john curl said:
Mr Curl,
any plans of reducing the NFB even further in future designs ?
The continued bashing of Ayre designs shows rather bad taste, imo.
Considering Mr Hansen's track record alone, he should have earned some credit.
The pro guys overhere do show professional courtesy. (whether it's noblesse oblige or honour amongst thieves)
i'll likely be the only one with an oldy Cordell EC amp and an ExtremeA.
I'm not gonna tell you how it sounds, Mr C.
john curl said:
John, in all seriousness, I think that just about every class A/B amp I’ve ever made would have better open loop linearity than that Ayre design – especially so if I cranked the bias up for class A at similar levels. I’ve never directly measured my amps open loop, but I hardly need a sixth sense to work this out. To make comparisons, all I need to do is to examine the THD plots for the Ayre, and, picking points at specific frequencies, extrapolating from what I know my negative feedback factor to be. I may just be an “amateur” in the eyes of some, but this isn’t rocket science.
Also, considering your doubt over the worth of negative feedback, I’m surprised that you don’t run your amps open loop.
Cheers,
Glen
John Curl
--- many serious audio designers, after designing high feedback power amps over the decades, make an effort to make amps with limited or no global feedback.---
I noticed that. I also noticed that some audio designers (Kaneda, Hiraga) after decades of using high slopes (18 dB/o) in their crossover designs discover the "virtues" of 6 dB/o slopes. Were all the serious designers wrong in their youth ? Has the sound changed ?
I think that the late discover might probably be due to an aural preference in order to compensate for hearing being less accurate : the first and most pronounced effect of low feedback designs and 6 dB/o slopes is more harmonic and intermodulation distorsions, which means more level in the treble region.
--- many serious audio designers, after designing high feedback power amps over the decades, make an effort to make amps with limited or no global feedback.---
I noticed that. I also noticed that some audio designers (Kaneda, Hiraga) after decades of using high slopes (18 dB/o) in their crossover designs discover the "virtues" of 6 dB/o slopes. Were all the serious designers wrong in their youth ? Has the sound changed ?
I think that the late discover might probably be due to an aural preference in order to compensate for hearing being less accurate : the first and most pronounced effect of low feedback designs and 6 dB/o slopes is more harmonic and intermodulation distorsions, which means more level in the treble region.
darkfenriz said:
Thanks for your interest, Adam. Soon, hopefully. Right now I am struggling to prepare for the workshops we are presenting at HE2007 in New York in May (www.he2007.com). One thing I'm almost ready to post is a description of the Peak-Average display that we use in the workshop to show crest factor. Another item is more detail on the Distortion Magnifier. Yet another is on the EQSS subwoofer we will be using at the workshops.
Cheers,
Bob
G.Kleinschmidt said:Yikes! Well that leaves me earless– I’ve got two class A amplifiers on the go.
Hi Glen,
If you promise to bring to these heater amplifiers not in large quantities to market, I'll grant you a reprieve.
Huh? 512w rms, pure class-A, and only 350W idle? Not >1024W idle?
I haven’t finished building it yet and it looks like I’m going to have to take out a second mortgage on my house in order to buy a distortion analyser good enough to measure it.
Indeed, an AP is rather expensive. I solved this problem in a different way, by means of a professional sound card, a bank of passive filters and a S/W program (written by myself) to subtract the fundamental and distortion produced by sound card itself. Moreover. This progarm averages the residual over million cycles or so. As result, I can see the distortion products, that would otherwise completely immerse in noise, down to a level of -140dB.
Sure, in isolation, a class-A O/P stage is always better. But if one can build a class-B amp with 1ppm distortion I don't bother to bring it further down to say 0.3ppm using class-A.
My other design is just a simple little one that I just knocked up for a bit of fun, for driving an old pair of efficient PYE speakers I’ve got in my study. I noticed that you do spice simulations by request – care to do a despicable class A? Pretty please?
Cheers,
Glen
"I noticed that you do spice simulations by request"
But for you I'll make an exception, although it took me 3 hours to capture your schematic.
Hopefully, you don't mind that I'have changed your design slightly, because I don't have models of MJL19114/4 and I have R28/29 given a value of 47R and C5/6 150pF, as they were unspecified (see attachment). These values give a unity gain frequency of the global FB loop of 1.2MHz. Was that your intention?
The results:
THD is simmed at 20kHz and 1V below clipping as well as 1% of full power. BW = 100kHz. R67 = 10M -> TMC disabled.
Class-A, Iq = 1.5A, RL = 8R
Vo = 22Vpp: THD20 = 17ppm
Vo = 2.2Vpp: THD20 = 0.34ppm
Class-A, Iq = 1.5A, RL = 4R
Vo = 22Vpp: THD20 = 70ppm
Vo = 2.2Vpp: THD20 = 1.7ppm
For comparison, I also simmed the class-B version
Class-B, Iq = 50mA, RL = 8R
Vo = 22Vpp: THD20 = 30ppm
Vo = 2.2Vpp: THD20 = 42ppm
Class-B, Iq = 50mA, RL = 4R
Vo = 22Vpp: THD20 = 130ppm
Vo = 2.2Vpp: THD20 = 131ppm
As we can see, Class-A, particularly at low levels, is by far superior.
The end of the story? No! Now I did the same with Transitional Miller Compensation (TMC) enabled:
Class-A, Iq = 1.5A, RL = 8R
Vo = 22Vpp: THD20 = 11ppm
Vo = 2.2Vpp: THD20 = 0.3ppm
Class-A, Iq = 1.5A, RL = 4R
Vo = 22Vpp: THD20 = 21ppm
Vo = 2.2Vpp: THD20 = 0.4ppm
Class-B, Iq = 50mA, RL = 8R
Vo = 22Vpp: THD20 = 11ppm
Vo = 2.2Vpp: THD20 = 0.9ppm
Class-B, Iq = 50mA, RL = 4R
Vo = 22Vpp: THD20 = 29ppm
Vo = 2.2Vpp: THD20 = 35ppm
Class-A is still better, but the differences are markedly smaller. Compared with class-A without TMC, class-B with TMC does quite a good job.
Glen, if you are interested in more fugures (like IMD), please, let me know.
BTW, I lost your e-mail address.
Cheers,
john curl said:
Hi John,
If I understand you well, local FB is OK, but applied to more than one stage it's suddenly a crutch? I find this a rather curious theorem that needs to be substantiated with scientific (not sonic!) arguments.
And why you make an exception for Halcro? Is Bruce Candy the one and only guy on earth who is allowed to use global FB?
Cheers,
BTW I'm not a amateur, rather an EE.
Estuart, being an EE doesn't make you a pro in audio amp design. You have to win some 'races' first. For example, do you think that Chas Hansen isn't aware of the output stage internal loop, first done by Hawksford, and now promoted by Cordell on this website?
He not only knows about it, but he has tried it. However, he told me that it did not make the amp 'sound' any better, so he decided not to employ it.
For us, sound quality is the sole criterion for real success.
He not only knows about it, but he has tried it. However, he told me that it did not make the amp 'sound' any better, so he decided not to employ it.
For us, sound quality is the sole criterion for real success.
