A lead lag can be useful for compensating the Miller loop. Other than that, not much good.
Last edited:
Not on its own, agreed, but apparently pretty important. Get it wrong and oscillation becomes a very close friend. As a bonus of compensating the miller loop you can move the miller take off point to the pre drivers (in a triple OPS) rather than VAS/TIS output.
I can only comment from a stability point of view in real life at present. Don't have any THD measurement equipment yet. Could comment on the sound but words are more or less meaningless.
Last edited:
PSRR is not a problem, just RC(10 ohm, 220uF) power line filtering of the gain stage is good enough, PSSR is over 70 dB up to 500 MHz. Use of capacitance multiplier will enhance that.
I think that use of cap multiplier type of the power regulator(like I used in my VFA TT amp with all needed protection incorporated) is good idea, you need to sacrifice about 5 V on both power polarity and I think it's worth as the power loss is one tenth of the power amp losses.
dadod,
if you use a real world cap, you wont get that kind of performance I am afraid - once you get above a few 100 kHz, trace inductances become a very big issue - I know this from practical experience.
All my sims show about 50dB PSRR with a conventional CFA structure. Adding simple filtering helps at HF, but again, with real world caps, its only good up to a point.
VFA is much better in this regard at LF because the loop gains are higher. But, at HF they are no better. Beyond the closed loop -3 dB bandwidth both topologies are similar, although compensation can make an important difference (see Harry Dymond AES paper for example).
Cap muliplier is a good solution . . . but compared to AFEC its not as elegant. You need 2 transistors, higher supply rails (or have to lose some output swing), filter/decoupling capacitors and associated resistors. My practical experience on the VFA e-Amp tells me you can get 30-40 dB up to a few 100 kHz, which is quite good IMV.
The challenge: is there a really simple way to improve it - say for example by cross coupling the rail noise, or using passive feed forward?
if you use a real world cap, you wont get that kind of performance I am afraid - once you get above a few 100 kHz, trace inductances become a very big issue - I know this from practical experience.
All my sims show about 50dB PSRR with a conventional CFA structure. Adding simple filtering helps at HF, but again, with real world caps, its only good up to a point.
VFA is much better in this regard at LF because the loop gains are higher. But, at HF they are no better. Beyond the closed loop -3 dB bandwidth both topologies are similar, although compensation can make an important difference (see Harry Dymond AES paper for example).
Cap muliplier is a good solution . . . but compared to AFEC its not as elegant. You need 2 transistors, higher supply rails (or have to lose some output swing), filter/decoupling capacitors and associated resistors. My practical experience on the VFA e-Amp tells me you can get 30-40 dB up to a few 100 kHz, which is quite good IMV.
The challenge: is there a really simple way to improve it - say for example by cross coupling the rail noise, or using passive feed forward?
Quick 'executive' summary of what we have acheived so far in this thread:-
1. Demonstrated 2 CFA designs (dadod and bonsai) that acheive low single digit ppm distortion performance at high power at 20 kHz, easily matching VFA in practice (sub 1ppm is a pipedream - lets not waste our time on this nonsense). We settled on ~3MHz as the upper ULGF to demonstrate this, and acknowledge that for a practical output triple amplifier, this figure would need to be 1.5 MHz or lower. This will increase distortion, but it will still be sub 10ppm
2. Shown that the amplifier complexity is about the same as competing VFA topologies; Separately, other CFA designs on the forum also demonstrate this - e.g. LC's VSSA, JLH 10 watter (from 1969)
3. Investigation in to the TIS lead us to conclude that a simple beta enhanced circuit is all that is needed to deliver very low distortion
4. The designs inherantly have high slew rates and bandwidths and this is set by the front end gm - the feedback resistor sets the SR and the loop bandwidth
5. For discrete audio power amplifier applications, we have shown in sims that we can apply Miller, TMC and TPC comp schemes, which helped us acheive the distortion performance mentioned above
6. We now plan to do more investigation on CFA PSRR
1. Demonstrated 2 CFA designs (dadod and bonsai) that acheive low single digit ppm distortion performance at high power at 20 kHz, easily matching VFA in practice (sub 1ppm is a pipedream - lets not waste our time on this nonsense). We settled on ~3MHz as the upper ULGF to demonstrate this, and acknowledge that for a practical output triple amplifier, this figure would need to be 1.5 MHz or lower. This will increase distortion, but it will still be sub 10ppm
2. Shown that the amplifier complexity is about the same as competing VFA topologies; Separately, other CFA designs on the forum also demonstrate this - e.g. LC's VSSA, JLH 10 watter (from 1969)
3. Investigation in to the TIS lead us to conclude that a simple beta enhanced circuit is all that is needed to deliver very low distortion
4. The designs inherantly have high slew rates and bandwidths and this is set by the front end gm - the feedback resistor sets the SR and the loop bandwidth
5. For discrete audio power amplifier applications, we have shown in sims that we can apply Miller, TMC and TPC comp schemes, which helped us acheive the distortion performance mentioned above
6. We now plan to do more investigation on CFA PSRR
I am tempted to build this amp as I am in possession of four pairs Hitachi laterl MOFETs.
