High Gain Tube Preamp Hiss

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello everybody. I'm a "tube noob" and having trouble with this first amp. It sounds great, and all the controls respond as they should.

Only problem is that at max gain and max volume, there is nasty hiss from the speaker, more so on the OD channel.

Experimenting with caps, it has been found that placing the following across the specified terminals (grid to cathode), the hiss is reduced considerably, but the upper edge of the tone is compromised, which is undesirable.

Affecting both channels:
V3 7-8 470p
V4 7-8 270p

Norm:
V1 2-3 270p

OD:
V2 2-3 270p

The tubes are all 6N2P-EV
Pre 1 .JPG
Pre 2 .JPG

Any ideas or pointers much appreciated.
 
Last edited:
I do also notice that there is a lot of HF gain ... the cathode bias capacitors are very low value. Normally use 22 to 47uF.

Marshall uses 0.68u. It's worth an experiment, though.

I also have regulated DC heater supply, which is currently not using fast recovery diodes, but they are bypassed with 0.1u. Moreover, because voltage regulation at full power is so poor, I had to increase the dropout margin considerably. This is causing the rectifier bridge and the regulator to run rather hot. I was wondering if this thermal noise can make it into the signal path.

Thank you for this idea, too.
 
Hiss generally comes from 2 sources, resistor thermal noise particularly resistors on the grid (grid stops and grid leaks). This is proportional to the size of the resistor.

Random variations in grid current, this is best handled by shunting to 0V by use of low value resistance from grid to 0V (looking away from the grid back toward the preceeding stage). Again, low value grid stop and grid leak is your friend.

This is not a good design with respect to noise. The interstage attenuators are not agressive enough and the resistor values are too high. Also some of the grid stop resistors are too high, probably in an attempt to tame high frequency response ( they form a Low Pass Filter with the Miller Capacitance).

A better way is to use lower grid stop and grid leak to shunt grid noise to 0V and to handle the high frequency roll off by adding a cap from anode to 0V. Adding these HF roll off caps, particularly in the early stages (where there is a lot of gain down stream) should help.

If looking for a highish gain design that was done reasonably well (with respect to low noise design), have a look at the Carvin Legacy Lead Channel (just one example)

Cheers,
Ian
 
Last edited:
Thank you very much for you input, Ian.
Hiss generally comes from 2 sources, resistor thermal noise particularly resistors on the grid (grid stops and grid leaks). This is proportional to the size of the resistor.
All the resistors are over-size, overkill.

Random variations in grid current, this is best handled by shunting to 0V by use of low value resistance from grid to 0V (looking away from the grid back toward the preceeding stage). Again, low value grid stop and grid leak is your friend.

This is not a good design with respect to noise. The interstage attenuators are not agressive enough and the resistor values are too high. Also some of the grid stop resistors are too high, probably in an attempt to tame high frequency response ( they form a Low Pass Filter with the Miller Capacitance).

A better way is to use lower grid stop and grid leak to shunt grid noise to 0V and to handle the high frequency roll off by adding a cap from anode to 0V. Adding these HF roll off caps, particularly in the early stages (where there is a lot of gain down stream) should help.
How low?

If looking for a highish gain design that was done reasonably well (with respect to low noise design), have a look at the Carvin Legacy Lead Channel (just one example)
Cannot find schematic.

Ok, assuming it's not resistor noise, then it must be either tube noise, RFI, or noise from other electronics.

Testing for RFI suceptibility, turned everything up and switched a flourescent light on/off that was plugged into the same socket. A click was heard. A microwave oven made really loud clicks when it started and stopped.

I read here Spice and the art of preamplifier design, Part 1 a particular common mode power supply filter worked good. I have one but it was out of an ungrounded TV set. It is also where I learnt the RFI test.

This one has two "secondary" caps that are grounded in between. It also claims differential mode attenuation. http://www.mouser.com/ds/2/337/85103004-24804.pdf

My cabinet is completely enclosed and the preamp and sockets bottoms are fully shielded with copper lined box. I mentioned all this since it would be better to prevent RFI from coming in (if that's what it is) than to attenuate it within the circuit. Caps definitely kill highs and harmonics.

Here's one schematic http://www.carvinservice.com/crg/schematics/MTS3200%20REV-E.pdf

What do you think?
 
Last edited:
When I say resistor size I actually meant resistor value, higher resistance mean higher thermal noise.

Resistance from the grid back to 0V should be kept low - that means that a interstage divider of 470K + 1M would be better as a 220K + 470K or even 100K + 220K, since from the point of view of the following tube, the grid to 0V resistance is lower and so more grid noise will be shunted to 0V. Actually the 470K +1M pot in your circuit means that at max gain you have 2/3 of the signal passing through. A 470K + 500K pot or even a 470K + 220K pot would most likely be better (more minimum attenuation) - ? Do you ever use the last few degrees of that pot ?

