PI/NFB hum on a Laney GH100L guitar amp

[zkrx]

Hello guys,

I've been lurking for a little bit and have scoured the forum for related topics. I couldn't find what I was looking for however so I decided to create my own :).

I've had this tube amp for more than 10 years now. I'm its first owner. It is a guitar amp with four 12ax7 and four el34. You can find the schematic on the web easily:
https://www.thetubestore.com/lib/thetubestore/schematics/Laney/Laney-GH100L-Schematic.pdf

This amp had a power tube failure 10 years ago where the anode came in contact with the screen. It made a huge spark while rehearsing and killed the fuse. I've used it with only one pair of power tubes for years following that.

I decided to dust it off a couple of months ago, put in new tubes and set the bias voltage. I bought these tubes in 2013 and haven't used them since. I chose 4 EH preamp tubes and two matched pairs of EH el34. Here are the el34 characteristics as given by the supplier:

G1[-V]	H2C	S	G	G1
44	0	6.3	1.3	1.87
44	8	5.6	1.8	1.38
44	0	6.2	2.2	1.13
44	0	6.6	2	1.25

After having installed them, I realized that this amp had quite a bit of 50/100Hz hum. This probably comes by-design on this amp, 100W isn't exactly meant to be used as a bedroom amp. But I am now a bedroom guitarist, and I've decided to try and address that.

I've already chased down a couple of problems, notably:
- A 1k5 4W screen resistor was cracked open, probably following the tube short incident. I've replaced it.
- The master volume pot was faulty, changed it.
- I changed nearly all the important electrolytic caps, I don't think it had an impact (older ones were still good).
- I have 240Vac at home instead of the 230Vac this amp is meant to take.
- A good amount of hum seemed to come from the heater. It depended heavily on what preamp tubes I would use. I've modified a PC SMPS to output 6.2Vdc. I now use it to power my heaters (el34 included) and it seems to works fine. The hum diminished drastically.
- I've installed 1R precision resistors on one pair of el34, cathode side, so that I can measure the cathode current and deduce the idle anode current.
- I've set the bias voltage around -42V so that there's a 36mA anode current (Ik - Iscreen) flowing in those tubes at idle.

However, some hum remains, and it's still loud enough to bother when everything else is silent. This hum is present with v1, v2 and v3 removed (only the PI (v4) and the output tubes are plugged). If I remove v4 (power tubes only), it goes down to lower levels. It disappears completely when I finally remove power tubes.

I measure ~4.5Vac between any OT primary wire and ground. This makes sense since anode voltage is taken before the choke. (I measure ~100mVac after the choke). I've added two more 50uF reservoir caps in parallel with those already present at P51, making them 200uF in total, and the ripple voltage drops to ~1.5Vac. At that point, only with power tubes, hum is barely audible and nearly silent. However, it increases significantly when I plug back the PI (v4) tube. And I'm still trying to isolate why.

I've done some more experiments. With v1, v2, v3 removed (v4 to v8 present), desoldering the NFB wire from the OT secondary eliminates that PI hum. This means that the hum comes from the feedback circuitry.

With v1, v2, v3 removed (v4 to v8 present), the hum sound (very) slightly changes by rotating the presence knob. Hum level considerably changes when I enable/disable the resonance switch. The master volume has an influence on hum level too, even with the earlier stages disabled (and input to MV left floating!). At 0 (PI input to GND), hum level is high. As I increase the master volume, hum diminishes and completely disappears between 5 and 7. Turning the pot further makes the hum gradually reappear and it becomes loudest at 10.

Now here's my explanation:
Ripple finds itself injected back into the second LTP input (P27 on the schematic) from the OT through the NFB circuit. As it isn't common mode (this ripple isn't present on the first LTP input - or P24), it isn't rejected by the differential amplifier. When I turn the MV up, some ripple from the previous stages (or from the air when MV input is floating?) is injected into the first LTP input (P24). At this point, it becomes common mode and starts cancelling its image found on the second input (P27). Between volume 5 or 7, ripple levels are matched on both inputs (P24 and P27) and the amp becomes silent. Beyond, first input ripple (P24) takes over.

