Re: Re: Re: Re: Re: Re: Re: Re: Output impedance of Hafler DH220
Edmond,
I was just looking over your design again and noticed that R14 is in series with C15, don't you need a DC path to bias on D1, D4, Q7, and Q8, BE junctions? Or did I miss something?
Thanks for sharing your design, nice work!
Pete B.
estuart said:
Edmond,
I was just looking over your design again and noticed that R14 is in series with C15, don't you need a DC path to bias on D1, D4, Q7, and Q8, BE junctions? Or did I miss something?
Thanks for sharing your design, nice work!
Pete B.
Bob Cordell said:
Well so would I, but I don't promote the use of low NFB - unlike quite a few of those who can hear output inductances (and like high THD)...........
Bob Cordell said:
I have spent quite a LOT of time trying to make a good symmetrical differential input stage with current mirror loading, but I have not been able to avoid the need for a load resistor to define the VAS bias current without over-complexity with VAS servo amplifiers and the like. There are disadvantages to this scheme, but they can be avoided by designing the rest of the circuitry adequately. For instance in my simple class A amp:
http://users.picknowl.com.au/~glenk/12W.HTM
I use symmetrical differential inputs with I mirrors and forced current through a 1k resistive load to adequately define the VAS current. Those 1k resistors may look like a horrible compromise, but they really are not. Having push-pull drive to each VAS greatly improves performance over just a single-ended resistive pull up.
The transitional type of miller compensation is used, but I use a triple emitter follower output stage, so VAS output impedance isn’t a problem. Testing of the amp pretty much confirms the simulated THD of around 10ppm THD-20.
In my 512W amp:
http://users.picknowl.com.au/~glenk/512W.HTM
I have gone for dual complementary LTP inputs again, but have reverted to simple resistive loading (because I also have a dual complementary LTP VAS and symmetrical/bridged output stages) . You like JFET LTP’s and I like BJT LTP’s. Due to the much higher transconductance of BJT’s, I can get away with having much lower collector impedances, whilst still achieving an adequate amount of LTP gain and linearity.
Cheers,
Glen
Re: Re: Re: Re: Re: Re: Re: Re: Re: Output impedance of Hafler DH220
Hi Pete,
Your are quite right. Just remove the "spurious" C15. Due to down scaling of the drawing it's hardly visible that C15 is shorted. Also the dot at the node C5/R18 has vanished, so this node is also connected to C6..C9.
BTW, the schematic was only meant as an example of the basic idea. It's far from complete, no protection, no bias generator etc. and the BC327S and BC337S are "virtual" transistors. Replace them for example by 2SC3601 and 2SA1407 respectively.
Please, accept my apologies for all these inaccuracies.
Cheers, Edmond.
PB2 said:
Hi Pete,
Your are quite right. Just remove the "spurious" C15. Due to down scaling of the drawing it's hardly visible that C15 is shorted. Also the dot at the node C5/R18 has vanished, so this node is also connected to C6..C9.
BTW, the schematic was only meant as an example of the basic idea. It's far from complete, no protection, no bias generator etc. and the BC327S and BC337S are "virtual" transistors. Replace them for example by 2SC3601 and 2SA1407 respectively.
Please, accept my apologies for all these inaccuracies.
Cheers, Edmond.
G.Kleinschmidt said:
Hi Glen,
In the past I have showed you an example of a VAS servo (a common mode control loop), see:
http://www.diyaudio.com/forums/show...d=101745&perpage=10&highlight=&pagenumber=121 post#1207, which comprises 6 additional transistors. Not that much, regarding the 40 transistors already employed in your 12W design. So, I wouldn't call such a servo an "over complexity". Apart from more gain and better control of the VAS standing current, such arrangement is far less susceptible to a mismatch of the two Miller cap's. Even a few percent does upset the VASes. Maybe Bob is willing to explain again the story of "fighting VASes" or give us a link to his treatise on this subject.
BTW, so you have built and tested the 12W amp. Great! Can you give us some more THD figures? as I like to compare them with my simulations.
Cheers, Edmond.
G.Kleinschmidt said:
Hi Glen,
I agree completely with this statement, but I hope you don't think that my use of a single input differential pair means that I use a single-ended VAS. I always use a full-complementary VAS, but I drive it differentially from the single input differential pair. This can be seen in the design of my MOSFET power amplifier at www.cordellaudio.com under published papers.
Bottom line is that the full complementary dual differential pair input is a convenient way to drive a push-pull VAS, and looks nice and symmetrical on paper, but it is not the only way to skin that cat. I believe that my approach yields higher performance with perhaps slightly greater complexity.
Cheers,
Bob
the complementary diff/current mirror bias problem has been discussed before
Unstable VAS current in amp from Slone book:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796
I simmed a servo approach
optoisolator VAS bias for Comp Diff?:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=56860
Unstable VAS current in amp from Slone book:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796
I simmed a servo approach
optoisolator VAS bias for Comp Diff?:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=56860
G.Kleinschmidt said:
Hi Glen,
Thanks for referring me to this paper by Marshal Leach. I finally got around to reading it in detail. His scheme for eliminating the need for the coil by transitioning the HF NFB pickoff from after the output stage to before the output stage is interesting.
