2A3 Monoblocks

[Cressy Snr]

This is where it gets interesting for me:
http://www.tubecad.com/2008/08/blog0147.htm

John Broskie's blog from a few years ago can perhaps explain why my current Ultrapath amplifier does so well at nulling out mechanical hum, noise and
whatever was left after 2 stages of regulation, to give those inky backgrounds.

Because of the cascode's lousy PSRR, it exactly complements the Ultrapath output stage's similar characteristics, but in opposite phase, or something like that.

Looks like the best input stage for an amp using the ultrapath connection is in fact a cascode or alternatively a pentode. Triode input amps are far noisier with an ultrapath output stage.

Broskie shows several circuits using a cascode input with some cathode-to- grid feedback on the output stage. These, he reckons make the best use of ultrapath configurations.

He does not show an ultralinear cascode with ultrapath output stage, but it certainly works very well for me.

[Cressy Snr]

And this is the end of the line for this particular project.

The schematic:

[Paul Barker]

I would just try one more thing if possible. bypass the output cathode with 200uF

(-3db 5hz given 1/r = 1/rz +1/rk where rz = (ra + rp) / (mu + 1) then C in Farrads = 2pi.f.r.) 2hz would be better design goal but I have kept it to sizes which would match your 30uF cap and commonly used caps. For 2hz you would need 500 bypass cap 100 ultrapath cap.

As it stands with the ultrapath at 30 (38 is ideal 200/mu+1) you have an opportunity to try the concept and see if you like it..

See what you think.

If you didn't have a shunt power supply and you used AC heating on the output valve you would like it. But with a smooth dc filament supply and a regulated power supply it may be guilding the lilly. The purpose of the scheme was to both cancel noise on the filaments and cancel power supply noise. It was very effective in the days we used ac heating (which I still use because I am lazy and the hum doesn't bother me, and anyway it sounds at least as good as the most exotic dc scheme you care to name)

[Cressy Snr]

Hmmm...

I tried your final mod Paul and it is quite an interesting situation we have.
I would swear there is a slight increase in the presence of low-level detail but that might be wishful thinking. It is certainly no worse than the un-bypassed cathode.

Either way I think maybe the 2A3/6B4G has reached its ultimate expression in this circuit, well it has for me at least.
We'll what Nick thinks tomorrow, as I'm taking the amps to his place for a listen whilst I collect the 10K output transformers he has kindly offered to loan me.

Steve

[Cressy Snr]

Late last night I further developed Paul's suggestion of bypassing the 6B4G's cathode.

Broskie (August 2008, "Ultrapath - Ultra Simple...Except That It's Not") writes:

We do want all the power-supply noise present at the triodes plate, grid, and cathode. Once again, an output transformers primary only reflects delta, differences across its leads. If the voltage across the primary and the current through the primary remain constant, the secondary cannot see the power-supply noise. So what had been the cascode and Ultrapaths main liability, poor PSRR, has now become an asset. A grounded-cathode amplifiers better PSRR figure would work against us in this amplifier, as would a grounded-grid amplifier. On the other hand, a pentode-based grounded-cathode amplifier would work as well as the cascode circuit, as the pentodes super-high plate resistance also results in a truly poor PSRR.


And he suggests this variation on the bypassed cathode resistor, showing the noise paths in the cascode/output stage circuits.



As we can see there are two signal/noise paths, one is ground referenced and the other B+ referenced.

He goes on tho explain :

The input signal is ground referenced, as is the signal source. But once the signal become amplified by the input cascode stage, the signal effectively becomes B+ referenced (because of the cascodes negligible PSRR). The output stage is entirely B+ referenced, as the output tubes cathode, grid, and plate see all of the B+ noise and voltage drifts. This dual-reference arrangement can work quite well, as long as we strive to maintain the purity of the two reference territories. In other words, we must mentally block out the two reference domains and work to keep each isolated from the other. For example, if an input stage with much better PSRR were used or if the output tubes cathode were replaced by a zener diode, the overlap would be huge and amplifiers performance severely compromised.

