Hullo all,
I'm working through Morgan Jones's books and Steve Bench's webpages, using as my example a 6S33S from which I got ~10% distortion at the second harmonic and 35% from the 6S41S. I've seen a number of amplifiers around the Infobahn with these valves in and was intrigued about their predicted performance.
I've checked and re-checked the calculations and got the same answers, so my first question is are the people who are listening to these amplifiers great fans of distortion or am I getting entirely the wrong end of the stick?
Secondly, though websites like GoodSoundClub recommend 1.2kOhm loads and around 220V/150mA quiescent operating values for the 6S33S (which results in these figures) they'd surely improve with higher loading and voltage?
Note I'm only talking about the figures here as an example to improve my understanding because of course everyone likes different sonic characteristics! However surely figures as ridiculous as 35% are way, way out if people are listening to these and saying they're accurate?
I'm working through Morgan Jones's books and Steve Bench's webpages, using as my example a 6S33S from which I got ~10% distortion at the second harmonic and 35% from the 6S41S. I've seen a number of amplifiers around the Infobahn with these valves in and was intrigued about their predicted performance.
I've checked and re-checked the calculations and got the same answers, so my first question is are the people who are listening to these amplifiers great fans of distortion or am I getting entirely the wrong end of the stick?
Secondly, though websites like GoodSoundClub recommend 1.2kOhm loads and around 220V/150mA quiescent operating values for the 6S33S (which results in these figures) they'd surely improve with higher loading and voltage?
Note I'm only talking about the figures here as an example to improve my understanding because of course everyone likes different sonic characteristics! However surely figures as ridiculous as 35% are way, way out if people are listening to these and saying they're accurate?
There are audiophiles who object to the use of the term 'accurate'. Baffles me!
Two possibilities:
1. You have misunderstood the charts, and the distortion is not as bad as you think.
2. You are about to tell the Emporer that he has a wardrobe malfunction.
I am not familiar with either valve, so I am not in a position to judge.
Two possibilities:
1. You have misunderstood the charts, and the distortion is not as bad as you think.
2. You are about to tell the Emporer that he has a wardrobe malfunction.
I am not familiar with either valve, so I am not in a position to judge.
http://www.mif.pg.gda.pl/homepages/frank/sheets/018/6/6S41S.pdf
Well looking at the curves, there seems to be two distinct areas of operation, with grid voltage under a 100 volts minus, and over 100 volts minus. If you operate the tube so that parts of the signal go over this line, you will have massive distortion, just as you calculated. However, if you stay firmly in the lower region, it seems to be quite linear. At least not horrible by any means.
It seems to be a relatively low voltage tube.
Well looking at the curves, there seems to be two distinct areas of operation, with grid voltage under a 100 volts minus, and over 100 volts minus. If you operate the tube so that parts of the signal go over this line, you will have massive distortion, just as you calculated. However, if you stay firmly in the lower region, it seems to be quite linear. At least not horrible by any means.
It seems to be a relatively low voltage tube.
What power level, load impedance, operating point and supply voltage ?
I think most big (DIY) SE amplifiers without negative feedback are performing very poorly with respect to distortion.
But like most "home made" amplifiers, those sound perfect to people who have built those. I have seen many such examples.
In the 6S33S's datasheet the normal operation point is 120V. However it seems accepted that this is for voltage regulation (its original purpose) so people seem to be running it at ~220V.
For the 6S41S I chose 150V/120mA (rather arbitrarily) because it seemed to be a tangent to the power curve and in a middling position with some headroom to change the load a little. I went for Mr Bench's "double the quiescent voltage" to get the tentative line, which was (as with the 6S33S) about 1.2Kohm. However after getting 35% I wondered if it were worth calculating for another line!
MrCurwen: after -100V the curves go in -20V increments rather than -10, so I think they're actually pretty similar in separation.
Artosalo: You're probably right - I've never measured the distortion on my amplifiers, but then I never designed them and just followed a diagram. Now I've an oscilloscope I'd rather like to do some tests!
For the 6S41S I chose 150V/120mA (rather arbitrarily) because it seemed to be a tangent to the power curve and in a middling position with some headroom to change the load a little. I went for Mr Bench's "double the quiescent voltage" to get the tentative line, which was (as with the 6S33S) about 1.2Kohm. However after getting 35% I wondered if it were worth calculating for another line!
MrCurwen: after -100V the curves go in -20V increments rather than -10, so I think they're actually pretty similar in separation.
