YAP - Yet Another PowerAmp

Status
Not open for further replies.
Jump to Latest
#104
HEC vs NFB

andy_c said:
There is an important difference from a practical point of view. For a conventional feedback OPS, the bandwidth is constrained to be approximately the same as the ULG frequency. For an HEC stage, the bandwidth has no such constraint. It's instead constrained to be approximately the same as without HEC - normally a much higher number than the achievable ULG frequency of a feedback OPS.

Granted, its distortion will be no better (and maybe somewhat worse) than a conventional FB amp with the same ULG frequency, but its bandwidth will typically be appreciably higher. This has implications when it's placed inside of the feedback loop of the amplifier as a whole.
Click to expand...

Sigh. L’histoire se répète. The pros and cons of HEC and NFB have already been discussed last year.
Anyhow, the output devices are the prime components who limit the bandwidth. Although NFB requires a few more trannies in the signal path (preferable high speed types), in practice, the closed loop bandwidth as well as the phase response will be almost exactly the same. (Ovidiu measered 10MHz!!!).

Furthermore, also HEC requires frequency compensation, just like any other ordinary NFB circuit

I consider HEC as an interesting academic curiosity, or saying it more bluntly, as a weird kludge, typically embraced by NFB haters who erroneously think that HEC has nothing to do with NFB.
 
#105
Re: HEC vs NFB

Edmond Stuart said:


Sigh. L’histoire se répète. The pros and cons of HEC and NFB have already been discussed last year.
Anyhow, the output devices are the prime components who limit the bandwidth. Although NFB requires a few more trannies in the signal path (preferable high speed types), in practice, the closed loop bandwidth as well as the phase response will be almost exactly the same. (Ovidiu measered 10MHz!!!).

Furthermore, also HEC requires frequency compensation, just like any other ordinary NFB circuit

I consider HEC as an interesting academic curiosity, or saying it more bluntly, as a weird kludge, typically embraced by NFB haters who erroneously think that HEC has nothing to do with NFB.
Click to expand...


Edmond, Edmond, you continue to be so misleading in your characterization of HEC. I know you think it is a fraud, but others don't. I am certainly not a NFB hater and I think HEC is a valuable technique. This does not mean that I assert that it is the only way to get very low distortion; some of the other approaches, including some of the excellent work you have done, appear to do just as well. I have also always been open-mided about viewing HEC as a form of NFB and needing of compensation. I dare say I was probably the first to point that out in my original paper. But I have also defended the value of alternative ways of viewing HEC.

Why are you so bitter and close-minded about HEC? Is it the use of the word "correction" in its moniker that drives you nuts? It was invented and named by a Brit, after all.

Bob
 
#106
Hi Bob,

I'm NOT bitter and close-minded about HEC, rather annoyed. Both techniques performs equally well (that is, if precision resistors are applied in case of HEC).
I'm only getting mad when someone, after the exhaustive discussion and thorough analysis in the past, still claims that HEC is better than NFB, which is definitely not the case.

Cheers,
Edmond.
 
#107
Edmond Stuart said:
I'm only getting mad when someone, after the exhaustive discussion and thorough analysis in the past, still claims that HEC is better than NFB, which is definitely not the case.
Click to expand...

I'm not sure if you're talking about me here, but I'm not actually claiming HEC to be better. The need for precise resistor matching and the trim pot to minimize distortion with HEC is, frankly, a PITA, as is the need for boosted front end supplies.

However, if I want to reduce risk by reducing the ULG frequency of the output stage, with HEC I can do so without reducing the OPS bandwidth. That's not possible with 1DOF conventional feedback*. Does it matter? Seems like that depends on the implementation.


*traderbam did show a block diagram of a 2DOF conventional feedback system that could also do this, but I couldn't figure out a good implementation. Could be my loss.
 
#108
andy_c said:

However, if I want to reduce risk by reducing the ULG frequency of the output stage, with HEC I can do so without reducing the OPS bandwidth. That's not possible with 1DOF conventional feedback*. Does it matter? Seems like that depends on the implementation.
Click to expand...

Just another small detail; while I was successfully able to experiment and breadboard Bob's HEC as part of the PGP project (pictures are on my web site under History) the YAP OPS was impossible to stabilize, on breadboard, over about 3-4MHz ULG. Also the YAP OPS PCB layout is way more critical. As strange as it may look, HEC is closer to a quick win compared to anything similar to the YAP OPS.
 
#109
That's interesting. I remember reading your pages a while ago, where you were talking about breadboarding and that it worked out well for you when developing the PGP. But it sounds like you have found the limit of breadboarding, at least as far as power amp output stages and their attainable ULG goes.