[zkrx]

Other stuff I've tried:
- Moving the NFB wire. Hum doesn't seem to change. Implies that it isn't induced by surrounding magnetic fields.
- Bypassing the NFB portion of the PCB by unsoldering component leads and connecting directly to the OT secondary with a wire. Hum still there with the same behavior.

Thereby my questions:
- Does it all make sense?
- Should I try to get rid of the hum by:
    - filtering the pre-choke supply further (with a resistor - RC filter)?
    - adding individual bias pots for each power tube so that I can precisely balance them and hopefully eliminate that ripple before it spreads through the NFB?
    - adapting the LTP tail resistor (R32, currently 10K) to better reject that ripple at the PI stage? This assumes that there is some ripple present on the first input when volume is 0 (GND). Otherwise, ripple could still be injected somehow.

I'm a bit surprised because I found NO mention on the web of ripple noise spreading from the OT to the PI input through the feedback loop. And OT supply without choke filtering seem common with push-pull topologies.

Thanks for reading

[shooter]

desoldering the NFB wire from the OT secondary eliminates that PI hum. 

so with that condition and nothing plugged in, the amp is quiet?
IF so, does it stay quiet when you crank up gain/volume?

[sluckey]

Be sure the OT secondary has a good connection to ground.

OTOH, is the hum objectionable in a room that can seat 500 people and also has a dance floor?

[PRR]

Welcome.

Much of your work seems right. I would urge *four* 1 Ohm resistors to be sure you don't have a dead output tube; a 2:1 unbalance is sure to inject PS buzz. The NFB connection should REDUCE buzz, so something IS wrong there. Does signal gain go up or down or the-same when NFB is connected?

Your theory of buzz-path through OT fails because the two sides of the plate winding should cancel. Or not, if the two sides are not reasonably equal (bad EL34).

And as Sluckey is saying: hundreds of hep-cats(*) in the room will absorb or overwhelm noise which is punishing in the house. (*)One acoustician wrote that dead cats will help many room problems by simple absorption, but cat-lovers object, anyway hep-cats pay cover-charge, cats do not.

[zkrx]

Thanks for helping out. Here are my answers.

so with that condition and nothing plugged in, the amp is quiet?

Yes, silent. Remaining hum comes from pre-choke ripple and is barely noticeable since I've added 2 more filter caps.

IF so, does it stay quiet when you crank up gain/volume?

Yes, it does. At some point, hiss comes in from the preamp, but this is expected and perfectly fine. And approaching full volume, another source of hum, louder, takes over. Also, there's a point along the MV curve where the hum decreases (or even disappears completely). I believe that I bring ripple to the signal input of the PI and therefore both input ripples cancel out.

Be sure the OT secondary has a good connection to ground.

I've triple checked and all seems good on that front. It take ground from the chassis, but I don't think this is a problem. I've also compared with pictures of the same amp that I could find on the web.

OTOH, is the hum objectionable in a room that can seat 500 people and also has a dance floor?

This is why I said this is probably more of a design "problem". This amp clearly isn't meant to be played in a bedroom. Most users won't even notice it. Between one's earplugs, a 400W Ampeg SVT-4 on one side and a drummer banging in the background, this is hardly an issue. With that being said, this would be annoying in a studio situation. Live, these are usually mic'ed and the sound guy would have to gate that during quiet passages one way or another. Perhaps my expectations are simply too high, but this makes for a good challenge anyway. Besides, I've never played in front of dead cats .

So I went ahead and soldered the two missing 1R onto the other pair. I measure the same cathode current on all 4 tubes, that is 40mA +- 1mA. I also tried another PI tube, same thing.

I took several DC readings around the PI with no input signal. I am not familiar with tube circuits, but nothing strikes me as obviously wrong. Voltages are a bit higher than what the schematic says, but Laney seems to have changed the 25W resistor between P54 and P55 from 22k (schematic) to 10k (my amp), so that must explain it.



I also measured the DC resistance of each side of the OT primary and got 17R6 vs 16R3. I don't know how this OT is physically wound, so I'm not sure that I can infer anything from that.

