I can find the reference about CFAs "don't suffer from TIM" in the audio context. TIM distortions are generated when the input signal slope approaches the amplifier slew rate. It is true that CFAs may have much larger slew rates compared to VFAs, however to state that CFAs are generating less TIM for audio purposes, that's a stretch by any metric. VFAs can easily be build with slew rates that avoid any measurable audio TIM distortions.
The rest is the standard "open loop bandwidth" nonsense.
Intermodulation Distorsion with real musical signals and Harmonic Distorsion Components are manifestations of how is a signal is deformed by a circuit. As they are strongly are related to Total Harmonic Distorsion which is easily measured, I see obtaining the best performances for THD is wholly justified.
I am still looking for facts that prove that other performances like maximal slew rate or large open loop bandwidth are tokens to a better reproduction.
That said, some CFAs are certainly excellent.
But their simplicity is only apparent. I know Hiraga's class A from day 1; because of its huge power supply, I never could consider it as a simple design.
Designing circuits with intrinsic high PSRR is smarter and more economical than to rely on bulky power supplies and decoupling.
Originally Posted by Bonsai
The trade off between the conventional CFA front end and the VSSA is that the conventional CFA front end can be made all DC coupled and with a single set of resistors setting the gain - some people would see this as a distinct advantage.
At last! A definite technical win for Diamond Input over simple VSSA type 🙂
Also probably more linear when fed by a high impedance source.
Can you provide numbers ?
Intermodulation Distorsion with real musical signals and Harmonic Distorsion Components are manifestations of how is a signal is deformed by a circuit. As they are strongly are related to Total Harmonic Distorsion which is easily measured, I see obtaining the best performances for THD is wholly justified.
Not when the other parts of the signal chain are producing orders of magnitude more distortion. Again, in the best exemplars of both types, the differences are in single digit ppm.
I am still looking for facts that prove that other performances like maximal slew rate or large open loop bandwidth are tokens to a better reproduction.
Ditto your claim that a 2 ppm VFA is better than a 4 or 5 ppm CFA
That said, some CFAs are certainly excellent.
But their simplicity is only apparent. I know Hiraga's class A from day 1; because of its huge power supply, I never could consider it as a simple design.
Designing circuits with intrinsic high PSRR is smarter and more economical than to rely on bulky power supplies and decoupling.
I see VFA designs with cap multipliers now delivering 80 or 90 dB PSRR. You can do the same for CFA, or even more simply use AFEC (and get DC servo as well). CFA PSU's don't have to be bulky - that may have been specific to Hiraga's designs.
Not when the other parts of the signal chain are producing orders of magnitude more distortion. Again, in the best exemplars of both types, the differences are in single digit ppm.
I am still looking for facts that prove that other performances like maximal slew rate or large open loop bandwidth are tokens to a better reproduction.
Ditto your claim that a 2 ppm VFA is better than a 4 or 5 ppm CFA
That said, some CFAs are certainly excellent.
But their simplicity is only apparent. I know Hiraga's class A from day 1; because of its huge power supply, I never could consider it as a simple design.
Designing circuits with intrinsic high PSRR is smarter and more economical than to rely on bulky power supplies and decoupling.
I see VFA designs with cap multipliers now delivering 80 or 90 dB PSRR. You can do the same for CFA, or even more simply use AFEC (and get DC servo as well). CFA PSU's don't have to be bulky - that may have been specific to Hiraga's designs.
*musings*
I think we have to make CFAs from now on. Then we can bicker about which output devices sound better since their distortions are more pronounced 😀
Edit: On a more serious note: One can never have low enough THD because compensation for stability will always be at cost of THD.
I think we have to make CFAs from now on. Then we can bicker about which output devices sound better since their distortions are more pronounced 😀
Edit: On a more serious note: One can never have low enough THD because compensation for stability will always be at cost of THD.
Last edited:
Why cant you simply inject feedback as current into an LTP in order to compare the two. It should be quite straight forward to make an AC injection below the current mirror.