I must admit, I am amazed at how high the ULG of YAP is. I simply didn't think it was possible.
 
#110
andy_c said:
I must admit, I am amazed at how high the ULG of YAP is. I simply didn't think it was possible.
Click to expand...

Actually the CLG of the whole amp is around 600KHz, set by the input RC filter, which also sets the slew rate. The ULG of the whole amp, without the input filter, Zoebel, output coil, is around 2.5MHz, set by the front end (opamp+transconductance stage).

Take a look at the OPS 5V/500KHz response in a 100nF capacitive load http://www.diyaudio.com/forums/showthread.php?postid=1575583#post1575583

It's fast, isn't it? 😱 The slew rate is around 270V/uS (non slewing, of course) at full output. Talking about the diamond buffer being slow 😀 😀 😀
 
#112
andy_c said:
Oh, yes I meant the YAP output stage. Yes, it is quite fast - CFB at work.

Did I get the gate stoppers right at 150 Ohm? My eyesight is terrible. Also, what are the 27k resistors in the OPS for?

Oh, and was it possible to stabilize the OPS without the zoebel?
Click to expand...

150ohm for N ch and 100ohm for P ch. I could probably go a little lower... It is stable without Zoebel and output coil, but I like staying on the safe side.

27k - mostly they are for avoiding false protection triggering on fast transients. Experimentally, i've noticed they also help minimizing the impact of the Q4 Q19 Q5 Q21 switches Cob on the OLG. Switching is also slightly improved.
 
#113
AKSA said:
Knock it off, you guys.
Accept that opinions differ, and leave it at that.
Click to expand...

Why 'knock it off', AKSA ????
Are we telling some things you do not like to hear (or more correct read).
I think YOU ar getting a bit emotional, at least you are subjective 😉
And what is wrong with emotions?
It is one spice, the salt in any fair & honest discussion.


'Why are so many audiophiles beguiled by SETs, particularly DHTs,
when the measured distortion is so outrageously high?'
There just HAS to be a reason for it. Hugh

What do you think is the reason, Hugh Dean?
You have any ideas .. speak out! If you dare ....... 😀
Or I may tell you my true answer .. and you will not like my opinion .. about 'audiophiles' .....


Think one fair question is:
---------------------------------------
Why do majority of Hi-Fi interested people prefer Transistors/IC-Chips Amplifiers?
Like most music listeners & companies selling Amplifiers.
Be it Class-D, MOSFET or BJT,
be it HEC, Cordell or plain GNFB feedback.

Not to mention most AKSA Amplifiers KITS 😀
-----------------------------------------------------------------------


syn08
Thanks for both schematics attached 😎
Current versions of the Output Stage and the Front Stage.

Lineup - Audio Regards!
 
#114
What could be done better...

It is fair to reveal a few small issues with the OPS.

1. I was overestimating the dVgs/dT of the output MOSFETs. It is in fact around 5mV/K and hence they are thermally a little overcompensated. The bias is set to 180mA and drops to 150mA after one hour at half power (worst case dissipation).

2. I was hoping that the same tube 2SK1530 and 2SJ201 are more or less already matched. Wrong guess... I got them from Digikey but the dispersion is pretty high, so in an ideal world they would need sorting and matching. Given the price of these critters, I've decided to keep them as they are, +/-20%

3. A design error. Under normal circumstances, the feedback resistor R52 dissipates very little power. However, under faulty conditions, after the protection kicks in, R52 "sees" the whole output voltage swing, as at the front stage output. Hence, the power dissipation increases significantly and the resistor quickly burns up. I am looking after a noninductive radial power resistor that would fit on the PCB, so far with limited success...

I'll report back with other deficiencies as I discover them... So far I have no good reason for another PCB revision.
 
#115
Re: HEC vs NFB

Edmond Stuart said:
I consider HEC as an interesting academic curiosity, or saying it more bluntly, as a weird kludge, typically embraced by NFB haters who erroneously think that HEC has nothing to do with NFB.
Click to expand...

Functionally, HEC is NFB. So it is erroneous to contrast them in functionality terms. As a NFB implementation, nominally a servo circuit, HEC allows the possibility of a perfect integrator in the FB loop transfer function. However, this is at the expense of the need for fine adjustment. A conventional servo has identical characteristics and avoids adjustments, although the LF loop gain may top-out sooner. But having enormous dc gain is unecessary and is a poor trade off for the burden of adjustment.

I think you are absolutely right: HEC is an academic curiosity. I also suspect many of it's advocates have failed to understand what they are advocating, or have chosen not to.
 
Status
Not open for further replies.
Top Bottom