Artosalo: You're probably right - I've never measured the distortion on my amplifiers, but then I never designed them and just followed a diagram. Now I've an oscilloscope I'd rather like to do some tests!
I was looking at these valves originally because they look rather exciting (and come on, why have your music made by little light bulbs if they don't look a bit out of the ordinary?) and thinking about push-pull. If my distortion figures are accurate (indeed DF96, some people dislike that idea! I think a wardrobe malfunction might be on its way, even if we allow for non-ideal speaker loading) then surely P-P is the only way to go?
Another question is are the distortion figures only tenable at high output power levels? If the volume were low or the speakers sensitive; would they be correspondingly decreased to the point of not mattering, or do they have a proportional relationship to output power?
Another question is are the distortion figures only tenable at high output power levels? If the volume were low or the speakers sensitive; would they be correspondingly decreased to the point of not mattering, or do they have a proportional relationship to output power?
It seems likely that 1k2 is a rather steep load line for this valve in spite of the fact that its dynamic impedance is low; MJ uses 2k for the SE design in his 3rd edition using a regulator triode. If you are using the formula to estimate then it is easy to see that this will decrease the numerator and increase the denominator. These regulator valves probably work best as Push-pull or better still,a power differential stage.
I have seen some designs around where the load impedance is 600 ohms. However I would have thought that for a similar operating point that would make the anode voltage very high at switch-on? I noted that Jones's distortion figures in S-E were high (7% 2nd harmonic I think), though I think he specified those were at maximum power (and "typical for the genre" I believe he wrote!) The book's not by me at present so I don't know if he measured or predicted them.
It seems from Bench's loadline examples that instead of a direct summation in P-P, the odd-order harmonics decreased slightly. This would mean that the lower values I calculated for 3rd harmonic on the 6S33S (around 1% I think) would be less noticeable. However again confusion arises as the 6S41S (which I believe is the single triode predecessor of the internally-paralleled 6S33S) got a 3rd harmonic result of -78.7%! I think I can assume that would be unpleasant to listen to.
I'm going to recalculate the 6S41S for a higher voltage quiescent point (more comparable with the example I tried for the 6S33S) and also for a lower load impedance.
DF96: I'm using Jones's labour-saving idea here that for P-P planning a good loadline for S-E can be assumed to work, instead of plotting compound curves that are only approximating the varying load anyway, so does the distortion continue in a similar progression for P-P or do the S-E results go out of the window beyond the obvious even order cancellation?
It seems from Bench's loadline examples that instead of a direct summation in P-P, the odd-order harmonics decreased slightly. This would mean that the lower values I calculated for 3rd harmonic on the 6S33S (around 1% I think) would be less noticeable. However again confusion arises as the 6S41S (which I believe is the single triode predecessor of the internally-paralleled 6S33S) got a 3rd harmonic result of -78.7%! I think I can assume that would be unpleasant to listen to.
I'm going to recalculate the 6S41S for a higher voltage quiescent point (more comparable with the example I tried for the 6S33S) and also for a lower load impedance.
DF96: I'm using Jones's labour-saving idea here that for P-P planning a good loadline for S-E can be assumed to work, instead of plotting compound curves that are only approximating the varying load anyway, so does the distortion continue in a similar progression for P-P or do the S-E results go out of the window beyond the obvious even order cancellation?
Last edited:
35% seems excessive. There's probably a mistake in there somewhere. You also have to consider that the graphical estimates for harmonic distortion lead to high side errors as these methods depend on at least one simplifying assumption: that the harmonics are either in phase or anti-phase to the fundamental. Isn't necessarily true.
I had an estimate of 5.0% H3 with PP 6BQ6 finals. The measured result was 2.98% H3. Doubling the bias current dropped that to an estimated H3= 0.18%, even though that busted the spec for Pd. Open loop, the cooler biasing and higher H3 was sonically inferior to the hotter biasing. (Being that the 6BQ6 is a TV HD type, the spec sheet ratings are very conservative, and biasing hotter than spec for the less demanding audio final use was of no consequence.)
These graphical methods give rough estimates for actual performance, which you can determine only by building and listening. Spend some time with it running open loop before deciding how to proceed, and what levels of NFB are likely to be needed to clean up the sonic defects.
Well at Zp = 3.3k, quiescent point 190V/100mA, 2nd harmonic is 8.5% and 3rd 2.85%. Rather improved! Grid bias -70V as it seems to give an even swing. Does that seem a reasonable approach?
