[patrick-turner]

On Tuesday, 6 November 2012 02:46:34 UTC+11, John L Stewart wrote:

snip for brevity,

The THD may go down but what about the higher order harmonics & IM? It has been shown by many tests that these are the objectionable components in the reproduced sound. Cheers to all, John

Well, you are possibly about correct about IMD being worse sounding than THD. All music generates many harmonics of fundemental F produced by instruments. An amplifier or speaker adds to the THD but is not all that bad itself. If a saxaphone's 3H is increased by 1% then the tone won't seem to have changed much. But IMD products, ie, the sum and difference between various F are often quite an-harmonic, and their presence can have a far greater effect.

If you have an OP tube which makes 5% of 2H near clipping, then for 2H cancellation you need the driver to make 5% of 2H. That 5% cancels the OP 5%, but in fact the sum and difference between the driver 2H and fundemental are produced, ie, 1H and 3H, ie, more fundemenmtal plus 3H, and so you can get more 3H than was present to begin with even if you had a perfect driver. But 2H + 5H yields 3H and 7H, and suddenly you have lots more H than you started with although their levels tend to be lowish, and which are called second order. AFAIK, there is no need to worry about their production where the amp had low THD without any FB, as is the case with most well made class A amps used where they are meant to be used at average power 1/10 clipping power. One can always analyse to find just how many H a single 6SN7 might produce even when set up optimally. One may find H extending to 10th H. So what? We still must use devices to amplify sound, and you can't avoid the H production by one weay or another. RDH4 gives a guide to IMD significance and cites that IMD is around about 3 times the % of THD where you have the standard old IMD test where there the output V0 has say 4V of 80Hz and say 1V of 5kHz. So if THD was say 1% at 4V of 80Hz, then expect 3% IMD relative to 5kHz levels. One can display the 5kHz tone on a CRO, and without any 80Hz by means of using a high pass filter, and then you can see what % depth of amplitude modulation exists, ie, presence of sidebands at 4,920Hz and 5,080Hz. Neither of these F are harmonious with 5,000Hz, and one may think they'd sound bad, but IMD also is generated within instruments, and our ears seem used to such slight amplitude modulation of high F tones by the usually much larger amplitudes of bass tones.

Anyway, All I know is that if you have an SE amp which measures well without using much GNFB, it usually sounds just fine at levels for which it has been designed for. I have found that the best way use multigrid tubes to get basic linearity is to employ CFB windings on the OPT, and not to rely solely on the triode connection for FB via the screen, which is of course the same as the triode internal NFB in a real triode such as 300B. With say 20% CFB with a multigrid and SE amp, THD just under clipping can be routinely less 2% with correspondingly low IMD.The CFB converts the OP stage to have quite low voltage gain like as if it was a triode. So the driver stage which is usually a signal triode then has to make a substantial amount of drive voltage, say 60Vrms and so it also naturally generates around 2% THD even if fairly well set up. So some substantial 2H cancelling occurs inevitably. Usually an OP triode tube at near clipping may make far more 2H than the driver tube could ever make unless someone has set up the driver with a low value of anode load to provoke production of 2H.

I've long since given up being too concerned about the unavoidable issues of THD/IMD in tube amps. All one can do is minimise them "legitimately", and I find 99.999% of everyone likes the sound. And if they change to something solid state and THD becomes 100 times less because an extra 40dB NFB is used, they don't hear the slightest betterment, and in fact are likely to dislike the sound.

As I said previously, I working on drastically reforming a pair of Jadis 300B SE amps. The owner is currently using Atmosphere OTL with sensitive speakers and it will be interesting to see if he finds the Reformed Turnerized Jadis to be better. Perhaps he'll find it different, rather than better/worse, but my main function has been to make the Jadis work properly, quit smoking, and stop being a sulking and silent and expensive French loungeroom companion. I have thought of sending the original Jadis OPTs to Jadis in France but methinks they'd take no notice at all, and pretend i didn't exist, and probably never admit they ever could have forgotten to use an air gap. I'm trying not to be unkind. Anyway, Jadis are free to take my advice in future when I post up my reforms for all to read. BTW, the Jadis have CLC filtering for main 300B B+. The choke used is a 20mm stack of 20T E&I, and C are 2 x 470uF/500V. Nice caps, but shame about the choke. The choke air gap was 10 times too large, reducing L value to a negligibly effective amount. I was able to re-gap the little choke to give nearly 2H, and thus it is now useful. PSU resonance is now 5Hz, Q not too high, and just passable.

I could perhaps use the Jadis OPT for a future parafeed amp because available wasteless winding Z ratios are 888:1 and 222:1. So 888:1 gives 5k0 : 5.6r, and Lp is high enough with no air gap at 88H, and RwP = 76 ohms only, so losses are low enough. The OPT would them be useful for an 845. But with 222:1 I get 1.25k : 5.6, OK for 3 x KT88/6550 or 4 x EL34 in triode for about 30Watts. Using 2 x KT120 in triode also comes to mind, because one may idle the KT120 at Pda at 35W and possibly 40W because Pda rating = 60W. A single KT120 is potentially able to do what 2 x 6CA7/EL34 can do in SE.

But KT120 triode Ra is not really all that low to allow a low RLa value, and so RLa must be high with probable Ea at +500Vdc, and Ia up to say 70mA, with load for one at 5k0, giving PO = 12 Watts. Theroetically 2 x KT120 in parallel should give 25Watts into RLa = 2k5, just right for a Hammond SEA1627, but not that easy with the old Jadis OPT because there are just not enough sensible and available load matches.
Patrick Turner.