EFK 4/12/2018 5:30 PM
Preamp DC filaments and capacitance
I'll try to make this short and concise but probably won't, sorry!

The amp is running two channels into a paraphase. 6J7G channel 1, 6F5G channel 2, 6Y7G paraphase. Initially I wired everything on AC and while the filament hum was not bad, the 6J7 channel was a bit too 'hummy' for my satisfaction.

So I shotgunned with the 6V battery for the preamp, perfect silence.

I know I should be looking at the regulated thing for DC but I've had very good luck in the past with the big diode bridge and capacitor approach.

Lifted the 6.3 vac center tap and used two 100 ohm resistors on the AC side (power tubes).

Ran the AC to the bridge and initially used an extra diode coming off the positive side for a bit of voltage drop but the DC voltages were still too high. All I had on hand for resistors were a bunch of .1 ohm 5W wire wound. So, I used one .1 ohm resistor on each side of the AC feeding the bridge (might offer a smidge of cushion at startup?) and one .1 ohm resistor on each side of the bridge feeding the tubes. DC voltage is now perfectly in range.

Used a 15K uf cap across the bridge, then began to experiment with grounds.

Ground reference: long story short, the best hum reduction came by using the dual 100 ohm resistors on the AC side, and a pair of .68 caps on the DC side, both resistors and caps running to the same isolated point, and that point being the top of the power tube cathode resistor. So AC and DC filaments are both floating on the cathode voltage, about 24V.

Measuring the hum with a Fluke at the speaker, 6F5 is literally inaudible. I measure 3-4 mV AC of 120 hz and this is the with the 6F5 channel volume completely cranked (of course, shorting input jacks). I'm very happy with that!

However, hum on the 6J7 channel is louder (with volume cranked) at 15 mV AC, and it is definitely 120 hz hum.

I know the 6J7 is a bit gainier, but I have the 6F5 set up pretty darned hot and that sucker is almost dead silent at max other than some ubiquitous white noise.

Now, it gets weird to my mind. I thought I'd test it out by tacking on an extra 15K uF cap to see if it helped, for a total of 30K uF across the bridge, and the hum got LOUDER. It jumped up to 23 mV AC, and believe me, it's an audible difference.

I've also tried different 6J7G tubes - I have a bunch - and the results are the same across all of them. Tried some EF37A which I thought were supposed to have better internal filament design and the results are still the same.

I know I'm going to get the "go regulated" comment but I just wanted to work with what I had on hand, so what am I missing here? Why would *adding* capacitance (added to the same points across the bridge, I might add) cause the hum to get louder? Given the current draw of the combined 6J7, 6F5 and 6Y7 (or 6N7 when I use one of those), the 15000 uF would probably be considered marginal and one would expect to use more, I would think?
 
mozz 4/13/2018 6:40 AM
I wouldn't go DC till a last resort. I've lifted the 6.3vac center tap and used the 100 ohmers a few times and that was more than enough to quiet it up. Layout usually is the cause. Was the original hum 60 or 120? Does the 6j7 have a pin for case?
 
EFK 4/13/2018 6:57 AM
I did try it on AC first, it was simply too much 60 hz filament hum especially with the 6J7. Didn't matter whether filament winding was grounded through center tap or through the pair of 100 ohm resistors. I tried a humdinger also, but not enough hum reduction for my taste so I went to DC. These are old glass tubes with the top grid cap but they are completely encased within old style tall metal tube shields, all grounded. The hum is now 120 hz and is definitely coming from the filaments of the 6J7. I've got the DC voltage right at 6.3-6.4 volts. The hum with the 6J7 channel fully maxed at volume is actually not bad, but compared to the 6F5 (which is silent) it's noticeable and it's bugging me.

I can't figure out why adding more capacitance causes the filament hum to get louder - it seems backwards!
 
mozz 4/13/2018 7:06 AM
Yeah i have 6j7 in a prewar supro and a film o sound, don't remember if they were g glass, but the metal do have a metal cap over the grid. Is the cathode bypassed? Is pin 1 grounded?
 
alexradium 4/13/2018 7:13 AM
you also have to consider the switching noise of diodes,i guess that's what you are hearing the most.
No matter how you try to filter that,using fast diodes that cost a lot and ceramic capacitors,the easiest and least expensive solution is regulation,either with LM3xx chip or NPN transistor with a zener,just provide adequate heatsink.
 
EFK 4/13/2018 11:04 AM
Yes, cathode is bypassed and pin 1/shell is grounded, although with the old glass coke bottle tubes there is usually not an actual pin 1 present since there is no shell to ground. The add-on goat type tube shields completely encase the tube and provide the rf shielding and they're grounded to the chassis via the attachment screws.