Now below is my version. The 300K grid resistor has been split in two, and the 220uF bypass cap AC couples the B+ to cathode noise path to the grid, via the 150K resistor at the top of the chain.



Let's just say it's good shall we?

[Paul Barker]

Yes I saw what Broskie adapted it to. In your case because of the cascode this makes a lot of sense.

The only refinement which sticks out like a saw thumb, is to replace the pair of 150k resistors with a pot which you adjust for best outcome. OK you could then put fixed resistors in. This would allow for fine tuning where the two caps are not exactly right proportions, as it voltage devides the noise/correction signal.

[IslandPink]

Steve
I'm struggling to see why the ultrapath mods help when you have a power supply that is both series and shunt regulated . How much noise was there on there to begin with ?

Also, surely the signal return path through the shunt valve is cleaner than any ultrapath cap you could use on the output stage ?

Does it sound better in all respects ? - which parts of the presentation got better ? Just wondering if you have an explanation for this , it's certainly an interesting result.

It's not that I don't believe you, I'm just confused, and thinking what would be the best approach if I eventually have a shunt-reg power supply on my multi-channel amp .

[Nick]

I heard them this morning, they are very very nice, as I would expect. Mu only comment to Steve was that they were maybe missing a tad of low end below 50hz. But I think that would need some expensive output TX's grafting on the end. Min and top is just lovely they play complex music (I tried hard) without getting in any way confused.

[Cressy Snr]

Steve
I'm struggling to see why the ultrapath mods help when you have a power supply that is both series and shunt regulated . How much noise was there on there to begin with ?

Also, surely the signal return path through the shunt valve is cleaner than any ultrapath cap you could use on the output stage ?

Does it sound better in all respects ? - which parts of the presentation got better ? Just wondering if you have an explanation for this , it's certainly an interesting result

Well it's certainly a conundrum isn't it.

When I was at Nick's this morning we were discussing the very same contradiction.

One thing the ultrapath mod did straight away was to quieten the mains transformer down to almost silence. Now the regulators would have blocked the electrical noise obviously, so it shouldn't have mattered one jot
whether there was mechanical transformer noise or not. But nevertheless the virtual silencing of the mains transformer was there and repeatable.
The question then has to be, was whether transformer mechanical hum was being transmitted through the metal chassis to the directly heated tube filament, setting it vibrating in sympathy, causing noise to be injected at the output stage, effectively "shorting out" the regulators and introducing intermodulation effects in the output signal.
The fact that the soundstage became wider, clearer and "blacker" despite there being regulators already there would seem to support this theory.
So could tube dampers have achieved the same effect? who knows.

The second theory we had was about the improvement caused by the rerouting of the AC noise path as anode to cathode to grid, rather than anode to cathode. Could this have been some sort of signal feedback mechanism happening alongside the noise path routes through the amp so that we had some sort of SE version of Steve Bench's harmonic equalizer going on in the cathode/grid path.

Whatever it is it certainly works. The amps are even better than they already were at unravelling complex mixes, presenting them as a coherent musical whole without blur muddle or hardness.

As Nick says, the bass end is curtailed at the very bottom, compared to the Pass F5, which we heard and probably Nick's 211, which we did not.

As Nick has inferred, getting that half octave of bass extension at the very bottom will probably cost more in iron than the whole amp setup put together C'est la vie and all that jazz.

[IslandPink]

Steve
A late-breaking question here, with relevance to my recent activity :
When you showed this amp at Owston about 18 months ago ( was it ? ) on the legendary Sunday , were you using the DIYHiFiSupply filament modules on the 6B4G's ?

[Cressy Snr]

Steve
A late-breaking question here, with relevance to my recent activity :
When you showed this amp at Owston about 18 months ago ( was it ? ) on the legendary Sunday , were you using the DIYHiFiSupply filament modules on the 6B4G's ?

Yes, the DIYHiFiSupply modules were used on the 6B4G filaments.