High gain noise management

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all,

I have built an amp that basically runs 3 12ax7 halves flat out into eachother. It has a good amount of gain but it is noisy ( hummy and ofc hissy)

What I am trying to understand is this: what if I add resisitive dividers per stage but add another one .. would this be quieter?
You throw away noise with a divider but you also throw away signal, so signal to noise ratio stays the same, correct? So looking at it like this nothing would change. And the extra stage would add hiss/hum also probably.

Most high gain amps I see are based on Soldano's work and they use a lot of dividers and are quiet but maybe the dividers are more for reducing blocking distortion.
 
AX tech editor
Joined 2002
Paid Member
The noise performance is dictated by the first stage. The usual routine for lowest noise from multiple stages is to give the first stage the highest gain possible. That will give it the lowest input-referred noise.
Try something like current source loading.
And then see if you can cut back on the number of stages. Do you really need three?

Jan
 

PRR

Member
Joined 2003
Paid Member
... cut back on the number of stages. Do you really need three?
...A modest amount of feedback, either local or global, might pull this back from the brink.

Gentlebeings-- This is a guitar amp. MORE MORE MORE gain!! No NFB, we want the sweet(?) sound of 12AX7 abuse!!

Attaching the more obscene part of Carvin's vile plan.

The eleven cascaded stages, with interstage loss, might be a good FM limiter. I'm starting a sim of this plan but both my fingers and my sim are tired tonight.

edit-- sim chokes on 8 stages (eval version). Here's 4 stages, drive levels 10mV to 1V peak.
 

Attachments

  • Carvin-Quad-X-Amp-chain.gif
    Carvin-Quad-X-Amp-chain.gif
    36.8 KB · Views: 112
  • Carvin-Quad-X-Amp-4-chain.gif
    Carvin-Quad-X-Amp-4-chain.gif
    41.1 KB · Views: 105
Once you've solved all your other problems, stop expecting low-noise performance from common-or-garden variety tubes. The 7025 is the first stop as it's basically a better 12AX7

And while I think of it,
a) bandwidth limit your first (noisy) stage. Hard.
b) clean all the crap out of the ground and power for that stage.

Those are the easy things. The hard things are all the stuff we know about signal routing and cross talk.
 
Last edited:
Hi all,

I have built an amp that basically runs 3 12ax7 halves flat out into eachother. It has a good amount of gain but it is noisy ( hummy and ofc hissy)

3 12ax7 halves give 100*100*100 gain which is 1,000,000 !
So not surprising you are getting massive noise.

Screen input signal. Ground plane on your pcb.
Keep transformers and high volts ac away from audio signal etc etc
Keep heater wires away from audio.
 
The eleven cascaded stages, with interstage loss, might be a good FM limiter. I'm starting a sim of this plan but both my fingers and my sim are tired tonight.
Fascinating.

I'm contemplating how to get a "hyper-clean" sound and this is fascinating. Particularly the sets of pairs (each pair having a "normal" stage; an attenuator; then a cold clipper) and the absence of cathode bypass caps. I wonder if the bias shift (in your simulation) has an audible effect?
 
Last edited:
Account Closed
Joined 2010
Now i remember that I built(11 years ago) only the 5 stages circuit and compared with many 2 stages high gain circuits, like Mesa(soldano-slo100), it had no noise but a considerable level of drive...I never had the balls to clone the 11 stages...hopefully 2 X "nothing"= " nothing" +3db...everybody liked more the Mesa(soldano) sound for drive though...the main difference was in the voltage supply, Carvin had 3 times lower supply than Mesa...so it worked a bit as a starved anode too...
 
Last edited:
Account Closed
Joined 2010
here you have the channel 3 sound(5 triodes):
YouTube
YouTube

here you have the channel 4(11 triodes)
YouTube
each channel has multiple mods and variations.
Unfortunately they put so many drive mods on an amp who had one of the best clean ever...Clean players aren't really dead about drive...so they discontinued it for the right reasons.
A small review that tells a lot to an electronics engineer:
"Channel 4: Grosse distortion, but not Hi-Gain so far. Good sound, quite versatile, especially thanks to the équalo. A good distortion lamp, but ultimately lacks personality"
as channel 4 lacks high gain usual hum...the hum is the one that makes the high gain "recognizable high gain", Carvin has no hum in high gain...just distortion and a lot of it!
 
Last edited:
as channel 4 lacks high gain usual hum...the hum is the one that makes the high gain "recognizable high gain", Carvin has no hum in high gain...just distortion and a lot of it!
Agreed. IMD from power rail recharging/screen IMD seems to be a key component of the high gain power chord, even if they don't allow as much sag as earlier "mid gain" amps.

In the case of my "clean" chain I think I'll need each pair set closer to "clean" but with similar gain/attenuate so that all even stages clip at the same time, first into cut off (for lotsa 2nd) and then it's a race between the odd stages and grid cutoff.

I think.:eek: Maybe I want grid bias in the odd (warm) stages first :eek:
 
Last edited:
Account Closed
Joined 2010
In the case of my "clean" chain I think I'll need each pair set closer to "clean" but with similar gain/attenuate so that all even stages clip at the same time
not really true...Look at how Mesa(dual rectifier) implemented the Soldano slo100 gain stage vs Hughes and Ketner(Triamp)
Mesa used the 2nd and 3rd gain triodes in one valve , Ketner used them in separate valves.
The difference is that H&K had a higher output because using those two stages in one valve, there was some kind of negative feedback, while H&K needed to add an external feedback at higher frequency to keep it clean, though the gain is definitely higher just by using those two stages in different valves.
I built both versions and that is how i found out...
Carvin preamp has a lot of low quality semiconductor on the signal path too...
 

PRR

Member
Joined 2003
Paid Member
Fascinating. ..... I wonder if the bias shift (in your simulation) has an audible effect?

The bias shift is done before 10mS. At guitar time-scale and pluck transient, I suspect bias-shift has "no" audible effect.

I would not say the chips are low quality. They are low-spec by audiophile standards, true. However at guitar level and bandwidth, the '558 have been a mainstay of guitar electronics. They work. I might chose a different chip, not sure; but it isn't a bad choice.

Anyway if those tubes do anything they ought to overwhelm the chips.

all even stages clip at the same time

My suspicion is that you want "layered" clipping, not all-at-once. However only the breadboard can tell for sure.
 
Account Closed
Joined 2010
the njm4558 is widely used and appreciated for its trashy sound in Polytone where some people try opa2134 for a cleaner sound...but here we have countless 4558 in series everywhere.It sounds like a "fine choice"...Now...for a drive circuit you don't want a trashy nonlinear stage before the drive circuit because you won't be able to trim the final products in the eq section.There are lots of fet switchers too and then.... should we wonder why people say that the drive section lacks dynamics?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.