New design humming - help requested

[rootz]

I loved the Benson Monarch reverb since the first day I played one. Simple, great sounding amp. and I love the über simple reverb. But I wanted a bit more: tremolo. So I thought, how about a Tremolux 5e9a kind of tremolo? Works on the paraphase cathode, so should work in a Monarch. I figured it should sound like a bias wiggle trem, as it pushes the Pi into cut off and back. Great, I like that!

I went on to design a PCB for a small chassis. Have done quite some PCB's years ago for pedals and tube hifi. I thought I could do it again, but I might have been mistaken about that.

The amp hums!!

I get a 320mV pp 100Hz signal at the 8 Ohm output with every dial on 0 on the scope. My Fluke 117c measures around 60mV RMS at the same output (yes, those deviate, not sure why, not that relevant at this stage).
First step: pulling tubes. Only pulling the PI quiets the amp (and the output tubes obviously). So the problem seems to be at the PI. Grounding the grid of the first triode does not influence the hum in a meaningful way. Changing the PI tube for another neither. Output tubes perfectly matched, at least at idle. PI not perfectly balanced, but as good in another silent Monarch I built some time ago.
Grounding the left output tubes grid totally quiets the amp. On the right side grounding the grid halves the hum signal at the output. I still don't get why that is. Grounding the grid of the second triode of the PI also quiets the amp. So it looks like the hum is injected after the grid of the first triode, but before the grid of the second one? That would leave the anode or cathode of the first PI triode as injection points.

What did not mitigate the issue:
- Pulling tubes other than the PI (or outputs);
- Totally disconnecting the tremolo triodes;
- Reflowing solder connections around the PI and trem tubes;
- Replacing the PI cathode resistor;
- Connecting different ground points with clip leads (trying to find ground loops);
- Using perfectly matched output tubes;
- Moving wires around (short wires, so not much movement possible).

What is strange:
- PI anodes are at 152V and 160V with a known good Philips ECC83. I expected a much higher voltage, more in the range of 185-190V. Other preamp voltages are within 5% of calculations and simulations.
- Hum seems to vary with B+ voltage. With a JJ 5Y3 I get less hum compared to a TAD 5Y3. The TAD gives a 40V higher B+, so also more current draw in the PA. More ripple, more hum?
- B+ carries a 7.5V rms ripple. That is a bit much, but shouldn't be a problem in a PP output stage.

What does work:
- Turning up the Depth pot lowers the hum considerably. It also shifts the operating point of the PI up the load line towards cut off, so that is expected when the PI is the hum source;
- Disconnecting the PI cathode resistor and bypass cap and bypassing them of the pcb with equal components halves the hum.

There is a clue hiding in the PI cathode changes, but I'm not sure what it is. Why are the resistor and/or cap apparently picking up hum on the PCB, but not when mounted off board? And what steps could I take next? I'm happy with halving the hum, but need at least a 90% reduction to have a useful design.

Layout and schematic added. Hope you guys can take a look and give me some more clues. Maybe the grounding is just too messy and prone to hum problems?

[xtian]

Examine your grounding scheme. I can't suss it out from your drawings.

[bepone]

In first line without further looking,
pcb design is not ok and needs general redrawing (and reconnection)

[R.G.]

Yep, PCB grounding scheme is the most likely culprit.

It may be the PCB combined with how your connections to chassis ground is done. I can't figure it out from the diagrams either (no part designators shown), but one possible issue might be grounding the tremolo depth pot (and in fact the whole tremo circuit) partly to chassis ground. Depending on what has been done with chassis ground, that could inject ground hum into the PI. It's not clear how the tremo switch ground and tremo speed control ground is done. If these go to chassis, not back to the circuit on the PCB, you could be picking up hum to be amplified by the tremo circuit and injected into the PI.

There is a whole litany of grounding guidelines regarding chassis and circuit grounds that could be involved. On the PCB, it would help to make sure that the first filter cap (C26 in the schemo) negative terminal is where the PT CT connects. Then take one trace from there to a star ground polygon. From the star-ground polygon run ground traces to circuit clumps by section. Don't ground tone/volume controls to chassis, run a wire back to the PCB star point. In general, use the chassis as a shielding shell, not a conductor. Things like that.

[rootz]

Thanks all for your contributions. I think I've found the biggest culprit already and as expected that was in the grounding scheme. I grouped trem, screens, B+ and output tubes cathode grounds together. Those went to the main ground point near the power transformer. That is where I normally ground the B+ caps and centre tap in a Dumble build. Reverb and V1 go to the ground point near the input. Apparently this grounding scheme was a mistake and I think the high ripple current at the main filter cap polluted the whole ground, including the PI ground.

So I looked at the 5E9A Tremolux from Fender. In that amp everything but the output cathodes, B+ cap and centre tap are grounded near the input jack. That is a big difference to my PCB! So I cut the ground trace between B+ filter cap and the rest and reconnected according to Fenders schematic. Boom! 20 fold reduction in 100Hz hum. Now 3mV instead of 60. That is a lot better, but still not close to what could be achieved. Yesterday evening I found out that the cathode resistor and cap of the PI picked up some hum. Mounting them of the board lead to a 30% reduction in hum. So I made that arrangement permanent and added a shielded wire from the depth pot directly to the cathode of the PI. And again Boom! Silent! 0mV residual hum.

