Re: NFB windings, was there a US style and UK style?

On 05/21/11 22:03, flipper so wittily quipped:
I'm not really a 'fan' of the paraphase but the battery filament tubes
I was using in Batman didn't leave much choice. The pentode-triode
(and diode I'm not using) share the same filament in the 1D8GT and
that ruled out anything under the 'cathodes'.

ugh, yeah probably no choice. even a long tail wouldn't work well unless you can somehow CCS both the filament AND Ik at the same time [unlikely]. So definitely limited in what's possible.

the basic circuit design is kind of like the way a differential amplifier circuit works, where the sum of the inverting + non-inverting must become zero when the circuit is in balance. The error is then infinitely amplified, and NFB re-balances everything to create a reliably amplified signal (or that's the theory). So when the output of the 'thing' that produces the inverting signal is summed with the original, this creates the 'error' signal. Ideally, you have zero error when the circuit works. Unfortunately, low gain triodes will result in higher error, so you'll probably need something to adjust balance on it. Hence... (your additional comments)

I call mine a 'hybrid' paraphase because, as mentioned, the triodes
don't have a lot of gain so it starts off looking like a plain
paraphase where you tap a 'small' signal off the first phase to drive
the second. But rather than just a tap I also link the second phase
into it, like a floating paraphase, so the tap 'predisposes' the
signal level and the feedback works around that predisposition.

I'm guessing this corrects for what I mentioned above. 'Tricky' stuff.

If the tubes are identical and exactly as spec'd for bogey then
everything gains and sums up to what's expected so there's,
essentially, 'nothing for the feedback to do'. But if either are off
then the summation doesn't 'come out right' and tends to push the
second triode toward a match.

and that's another problem with paraphase - you need matching 'things' and component aging isn't compensated for in the least.

It isn't very much NFB, maybe 9 dB, but, hey, why not? Actually, I'm
not quite sure how to calculate it but there's a 3 to 1 voltage
difference with R2 connected there ('closed loop') vs not so I figure
that's 9.5dB.

http://flipperhome.dyndns.org/Batman.htm

pretty cool. FYI something LIKE THIS is ideal for generating "tube amp distortion" without wasting a lot of power. You could use an 8 ohm dummy load and overdrive it into a different amp (which then gets you the volume). /me getting ideas.

The amp is still dominated by 2's harmonics but that's from the open
ended first stage.

a tetrode? Ok, just watch out for the reverse-resistance range at low current. 100k on the plate may suggest that you're already in that slot, hence the 2nd harmonic distortion [which may get even worse]. You might halve that and adjust bias for twice the current. OK gain may be lower, or then again may not. You're probably dropping 40v or so across the 100k which is .4 ma so 47k and .8 ma should get you a bit more linearity. [I'll leave it up to you to tweek the bias on it]. Ep too low is also likely to create problems. My rule is to pick a voltage that's about 1/2 B+ for the 'rest state', using a plate current that's reasonably linear for an easily obtainable G1 bias. Anyway, if you can put in test signals you might want to do that to generate your own curves, maybe 3VDC with a potentiometer and manual readings (or storage scope if you have one). /me reminded of the $89 pocket scope I bought a few months ago - very nice, goes up to 1Mhz.

Down side for cathodyne is that you have to make sure that the
Rp-Rk never drops below around 50V or so (depending on the tube) in the
worst case voltage swing.

yes. using high B+ helps. If you've got power tubes that need >400V that's not too difficult. You can add an R+C filter, drop down to maybe 380V with a relatively large R and make that the supply for the cathodyne part. I like to split the voltage so that the G1 bias is less than the voltage drop across Rp or Rk at steady state, and is also SIGNIFICANTLY less than 1/5 of the supply voltage. So then the best case voltages for 60V negative bias (on the power tubes) would be 70V on the cathode and ~310V on the plate (steady state, B+ 380), with 130 on the cathode and 240 on the plate at the 'highest current' point. Unfortunately this will have higher 2nd harmonic distortion due to the wider current swing. Anyway, Rp - Rk is 110, which is more than enough, so this could be tweeked a bit more, maybe 90V on the cathode and 290 on the plate at steady state, then Rp - Rk is 80V, still well within the linear range for a typical triode. Then you just pick resistors that match your desired current, and make sure the G1 volts at steady state is around Ek + bias. But if this is directly driven from a prior stage then you must increase Ek to an even higher amount to prevent Ep on the driving stage from getting too low. So now you tweek it up a bit more, with steady state Ek at 110 (~50V min on driver tube plate), and 280V Ep, with 'worst case' at 170V Ek and 220V Ep (now down to 50V), and you're "there". Current only has to change by a fraction of the quiescent value, you can direct drive it from the prior stage, and if you choose the right parts (and maybe even regulate the B+ for the pre-amps) it self-balances. Anyway, point well taken on min Ep - Ek.

At that point I'd do the 'long tail' thing and match the Rp's, and
use a constant current source.

Neither Williamson nor Baxandall had transistors to conveniently
create a CCS.

you could do it with tubes. Ground grid, put Rk that sets the current to what you want, plate in series with long tail cathodes, and use a voltage divider (filtered) to bias the grids. But then Ek is much higher than it would be with a transistor, since you could drop down below 2v and still get your CCS (thereby using ground potential for the grid bias). Cheating and using germanium transistors you could go even lower.

It also depends on whether it matters enough to bother with. PP gain
imbalance manifests predominately as 2'nd harmonic and, remember,
that's 'the first to go' with NFB.

interesting point. yeah, you can tolerate a bit of imbalance. But the result is likely to be a reduction in output power, or uneven clipping when overdriven.

.

Relevant Pages

  • Re: 6SN7 amp trouble?
    ... the plate will go positive. ... voltage supplying it. ... So you probably don't want quiescent bias that lands the plate at ...
    (alt.guitar.amps)
  • Re: SE fixed bias VibroChamp works; toob math question
    ... There are two readings, bias pot moved ... bias voltage range seems to be wrong. ... Pout is different from max plate ...
    (alt.guitar.amps)
  • Re: Tube matching?
    ... It will produce bias voltage from -16.85 to about -2. ... with plate current fixed at 45mA. ... octal tubes. ...
    (alt.guitar.amps)
  • Re: Traynor amp failure...need urgent help!
    ... amp and not a cathode biased amp. ... Start your own amp bias thread... ... the plate voltage, did we?). ...
    (alt.guitar.amps)
  • Re: Collins 32V-3 HF Transmitter NICE!!!
    ... Or RMS voltage times RMS current. ... The basic fact that I was not aware of is that there is an apparent conflict between the relationship between plate current and plate voltage if you look at the curves that show plate current as independent of plate voltage and then ask how do you get, on modulation, a peak input power four times the unmodulated input power so you can get a peak, on modulation, output power that is four times the unmodulated output power. ... It is just that the "direct relationship" will be different and anyone who compares the characteristics of two or more tubes will see that. ...
    (rec.radio.amateur.boatanchors)