Noise, and how to minimise it, in a small signal pentode circuit?

[jimboyogi]

Hi again,

I am designing and building an amp for a friend. The amp uses a gain stage coupled to a cathodyne PI, and a LarMar style post PI MV using a dual gang pot.

Last week I took out the 12AX7 that I had in there, and subbed in a 6U8A/ECF82, which I have a stash of.

Sounds great IMO, the pentode giving a voltage gain of ~ 150, which is able to overdrive the PI well. But the pentode is giving a fair bit of hiss, which is very audible at the output when the MV is turned up.

My question is this. Is there anything that I can change in the circuit to reduce the noise, or is this just the reality of a pentode producing this much gain? I have tried half a dozen different 6U8As, and the hiss is virtually the same for all of them.

Thanks for any and all help.

6U8A gain:PI DC.pdf

[Merlinb]

That stage is quite late on in the amp, so it is unlikely that the noise is coming from the pentode itself. The 220k mixing resistors will generate more noise than that!
For a start, I would reduce the input grid stoppers (the input of the preamp that is) to 10k or something, assuming they are currently 68k.

[jimboyogi]

Thanks for you reply Merlinb.

If I was a lot quicker at putting circuits into jschem, then i'd show the entire circuit, which would provide more information!

The input valve is a 12AX7. 1/2 is the input preamp (using a 10K input grid stopper) feeding the high Z bandmaster TS (from your pre-amp book), then a 1M "gain" pot. This feeds the 220K mix resistor at the pentode input.
The other 12AX7 1/2 is the reverb recovery stage, feeding another 1M "reverb" pot, then to the 470K mix resistor.

The mix resistors are dissimilar to adjust the relative signals of the dry and reverb, and the values are high to give decent isolation of the 2, allowing a strong reverb signal even when the dry signal is turned down on the "gain" pot, and vice versa.

The 6U8A gain stage/PI seems to be where the noise is being generated, as the noise is there when both the "gain" and "reverb" pots are all the way down.

Would a substantial amount of this noise be coming from the mix resistors? If so, are there any suggestions as to how to reduce their value, while still giving good isolation, without adding in any extra valves?

[jimboyogi]

OK, so when home from work, I tried out whether the 220K and 470K mix resistors (metal film) were causing noise.

First I turned "gain" and "reverb" pots all the way down, so that their wipers were grounded. Then I used a wire loop, and shorted across the 220K mix resistor, effectively grounding the pentode grid.

Most of the noise disappeared!

The remaining noise is very liveable with. So the next step for me is to find another way (other than large value mix resistors) to mix the dry and reverb signals together before the pentode grid, while still retaining enough isolation between their respective level controls/pots.

Any and all suggestions welcome!

[soundguruman]

capacitor, such as 7 pf, between plate and grid.

[Merlinb]

So the next step for me is to find another way (other than large value mix resistors) to mix the dry and reverb signals together before the pentode grid, while still retaining enough isolation between their respective level controls/pots.

I can't see an easy way of doing this, at least not without the whole schem. Maybe there is some clever way to inject one of the signals into the cathode or screen grid directly, without the need for a series resistor...

[jimboyogi]

I'll try and put up the whole schematic. It will take me a while, but it's all good practice.

The screen or cathode idea is very interesting. I actually started another thread here 2 days ago, to discuss bypass capacitors on the pentode screen and cathode in relation to this circuit, and the conclusion (thanks Gingertube!) was that for minimum noise, I needed to bypass both. If they are both bypassed, then am I able to use them as signal inputs?

One more thing. I tried changing the mix resistors from 220K/470K to 100K/220K. The noise contribution was audibly lessened, but when I shorted the 100K to ground the noise almost disappeared again. I am absolutely astounded by the noise contribution that these resistors can make in series with the grid. Especially since resistors of these values are used in series with the grids of numerous classic designs.
Is this the same for all valve types, or are pentodes particularly sensitive to grid series resistance, or is it just because of their relatively high gain?

[Steve Conner]

You might want to check whether the pentode might be oscillating. Try a grid stopper resistor right on the socket pin.