Then it should be possible to compare apples to apples.
I have modified a version of VSSA or SSA to AC feedback. Result is better DC stability and to my surprise a better Slam and focus in the base. With AC feedback its possible to lower the Feedback ratios to make the current feedback more current and less voltage, as the signal is modulated over lower gnd resistor values (a factor 10) I use 10 and 300 ohm to get the desired gain. (2X22mF/10v for each branch)
Then it should be possible to compare apples to apples.
I have modified a version of VSSA or SSA to AC feedback. Result is better DC stability and to my surprise a better Slam and focus in the base. With AC feedback its possible to lower the Feedback ratios to make the current feedback more current and less voltage, as the signal is modulated over lower gnd resistor values (a factor 10) I use 10 and 300 ohm to get the desired gain. (2X22mF/10v for each branch)
Hi Scott.... will you please put up a schematic for everyone so we know exactly what you mean.
-Thx-RM
Last edited:
Hi kgrlee,
I think by minimizing device count to avoid complexity is a bit of a cheat. You see, when you change the CFA into an LTP version, using the blameless topology, you basically ditch an important bit for the VFA: The symmetrical push-pull VAS action inherent to a symmetric CFA. Either a complementary LTP or a single-ended input push-pull VAS would level that playingfield. And such topology would allow for various compensation schemes easily.
Either way, there's no reason to think that amps like the VSSA wouldn't sound good. I'm positive they do 🙂
OK, but it will have to wait till tomorrow. Frankly it should be obvious, Input voltage applied to a transconductance, current out delivered to a gain node, the cascode just redirects the current so you can make a circuit with a +- power supply that has an output symetrical about ground to eliminate capacitive coupling.
This is not like common base stages in the RF world where you match terminate into the emitter and have gain to the collector, this is simply current into the emitter coming out the collector with the base current lost.
Last edited:
Hi Scott --- that will be great. There can be be no misunderstanding your description with a drawing. And,in the future, it can be refered to.
Thx-Richard
Thx-Richard
Good old Yahoo mail is refiusing to load diyaudio emails, and suggests visiting a help site to report the problem --- except the highlighted link provided doesn't do anything. Oh well I sure it is not just me (and since you asked, it is browser independent).
However: I shall present soon, probably starting another thread, a CFA gm core with the advantages of very small thermal distortion with noninverting signal swing, and reduced inverting-terminal current noise. Devotees of minimal parts designs will as usual not be pleased. To not confound things with other effects the initial presentation will use ideal current mirrors, current sources, and an ideal output buffer. Real ones will certainly degrade the predicted performance with the ideal ones, but how much remains to be determined. The -3dB bandwidth with the ideal mirrors and output buffer is about 50MHz with a closed-loop gain of 26dB, using garden-variety bipolars whose f sub t at 5mA and 10V is only about 300MHz. Again, the real mirrors and output buffer will likely spoil this. Still in all, it seems promising.
EDIT: mail is loading again
However: I shall present soon, probably starting another thread, a CFA gm core with the advantages of very small thermal distortion with noninverting signal swing, and reduced inverting-terminal current noise. Devotees of minimal parts designs will as usual not be pleased. To not confound things with other effects the initial presentation will use ideal current mirrors, current sources, and an ideal output buffer. Real ones will certainly degrade the predicted performance with the ideal ones, but how much remains to be determined. The -3dB bandwidth with the ideal mirrors and output buffer is about 50MHz with a closed-loop gain of 26dB, using garden-variety bipolars whose f sub t at 5mA and 10V is only about 300MHz. Again, the real mirrors and output buffer will likely spoil this. Still in all, it seems promising.
EDIT: mail is loading again
Last edited:
When I make a recording, I note the settings of the Soundfield Control Box and the recorder. This gives me an spl calibration to within about 1dB.
I'm one of the crazys who thinks recording he's made with microphones he's designed are the most important to assess reproduction 🙂 I like to play these back at the correct level and often check this with a calibrated B&K mike. But one of the party tricks of good recording engineers is that they can set playback levels to within 1dB of the original for a good recording using simple mike arrangements.