I have attached what I have of the Carvin Legacy schemo. Note that they put the high frequency roll off caps across the anode load resistor. This is identical to putting the cap from anode to 0V but the anode to 0V way of doing it has better inmunituy to any noise on the power supply itself. Also note that while not particularly low, the resistances back to 0V when looking away from the tube grids is never more than 470K. There aare probably better example circuits to highlight what I said but this was one which came to mind and I knew I could find.

Cheers,
Ian
 

Attachments

  • carvin_legacy_vl100_vl212.pdf
    116.9 KB · Views: 134
With the 500K OD gain pot, the last 10 degrees are practically useless.

I like the idea of bypassing to ground with the anode caps, if that will reject any PS noise.

Thanks. I got you on the resistor values. Less resistance = less noise.

I also noted with the Carvin the non-traditional input scheme. No gridstopper at all, and the bypass cap on the low value gridleak.

My noise starts at or before V1B, since hiss is attenuated with 6.8n at the anode. That means the input and/or the valve or its wiring itself. I will start in that area first, since it affects everything down the chain.

Also noteworthy are the massive values of the cathode resistor bypass caps. I tried inreasing those to no avail. My research found higher values than 1uF simply allow more bass through; a somewhat moot point in a lead amp, in my opinion. I have a 10uF film cap and it is physically huge.

This problem is bothering me so much I refuse to play!
 
The OD Gain Pot is VR1? The last 10 degrees of a log pot is at least 1/2 the total resistance. Change the 500K pot for a 220K or even 100K and at the same time you can reduce that 470K grid stop as we are no long overdriving that next stage so badly. Drop that to 47K or even 22K.
Both these things also will help to shunt more grid noise from that next stage because the resistance to 0V looking away from that grid is now much less.

Thed Input Stage is of course the most critical. Yopu want to run it at reasonably high current (say in the 0.7 to 1.0 mA range) whilst keeping grid current low (so that noise from grid current fluctuations is also low). That means keeping the B+ to the first stage high. You want at least +200V on the anode of that first stage, a bit higher is even better. The B+ decoupling resistor for this stage does not have to be right at the end of the chain, you can connect it back to the start of the chain if you like (so it is in parallel with the rest of the chain) in order to keep B+ high.

Just some random shots at the barn door.

Cheers,
Ian
 
The OD Gain Pot is VR1? The last 10 degrees of a log pot is at least 1/2 the total resistance. Change the 500K pot for a 220K or even 100K and at the same time you can reduce that 470K grid stop as we are no long overdriving that next stage so badly. Drop that to 47K or even 22K.
Both these things also will help to shunt more grid noise from that next stage because the resistance to 0V looking away from that grid is now much less.

Thed Input Stage is of course the most critical. Yopu want to run it at reasonably high current (say in the 0.7 to 1.0 mA range) whilst keeping grid current low (so that noise from grid current fluctuations is also low). That means keeping the B+ to the first stage high. You want at least +200V on the anode of that first stage, a bit higher is even better. The B+ decoupling resistor for this stage does not have to be right at the end of the chain, you can connect it back to the start of the chain if you like (so it is in parallel with the rest of the chain) in order to keep B+ high.

Yes, VR1 is the OD gain. If you look at the schematic above, OD runs right across the bottom. So V1B is the first input stage, V2 is two overdrive stages and then out to V3B which is followed by the DCCCF.

The normal channel starts with V1B then goes to V1A (around V2), then to the next stage V3B just like the OD. Note that LDR1 is essentially a dead short when in Normal (~200R). This stops the signal from going anywhere.

The B+ is already parallel feed for V1 and V2 and designed for 185v at V1. They both are taken from C which is a little higher. They both use 15k and 10u F&Ts.

Well, I spent a month finding the nastiest oscillation. It was due to a ground loop that made every little imperfection 100x worse than it would have been, but I learned much from the experience. Everything is critical in high gain. Also, I have read it's better to distort the waveform gradually among many stages. In fact, the Carvin Lead channel does just that!

What's another month or two for this problem?
 
Last edited:
High gain tends to give you hiss, because it's amplifying the noise as well as the signal.

A valve specifically designed for low noise was the EF86, which was a low noise pentode, intended for the front end of tape recorders and mike preamps.

Noisy triodes are usually used in guitar amps because hiss and noise isn't really a problem, if you want high quality use solid state instead of valves.
 
Thank you Nigel for your interest and insight.

High gain tends to give you hiss, because it's amplifying the noise as well as the signal.

A valve specifically designed for low noise was the EF86, which was a low noise pentode, intended for the front end of tape recorders and mike preamps.

Noisy triodes are usually used in guitar amps because hiss and noise isn't really a problem, if you want high quality use solid state instead of valves.

