7th harmonic distortion from PP pentodes depends on bias?

[ray_moth]

In modeling a PP pentode-mode Class AB1 EL34 amp with fixed bias, using LTspice, I've found that seventh harmonic distortion appears markedly worse with low quiesent curret, i.e. if it's biased too cold.

Biased close to the dissipation limits of the OP tubes, the distortion spectrum has the desirable 'waterfall' pattern (for odd harmonics only - the even harmonics cancel out in this 'ideal' model). But as the quiescent current is reduced, the 7th harmonic distortion starts to become more and more pronounced and the 'waterfall' pattern is ruined. Other odd harmonics (3rd, 5th, 9th, 11th) seem to be relatively unaffected by the bias setting.

I know it's been reported by many practitioners that the nearer you bias toward Class A, the better it sounds. I also know that 7th harmonic distortion sounds nasty. Put the two facts together and the modeling results I got seem to make sense. However, this IS only a model and I'm not sure how seriously I can take it. So, I wondered if anyone has found a relationship between bias setting and 7th harmonic distortion in the real world?

[Miles Prower]

Originally posted by ray_moth
However, this IS only a model and I'm not sure how seriously I can take it. So, I wondered if anyone has found a relationship between bias setting and 7th harmonic distortion in the real world?

I don't know about h7 in particular. However, with the project I built with 6BQ6GTBs, the usual bias setting, Pd= ~0.8Pmax, was considerably towards Class B operation. These being horizontal deflection types, I found I could spec bust, run 'em at Pd= 18W (12W rated, but that's for HD duty) and not get into red plate territory. Getting them further towards Class A definitely improved the sonics by reducing cross-over distortion.

That's probably where your h7 is coming from. X-over tends to generate high order harmonics.

[ray_moth]

Could be. The other thing I wondered was whether this was a particular vice of pentodes, as opposed to beam tetrodes. I'll have to try it with a 6L6 model and appropriate OPT - I use 3.5k P-P for EL34 but 6L6 needs more like 5.6k P-P for equivalent operation, AFAIK.

[kenpeter]

I get FFTs all over the place with LTSpice.
Wouldn't take anything the analysis shows too seriously.

Over on Nelson's side of the forum I had noted how
much LTSpice favored an imperfect plate bootstrap
over an actual constant plate current.

It gives funky results when impedances get too high.

I've also had troubles with windings over 10H.

Unless you are going to model ALL the things that
go on in a transformer, set the coupling factor at 1.

[Miles Prower]

Originally posted by ray_moth
Could be. The other thing I wondered was whether this was a particular vice of pentodes, as opposed to beam tetrodes. I'll have to try it with a 6L6 model and appropriate OPT - I use 3.5k P-P for EL34 but 6L6 needs more like 5.6k P-P for equivalent operation, AFAIK.

This probably won't settle anything. Different types give different results. Doing the twin-T test with the 807 gave results in line with the promised THD spec. However, the residual distortion had a significant amount of high order harmonics. Open loop, those sounded really nasty with some program material. Because of that, I included both local and global NFB in the 807 amp design, and this was the recommendation of the developer of the type: O. Schade of RCA.

6BQ6GTBs, OTOH, gave an estimated THD of 5.0%, but measured more like 2.98%. The residual distortion was almost pure h3. Consequently, these sounded much better when run open loop. The sound was more like overly "aggressive" without any pentode nastiness unless you cranked 'em almost to the point of audible distortion. You could consider the 6BQ6GTB to be a more powerful 6V6. The design here didn't include local NFB, just gNFB to take the excessive edge off the sound and to improve woofer damping.

That pretty much agrees with listening tests. H2 and h3 aren't too sonically detrimental, but higher order harmonics are definitely dissonant and nasty to listen to. As for why you get these results, who knows?

Modeling is, at best, an estimate, and I don't put too much trust in those. Better to spend some time listening -- at least a week -- while running open loop before deciding what your final design will actually need to improve the sonic results.

[jon_010101]

Beam tubes tend to have different distortion characteristics, including supposedly reduced high-order harmonic distortion compared to conventional pentodes. At least, according to Spangenberg (IIRC). The reduced screen current is another bonus. They are also slightly less sensitive to improper loading (curves more "squared off"), unless driven to the left side of curves where hell starts to break loose

Of course, at extrema (low current or low plate potential), it becomes highly tube dependent. Compare overall plate curve shapes of 6AQ5 with 6CZ5 (slightly-sweeter beam tube with similar characteristics), with 6BQ5.

[Merlinb]

Might this be related to the increased crossover distortion caused at colder bias settings, rather than the valves themselves?

[ray_moth]

Yes, that's what Miles suggested in Post #2 above. It could be, I don't know. There's certainly no crossover notch visible in the waveforms but, as it's only a sim, I'm not sure how far I can trust what it seems to be telling me.