I thought about switching noise but it's never been a problem before with other octals i.e. 6sj7, 6sl7 etc. And as I mentioned, the 6F5 (triode) is perfectly silent. I'm wondering if there is something about the 6J7 (pentode) filaments that make them more susceptible to feeding noise into the audio path?
 
Mike Sulzer 4/13/2018 11:51 AM
If adding more capacitance makes the hum get louder, but you have no hum when using a battery, then I think there is a grounding issue involved. (By the way when you say "hum", I assume mean 120 Hz (or 100 Hz) and low order harmonics, not the buzzing sound of higher harmonics.)
 
EFK 4/13/2018 2:20 PM
I thought about grounding, but it would have to be a ground issue somewhere in the filament string - I don't see how a battery running the filaments would affect a ground issue otherwise. The main filter grounds, PT center tap, output jacks (isolated) and screens are grounded together near the PT on their own ground lug; power tube cathodes also run to this lug.

Preamp and paraphase are all grounded on the other side of chassis at the inputs (isolated). I did separate out the paraphase at one point to try it out, but it made no difference.

AS I described above, filament grounds as follows: two 100 ohm resistors on the AC side, and two .68 caps on the DC side, all of these running to an isolated tab on a tab strip, then the ground reference run to the top of the power tube cathode R. I tried initially grounding on AC side only, and then on DC side only, but it was louder both ways. I also tried initially running the ground reference back to a chassis point - as opposed to the cathode - but the hum was still there and since grounding the string through the cathode R made no difference I left it there.

About the only thing I haven't tried is grounding it through the DC side only with resistors or a humdinger and leaving the AC side floating.

I'm not sure if I can ground it through the DC side with resistors (as opposed to the caps) while leaving the AC side referenced through resistors as well. As I understand it from one of Steve Connors old posts, both AC and DC sides can be ground referenced as long as they are both symmetrical in relation to ground.
 
mozz 4/13/2018 2:45 PM
Short the input grid to ground. Does the top cap have the metal shield? If your gain is too high might also be a problem. What's the screen to plate resistor ratio?
 
EFK 4/13/2018 3:22 PM
Yes, the entire tube is encased in a tube shield and the grid cap fits through a small slot in the removable top portion of the shield. The grid wire is grounded cable, input jack is a shorting jack so the grid is effectively shorted. Gain is admittedly high, but not not crazy high. 300V supply voltage - 270K plate R, 1M screen R w/ .1 bypass cap, 1.5K cathode R w/ 100 uF bypass. Initially I had a 2.2K cathode R but dropped back to 1.5K, no difference either way.
 
Mike Sulzer 4/13/2018 4:11 PM
If I understand correctly that you have two separate connection points to the chassis, I think that can cause problems.

Quote Originally Posted by EFK View Post
I thought about grounding, but it would have to be a ground issue somewhere in the filament string - I don't see how a battery running the filaments would affect a ground issue otherwise. The main filter grounds, PT center tap, output jacks (isolated) and screens are grounded together near the PT on their own ground lug; power tube cathodes also run to this lug.

Preamp and paraphase are all grounded on the other side of chassis at the inputs (isolated). I did separate out the paraphase at one point to try it out, but it made no difference.

AS I described above, filament grounds as follows: two 100 ohm resistors on the AC side, and two .68 caps on the DC side, all of these running to an isolated tab on a tab strip, then the ground reference run to the top of the power tube cathode R. I tried initially grounding on AC side only, and then on DC side only, but it was louder both ways. I also tried initially running the ground reference back to a chassis point - as opposed to the cathode - but the hum was still there and since grounding the string through the cathode R made no difference I left it there.

About the only thing I haven't tried is grounding it through the DC side only with resistors or a humdinger and leaving the AC side floating.

I'm not sure if I can ground it through the DC side with resistors (as opposed to the caps) while leaving the AC side referenced through resistors as well. As I understand it from one of Steve Connors old posts, both AC and DC sides can be ground referenced as long as they are both symmetrical in relation to ground.
 
EFK 4/13/2018 5:51 PM
Mike are you talking about the filaments or the general signal grounds? The filaments are referenced entirely through one pair of output tube cathode R (there are 4 output tubes, each pair sharing a common cathode R), no other connection to chassis.

There are two connection points to chassis for signal grounds, but when trying to troubleshoot this, I lifted the ground lug at the input jacks (this grounds everything from the paraphase up through the preamp and inputs) and temporarily jumped it back to the main B+/screens ground and it made no difference at all.

I also wondered if the 6J7 was being overly sensitive to the magnetic field of the choke (choke is all the way at the other end of the chassis, though). This is a choke input amp and when running, if I stick a piece of iron anywhere near the choke laminations, it clings to it with a serious magnetic pull. Given the choke wires were long enough, I unmounted the choke from the chassis and moved it away an inch or so to see if there was some magnetic or mechanical hum being picked up, but again, it made no difference.