I plugged a guitar and tried playing with and without NFB to compare. It's hard to tell by ear (it takes time to discharge caps, double check they effectively are, resolder the wire), but without NFB, volume is similar. A bit louder, if anything. Only hum is (almost) entirely gone. I also tried cranking up the amp a bit. Nothing weird happened, no squealing or anything. The amp played as usual.

My best guess is that the output stage isn't perfectly matched and filtering is a bit lacking before the choke. Some ripple goes through the output. NFB then brings it to one side of the PI where it enters the power section again. At 100W, this amounts to a considerable volume.

I will start poking with my scope to get a better insight.

[zkrx]

OK, I've got the culprit: ground loop between PCB ground and chassis ground. The 480V rectifier shares its ground with the entire PCB. The latter contains all the preamp stages and NFB circuitry. You can find pictures of the amp's interior easily on the web for those who are curious. The PCB is at the bottom of the pics, where the front panel potentiometers are. The NFB circuitry is on the right, just before the diode bridge:



I'll post some pics of my troubleshooting just for fun. For what's below, the scope is grounded to the chassis and probes are AC-coupled. Refer to the GH100L schematic:
https://www.thetubestore.com/lib/thetubestore/schematics/Laney/Laney-GH100L-Schematic.pdf

CH1: P32 (EL34 grid, bottom half)
CH2: P31 (EL34 grid, top half)
NFB: removed
description: Here, the NFB wire is removed. There is no buzz from the speaker. The scope shows common-mode hum (50Hz) on all grids that is tamed by the CMRR of the output section. We can barely hear it in the speaker. This is most likely induced by the power transformer's magnetic field and is the reason why I had to remove the transformers from my Yamaha active studio monitors, but I digress.


CH1: P32 (EL34 grid, bottom half)
CH2: P31 (EL34 grid, top half)
NFB: connected
description: With NFB soldered back, we can hear buzz from the speaker. The scope shows that a 100Hz component is modulating the signal if you squint hard enough (sorry, should have averaged that!). This happens on one half only (the one where feedback is applied), hence why there's no rejection from the output stage.


CH1: C13, MV-side (PI signal input, before coupling cap)
CH2: C14, NFB-side (PI NFB input, before coupling cap)
NFB: connected
description: here are both LTP inputs. Buzz is clearly apparent on the NFB input.


CH1: P32 (EL34 grid, bottom half)
CH2: C14, NFB-side (PI NFB input, before coupling cap)
NFB: connected
description: this one is just for fun. It shows the NFB input to the LTP on one channel and the corresponding input to both EL34 grids on the other.


At this point, I re-verified that everything was correct on the push-pull and output transformer sections, but couldn't find anything wrong. I became suspicious of a ground loop on the NFB path. Looking at the schematic, there aren't many places where signal goes to ground: the EL34 cathodes, R44 in the NFB path, the presence switch (which I've removed already) and the signal input of the PI coming from the preamp (preamp -> tone stack -> MV -> PI). My scope being grounded to the amp chassis, I decided to probe R44's ground point on the PCB. Sure enough, here is our buzz.

CH1: P32 (EL34 grid, bottom half)
CH2: R44, ground side
NFB: connected
description: Buzz is apparent on the PCB ground. This is where it comes from after all. The B+ bridge rectifier must be polluting it.


With that in mind, I decided to confirm my deduction with a simple test: move all ground points of the output section to the chassis, like so:

With both grounds to the chassis (instead of the PCB ground trace), the amp becomes quiet, just as it was without NFB. Interestingly, with only R44 grounded to the chassis, the buzz becomes LOUD. Probably because some buzz is also present on the signal input to an extent. It means that there was some common-mode after all, just not 100% (and not visible on the scope).

There it is. I hope this may help somebody in the future. Perhaps a Laney owner will find this helpful.

[zkrx]

Now I'll be looking for a practical solution to this. I'll probably try to ground the diode bridge to the chassis first, so that the PCB ground trace remains free of buzz. Hopefully this won't mess with the output section. Any suggestion welcome.