume that the heater supply would cause the problematic magnetic fields because of much higher currents, but here we are.

I was surprised too. I was prepared to move the heater filters and regulator out of the main PCB as well.

I also wonder why you didn't get different hum figures when you disconnected the other channel tubes. That would reduce the currents and therefore the hum too. I'm puzzled.

actually, isn't that a clue that it was indeed the B+ bridge rectifier that caused the noise? amount of current drawn through it would not affect the noise it leaks to the signal ground. G9 has a bit unorthodox B+ layout, with the big filter caps outside the PCB, and bridge rectifier ground tab physically *after* them. Maybe that makes it leak to signal ground.

The really interesting thing here is how much more edcor reacted to that hum, and how those spikes seemed to "stand" on the transformer noise floor. I wonder why that was as well.

Why? I mean, Lundahl has a proper mu shield, edcor not, right?

The edcor shields are actually adequate, now that I see only the PSU transformer induced hum.

The shields would not have any effect on those images (on the 100hz and multiples) since it was dirty power afterall. Somehow the edcors were more subject to it.

 

[1954U1]

Fantastic..
thanks a lot for all these details, thats going very useful for all of us with G9s.

Now, found that the culprit is related to the high voltage rectifier-regulator placement, I think at these 2 hypotesis:

1] dirty power.
but.. why your Fluke didnt catch the ripple?
maybe auto-range feature was enabled and this made it going mad?
[I have a Fluke 187, but thats never happened to me.. so far at least I thought so..]

2] electromagnetic induction from diodes-caps-regulator.
but.. do it was enough to move them away some cm..?


The mystery continues.. but at least we now know a way to solve the problem

 

[Kingston]

2] electromagnetic induction from diodes-caps-regulator.
but.. do it was enough to move them away some cm..?

how about half a meter? now that I have the B+ PSU separated, I can even touch the transformers with it. There is absolutely no noise. I really doubt it was electromagnetic induction of any kind.

As for fluke not catching the ripple, I have no clue.

Now I regret not doing this under heavy oscilloscope instrumentation. Seems there might have been more to learn from this.

 

[maxwall]

Kingston

great info that will be useful to troubleshooting for more projects than just the G9

could you snap a shot of your work with the relative distances of the boards away from the main PCB see we can visualize your scenario.

 

[1954U1]

2] electromagnetic induction from diodes-caps-regulator.
but.. do it was enough to move them away some cm..?

how about half a meter? now that I have the B+ PSU separated, I can even touch the transformers with it. There is absolutely no noise. I really doubt it was electromagnetic induction of any kind.

As for fluke not catching the ripple, I have no clue.

Now I regret not doing this under heavy oscilloscope instrumentation. Seems there might have been more to learn from this.

Half a meter?
I dont understand..
My quoted question was indeed an argument against induction hypotesis. 
[also if it remains the coincidence of Edcor's lack of shielding & more noise]

Anyway, I've understood that your PSU veroboard is really near to audio PCB,
So yes I agree, its time now for someone to do some tests with a scope and a not-modded PCB..

 

[steppenwolf]

Well, maybe it wasn't the B+ to be dirty, but the ground. I mean you have heavy ground currents pulsing into the ground traces really near sensible audio traces and transformers.
By placing the rectifiers away from the board and (at least I assume) connecting the groud carrying those pulses to the star ground you might have eliminated the dirty ground.

Best,
Stefan

 

[Kingston]

could you snap a shot of your work with the relative distances of the boards away from the main PCB see we can visualize your scenario.

I would have done that already if I had a camera for it. That's why I have only posted software screengrabs this far.

But there's not much to see in the layout. Just a standard G9 PCB (B+ area unstuffed) on the far left side of the case. Then two PSU toroids on the far right (haven't decided their final placement yet, maybe they will be on top of each other, maybe sideways, gotta test). Then in the between of these is the extra 5cmX10cm veroboard with the B+ with large caps. The placement of the extra B+ board does not affect noise at all.

I haven't decided where to bolt the output edcors either. They'll probably end up on the right rack side, kind of on top of the input trafos on PCB.

 

[qwerty8787]

Is the spare PSU pcb good for making the external board? the one that's coming with the other pcb's.

I saw that there are shunts between some copper tracks....not good

 

[Kingston]

I haven't tested it. I didn't get the spare board with my PCB set and just built the B+ part on a veroboard. Not awfully clean looking but works.

 

[keninverse]

Is the spare PSU pcb good for making the external board? the one that's coming with the other pcb's.

I saw that there are shunts between some copper tracks....not good

Pardon the ignorance but what exactly is a shunt path/track on a pcb?

I should explain...I did exactly what you're suggesting.  I used the extra psu for the HT rail.  I haven't ran it through a spectrum analyzer but to my ears the 100hz is almost completely non-existant.

 

[qwerty8787]

I mean that I found two paths connected to each other for mistake! maybe a printing error, dunno.
It makes me think about double checking the main board PSU, tracks are very crowded and I still have that 100Hz problem...maybe some accidental soldering  ???