"NO advantage" to this alien "controversial feedback" ??? Hmmm...
Putting aside the bad PSRR , one could get high end performance with
a 4 device IPS/VAS. Can't do this with a VFA , the "stripped down"
3 device LTP/bootstrapped VAS LIN can only do .05% THD ... if that.
At 5 to 6 devices (LTP/CSS/beta enhancement), the LIN may JUST
match a Peeceebee.
Only at "full blown" level .... (LTP/cascode/CM/2 CCS/beta enhancement),
will a blameless match either the NX or (enhanced)VSSA designs.
So , a "CFA" can reduce device count in the IPS , but require "support"
(regulation).
By reading what has been said about "the sound", the REAL advantage
is the very high slew rate. 80v/us and above (like the Badger), will be
described as " detailed , excellent soundstage".
Then , I still have 2 simulations that require NO compensation 😀 .
Could a CFA be "idiot proof" ?
OS
Putting aside the bad PSRR , one could get high end performance with
a 4 device IPS/VAS. Can't do this with a VFA , the "stripped down"
3 device LTP/bootstrapped VAS LIN can only do .05% THD ... if that.
At 5 to 6 devices (LTP/CSS/beta enhancement), the LIN may JUST
match a Peeceebee.
Only at "full blown" level .... (LTP/cascode/CM/2 CCS/beta enhancement),
will a blameless match either the NX or (enhanced)VSSA designs.
So , a "CFA" can reduce device count in the IPS , but require "support"
(regulation).
By reading what has been said about "the sound", the REAL advantage
is the very high slew rate. 80v/us and above (like the Badger), will be
described as " detailed , excellent soundstage".
Then , I still have 2 simulations that require NO compensation 😀 .
Could a CFA be "idiot proof" ?
OS
Zero, zip, nada, unless you are in the 500V/uS team.
- I would exchange anytime 500V/uS for a better PSRR.
- Complexity is a red herring.
- Anyway, if the .05% is correct, I thought that's all the audiophiles care for? And that the THD race is stupid?
So you are back to at least 6 devices.
Yeah, right, I've seen the same descriptions about the Wavac, 10% THD and a 200,000 squid price tag.
Nope, you are compensating an amp by using the devices parasitics. A big nono in my book, at least because it makes the whole thing more layout dependent. And BTW, you are happing for saving a couple of ceramic caps?
Anyway, you got a new toy, enjoy responsible 😀.
Member
Joined 2009
Paid Member
Agree, but even in a VFA the device parasitics and pcb parasitics are part of the compensation so it's neither good nor bad.
Although Hugh has pointed out before that device parasitics are not always as 'pure' as an external cap.
Right. On a pure theoretical point of view, it is not so good than using a real (linear) capacitance.
Not sure who's Hugh and what he means by "pure" (why am I smelling something fishy here?), but device parasitics are always deeply nonlinear. Which makes the whole amplifier compensation bias dependent. E.g. you may have it stable at zero output, but oscillating as hell with outputs towards the rails.
Depending solely on the semi-con device parasitic capacitance for compensation is probably not a good idea as these parasitic parameters, Cob for example, are often highly non-linear and can cause distortions or instability in dynamic conditions. I would try to use reasonable sized external capacitors and not to let parasitic parameters dominate.
Member
Joined 2009
Paid Member
Hugh = AKSA
I remember he noted that Cob is non-linear, it is a capacitor with a poor dielectric to say the least. So having low Cob and then adding external capacitance makes for a better compensation scheme. But even with VFA, you still have that Cob as part of the equation. A similar thing can be said for adding emitter degeneration resistors, they are better controlled than the internal Re of the device.
Try using some of the modern power MOSFETs, these little buggers will oscillate by themselves.
Last edited:
I know they will.... so it isnt a VFA nor CFA issue.
Also, I use one's which don't do this quite so easliy and dont have problems.
-RM
Last edited:
I'm a wise ol' man ... I would never do that. In fact ,
I will usually over- compensate to keep the magic smoke away. 😀
My point was the fact that a CFA CAN do this.
I alway's cover my butt .... If I don't like the "CFA" result , I have 3 proven
VFA designs to mate to that nice EF3 OPS I posted. I reverse engineered
(and built )the Luxman 120a (Hawksford VAS), but never heard it with it's preferred triple.
I'm on the Hawksford now (layout wise)...
OS
Attachments
Waly, presumably, there is some practical & useful info in these posts that will help us design better amplifiers, whether CFA or VFA.
Meanwhile .. in an attempt to draw the discussion back to useful matters ..
This is your explanation in #3238 of the circuit you posted in #3149 in reply to Edmond's in #3135 claiming ' less THD, better PSR (I think) and less IPS dissipation'
Let's be clear about this. Have you simulated your circuit and also something similar to Edmond's (#3238) and found these advantages?
Yes, although they are all visible by inspection (except a quantitative analysis of PSRR and distortions).
I'm tired to be summoned.
Not sure I understand your point about slew rate. Are you saying it's not an important (for audio) property of CFAs? My point is: anything over 1V/uS for each peak volt of output is a useless marketing driven number.
I said if the .05% is correct. Which IMO is not, it only shows an incompetent design. D. Self got 0.001% distortions with a classic VFA Lin topology (aka "blameless").
Quoting again Cherry:
To reduce distortions there are exactly three ways:
a) bias (e.g. class A)
b) distortion cancellation
c) negative feedback
It is funny to see how CFA people are throwing away the healthy concept of "first linearize then apply feedback" which is well illustrated in the LTP loaded with a current mirror, for the dubious (if at all, due to poor N/P device matching) distortion cancellation in the diamond buffer. Then in turn fighting for the last drop of loop gain to linearize the amp back to the best VFA values, only to realize that the CFA can't provide any extra loop gain compared to a VFA of the same ULGF. Which ends up with pushing the CFA ULGF up to non realistic values for a practical implementation.
But then I understand it's frustrating to accept that linear audio electronics is, for the last 20 years, into a dead end. Acoustics (speakers, rooms, propagation, etc...), I would put some money on that horse.
I think you have already stated there is zero advantage to CFA compared to VFA. Got it.
The only 'real' disadvantage I see of CFA is a pratical one - finding reasonably matched discrete N and P compliments. No problem with new IC processes. For DIY its much easier to find matched same type, esp N type.
I have made very low distortion amps without a large amount of fb or ULGF. Just need matched devices and the topology takes care of itself.
Thx-RNMarsh
Last edited:
This is absolute nonsense at every level - sub ppm distortion performance has been demonstrated in numerous designs on this thread.
I don't recall the subjective crowd booing CFA. Most of the naysayers were on this thread . . .
Depends on what you mean by "matching". Even in modern IC processes, n/p Vbe matching is at least one order of magnitude worse than n/n or p/p.
AFAIK, you made preamps and headphone CFAs, the discussion here is about power amplifiers. By far not the same thing, for low power a CFA DSL driver like the TPA6120 will drive any headphone set with ppm distortions, ultra low noise and a huge bandwidth. Building with discretes anything to match the TPA6120 is very difficult, if possible at all (not to mention the cost factor).
Don't you love when people are running out of technical arguments? No worries, I'm not hoping to spoil your marketing campaign.
About sub ppm CFA amplifiers, you just made my point. Based on the results on this thread, try a blameless with 6MHz ULGF and 4 pairs of output devices running at 50W/8ohm output with +/-56V rails and see what you get, you'll be surprised.
Have fun
