Couple of improvements - v0.7
OK, this one seems to be final for this design
Slightly modified input section to allow high performance low voltage BC5XX, switched to non-inverting servo, connected to low-impedance input and - most important - moved to double-drivers configuration - approach, similar to the one used in Alchemist Forseti APD15. Separate driver for each output transistor.
Roughly 30-50% lower distortion, same high stability margins and closed loop responses.
Working on test PCB layout...
Cheers,
Valery
P.S. A bit of a surprise for me - this double-drivers approach shows slightly wider open loop bandwidth = better 20KHz performance
OK, this one seems to be final for this design
Slightly modified input section to allow high performance low voltage BC5XX, switched to non-inverting servo, connected to low-impedance input and - most important - moved to double-drivers configuration - approach, similar to the one used in Alchemist Forseti APD15. Separate driver for each output transistor.
Roughly 30-50% lower distortion, same high stability margins and closed loop responses.
Working on test PCB layout...
Cheers,
Valery
P.S. A bit of a surprise for me - this double-drivers approach shows slightly wider open loop bandwidth = better 20KHz performance
Attachments
Last edited:
The output transistors see a lower source impedance and this is reflected as a smaller time constant with the transistor capacitance.
I choose distributed drivers for other reasons but found this a nice bonus.
Best wishes
David
The open loop UGF is indeed many MHZ. The sx-Amp is about 20 MHz IIRC and in uncompensated closed loop it's -3dB at 11 MHz (very peaky).
The loop UGF is typically 1-1.5 for a triple, and up to 3.5 for an EF2 using modern 30 MHz devices.
However, classic CFA can go much higher because their lower OLG and therefore loop gains mean the OPS pole can fall at or below the 0 dB gain line. In other words, you can get away with very light comp and still have a stable amp. This allows you have a very wide loop gain BW, if that's your choosing.
The minute you go for a high OLG CFA design, you have the OPS pole to contend with ( it will now lie well above 0 dB) - this will require some form of pole splitting or advanced comp just like a VFA.
The loop UGF is typically 1-1.5 for a triple, and up to 3.5 for an EF2 using modern 30 MHz devices.
However, classic CFA can go much higher because their lower OLG and therefore loop gains mean the OPS pole can fall at or below the 0 dB gain line. In other words, you can get away with very light comp and still have a stable amp. This allows you have a very wide loop gain BW, if that's your choosing.
The minute you go for a high OLG CFA design, you have the OPS pole to contend with ( it will now lie well above 0 dB) - this will require some form of pole splitting or advanced comp just like a VFA.
we usually talk about the feedback gain intercept frequency when we have high minimum gain amps like audio power amps
sometimes called unity feedback gain frequency, or unity loop gain frequency - when the amp open loop gain == the feedback gain network attenuation
literal unity gain frequency isn't often looked at
and the "speed advantage" of CFA or "flat loop gain" is relatively small - it is high voltage driver and output device speeds that are most often limiting
"flat loop gain" is a poor system choice - if you have even 20 dB of loop gain you only "win" 5.7 degree of phase margin compared to having high DC gain/sloping loop gain of 1st order all the way down to Hz
the phase margin cost of sub 20 kHz gain "corner" associated with high loop gain is negligible near 2 MHz loop gain intercept
and the excess loop gain/feedback factor reduces all errors - including phase errors, TIM/FM IMD... by the gain at the error's frequency - audio frequency IMD caused by higher frequency nonlinearity and difference tones gets reduced more by sloping "low bandwidth" high loop gain - potentially lots more
sometimes called unity feedback gain frequency, or unity loop gain frequency - when the amp open loop gain == the feedback gain network attenuation
literal unity gain frequency isn't often looked at
and the "speed advantage" of CFA or "flat loop gain" is relatively small - it is high voltage driver and output device speeds that are most often limiting
"flat loop gain" is a poor system choice - if you have even 20 dB of loop gain you only "win" 5.7 degree of phase margin compared to having high DC gain/sloping loop gain of 1st order all the way down to Hz
the phase margin cost of sub 20 kHz gain "corner" associated with high loop gain is negligible near 2 MHz loop gain intercept
and the excess loop gain/feedback factor reduces all errors - including phase errors, TIM/FM IMD... by the gain at the error's frequency - audio frequency IMD caused by higher frequency nonlinearity and difference tones gets reduced more by sloping "low bandwidth" high loop gain - potentially lots more
Last edited:
Vzaichenko that is a cool circuit but it's not simple
Andrew, it was rather simple in the very beginning, now - I agree - it's not that simple, but still "cool" and "clean" enough
At the same time - definitely more mature, rather compact in terms of PCB size and easy to build (at least, I try to keep it this way
)...Hi jcx, would you have any insight on phase requirements where the OLG meets the closed loop gain? If, for example, phase margin at that point would be low, but restore to decent levels at the ULGF, is that still acceptable? Or would one want to obtain at least some margin at that point? In my simulations, this doesn't seem a too big issue when the phase returns from 180deg close to this equity point. I haven't been able to find a definitve answer or info on this.