I 'feel' fairly certain that this hum is coming from the DC filaments somehow. Maybe the 6J7 is just an especially sensitive tube. I should give some other similar tubes a try, something like a 6L7 or 1612/1620 or something like that. I'll have to do some digging through the tube data...

I very much appreciate the replies everyone!
 
trobbins 4/13/2018 6:41 PM
Is this just one filament 6V winding being used to AC power all valves, except the input 6J7 which is supplied from rectified and filtered DCV but using the same heater winding as AC input to the rectifier?

Also just to clarify your initial testing with AC heater supply to the 6J7, did you use a pot tuned humdinger referenced to an elevated DCV (eg. output stage bypassed cathode bias) ?
 
EFK 4/13/2018 7:40 PM
This is a single 6.3 VAC 6A winding, powers four power tubes on AC, then is rectified to DC and powers the 6J7, 6F5 and 6Y7.

When I initially had all tubes on AC, I first tried just the filament winding center tap to ground, then I tried a humdinger. Unfortunately I did not reference the humdinger pot to an elevated voltage at any point in that test, although I'm not clear on whether that may have helped. Certainly no cathode followers in this amp.

I know this method of rectifying the same AC winding to DC is not considered ideal, but I've used it in the past with a wide variety of octal preamp tubes and never had any issues. I've not used it with 6J7 in the past, hence the reason I'm wondering if for some reason 6J7 filaments are particularly sensitive or something. And, I would have expected the additional smoothing capacitance I added to reduce the 120 hz hum rather than increase it measurably.

I can live with what I've got now, it's just that the 6F5 on the same rectified DC filament string is so remarkably silent, I am confused as to why the 6J7 channel would hum like this. And both preamp tubes are running off the same filter node, both grounded to the same point.
 
mozz 4/13/2018 8:03 PM
Swap in a metal 6j7 if you have one. I would make up a 6sj7 to 6j7 adaptor and try that also, it might narrow it down. If all else fails, pick a amp and use all those 6j7 values. Have to taken out the 100uf and tried smaller? Higher screen voltage will be lower gain i think. Maybe you are pushing the tube beyond its gain limits. Last suggestion is extra node just for that tube. Here is my filmosound schematic, my 6j7 values are not far from yours. I've never measured gain of that input stage though.
[ATTACH=CONFIG]48457[/ATTACH]
 
trobbins 4/13/2018 8:52 PM
Without photos, could you clarify some topics please:

Do you have a 6J7 grid stopper right at the top cap, and then use screened cable from the top cap to the grid leak, and ground the screen at the grid leak and cathode bias 0V common node ?
Do you have the standard steel can shield on the valve?
(Some example photos in this restoration https://www.dalmura.com.au/static/Ph...0amplifier.pdf )

What 6J7 cathode resistance and bypass cap is in circuit, and what cathode bias voltage is present? You may want to try increasing the bypass cap, to better short the cathode to 0V, or if the voltage is practical, then use a LED for cathode bias.

I'd try again to use the humdinger on the DC side (rather than across the winding), as 0.68uF is circa 2kohm impedance, so the DC heater terminals are still bouncing around at 120Hz with respect to the 6J7 cathode, and so the leakage capacitance and resistance between 6J7 heater and cathode has still got quite a high 120Hz signal across it, and that may not be completely bypassed by your cathode bias bypass, and so presents itself as a cathode to 0V hum signal, and hence as an input grid hum signal.

You may want to also try reducing the heater DC filter cap to say just 10uF - a volt of ripple is still much closer to DC than 6.3V of '60Hz ripple', and if that reduces the higher frequency rectified pulse harmonics flying around then that could be noticeable.
 
EFK 4/14/2018 6:54 AM
Yes, the preamp 6J7 and 6F5 are completely encased in can-type tube shields like in the linked article. The grid caps are ceramic with the grid stoppers soldered directly to the 'clamp' inside each ceramic cap, and there is screened/shielded cable running directly from the input jacks through the chassis and up to the grid caps. Each can-type shield has a top cap that fits over top the whole grid cap arrangement, so each preamp tube is entirely shielded from base to top. I ground the grid wire shielding to the chassis but not to any signal grounds, as it's shielding for RF and 'junk' much like the chassis itself or the tube shields and I believe I read one of Aiken's papers long ago that advocated for keeping any shielding grounds separated from signal grounds. At least, I think it was Aiken.

I've tried both 2.2K and 1.5K cathode R, both using a large 100 uF bypass cap. There doesn't seem to be any hum difference with either cathode resistor. I do have some red led's and they would probably provide a close enough voltage, so I can pretty easily give one of those a try.