You could also try scaling the resistor values to reduce thermal noise. So for example, change the gain and reverb pots from 1M to 100k, and change the mix resistors to 22k and 47k.

The third possibility is that you have a bad batch of ECF82s with excessive grid current.

[jimboyogi]

Thanks soundguruman, I'll give this a try also. I've got some 10pF silver micas that I can try.

[jimboyogi]

Thanks for the reply Steve.

I've been scoping the pentode, and I have not seen any sign of oscillation. When I scope the cathode output of the cathodyne however, the aerial that the scope lead creates gives me something in the 50KHz range!
I will try to put a grid stopper on the pentode grid though, especially on the actual build. ATM the amp is just a mess in my prototype rig chassis.

With your suggestion of scaling the resistor values. I am chasing as much gain from the circuit as I can get, and I'm concerned that if I reduce the values of the gain and reverb pots, then the previous gain stages will be heavily loaded, and the gain reduced.
It's interesting in this amp that the only noise that is really obvious is that coming after the gain and reverb pots. When either of these is passing signal through to the rest of the amp, then the SNR seems very good. But with these 2 pots all the way down, and the master volume all the way up, that's the noise that seems out of place, and thats what virtually disappears when the mix resistors are shorted.
I will try lowering the value of the mix resistors as you suggested, when I lowered them from 220K/470K to 100K/220K I did notice a reduction in noise, so by continuing to scale them down, hopefully I can get the noise figure down satisfactorily.

Re. the valves, I've got a fair stash of these NOS, from various sellers, and a big range of brands. I'll get rolling and see what I can find out.

[jimboyogi]

I haven't tried all of the changes/schematics that I promised, but in my defence I did have to work through the weekend.

But what I have tried had interesting results.

First I tried further scaling down the mix resistors to 10K/22K. Using a 10K series resistor for the dry signal, there was only a small amount of extra noise generated. When i shorted the wire loop across it there was an audible, but small, decrease in hiss. But the noise level from the 10K is acceptable.
However using these 10K/22K mix resistors, the 2 "channels", dry and reverb, were way too interactive when using the "gain" and "reverb" pots. The biggest hurdle was the dry signal volume loss when the "reverb" pot was turned all of the way down. This set-up wasn't really usable.

So I decided that having the dry and reverb signals as 2 fairly independent channels could be sacrificed, for the benefit of a lower noise floor.

I replaced the 22K mix resistor with a 100K, so the ratio is now 10K/100K. This has almost eliminated any interactivity between the dry signal volume, and the "reverb" pot settings. But there is a great deal of interactivity between the reverb signal volume and the "gain" pot setting. So much so that the "gain" pot now controls both the dry and reverb signal volumes simultaneously, almost as if the 2 channels were mixed prior to the pot. The "reverb" pot now acts to control the mix of dry and reverb.

It sounds great IMO, and now the controls function like the expected classic amp reverbs, which is probably more user friendly that what I had previously.

[tubeswell]

or between plate and cathode?

[jimboyogi]

I've done some more work on this, and have attached an updated schematic.



Firstly, even though reducing the series input resistor for the pentode grid to 10K reduced the hiss a lot, I wanted to reduce it even further. So I removed the pentode cathode bypass cap. This increased hum a little, but reduced the hiss, probably just because the gain was reduced a bit. Then I introduced NFB to the pentode cathode. I tried 6dB. The noise was reduced so greatly that you cannot hear it, even with the MV on full. This was big progress.

Next up, I was struggling to get a good clean sound from the amp, even with NFB, MV on full, and the "gain" control on low. I tried to find the problem in the pentode, changing values, scoping, and listening. But I couldn't get what I wanted.

So I looked at the input mixer to the pentode. The way I had it, if the "reverb" pot was turned all the way down, then the AC load to the preamp input gain stage was only 100K//1M, which I suspect was loading it heavily and causing quite non-linear operation.

So I rewired the input mixer to what is in the updated schematic. Now I'm getting more of the dry signal into the pentode, and it sounds much more linear/clean.

If anyone can spot any problems/downsides to this set-up then please let me know!