P.S. sorry for the somewhat off-topic
Last edited:
Sure, if you are talking about pole splitting and a ULGF of 2 MHz you are absolutely correct.
But I don't need to pole split in a low feedback CFA.
The result is the loop UGF can locate much higher. If you prioritize PM, and not a frequency, you can close the loop at 3 or 4 MHz and still get a stable amp.
I am not going to get into a discussion about the benefits or not of feedback. We all know what they are, suffice to say that I cannot hear the difference between 0.1 % distortion and 1 ppm and neither can anyone else - and especially so if we are talking about speaker distortion - as for vinyl, let's not even go there . . .
Maybe when you're talking about purely harmonic distortion, it's still a somewhat 'musical' distortion (apart from the higher odds). X-over is a different beast and I am convinced that this type of distortion is definitely noticable. You won't notice an immediate difference, but if you know what to listen for it becomes distinguishable after a few minutes of listening. It sounds 'annoying' without really being able to pinpoint it in technical terms. Exaggerating x-over (pure class B) makes it instantly detectable by listening to the right parts in voice/music. Again, it does not sound malformed to your ears, but when you know what a proper original sounds like, you'll pick it up. At least, I do/can and I verified this for myself. Then you have speaker loads that turn your 10ppm amp into a 1% distortion generator and that to me is a receipt for claims that amps definitely can sound different from each other over prolonged listening periods due to their distortion profiles under real loads. But alas, that's drifting away from the orginal discussion...that I cannot hear the difference between 0.1 % distortion and 1 ppm and neither can anyone else - and especially so if we are talking about speaker distortion - as for vinyl, let's not even go there . . .
(suffice to say one requires quality speakers to pick up these finer subtleties, like high-end dome tweeters or horns)
Last edited:
Well, I would put thinds in the following order - ok, arguable in some points, but roughly goes like this:
1) Power supply. Very important for sound quality. After all, the amp circuit just controls the power, coming from the PSU to the speakers. But first of all - we need a steady, solid power to start with.
2) The way the output stage is arranged, including pure "implementation" matters - wiring, grounding, Zobel/Thiele networks, PCB traces, etc.
3) Overall circuit stability. All the HF garbage, ringing, short-term oscillations influence the sonic qualities pretty heavily if not eliminated / addressed properly. Square wave responce - possible overshoot is not good.
4) Somewhere here goes cross-over distortion (if any), transients handling (causing inter-modulation), other rather noticeable effects and artifacts. Slew rate importance is also somewhere here.
5) Finally, overall linearity - THD, harmonics profile, closed loop frequency/phase responce.
Well many of these things are inter-related, but still... this is my rough view.
1) Power supply. Very important for sound quality. After all, the amp circuit just controls the power, coming from the PSU to the speakers. But first of all - we need a steady, solid power to start with.
2) The way the output stage is arranged, including pure "implementation" matters - wiring, grounding, Zobel/Thiele networks, PCB traces, etc.
3) Overall circuit stability. All the HF garbage, ringing, short-term oscillations influence the sonic qualities pretty heavily if not eliminated / addressed properly. Square wave responce - possible overshoot is not good.
4) Somewhere here goes cross-over distortion (if any), transients handling (causing inter-modulation), other rather noticeable effects and artifacts. Slew rate importance is also somewhere here.
5) Finally, overall linearity - THD, harmonics profile, closed loop frequency/phase responce.
Well many of these things are inter-related, but still... this is my rough view.
I'd agree with that order roughly, the PSU being the most important (as well as on-board decoupling of outputs). What's worth a low THD if the outputs can't deliver what the amp wants them to deliver (and subsequently 'oversteers' the output, and then immediately trying to compensate for that again). I'd lump 4 and 5 together, all the effort to lower THD also lowers IM etc as jcx said.
Last edited:
Well, those caps just move both curves (gain and phase) to the left - roughly 1MHz, so I don't see too much influence on overall PM/GM. However, they prevent the CFP pairs from potentially oscillating locally - as you just mentioned...
PCB!
OK, here comes the layout...
8.5 x 3 inch, double-sided. Not really simple any more
However - still pretty cool
Placing the order tomorrow... takes about a week to produce.
And then we'll see...
P.S. Big ones are actually under the board - 3D renderer just can't show them the right way Anyway, gives an impression.
OK, here comes the layout...
8.5 x 3 inch, double-sided. Not really simple any more
However - still pretty cool
Placing the order tomorrow... takes about a week to produce.
And then we'll see...
P.S. Big ones are actually under the board - 3D renderer just can't show them the right way
Anyway, gives an impression.Attachments
Last edited:
Vzaichenko that is a cool circuit but it's not simple
Compared to my upcoming "transimpedance stage"(TIS) this is "bombed out simple"
.OS
You should have the same ABSOLUTE perfect results as member Thimios
had with a 1 stage CFA.
Using the "super VAS" , perfect clip ... impressive slew.
This is also a SOTA CFA design.
OS
Hope so
For me it will be very interesting to audition the double-CFP, especially with such a fast front-end...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.