Actualy I intended to redesign one of my JLH 80W mofet amp(I've got two) and use existing regulated(fixed output voltage) power supply.
I desided not to use OPS with diamond buffer, not good behaviour with TPC I used here where it includes OPS in it.
BR Damir
Actualy I intended to redesign one of my JLH 80W mofet amp(I've got two) and use existing regulated(fixed output voltage) power supply.
I desided not to use OPS with diamond buffer, not good behaviour with TPC I used here where it includes OPS in it.
BR Damir
Attachments
Interesting. I tried a comp like this in sims but couldn't get it to work with low THD, you obviously have.
Have you tried magicbox's idea of not using gate zobels and just have a small cap directly soldered between drain and gate.
Yes I tried that too and it works, just with RC to the ground I get better PHM(in simulation).
Hard to find, yes, expensive, yes, but you can use Thoshiba laterals 2SK1056 2SJ162 or of other producers as in VSSA thread.
I would had wrote "less or equal complexity", because, every time i compared, in sims as well as real world, i can have better results, apart PSRR, just replacing LTP by a single CFA input stage.
About PSRR, i strongly insist to use separated power supplies (regulated or cap multiplier) for everything else than power devices. Why to target sub 10 ppm of HD if you have hum and rectifier's noises 100X higher ?
Nice.
May-be it should be interesting to build your output stage on a separate board, in order to compare the sound with very simple CFA input stage (and VFA) in real world ? The occasion to get an idea of real HD impact, as well as slewrate witch will be better with simpler design.
Member
Joined 2009
Paid Member
It seems that I missed a lot of pages in this thread since my post (#18) - but from reading a few pages now it doesn't seem that my conclusion has changed. So I'm really interested to see what happens from the listening tests. I think the differences will come from how well each author can practice his/her art. A badly implemented LTP is readily beaten by a well implemented Singleton, and vice versa. If tubes and transistors are different religions, perhaps CFA and VFA are schisms within the transistor camp....
Hi Esperado,
I would like to stress something about VFA and CFA. First I am far from any kind of expert designer either of those types. I am more confident in VFA design as I designed and built some of this amp type and I am very satisfied with the sound specially my TT amp. It is a bit complicated but distortion(in simulation) is very low and quite easy to get there.
To design a CFA amp is harder for me, have to learn new tricks, and what I learned up to now tell me two things: first, input stage with diamond or Baxandall pairs distorts less then LTP but it’s not easy to get high open loop gain, and if you get there it loses some of the advantages over VFA, and second is that input gain stage is more sensitive to type of the OPS.
In my opinion it’s almost impossible to distinct the sound of good design CFA from good design VFA.
BR Damir
I agree on points 1. As you have probably seen, each added tranies brings additional poles (phase turns), pretty hard to compensate. Reason why, at the end, i tend to prefer simplest CFA designs.
On point 2, i tend to disagree. But it is a very subjective position, on my own, although without preconceptions.
So i'm very interested to read your comments, once you'll have build your own CFA amp. I founded there is something different *by nature* in the way CFA present the musical landscape. (of course, space for lot of exceptions). And i'm not the only one.
I wonder if the damages caused by this research of ultra low THD don't have kill the natural benefits of CFA topology.: High slew rates and stability, including no overshoot.
Of course, you will spend some time to tune a low pass filter at the input of your amp, for the best listening results, before to play real music.
I'm very tempted to build your design too, in order to figure out if any advantages (or disadvantages) to have ultra low HD numbers, and if what i feel as the "CFA character" resisted to this complexity... Or not.
In other words, if "it’s almost impossible to distinct the sound of good design CFA from good design VFA." is for true.
Of, if you prefer, if what i like in CFA amps is just their defects...or not ;-)
Last edited:
I used to have a JLH's amp with a regulated JLH power supply which switched off in presence of a DC default at the output.
Mine has the power supply entering in protection mode much too often.
I think regulated power supplies are a very unuseful feature for a power amp, bar at the developing stage.
Is your JLH's amp circuit the one which had a very curious differential VAS stage loaded by two CSS's instead of a current mirror ?