SET amplifier Q
On Monday, 5 November 2012 14:03:31 UTC+11, Phil Allison wrote:
"John L Stewart" 'Phil Allison Sounds like by either intent or serendipity the 300B and 6SN7 2'nd harmonics are canceling.- ** Bingo !! Seems an eccentric Russian gent was responsible for the circuit design. There was an article on 2H cancellation in SE amps published about 10 years ago, I think in Glass Audio. Used the curvature of the driver to cancel curvature in the OP Tride. One would think this possible at only one level on the output. ** More testing with the FFT feature on a Rigol scope: 3H is dominant from about 9 watts through to 21 watts. From 1watt to 4 watts, 2H dominates, then 4H, then 3H. If load R is changed to 16 ohms, 2H dominates all the way. 2H cancellation is occurring over a fairly wide range, but it is not complete. I can swap the 6SN7 with almost no change. Neat eh ? .... Phil
Its actually possible to exploit the set up of a driver tube to cancel 2H of the SE OP tube. But while maximum 2H cancellation plus probable 4H cancellation can be arranged for the nominal load to be used, the cancellation may not be much when RLa is above of below nominal RLa. Roughly the same Vg input to 300B is needed over a range of RLa and a typical SE triode driver makes roughly the same 2H for most RLa values. But 2H is most severe in 300B when RLa is low, and then least when RLa is high. Some 3H is generated by intermodulation process and you can end up with higher 3H than you'd have if the driver triode was more linear. I found the cancellation of 2H with SE tetrode & pentode amps with about 15% local CFB could reduce THD to less than the same OP tubes used in a PP circuit. But with the multigrids, 2H is highest with low RLa, then as RLa is raised it nulls out at some centre RLa value even if the driver is perfectly linear, and then for RLa above the centre value 2H increases and its relative phase is opposite to low RLa values. Therefore the 2H of the driving triode adds to that of the multigrid OP tubes. But after spending much time studying this phenomena of THD cancellatiuon, I decided it was not worth deliberately setting up a driver tube to be slightly non linear to give maximum cancelling, and I concluded its better to make the driver as linear as possible no matter what driver and output tubes may be. I found that the CFB OPT connection forced THD production to be very much lower than for triode or UL SET, and then you'd still get some 2H cancelling with low RLa so the end result was THD equal to many class A PP amps over a useful range and for a similar amount of low GNFB applied.
Nearly all SE amps with 300B are ideally used where average PO 0.1 x max clipping PO so then average THD 0.2% and mainly 2H and lower 3H. This assumes you may have 10dB GNFB. But in fact if you remove the GNFB the sound change is very subtle if any can be percieved at all. I have had ppl say, "Well, now we have disconnected the NFB it sounds the same" - even though N&D has risen maybe +3 times for a given volume level. If RLa for the 300B is say 3k6, and Ra at the bias point is 1k2, then DF = 3 without any FNFB, not to be sneezed at. Ppl will argue the GNFB either gilds the lily, or makes it into pumpkin. My conclusion is that more ppl like adequate GNFB well applied than dislike it, citing better bass tighness and a creamier top end, ie, the NFB allows the tubes to blossom, so good GNFB is natural organic electronic fertilizer....
Some amps such as the EAR509, a 100W Paravicini PP concoction with a total of 44dB NFB including the unity gain OPT, does not sound as well as something with less NFB and far lower PO but with much more pure class A1. I think its pointless playing silly tricks with nearly all class B and huge amounts of NFB. But hey, that's what is done with OTL amps, and ppl rave that they can't hear the veil of the OPT. But they don't understand the tubes work mostly in class B and with terrible load matching and with huge amounts of NFB.
There is truth in that GNFB can increase the THD spectra so creating higher H that were not present when no GNFB is used. But the effect seems to rear its ugly head when the THD without NFB is high at say 10%, and not much NFB is applied, say 10dB, and 10dB is quite typical for triode amp GNFB, so it can be said use of this FB in a triode amp invites sonic horrors. There is a point where even if open loop THD = 20% before clipping, and there was still sufficient open loop gain, then increasing NFB to say 26Db will usually reduce THD to 1%, and as THD is about proportional to Vo, then at 1/3 full Vo and at 1/10 full PO, THD should be say 0.3%. But its not ideal though to start with 20%. Its better to have open loop THD 2% if possible and then the figures and sound all get better without huge dollops of NFB, and the mild 10dB GNFB does not make things sound worse. Along with applying GNFB come the rigours of making sure the amp is stable and does nt have a peaked response just below and above the AF band. One does not want LF peaking to upset dynamic DC core magnetization. I suspect many pp hate NFB because it forces them to think of the amp as an active bandpass filter with NFB, and think about Nyquist stuff.
There have been attempts to cancel 3H production of most PP amps where typically wave peaks are flattened slightly. Therefore one could make a driver PP amp stage which produced oppositely phased 3H, ie, peaks on waves became "more pointed" I tried all this about 12+ years ago and concluded it was a totally hopeless design aim. Just make it linear, said my brain, without tricks.
I also tried positive current FB and its easy to get a negative output resistance. But THD rises even though Ro becomes negative. And if there is a shorted output, the +current FB goes sky high and amp violently oscillates. No sir, no tricks, right?
Patrick Turner.
|