DC Imbalance with the LTP

[rlaury]

Hi Group:
I've just spent the last few days searching the forum for some answers on this subject but never really found a concrete answer.
I'm using a 12AT7 for the phase inverter with a simple DN2545 CCS set for 8mA.
I have near perfect AC balance but the DC balance is off by 6-8 volts.
I'm capacitor coupled to the output stage (KT-88's) so the DC shouldn't matter. The amp already has pretty low THD of .18% @ 55 watts into 8 ohm. I've seen examples of using a pot in the cathode of one tube to adjust the DC balance BUT......

Why worry about the DC balance if the AC balance is near perfect??
I've also read that using a cascode CCS may bring them into DC balance
but my question is why bother if the AC balance is correct???
Any and all opinions appreciated.

RonL

 

[SY]

A cascode CCS won't get you any significant improvement in DC balance- that's a function of section to section matching. 6-8V is pretty good.

And you are absolutely right- it matters not a bit.

 

[rlaury]

Thanks SY That's what I wanted to hear.

RonL

 

[Miles Prower]

Cascode CCSs help break up stray capacitance to ground that will compromise the "CCS-ness" at high frequencies.

 

[rlaury]

Thanks Miles.
I'll be running some test at 10 - 20 Khz this week. BTW any advantage to have some small R's between the cathodes and the CCS junction?? ARC seems to do this but PSPICE shows no difference in distortion or balance

Thanks
RonL

 

[artosalo]

It seems that your 12AT7 has dissimilar halves. With similar halves and CCS the dc balance should also be perfect.

The amp already has pretty low THD of .18% @ 55 watts into 8 ohm.

This is obviously with global NFB ? If so, then it does not tell much about the LTP and its balance. More important is to know what is the distortion from both outputs of the LTP with max. required output level and with real load resistance (Rg of output tubes).

I have observed in practice that it is not essential to get fully equal output levels from LTP. Instead it is essential to get fully equal output levels from the output tubes. Therefore I prefer adjustable output levels from the LTP. Then you can also use not so well matched output tubes.

 

[SY]

Cascode CCSs help break up stray capacitance to ground that will compromise the "CCS-ness" at high frequencies.

The fortunate thing is that it's at the cathode, which is a low impedance node. So one needn't get too fancy with the CCS- even a simple 1 transistor circuit will give essentially perfect balance even with wildly mismatched tubes.

 

[piano3]

Sy, I would like to clarify my understanding(or lack of); in general,in a LTP,the DC balance does affect the PSRR or not ?

 

[SY]

Why would it? It's the AC component from the power supply that you care about.

edit: The quality of the CCS will affect common mode rejection and hence power supply rejection. But in practice, the signal levels are high and even a simple CCS will get you at least 40dB of rejection.

 

[piano3]

Thank you Sy, It is just that I was remembering a section in MJ about CMRR of different topologies and that if the valves were not matched in a diff. pair then CMRR would be degraded.

 

[SY]

It's the mu that needs to be matched to optimize CMR (CMR is proportional to mu1*mu2/(mu1-mu2)). Higher mu also gives higher CMR.

 

[piano3]

Ah yes, of course, whereas it is mainly mismatch in gm that causes the difference in DC between the outputs, is that right? But how does one easily match mu?

 

[SY]

No, you can have a gross gm mismatch and still have perfect AC balance- the CCS guarantees that. DC mismatch can be caused by anything from slightly different cathode emissivities to variations in physical alignments.

You can try to match mu, but it's generally not worth it for the extra couple of dB of CMR. To measure it, set up a jig that applies a constant DC voltage between cathode and grid (I use an LED from cathode to ground), load the plate with a CCS, then measure the gain.

Mu is generally pretty consistent from section to section of dual triodes; 10% is pretty much the worst you'll see with any reasonable quality tubes.

 

[piano3]

Thank you Sy, I understand much more clearly now.

 

[DF96]

DC imbalance in the LTP can be caused by grid current, if the second grid is fed from the first grid by an RC network (e.g. Mullard 5-10). This can of course happen with any valve if it is a bit gassy. For the ECC83 it can also happen of the anode voltage is too low, as the grid-cathode voltage then wanders into the grid current region.

 

[Miles Prower]

Thanks Miles.
I'll be running some test at 10 - 20 Khz this week. BTW any advantage to have some small R's between the cathodes and the CCS junction?? ARC seems to do this but PSPICE shows no difference in distortion or balance

This is more of a solid state practice when using BJTs. BJTs have gm to burn, and so all that is required is to increase the current a bit for no loss of gain when including emitter degeneration. Triodes don't have all that much gain to begin with, and, unlike transistors, cathode degeneration not only results in a loss of gain, and gm -- important since LTP balance is proportional to: 1/(RtailGm) and you need all the gm you can get. It also causes an increase in effective plate resistance, and therefore makes the plate resistance a larger percentage of external plate load, and could increase distortion and/or change the harmonic distribution to something less sonically desirable since triodes produce more distortion the more heavily they are loaded.

If PSPICE is showing no difference, then it's likely your cathode degeneration resistors aren't big enough. For hollow state design, I'd forget about it. The only reason to include any resistance in LTP triode cathodes is to balance out DC offset. It's already been established here that it's not necessary unless you're DC coupling and/or making an instrumentation amp that needs to go down to very low frequencies.