Help with Shunt Regulation and Noise

[dsavitsk]

I am working on a new headphone amp that is shunt regulated ultrapath design. The tubes are triode strapped 7788's. The CCS has about 31mA going through it, the regulators are 0D3 (153V) and the tubes have about 16mA of current.

I built it as per the schematic below, and there was quite a bit of hiss. I bypassed the regulator with a 0.1uF cap which didn't do much. I also changed tubes as well as regulator tubes with no change. I then bypassed the cathode resistor with a large electrolytic and the hiss went away -- and the bass is better.

So, in some sense, problem solved. But, in a better sense, I am not sure that I am sure what is going on. I assume that the bypassed Rk shunts noise that the ultrapath cap is delivering to the cathode to ground. Since the tube's mu is high, this was presumably being amplified by a lot, and hence the hiss.

So, my real question is whether the ultrapath cap is actually doing anything. That is, does the 0D3 present a low impedance signal path and is that where the signal path is, or is the output signal loop through the ultrapath cap like intended?

And, another question, the CCS for now is just a single IXYS chip. Is this presumably the source of the hiss -- would a better CCS design cut down this noise a bit.

[SY]

Can you fill in some values?

[kevinkr]

Hi Doug,
Without knowing more about your actual component values it's a bit of a guess, but I would first suspect gained up johnson noise from the cathode resistor itself, and then I would suspect noise from the gas tubes being injected into the cathode circuit of your 7788s.

When I first started working on my 26 dht based pre-amp design I used an 0D3W as the source of regulated plate voltage - to say it was noisy was a marked understatement. Lots of hiss, and this was in a transformer coupled application with fixed bias applied to the grid of the 26. The solution was an active voltage regulator rather than the original gas tube employed.

[kevinkr]

Hi Doug,
Slightly off topic, but I suspect that you probably don't want to use an unbypassed cathode resistor in this application because it appreciably raises Rp..

(Edit: True even with an ultrapath cap because of its limited effect at low frequencies - does help the distortion problem I mentioned below.)

Think something like Rp + Rk(mu+1).. Ouch..

Led bias or fixed bias are alternates to the bypassed resistor you could try. I've had BIG problems with the sonics of small cathode resistors bypassed by large electrolytics - requiring additional smaller EL and film caps to make it sound ok. I haven't investigated to the root cause, but suspect DA, DF, ESL, and even the mechanical construction of the cap - in any event it is quite audible, and can be observed on a scope with burst waveforms. HF transients sound very nasty.

[dsavitsk]

Originally posted by SY
Can you fill in some values?

Attached with values, and with the changes I made to eliminate the noise.

I am really unsure of that 10u cap -- nobody seems to know how to actually calculate the ultrapath cap value with some people saying to consider Rp + transformer primary Z, and others saying things about Rk, and others saying to just try values until it works. I find this frustrating as can be.

Originally posted by kevinkr
Hi Doug,
Without knowing more about your actual component values it's a bit of a guess, but I would first suspect gained up johnson noise from the cathode resistor itself, and then I would suspect noise from the gas tubes being injected into the cathode circuit of your 7788s.

I essentially replaced a linear mosfet based regulator with the 0D3. The linear reg was pretty silent, so I am guessing the gas tube. And, this is such a high mu tube, that it does seem that any noise is a big issue. I think I may need to investigate some other regulation methods in the future.

Originally posted by kevinkr
Hi Doug,
Slightly off topic, but I suspect that you probably don't want to use an unbypassed cathode resistor in this application because it appreciably raises Rp..

...

I've had BIG problems with the sonics of small cathode resistors bypassed by large electrolytics - requiring additional smaller EL and film caps to make it sound ok.

I first tried some cheapo electrolytics (820uF), thinking that they were out of the signal path and wouldn't matter. I am, as you may have gathered, cap obsessed -- or rather, obsessed with ridding the signal path of them. The sound was clear, but as you say, harsh on top. I replaced them with some blackgate NX 680u caps which improved things quite dramatically in general, though it did reduce the bass a little. This is why it seems like the "ultrapath" is not the only signal path as these caps do seem to matter quite a lot,.

I may try some fixed bias methods as well, but for now, the $1000 question (actually, the $29.00 question) is whether those electrolytics should be increased to 1000uF (BG's seem to jump from 680 to 1000) or if they need to be mu times larger than the ultrapath cap. Maybe that is only for cancelling PS hum?

[Zen Mod]

temporary replace CCS with resistor , just to see it is culprit ;

if yes- you can still decouple CCS from rest of circuit with smaller resistor, and put big cap to gnd ........ even if better solution (in this case ) is to make more silent CCS


EDIT :

if you cliick on my www button , and look for my WOT preamp - you'll see that I have almost same arrangement ........ but without CCS and without ultrapath cap .

my pre is dead silent

[SY]

Have you inadvertently built a relaxation oscillator? VR tubes do not like being bypassed, and you have effectively done so with the capacitor string in parallel with the (OK value) 0u1 cap.

[Wavebourn]

Did you connect an input ground as drawn, or directly to a rectifier?

[Wavebourn]

Originally posted by SY
Have you inadvertently built a relaxation oscillator? VR tubes do not like being bypassed, and you have effectively done so with the capacitor string in parallel with the (OK value) 0u1 cap.

If current through it is less than few mA it can oscillate. I usually assume more than 5 mA.

[dsavitsk]

Originally posted by SY
Have you inadvertently built a relaxation oscillator? VR tubes do not like being bypassed, and you have effectively done so with the capacitor string in parallel with the (OK value) 0u1 cap.

I hope not, but if I did it is either ultrasonic or subsonic. Noise without Ck is like white noise, so perhaps I need to get out a scope and see if anything looks sawtoothed.

But, it did occur to me that this is like a huge bypass cap to the regulator which is why I was wondering if the regulator acted as a low impedance path for AC to complete the current loop -- maybe I'll try removing the ultrapath entirely? I was basically looking at this schematic for the idea: http://nutshellhifi.com/Raven-MarkII.gif

Originally posted by Wavebourn
If current through it is less than few mA it can oscillate. I usually assume more than 5 mA.

CCS is at ~31mA, tube runs at ~17mA.

Originally posted by Wavebourn
Did you connect an input ground as drawn, or directly to a rectifier?

PS is in another chassis, so input ground is pretty much as drawn.