*grin*
Just got a loan of a Tek SG505 oscillator, plus AA501A distortion analyser, in a TM503B mainframe, as well as a HP3588A spec.an.
Now we're cooking with gas. The distortion analyser reads 0.0027% THD when I plug it straight into the oscillator.
Cheers,
Suzy (gonna have a busy lunchtime)
Just got a loan of a Tek SG505 oscillator, plus AA501A distortion analyser, in a TM503B mainframe, as well as a HP3588A spec.an.
Now we're cooking with gas. The distortion analyser reads 0.0027% THD when I plug it straight into the oscillator.
Cheers,
Suzy (gonna have a busy lunchtime)
Some results
These are measured results, into an 8 Ohm dummy load.
The minimum at 1KHz is 0.0013% (10W). 10KHz minimum is 0.0032% (20W). The amp is great to 120W (0.0015% @ 1KHz), then clips. The last two points are at 140W and 150W.
The lump in the distortion between 1 and 10W at 10KHz is an interesting feature. I haven't the foggiest idea what might be causing it.
Cheers,
Suzy
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
These are measured results, into an 8 Ohm dummy load.
The minimum at 1KHz is 0.0013% (10W). 10KHz minimum is 0.0032% (20W). The amp is great to 120W (0.0015% @ 1KHz), then clips. The last two points are at 140W and 150W.
The lump in the distortion between 1 and 10W at 10KHz is an interesting feature. I haven't the foggiest idea what might be causing it.
Cheers,
Suzy
You've got spectral analysis as well? To show weight of harmonics, to exclude the "+N" for lower levels and to see noise background.
i would like to hear about listening impresions by a third
party
distortion numbers mean nothing
cheers
john
party
distortion numbers mean nothing
cheers
john
PMA said:
Not as yet. The spectrum analyser I borrowed isn't much good at all. I can only get about 70dB SNR out of it, and that's nowhere near enough to see the harmonics.
The distortion analyser just gives a numeric readout. There's no information about what's in what harmonic.
From simulations, the distortion is mainly 2nd, 3rd, and 4th.
I think I'm still going to have to get a good soundcard to do a decent job of analysing it.
Though having said that, I'm fairly happy with the numbers I'm getting.
Now if I only had another clean source, I could do intermods...
Cheers,
Suzy
johndiy said:
My God, what have I been thinking all this time!
All this patient work ensuring the amplifier faithfully reproduces the input signal is clearly a complete waste of time. Clearly all I needed to do all along was bung some volts on the gate of a transistor, and "do magic" to make it "sound good". Stuff that the thing that I'm hearing bears no resemblance to what the artist recorded.
Oh, and I should put a good 10nF between my driver collectors and bases, to keep that sound warm! Don't want any of that terrible MOSFET mist creeping in now, do we.
Sorry dude, you're in the wrong thread.
Regards,
Suzy
One last plot
So I guess there's some slew rate limiting going on after 1KHz, which makes sense. Below 100Hz I'm probably seeing DC blocking capacitor non-linearities.
An externally hosted image should be here but it was not working when we last tested it.
So I guess there's some slew rate limiting going on after 1KHz, which makes sense. Below 100Hz I'm probably seeing DC blocking capacitor non-linearities.
I guess what he meant is that THD measurements don't correlate well with what you hear (although in my book any transistor amplifier with too much THD is masochism).
Schematic critique :
- unbalanced base resistors in the input pair : the pair will not be perfectly balanced, increases distortion.
- no driver before the output stage : compromises linearity as the very nonlinear input impedance of the output stage is driven by a high impedance current stage (increases crossover distortion among others)
- compensation isn't very easy (no Cdom because of the double balanced structure)
- the VAS diff pair doesn't have emitter degeneration resistors ; any imbalance will add distortion
I don't like this schematic. Have you read Douglas Self's book ? It is extremely interesting.
Schematic critique :
- unbalanced base resistors in the input pair : the pair will not be perfectly balanced, increases distortion.
- no driver before the output stage : compromises linearity as the very nonlinear input impedance of the output stage is driven by a high impedance current stage (increases crossover distortion among others)
- compensation isn't very easy (no Cdom because of the double balanced structure)
- the VAS diff pair doesn't have emitter degeneration resistors ; any imbalance will add distortion
I don't like this schematic. Have you read Douglas Self's book ? It is extremely interesting.
One can see fast rise of distortion above 2kHz, and this does not indicate to good sound.
Probably 100Hz - 1kHz range is at the limit of the measuring method?
Probably 100Hz - 1kHz range is at the limit of the measuring method?
johndiy said:
I agree, except that I would have followed the statement with "if it sounds crap".