For now, solid state for the pre is out of the question. I will get it as quiet as possible and have to settle for that. I found the "tone killer" caps' ill-effects can be remedied with tweeters. A large horn or two would be needed in conjuction with 4-12s, I believe. I used a 5" and a 2" bullet and it was ok. Basically, the combo restored what was lost. I had L-pads and adjusted to hear each individually, until one of them burnt up trying to find the hiss at 100% output. The one literally burned the cabinet wood.

I was wondering what that stink was. Also the power amp was cutting out.....
 
I don't quite see the reasoning in cutting the treble, then boosting it at the tweeters?.

It is a less-than-optimum way of attenuating the noise without losing the character of the amp. To "fix" the problem, we are taking the upper edge off the tone (all tones in fact), then restoring the harmonics with a more responsive driver in that range. So much for those who believe tweeters cannot be used in guitar amps. I believe there are some cabinets that utilize them. Maybe more for acoustic guitars. Isn't that one big reason why people buy Martins--for the harmonics?

I have found that, in life in general, it's always better to treat the cause rather than the symptoms. This is why I said it would be better to stop whatever it is, whether noise or RFI, at the source.

In the medical industry, it's common practice to treat the symptoms and not the cause. That's one reason why most everyone is sick in the USA.

So I started at the beginning of the chain. I know it's not the instrument, since everything is the same whether plugged in or not. I had reduced R2 (series input resistor to V1B) to 60k. Just a minute ago I paralleled it with a 10k and the hiss was greatly reduced. But the long leads picked up 60 cycle noise and some RFI (everything was open for this test).

There is an RFI cap across grid and anode which would have to be adjusted from 120p to 680p with a 10k. Or maybe eliminate it altogether like the Carvin. However, their scheme parallels a very low value grid leak resistor with the guitar pickups, imposing a considerable load, especially on high gain versions.

Reducing the value of the grid leak there did nothing.

This is where I learned that resistor can be reduced from the traditional 68k value: The Valve Wizard

Thank you very much for your comments.
 
If you have access to an oscilloscope, try if any of the amplifier's stage is self oscillating at an inaudible frequency, it also can give this sound, and also can disappear adding capacitor sin some points at the amp.

Good luck!

Thanks for that. I really wanted to pinpoint where the areas were, but the only scope I have is Visual Analyser. We can kill the noise with caps, but the tone suffers considerably.

I still don't know if it's Johnson noise, RFI, or ultasonic oscillation. I had the idea of using clamp on RFI filters, but they only attenuate a few hundred ohms. They might be helpful in isolating the area. When I had ground-loop-caused oscillation, I split a ferrite with a Dremel and glued the pieces to tweezers. When closed around one of the ground wires, the oscillation changed its tone.

I just tried reducing R2 to 10k and left the 1M grid leak alone, which did nothing. The Carvin setup made a radio receiver.

If a cap is added to either, grid to cathode, anode to cathode, or if the anode resistor is bypassed with a cap, all result in hiss attenuation.

I am tempted to try those pentodes, at least for V1 and V2 but it sounds like a can of worms for someone like me. Besides, I heard the new productions are no good, and I'm not spending $200 on any valve.

I found a lesson on local feedback for preamp valves. I will study that more. That was mentioned above, but I wasn't sure if he meant on the tube output section which I do not have.

I am exhausted.
 
Thanks for that. I really wanted to pinpoint where the areas were, but the only scope I have is Visual Analyser. We can kill the noise with caps, but the tone suffers considerably.

I still don't know if it's Johnson noise, RFI, or ultasonic oscillation. I had the idea of using clamp on RFI filters, but they only attenuate a few hundred ohms. They might be helpful in isolating the area. When I had ground-loop-caused oscillation, I split a ferrite with a Dremel and glued the pieces to tweezers. When closed around one of the ground wires, the oscillation changed its tone.

I just tried reducing R2 to 10k and left the 1M grid leak alone, which did nothing. The Carvin setup made a radio receiver.

If a cap is added to either, grid to cathode, anode to cathode, or if the anode resistor is bypassed with a cap, all result in hiss attenuation.

I am tempted to try those pentodes, at least for V1 and V2 but it sounds like a can of worms for someone like me. Besides, I heard the new productions are no good, and I'm not spending $200 on any valve.

I found a lesson on local feedback for preamp valves. I will study that more. That was mentioned above, but I wasn't sure if he meant on the tube output section which I do not have.

If you unplug V1 does the hiss disappear? - it's usual to only use a low noise pentode for the first stage.

Feedback reduces gain, and increases quality - presumably two things you don't want to do?.

However, removing the capacitors on the cathodes will give you local negative feedback, the capacitors also seem rather small? (although I haven't bothered doing the calculations), which would give higher gain at higher frequencies and make the hiss worse.

So try disconnecting the cathode capacitors one at a time and see what effect that has.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.