Peter Walker said that for a natural recording, there is only ONE correct playback level. If its played too soft, it doesn't sound further away. It sounds too small.
I concede I'm not typical and I like to sit up front for orchestral concerts.
There are famous Soundfield Mike recordings which overload the mike, Sony PCM-F1 recorder, 1000W/channel amplifier and 90dB/W speakers. But with appropriate speakers & amps, this can still be played at the correct level with no audible distortion.
Google 'The Garage Door'.
_______________________
But to put all this semi-liquid BS into perspective ..
For a 50W 8R, I'd pay really close attention to overload & recovery behaviour. This will be important to at least 28% of this august forum ... and at least 50% of people working at Wharfedale in da 80's.
I'd like 200W 8R for my own use at home. Won't quite play 'The Garage Door' properly but OK for music 🙂
For conducting Double Blind Listening Test on speakers, I'd want 1000W 4R and impeccable performance in all other respects. The unit I favour is now Unobtainium. Any suggestions?
Last edited:
You can look at it that way.
The way I see it that as the CFP input stage is more linear the error signal needed for correction is smaller. One can say that higher loop gain is desireable to linearise the outputstage especially since in power amps this stage is in class AB. Keep in mind opamps are class A.
This is the reason Texas says that CFAs have lower TIM than VFA. So each have its pros and cons.
The linearity cant be worse, with VSSA its not possible to have better THD figures compared to a diamond buffered one. You are also only looking at sims without taking into account temp effects, the VSSA bias stability is all over the place. Look at that thread and see how many builders mention this, thats the main reason lateral mosfets are used. The vas current varies considerable. Not desirable at all. There is a reason opamp manufacturers dont use that topology. It also has lower input impedance for the same Ic. Alas the current on demand feature is also crippled. I suspect this is the reason the diamond has lower THD. Further if we get to the stage in this thread where advanced CFA design is shown youll notice that the most notable ones come from improvements with the diamond buffer and not possible with VSSA structure.
Youre going to have a good surprise when you measure that amp for real. I doubt youll even manage 1ppm. Also its stability is marginal at best, not in the interest of many that would like to keep their speakers healthy over a long period. When I have time to compile a LT simulation I will show you sims where the vas idea I showed is implemented in a CFA and how it will eclipse your THD figures despite having lower OLG. Also stability and quality coefficient is near optimal.
Du.uuh! Give yus VFA fanatics an inch an u tak a mile. 😱
We concede the input Current Mirror ... and now you want a completely symmetrical VFA too. Next you'll want all components hand carved from solid Unobtainium by virgins just for VFAs 🙂
Wanna propose suggestions for a 'fair' playing field?
But there is no need for such complexity to get supa dupa VFAs. Here's my #499 circuit performance with C1/3 increased to 100p to make wahab happy. You'll note from the Loop Gain that further optimisation is possible but it's still 1ppm 20kHz 50W 8R.
But Magic, please post your ' single-ended input push-pull VAS'. That sounds suspiciously like a CFA to me. And I want to know how that 'would allow for various compensation schemes easily'.
I'm seriously interested as that might help me do better CFAs.
Again, this isn't about VFA vs CFA but should be about better CFAs
____________________
I'm saddened that really competent people like Waly are still sniping at trivial (mis?)quotes bla bla. How is any of this of use or even interest to anyone?
How about using some of those brains the size of planets to help us do better CFAs (or if you prefer 'IPS topologies which aren't LTP but may be loosely called CFA') ?
Bonsai, pleased don't be discouraged. You have posted some of the most useful things on this thread and I'm sure you have more to contribute.
____________________
Not displeased ... but grovelling at your feet in awe oh Guru Brad 🙂bcarso said:
Our thanks for strewing your pearls of wisdom among us unwashed masses. We shall strive to exercise our single working brain cell to understand your Word 😱
I know you posted the question early on but ... is thermal distortion really significant for power amp IPS?