Actually, I built one of these 6000 beyomeths back in 1988-89. It was the first kit amp I'd ever tackled and building it was not a simple task. I managed to get hold of those dual n ch fets (2n6811 ??) and I still have about 5 or 6 spares in an unmarked component tray - somewhere. To answer your question though, and distortion measurements aside, the 6000, or at least my 6000, sounds very very good. Actually I would describe the sound as very "nice". To pen a few Audiophile adjectives, there's barely a hint of grain, not a trace of metallic edginess and the from the mids to the top what you hear is sweet and fluid. If warmth is a mosfet trait then yes there is some but only compared with your typical high fb bipolar designs. Over the years I've compared it (in my own system) with many other power amps including a Krell KSA200 and an ME1500. While both of those amps do outperform the 6000 in certain areas, I felt that the 6000 had the sweetest and most grainless treble. The ME1500 does not use global fb so I was surprised that it could sound, at times, harsh. The Krell had a cooler and more clinical presentation of the treble, but was grain free, and the bass, well it had me wondering if the plateau bias cct might be somehow colouring it.
FTR, I also built the SC LD 100W class AB amplifier. It's not bad if you like an antarctic treble, but I for one don't. If that's more natural then I'll take the msg and tropical nutrisweet additives every time, no matter what the spec an is telling me.
BTW, I've also converted the the 100W SC LD amp to bipolar -mosfet cfp ouput (2SK1058 / 2SJ162), and so far no signs of any oscillating up to 100MHz limit of the cro. I haven't rebuilt the pc analyser as yet (it was dismantled) but will do in the next few days. My sound card has HDL2 / 3 products at -103dBr0 (24 / 48KHz) so it will give me some idea.
peufeu said:
Defensive response:
- Input is a JFET pair, so has no base. The gate resistors are within a few percent of one another. Simulation shows absolutely no difference when they're exactly balanced, so I didn't bother.
- A driver would have reduced the quiescent current requirement in the VAS. If I ever build one with half a dozen pairs of MOSFETs, I'll be sure to include one. Again, simulation with two pairs of MOSFETs showed no useful improvement by adding followers.
- Agreed. Compensation was a challenge. I could probably improve the THD at high frequencies by reducing the VAS miller caps and gate resistors (not to mention reducing emitter degeneration throughout), but I erred on the side of stability.
- Dunno what you're talking about. The only transistors in the whole circuit without emitter/source degeneration are the cascodes.
I do like the schematic. It works very well, IMHO. And no, I haven't read Self's book.
Regards,
Suzy
PMA said:
Yes but - in absolute terms the distortion is actually very low above 3KHz. The rise merely implies a diminishing feedback factor which is unavoidable if you want stability. Of more importance is the harmonic analysis of this residual. This also gives clues as to the origins of the distortion rise. With complementary outputs and diff inputs then it's usually 3rd harmonic with a little 2nd and 4th thrown in.
Hm. I wa speaking about this one :
http://www.atnf.csiro.au/people/Suzy.Jackson/ETI5000.jpg
did I miss something ? 😀
http://www.atnf.csiro.au/people/Suzy.Jackson/ETI5000.jpg
did I miss something ? 😀
peufeu said:
Aaah. That explains it. Sorry, this thread has wandered a bit. We've been discussing the AEM6000 for a while (see
http://www.littlefishbicycles.com/poweramp/index.html for my implementation of this design.
I agree 100% with your critique of the ETI5000 schematic.
Cheers,
Suzy
You should rename it Big Fish.
Craftwomanship article, topnotch board design.(and more humor than you care to admit)
Shame those designs weren't built like that 25 odd ago.
Craftwomanship article, topnotch board design.(and more humor than you care to admit)
Shame those designs weren't built like that 25 odd ago.
suzyj:
very nice work, and nice writing at the web site, too!
it's great to see a project start and finish on this board.
any chance for you to "borrow" a nice amp from a nearby shop or friends to do A-B comparisons?
mlloyd1
very nice work, and nice writing at the web site, too!
it's great to see a project start and finish on this board.
any chance for you to "borrow" a nice amp from a nearby shop or friends to do A-B comparisons?
mlloyd1
suzyj said:
well youre right?, faithful reproduction of music is ultra low dist
and super high rf speed!?, perhaps you have never heard of
nfb dynamics and how the sonic properties of an amp can vary
from one design to the next, besides people listen to amps to
see if they like the sound then they look at the numbers i am
talking about audiophiles and high-end here, from what ive seen
from the eti5000 design there is nothing special about it and if
it turns out to have some good performance it would be because
of pure magical coincendence when you have to say something
about faithful reproduction it all comes down to good listening
tests and not numbers that have been in the mainstream audio
design some decades now, today high-end is all about listening
and sound thanks to nfb dynamics and new experimental ideas
regards
aem 6000
Johndiy,
“Sounds nice” and “faithful” reproduction are two completely different things. How many blind ABC and AB test did you actually carry out with musicians who were to tell, which system was most faithful to a live performance (classical instruments and vocals are our subject)?