[rootz]

R.G and Bepone, yes this pcb needs some redesigning. It was made in quite a hurry, which is never a good idea.

Currently the CT connects directly to the chassis, but cathodes of output tubes and mains filter ground go there as well. I will change that to a single wire and add the CT to the PCB.

All pots are now on a buss. I will ground them on the pcb at the stage where they are in the schematic and run a single wire from the pcb to the chassis. Can that star point still be at the input R.G.? That seems pretty silent?

[rootz]

Missed a point of you R.G. Trem is grounded at the buss bar going to the star ground at the input jack. It is not directly connected to the trem components on the board. It is now silent. I hear no hum to speak of and also no ticking artefacts. Will rewire as you suggested, which would mean all grounds on pots go to their respective components on the pcb. I believe your main point is that that is the best practice, right?

[bepone]

R.G and Bepone, yes this pcb needs some redesigning. It was made in quite a hurry, which is never a good idea.

i have redo power section because there you have a lot of problems..you are mixing all kind of heavy currents with signal.. extend the board to the right side (yellow), make rectification there, follow this and dont connect anything to first cap C1.. start to connect output cathode gnds on C2 (supplying the screens), and also PI cap gnd going to C2 gnd .. chassis is not connected to C1 in any way.

CT PT wire keep twisted with tube rectifier anode wires, and after rectifier keep twisted CT with that rectified B+, until arrives to C1..after C1 you are ok
to the C2 connect resistor for Ub+ line to the PI as you want.. will be ok

pcb.png

[bepone]

all tremolo circuit should be grounded together with PI gnd , the same point, minus pole of +BB+3 capacitor

[R.G.]

I believe it's the best practice, yes.

Star grounding is not the only way to get a hum-free setup. It's just the only way that is known to work ahead of building and trying. I like the term I heard here - "constellation grounding". This is a handier way to wire grounds with very, very, very nearly the same low noise and predictability of star grounding.

The wire from the PT centertap is special. It has current pulses through it that are several time bigger than the total DC current going into the amp. You want those current pulses returned to the PT through their own wire from the first filter cap negative terminal. The star ground point can be separate from the first filter cap negative, as only DC with full-wave rectified signal currents return to the first filter cap negative.

It generally works to just bolt the input jacks to the chassis, but unless you make that your star point (a you said you do) then it leaves the input signal ground prone to any wobbling on the chassis ground. AC leakage from primary wiring or capacitive leakage in the PT can inject hum then. I try to make the chassis just a shield for RF noise and safety ground, and not run any signal grounds on the chassis. I do that by having exactly one wire or contact from the signal ground/star ground point to the chassis. That makes it impossible for the chassis to inject hum to the signal ground. RF shield grounding on the chassis is probably best done with a ferrite bead on the input jack signal wires and a 100pF or so ceramic capacitor to a chassis grounding point right beside the jacks. I have heard arguments that insulating the jacks from chassis and running a small ceramic cap from both incoming signal and incoming ground to the chassis beside the jack is better at shunting RF to the chassis/shield.

[martin manning]

It's simple. You have to dance with the one that brought you.

[rootz]

It's simple. You have to dance with the one that brought you.

I had to look this one up. English is not my first language. I hope I haven't been disloyal to you or anyone on this forum. If so, I'm truly sorry and will correct it if possible.

[martin manning]

Nah, it’s a joke, Rootz. The current has to go home with the filter that supplied it.

[rootz]

My head and that joke weren’t even close to one another Martin, sjees.

In most part I’ll be getting back to the comments later this weekend. But thank hou alle for replying!

Here’s one question: would you say Peavey does a good job wrt ground management in say the first iterations of the 5150 or the XXX? The grounds in those amps seem to ‘flow’ from input jacks to main filter caps. In many cases the grounds of V1 and V2 are separated from the rest by means of a 47r resistor it seems. Also neat: shielding input traces with ground traces nearby.

Thanks again, I’m working on a new revision of the PCB. Main goal: (apart from hum free) build up like an eyelet board, but with the filter caps at their respective stages.

[rootz]

I'm going to take a bit of a risk with this design. I'm not yet incorporating all feedback you gave me. If that is wise? Time will tell. I partly stick to the grounding scheme Fender used in their Tremolux.

I've had great luck with ground planes in pedals and my hifi amps. Quiet, stable designs. Some are against it for tube amps, I think mostly because you have to be cautious wrt clearances and capacitive coupling of signal traces to ground or other traces. Then again, ground planes can provide some shielding and have other benefits. Peavey seems to use ground planes, though mostly limited to the input stage.

On the new boards all pot grounds return to the ground plane. The trem network grounds near the cathode resistor and cap of the trem circuit. The whole preamp including the input jack will be grounded near the input with only the anode B+ filter cap, CT and output tube cathode grounded at the other end of the chassis. CT grounds on the PCB to the first filter cap. I believe this is what R.G. meant.

Clearances of high voltage lines to all other traces and planes are double the required ones from IPC-2221 guidelines. DC and (possible) AC were taken into account there.

Testing with PCB's is cheap, so I'm not too afraid to make yet another mistake. I'll keep you posted about the results and share the design for personal use if it works properly.