Balance in CCS Long Tailed Pairs

[Geek]

Hello SY,

And proof of the pudding: with 0.4V peak to peak square wave input, here's the output (4.8V p-p for both sides). The balance is about as good as can be eyeballed on the scope.

I see you are not using the ADD function on your scope. The imbalances are there and grow significantly as the Vp-p gets larger.

CCSed LTPs do NOT require any sort of tube matching or feedback or any other fancy scheme for AC balance, even with GROSS tube mismatches.

I disagree, allow me to demonstrate.....

I have created a test circuit with some good gain and an A/B switch to test between pure CCS and a CCS with CMFB compensation.

Scope was setup as per the test circuit and the Vp-p per side was brought up to 65Vp-p:

I used a 1KHz (it's the same at 5KHz) test signal and measuring the difference between the sides in straight CCS mode got the following result:

Vertical scale is 2V/division. We see an error of 6.8V between the sections.


Now let's flip the switch and feed some difference signal back into the circuit as a local NFB loop:

We are now seeing a difference of ~4V.... 1.7:1 improvement.


But as I said in a PM to you, the circuit really shines as the level ramps up.

Let's switch back to the straight CCS and ramp up the voltage to the onset of compression... BARELY the onset from the scope, which was 85Vp-p and let's see what the difference between the phases are:

We now have an error of 14V.


And switching in the CMFB circuit:

An error of 6.6V... a 2.12:1 improvement.

When using tubes with really mismatched sections, as mentioned on my forum referenced to you, this improvement goes up.

Cheers!

 

[Geek]

Hello Don,

The DC unbalance and AC balance with CCS tail seem pretty much as expected. There likely are some other reasons to keep the tubes more matched though. I would think distortion would increase if the two tubes are not operating from the same point on their characteristics. Not to mention some loss of headroom if the DC balance is off far enough.

Absolutely!

When I developed my circuit for the ST-70 driver, this improvement in balance translates to better power and more importanly, cleaner power.

Then there are some subtle issues with LTPs that I have never seen addressed anywhere. Fred N. used an interesting setup where the normally undriven side of the LTP was provided a matching inverted signal by a pos. feedback path from an LTP output (a resistive divider from opposite plate to ground to provide a signal for the slack grid matched in amplitude, but inverted, with the input). I have seen something similar in RDH4 I think. The idea would be to symmetrize the operation of the two tubes with AC. Otherwise the input pushes them off center one way always. The question here would be whether an improvement in distortion is gained despite the pos. feedback.

Yes.

Fred spoke in depth on that in rec.audio.tubes at one point and his RA-100 improvement was a spinoff of that and this one is a spinoff of that yet still.

Then there are fixups to LTP distortion like the critical tail resistance to null the 3rd harmonic. A slightly different resistance can be used to null out each of the other odd harmonics. If one were to use a non-linear tail resistance, like some thermionic diode for the tail (of a specific conductance), could one null out ALL the odd harmonics.

That can get out of hand quick

There's always tradeoffs too. When I played with that a couple years ago, you'll null one as another goes up until it's time for a drink.

And finally, one could really shoot for the moon with an error correction LTP splitter scheme. Here one would simply provide a second resistive divider from non-opposite and opposite outputs back to the driven grid input as well (beside the 1st divider for the slack grid).

Stabilization of such a circuit might drive one buggy


Cheers!

 

[Wavebourn]

All roads lead to Rome:

http://wavebourn.com/forum/download.php?id=119&f=7



Concertina drives symmetrically grids of LTP. LTP has linear wideband device in tail called Resistor; DC regimes of all tubes kept stable by feedback from long tail of LTP back to 1'st pentode's screen grid.

Simple and elegant.

 

[Geek]

Hi Wavebourn,

Interesting!

I have been meaning to play with such, as it cures my two dislikes of the concertina - no gain and imbalance in all but feather-light loads.

Cheers!

 

[Wavebourn]

This thingy drives GU-50 tubes with parallel feedback, 240K from their plates to plates of 6P15P tubes in LTP. It was a mistake to use 6F12P there: amplification factor is too high, so I did not shunt R6 by a cap.
In current version I use different tubes.

Conclusion: calculations must be repeated several times before building...

 

[Geek]

Oooo, kinda like my "TillyMutt" prototype

Driver:
http://geek.scorpiorising.ca/contrib/Geek/ST-70_CMFB+SRPP_v1.png

Output:
http://geek.scorpiorising.ca/contrib/Geek/Output_stage_triode-pentode_v1.png


It's what Planet10 has in his living room now.

Cheers!

 

[Wavebourn]

Oooo, kinda like my "TillyMutt" prototype

Not exactly, I use long tailed pentodes to drive low resistance of output tubes.

 

[Geek]

Got a full schematic and/or writeup on it somewhere?

Cheers!

 

[Wavebourn]

No full schematic, sorry.
800V B+, 10K P-P 100W Edcors,
GU-50 outputs,
driver as drawn,
+12V (filament for small tubes), -80V (bias), +400V (drivers), +270V (screen grids, bias) regulated.
I sense currents of screen grids by opto-pairs (optical compressors) to avoid clipping.

Here is some mixed Russian/English description: www.wavebourn.com • View topic - Pyramid-VII: home stereo vacuum tube amp

 

[Geek]

Oh, that work of art you're speaking of.

Cheers Wavebourn!

 

[planet10]

It's what Planet10 has in his living room now.

Indeed And thourougly enjoying it.

Next i gotta get Duo to stop MooseFET from being MooseFART. Manual change of inputs is a bit of a pain, and daunting for all but me (or i could get off my butt, put 6H23pi in Fornaco and use it. MooseFET is cuter thou.

dave

 

[Geek]

Next i gotta get Duo to stop MooseFET from being MooseFART.

It's under lifetime warranty for you, Dave... I'll do it.

All he has to do is pull the muting transistors with vise grips and done!

That's the **only** diff between yours and mine (besides the PS) and since I can't duplicate it on mine, it has to be it.

Cheers!

 

[Duo]

The Vreg still sucks...

 

[Geek]

The Vreg is straight off the National datasheet... blame them

Cheers!

 

[Duo]

With that big cap slugging the feedback loop down?

 

[Geek]

What big cap?... 10uF to ground after the 317 is nominal

You looking at the main filters?

The 10uF across the Vadj resistors? Nominal too (page 9, LM117 DS).

Cheers!

 

[Duo]

Can you show me the schematic for moosefet? Maybe email it to me?

 

[Geek]

Done!

 

[Duo]

I don't think the mute circuit is causing the problem. The exhibited sound was a hum, 120Hz, which occurred during brownouts as the line voltage rose back to normal. This is normally indicative of regulator dropout. Are you certain the transformer voltages are as listed in the schematic? I notice you list 30V after the rectifier but you have 15VAC and 12VAC being stuck in series after rectification which should be more like 38V.

The only beef I have with the 10uF cap on the FB loop node is that the response of the regulator to load changes is going to be on the order of 2ms although I don't think it would cause the problem we're experiencing.

As for the mute circuit, a pull-down on the bases of those transistors might be a good idea just in case. A far better mute would be implemented with a mercury wetted reed relay or JFETs (if the circuit impedance isn't too low).

 

[Geek]

I think we've drifted wayyy OT...

No relay for this particular model. Wetted relay would be more than the entire unit anyway. FFT didn't show issues with the bipolar.

FET, sure!

Maybe there's a bum cap or the reg's borked

Vin-Vo should be within the regs ability...

(I'm sure SY or Dave can split us off into our own MooseFET topic)

Cheers!