In the paper, he quoted, based on SPICE simulations, effective output inductances of 0.9 uH and 4.5 uH with and without his technique in place (both without coils). Interestingly, and probably as expected, the application of his technique effectively causes an output inductance to be electronically "synthesized" (for lack of a better term).
Of course, either of these numbers is greater than the numbers of the Haflers, but there are two things to point out. It appears that he has got the K factor for his MOSFETs wrong, so their transconductance at a given operating current is only about 60% of what it should be. This translates directly into a higher HF open-loop output impedance and a consequently higher simulated inductance. Secondly, he only has 26 dB of NFB at 20 kHz. When these two factors are taken into account, his output inductance results are pretty consistent with what the Haflers are doing.
Finally, it would have been nice if his paper included at least simulation of HF distortion with and without his technique in place. I'm guessing that the application of his technique may significantly increase HF distortion from to the output stage, perhaps on the order of the factor by which it increases output impedance at any given frequency. I suspect that there is no free lunch here.
Cheers,
Bob
Hi Bob,
With reference to the practice of taking NFB off before the power stage instead of after it, perhaps just that I also find it advantageous in the case of difficult loads. (I am talking, in my case, of a FCP output stage and some 29 dB of global NFB to take any remaining distortion above 3rd to below the threshold of audibility.) I found the total distortion to increase by only some 30%, still in the 0.002% region.
Although I find this more stable, I found no way to make an FCP output stage unconditionally stable with all loads. I did some 70 simulations, and do need an inductor in some cases (referring to the parameters that you outlined when you asked that need for coils be discussed).
In general, may I say that I read the past number of pages here with great interest (and learning! - up to now I have not tried mosfets), but I do retain a feeling that the "Look Ma, no output coil!" syndrome is a little overrated, as I have stated before. I have by no means the best design on the show, but everything indicate that my results are blameless, and I still cannot see why avoiding a few uH is anything worth spending a whole lot of energy on. Some designs (not referring to anyone in particular) seem to boast high NFB and damping factors of hundreds - to what avail? I would guardedly say that sometimes sight is lost of what will make a practical difference.
Regards.
With reference to the practice of taking NFB off before the power stage instead of after it, perhaps just that I also find it advantageous in the case of difficult loads. (I am talking, in my case, of a FCP output stage and some 29 dB of global NFB to take any remaining distortion above 3rd to below the threshold of audibility.) I found the total distortion to increase by only some 30%, still in the 0.002% region.
Although I find this more stable, I found no way to make an FCP output stage unconditionally stable with all loads. I did some 70 simulations, and do need an inductor in some cases (referring to the parameters that you outlined when you asked that need for coils be discussed).
In general, may I say that I read the past number of pages here with great interest (and learning! - up to now I have not tried mosfets), but I do retain a feeling that the "Look Ma, no output coil!" syndrome is a little overrated, as I have stated before. I have by no means the best design on the show, but everything indicate that my results are blameless, and I still cannot see why avoiding a few uH is anything worth spending a whole lot of energy on. Some designs (not referring to anyone in particular) seem to boast high NFB and damping factors of hundreds - to what avail? I would guardedly say that sometimes sight is lost of what will make a practical difference.
Regards.
Johan Potgieter said:
Johan,
Thanks for your observations. In some ways I think of the coil as being analogous to a seat belt in a car. 99% of the time you probably don't need it, but when you do, it is good to have.
Even if the audible effects are not well understood, it is good to keep the coil as small as possible, consistent with stability.
Cheers,
Bob
Bob Cordell said:
I don't think the intent of this paper was to eliminate the output coil. It was an attempt to achieve greater stability.
Bob Cordell said:
The HF distortion definitely goes up. I thought this was mentioned in the paper. However, as the 2nd harmonic of 10kHz is not audible, he didn't regard it as a "problem".
My slight reservation here:
If distortion goes up, that means poorer linearity, thus also increased intermod. distortion. This could mean audible products rearing their ugly spikes in the audible region even though the distortion order itself is super 20KHz. I wonder if any tests concerning this have ever been done - until then, I am a little nervous here.
Bob,
Yep - thanks.
If distortion goes up, that means poorer linearity, thus also increased intermod. distortion. This could mean audible products rearing their ugly spikes in the audible region even though the distortion order itself is super 20KHz. I wonder if any tests concerning this have ever been done - until then, I am a little nervous here.
Bob,
Yep - thanks.
Bob Cordell said:
In his conclusions he stated that the effect of raising the output impedance may be minimized if the feedback point is taken from a low-impedance point in the circuit. He changed this point in his DIY amp, relative to that in the paper. His DIY amp is tapped off of the predriver stage.
jcx [/i]the complementary diff/current mirror bias problem has been discussed before Unstable VAS current in amp from Slone book: [url]http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796[/url] I simmed a servo approach optoisolator VAS bias for Comp Diff?: [url]http://www.diyaudio.com/forums/showthread.php?s=&threadid=56860[/url] [/QUOTE] [QUOTE][i]Originally posted by estuart said:
Hi JCX,
My apologies for answering so late. It seems that you have solved the problem very nicely, that is, at the DC level. No problem if the gain of the two branches from input to VAS output, as well as the two compensating caps are equal. However, in your design the caps are only 10pF (btw, aren't they too small?), so, probably little or no "VAS fight".
Cheers, Edmond.