Optimizing input stage for low noise

[SoulFetish]

I would like to get a consensus on a couple of ways to optimizing a parallel input stage for the lowest noise.
I've draw a schematic showing two common configurations and I had a couple of questions concerning them. Here are the schematics:



(Wow, that image is a little bigger than I imagined. But, Juan, that embedded image trick works like a charm)

Anyway, the -3db reduction in noise is achieved because the noise (of interest) generated by the triode is random and uncorrelated, while the signal is (obviously) correlated and amplified by both triodes. As I understand it, the cathode is the predominant noise generator in a triode. I've read that the -3dB noise reduction is only achieved when the parallel cathodes uses separate cathode resistors. (shown in example 1) Yet, if they share a common resistor, then there's a 3db increase in noise. I wanted to confirm this. Also, would a fully bypassed cathode have any effect on this? (as shown in example 2)
The other issue relating to noise, is resistor noise generated in the input stage. Would there be any practical, or noticeable change in noise by using a wirewound resistor for the 1M and 1k resistors shown vs, say, carbon film?

[catalin gramada]

hi. the input capacitance goes up. it is effective not some subtile effect. You will lose hights, that a fact

[SoulFetish]

hi. the input capacitance goes up. it is effective not some subtile effect. You will lose hights, that a fact

Thanks for the response. The increased input capacitance isn't a problem. I'm going to use it to bring the LPF cutoff frequency down to 20kHz (give or take).
In fact, I need to increase the capacitance if I'm going to use the 1k grid stopper. Figure the stage has a voltage gain of around -58
Cin = Cgk + Cga(A + 1)
Cin = 1.8pF + 1.8pF(58 +1)
Cin = 108pF(2) = 216pF for both triodes.
With a 1k grid stopper that gives me a -3db cutoff of 736.8kHz. That's not really going to give me the sort of RF suppression I'm looking for. Adding a 51pF cap from each grid to anode will bring me down to a little over 25kHz. Or I can add a shunt capacitor from anode to ground (whatever works better layout-wise).

[catalin gramada]

Yes, maybe was not as dramatic as change to 12au7 for instance, but in respect with single section I felt a sensible difference with large 68k grid resistor (I think...was) and no compensation.

[Malcolm Irving]

Interesting question. I don’t have a complete (or certain) answer, but here are some thoughts which I hope will contribute to the discussion:

I think when you connect the two grids together (in circuit 2) you need to consider the grid stoppers in parallel (similar to what you have with the cathode resistors). So (IMHO) the grid stopper in circuit 2 should be 500R, if you want to compare ‘like with like’.

I think the cathode is a source of hum and buzz, due to unwanted capacitive and resistive coupling with the heater. A large value bypass cap (say 250uF) will help shunt that noise to ground. Along with the other usual techniques such as DC heaters, etc.

Usually, the main sources of hiss are the grid stopper(s) and the tube itself. But if the resistance of the grid stopper(s) is low enough to be comparable with the cathode resistor(s), then the noise produce by the cathode resistor(s) would come into play. (Noise from them has a similar effect on the grid-cathode voltage.) I don’t think combining them in parallel would make any difference however (for the reason mentioned in the next paragraph). Using wire-wounds could help a little, although I would be surprised if it made any noticeable difference compared to metal-films.

As a general point, any series or parallel combination of resistors will generate the same noise level as a single equivalent value (IMHO). I like to think that this is because, although the noise in two resistors is uncorrelated, the noise within a single resistor is also uncorrelated (there are so many electrons and atoms all working ‘in series and parallel’). I think the same logic applies to two triodes in parallel (i.e. the parallel triodes produce the same noise as an equivalent single triode of twice the mutual conductance and half the anode resistance).

Most amps short out the 1M grid-leak when nothing is plugged in, and when a guitar is plugged in the impedance of the pickups is lower than the 1M. So the 1M is a massive noise source only when a cable is plugged in with no guitar, but otherwise doesn’t have much effect.

[catalin gramada]

forgot to mention, somehow related with the subject, don.t know if is count: found grounding point to chassis is always better from a noise perspective to be done right at the input

[SoulFetish]

forgot to mention, somehow related with the subject, don.t know if is count: found grounding point to chassis is always better from a noise perspective to be done right at the input

catalin, you read my mind!
I was just going to post this photo and ask for some feedback on my power supply ground - to chassis connection. Do you guys think this is close enough to the input jack (so as to not have to couple the input jack to chassis via .01uF capacitor)?

[catalin gramada]

From my point looks allright. you should tie the bus end and jack ground (with a separate short wire) to those nut. Three lugs star washer better in case you decide to tie local decoupling star of first stage to the bus in the same spot

[SoulFetish]

From my point looks allright. you should tie the bus end and jack ground (with a separate short wire) to those nut. Three lugs star washer better in case you decide to tie local decoupling star of first stage to the bus in the same spot

My plan was to tie all my local grounds for the input stage to the filter cap feeding it. And run a short wire to a solder tab secured to the screw. I fear grounding my input jack with another wire to the chassis would cause a loop. The only current That should exist in wire-to-chassis connection is in event of a fault.

[catalin gramada]

I was talking about the same thing but it is not any loop, should explain...



better drawing than explain in my broken English

[SoulFetish]

I was talking about the same thing but it is not any loop, should explain...

[ATTACH=CONFIG]43011[/ATTACH]

better drawing than explain in my broken English

Okay, I see. No loop. Although there is still signal current in the chassis "connection". I'm curious as to the wisdom behind this method vs grounding the jack at the capacitor node?

[catalin gramada]

"Although there is still signal current in the chassis "connection" -

-so what ? The signal is "disturbed" by nothing but EMI and supply ripple. It is a lug there.The input signal goes from the grid to those point and return to the source, nothing run through chassis.(except you don.t plan to do another one or more chassis connections somewhere..)To complete the chassis connections the earthing is only wire you have to add)
The whole idea is to keep EMI out of signal path, so I choose the proximal point (means as close) to tie cable braided shield to chassis.
Finally if you best choice is to tie input jack ground directly to local star then put a capacitor (inductance free) between jack ground and chassis nut (never try it but have a lot of sense to do so -usually a good compromise when a floating signal ground is required.)
Cheers

[catalin gramada]

[Gingertube]

The grid stop contributes most of the noise for the input stage so reducing the "standard" 68K is a good idea.
However 1K is too low to absolutely guarantee fredom from parasitic oscillation. For triodes 8/gm is recommended minimum. For a 12AX7 that is 8/0.0016 = 5K. I use 10K as does the Wiz.
See what the Wiz had to say:
The Valve Wizard
Cheers,
Ian

[Chuck H]

...