If I lift the dual .68 caps off the DC side, and use either a humdinger or a pair of resistors, should I lift the dual 100 ohm resistors on the AC side? I'd like to try a pair of 100 ohm resistors on the DC side but don't know if I can do that while leaving the dual resistors on the AC side as well without burning something up! I have seen diagrams elsewhere wherein people have used the virtual center tap on the AC side and a pair of smaller (i.e. 39 ohm, if I remember correctly) resistors on the DC side, but I've never tried that.

I will also try reducing the size of the large cap across the bridge - I have a few different values between 15000 uF and 10 uF so I'll work my way down and see what happens!
 
EFK 4/14/2018 6:57 AM
I don't have one but I ordered one to try. The problem is that the bases on those metal tubes and the smaller glass GT types are larger diameter than the base on the glass tubes, and they won't fit through the tube shield base. I'm going to make a "socket extender" with an octal base and socket, so I can run metal 6J7 and 6F5 without having to remove the shield bases, just haven't gotten around to it. I have heard that the metal tubes are supposedly a bit quieter than the big old glass tubes.
 
EFK 4/14/2018 11:09 AM
Lower levels of capacitance cause the hum to get louder so it actually seems like 15000 uF is the sweet spot. Also tried a red led, cathode voltage stayed pretty close (1.8 volts vs. 1.6 volts w/ resistor) but the hum level was unchanged.

Because the 6F5 is so quiet, I keep focusing on the tube itself. I wonder if the 6J7 - being a fairly sensitive pentode - could be somehow picking up something from the chassis in a microphonic way? I would expect 60 hz in a situation like this but maybe there could be a source of 120 hz of which I'm not aware. I do not have the 6J7 socket isolated from the chassis with grommets, which I've often done with other preamp pentodes to soften microphonics.
 
trobbins 4/14/2018 7:16 PM
Perhaps redo the test with a direct short from input grid to cathode bias 0V node, as a benchmark. If you have a spectrum analyser scope (or PC and software), then that would be your benchmark noise+hum spectrum. That then allows assessment of the input wiring and grid leak and input socket shorting mechanism - as that input circuit scheme should give the same spectrum as the benchmark. If the benchmark has too much hum and noise, then the spectrum will identify the dominant contributor. Are you operating in pentode or triode mode - if in pentode mode then is the screen bypassed to the cathode?
 
EFK 4/16/2018 3:25 PM
It's running as a pentode, screen is bypassed to the same ground node as the cathode. Is there some advantage to bypassing the screen to the top of the cathode? In an old thread here, take a look at this diagram: [ATTACH=CONFIG]48510[/ATTACH] What is the purpose of the 1K resistor across the DC side of the bridge? Also, why use 100 ohm resistors on the AC side, but only 39 ohm resistors on the DC side? Why not just use 100 ohm off both?
 
trobbins 4/16/2018 7:13 PM
Pedantically, the screen voltage is being decoupled, and the screen voltage is screen to cathode. In the old days it would often have been advantageous to ground the decoupling cap to chassis rather than bring the negative terminal back to the cathode - those decoupling caps were often the largest part by miles in that area of circuitry as they had to be circa 0.5uF and 400-600V and so were the waxed paper style. The 1k would discharge the cap, but the 39 ohm humdinger will do that too. I wouldn't follow that graphic.
 
Dave H 4/17/2018 2:49 AM
[QUOTE=EFK;493168][ATTACH=CONFIG]48510[/ATTACH] What is the purpose of the 1K resistor across the DC side of the bridge? Also, why use 100 ohm resistors on the AC side, but only 39 ohm resistors on the DC side? Why not just use 100 ohm off both?[/QUOTE] Isn't one of those artificial centre taps redundant? Have you tried it with just the DC supply cap -ve grounded without using any artificial centre taps? That's how it is in my Valve Junior and it has no audible hum.
 
EFK 4/17/2018 5:18 PM
I'm not using a setup modeled after that graphic, I'm simply trying to understand the purpose behind the methodology.

I think I've got it as quiet as it's going to be without going to dedicated, regulated setup. It's certainly quieter than just running it on AC. I increased the dual .68 caps to ground (through the elevated output cathode voltage) to a pair of closely matched 10 uF caps. No particular reason, just playing around. With the volumes maxed out, input jacks shorted, I measure 5 mV of AC noise at the speaker on the 6F5 channel (pretty much all hiss, no hum that I can really hear) and 13.5 mV of AC noise on the 6J7 channel (this is 120 hz hum). This is listening to it and measuring it with a speaker right on the bench next to the chassis. When I actually hook it up to a speaker cab and stand a few feet away, it's actually a lot less noisy than any of my old MArshalls so maybe I'm being a bit anal. I think ultimately the 6J7 is just a super sensitive tube - especially when using the big st shape glass version - when run as a pentode and set up for a good deal of gain. I really appreciate all the help and education!