However Vmax is 560V, so an expensive H.T. would be necessary and the power has decreased to ~3W. Not being experienced with the kinds of figures people are working with, does this seem to be worthwhile continuing with? I have no idea what effect it will have on the sound.
With this kind of power, in P-P I suppose NFB would be able to cause a significant reduction?
Edit: M.P. your post appeared while I was writing this - I realised it was only an estimate to see if a design's worth pursuing but didn't consider the phasing as well! I'm mainly trying to get my theory of pat before starting building as I've not seriously considered the reasoning behind the circuitry before. Also this side of the pond we've an alarming paucity of local shops selling components so it's more difficult for those of us who haven't had the years to build up a scrap box to get our hands dirty testing. I'm seeing my grandfather next weekend though, so I'll see if he's ready to part with more of his spares... I'm looking forward to giving the 'scope a workout on this too, so it'll be interesting to see how the predictions correlate with the measured results. Your going ahead with such figures and working them down gives me confidence! Thanks for the advice on taking account of the intended purpose - I'll bear that in mind. These valves are cheap so some experimentation might be called for into their characteristics!
However Vmax is 560V, so an expensive H.T. would be necessary and the power has decreased to ~3W. Not being experienced with the kinds of figures people are working with, does this seem to be worthwhile continuing with? I have no idea what effect it will have on the sound.
With this kind of power, in P-P I suppose NFB would be able to cause a significant reduction?
Edit: M.P. your post appeared while I was writing this - I realised it was only an estimate to see if a design's worth pursuing but didn't consider the phasing as well! I'm mainly trying to get my theory of pat before starting building as I've not seriously considered the reasoning behind the circuitry before. Also this side of the pond we've an alarming paucity of local shops selling components so it's more difficult for those of us who haven't had the years to build up a scrap box to get our hands dirty testing. I'm seeing my grandfather next weekend though, so I'll see if he's ready to part with more of his spares... I'm looking forward to giving the 'scope a workout on this too, so it'll be interesting to see how the predictions correlate with the measured results. Your going ahead with such figures and working them down gives me confidence! Thanks for the advice on taking account of the intended purpose - I'll bear that in mind. These valves are cheap so some experimentation might be called for into their characteristics!
Last edited:
If you build a P-P by essentially bolting two Class A SE back to back then the distortion will behave in the same way as output level changes, except that you will have some cancellation of even-order distortion.
If you change the bias and anode load, to take advantage of P-P, then distortion can get a bit more complicated. You can end up with either crossover distortion or gm-doubling if the bias is wrong. Some people prefer gm-doubling over 'correct' bias. Nobody likes crossover distortion.
If you change the bias and anode load, to take advantage of P-P, then distortion can get a bit more complicated. You can end up with either crossover distortion or gm-doubling if the bias is wrong. Some people prefer gm-doubling over 'correct' bias. Nobody likes crossover distortion.
"It seems likely that 1k2 is a rather steep load line for this valve"
A more normal operating point is 600ohm load with 220V anode voltage and 200mA anode current. This operating point give about 13W at less than 8% distortion. Therw where quite many amplifiers built in Japan with this operating after Dakesue released his first amplifier, (he was the first to import 6C33C to Japan, (he renamed it EC33C). I have one amplifier designed by him with this operating point and Tango transformers.
All regulator tubes are not the same, 6C33C can carry about twice the current as the 6528 used by Morgan Jones so naturally load impedance will be different.
BR Hans
A more normal operating point is 600ohm load with 220V anode voltage and 200mA anode current. This operating point give about 13W at less than 8% distortion. Therw where quite many amplifiers built in Japan with this operating after Dakesue released his first amplifier, (he was the first to import 6C33C to Japan, (he renamed it EC33C). I have one amplifier designed by him with this operating point and Tango transformers.
All regulator tubes are not the same, 6C33C can carry about twice the current as the 6528 used by Morgan Jones so naturally load impedance will be different.
BR Hans
tubetvr - I think that remark was referring to the 6S41S. On the curves, the 6S33S operating point is way over the maximum power dissipation for the 6S41S. I get better results with the lower and higher impedances from my original 1.2k loadline though!
At 0.9k 2nd is 2.41% and 3rd is -1.81%. I'm just going to try with 600 ohms to see if that improves it still further (and I've seen OPTs for 600).
At 0.9k 2nd is 2.41% and 3rd is -1.81%. I'm just going to try with 600 ohms to see if that improves it still further (and I've seen OPTs for 600).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.