If you go for it you will see some disagreements. In some cases some musicians will vote for say system A, while for others system B would appear most faithful. That is mainly because our brains differ in subjective evaluation of what senses tell them (assuming that they are OK). To make things more interesting, there are people who in non-blind ABC or AB tests can choose a preferred system but then in blind ABC or AB tests cannot tell it apart.
I think many of us saw good optical illusions. What we see is not there and what actually is there we do not see. The same applies to hearing to a substantially greater extent than seeing as seeing is more important to survival than hearing.
Nevertheless, there is a notable and statistically significant correlation between measurable performance of audio equipment and more or less faithful reproduction of input signals. The most important in this chain are speakers and microphones, preamps are generally least noticeable, while power amps are in between. Of course there are aspects of sonic reproduction we do not measure as we do not yet know what to measure and how.
Same happens with our brains. Apparently, electronics of our brain work not only at biological level but even at subatomic levels about which we know almost nothing - if we knew much about biology of our brain!
Anyway, from my experience the most faithful amps are these, which are capable to reproduce complex signals without distorting them. That, among other things, implies high slew rates and wide bandwidth as all kinds of inter-modulation and crossover distortions are to be minimised. Importance of reducing intermodulation distortion was strongly emphasized back in 1960s by prof. Otala. His concept of a good amplifier was soon taken up by Electrocompaniet – a company, which produced the best (most faithful) amps at least until the mid 1970s.
Of course there are many ways to skin a cat so we have many good amps today and many more pretty average and some lousy ones as well.
Suzyj,
When it comes to THD I agree with David. The amp is not slow either as Dave Tilbrook designed it to minimize intermodulation distortions without compromising stability. In its times and today it is a capable amp. It still may be tweaked (fast caps – eg Sikorel or even top Nichicons would likely make a diference on transients plus some tweaks in the circuit) but audibly the most progress is achieved by going active.
Cheers,
Johndiy,
“Sounds nice” and “faithful” reproduction are two completely different things. How many blind ABC and AB test did you actually carry out with musicians who were to tell, which system was most faithful to a live performance (classical instruments and vocals are our subject)?
If you go for it you will see some disagreements. In some cases some musicians will vote for say system A, while for others system B would appear most faithful. That is mainly because our brains differ in subjective evaluation of what senses tell them (assuming that they are OK). To make things more interesting, there are people who in non-blind ABC or AB tests can choose a preferred system but then in blind ABC or AB tests cannot tell it apart.
I think many of us saw good optical illusions. What we see is not there and what actually is there we do not see. The same applies to hearing to a substantially greater extent than seeing as seeing is more important to survival than hearing.
Nevertheless, there is a notable and statistically significant correlation between measurable performance of audio equipment and more or less faithful reproduction of input signals. The most important in this chain are speakers and microphones, preamps are generally least noticeable, while power amps are in between. Of course there are aspects of sonic reproduction we do not measure as we do not yet know what to measure and how.
Same happens with our brains. Apparently, electronics of our brain work not only at biological level but even at subatomic levels about which we know almost nothing - if we knew much about biology of our brain!
Anyway, from my experience the most faithful amps are these, which are capable to reproduce complex signals without distorting them. That, among other things, implies high slew rates and wide bandwidth as all kinds of inter-modulation and crossover distortions are to be minimised. Importance of reducing intermodulation distortion was strongly emphasized back in 1960s by prof. Otala. His concept of a good amplifier was soon taken up by Electrocompaniet – a company, which produced the best (most faithful) amps at least until the mid 1970s.
Of course there are many ways to skin a cat so we have many good amps today and many more pretty average and some lousy ones as well.
Suzyj,
When it comes to THD I agree with David. The amp is not slow either as Dave Tilbrook designed it to minimize intermodulation distortions without compromising stability. In its times and today it is a capable amp. It still may be tweaked (fast caps – eg Sikorel or even top Nichicons would likely make a diference on transients plus some tweaks in the circuit) but audibly the most progress is achieved by going active.
Cheers,
- Status
- Not open for further replies.