Member
Joined 2009
Paid Member
I'm not sure the CFA is anything special, it became a hyped-up topic of chat on this forum the past year but out in the commercial world I'm not sure it has very much attention compared with, for example, advances in Class D and digital sources.
The thermal balance of the input pair is not the issue here. The thermal imbalance is in the push-pull VAS stage, and this is reflected as Ic imbalance in the input pair. (See APAD6 p205 & on) So far as I can see at present, only a current-mirror can give the close matching of input Ic's that is required.
Thermal coupling of the input pair is a valuable way of reducing DC drift due to air currents, but it can't hold the Ic's equal.
That's second harmonic rising steeply at 12 dB/octave, which is not a good look on a graph. It would be due to neglecting a mechanism which I think I have now thoroughly explained. Not the way forward, surely?
Member
Joined 2009
Paid Member
I must agree that the push-pull VAS does have a thermal issue - but my worry has been the fact that the VAS current is poorly controlled and when you use a simple Vbe multiplier the poorly controlled VAS current results in poorly controlled bias - which affects distortion in a Class AB output. I resorted to heavy degeneration of the VAS devices to reduce their temperature sensitivity and used a more complex (Hagerman) Vbe multiplier which is relatively insensitive to VAS current. By doing this the amplifier was thermally stable but the heavy degeneration robs OLG and decreases the effectiveness of the nfb loop in reducing OPS distortion.
I have no plans to use a push-pull VAS approach again because of the ill-controlled VAS current problem. The single VAS topology, in most cases, controls the VAS current nicely via the global negative feedback loop.
I have no plans to use a push-pull VAS approach again because of the ill-controlled VAS current problem. The single VAS topology, in most cases, controls the VAS current nicely via the global negative feedback loop.
Last edited:
I have a balanced design (the eAmp) which you can read about on my website (link below) that uses a beta enhanced VAS stage that certainly reduces LTP current differences to very low levels. The small signal transistors hFE is typically 500 and the high voltage VAS transistor 100. The LTP current is high at 5 mA per side.
Cool VAS is an option, but there's also the possibility of using current mirrors to control the current, if the current is CCS set in the IP- stage and then mirrored into the VAS, it can be controlled via a negative temp-co CCS. this is the way I have advocated it done, and how I do in my different versions of CFA's
Cordell and Groner (in his "comments", Figure 59, Page 48) propose a push-pull VAS à la Hitachi with a common mode loop to balance the currents in the input stage. This common mode loop seems to be a very old technique used in some integrated op-amps.
OK, you got me there. That was written some time ago, and it is only relatively recently that I realised that using "the Lin configuration" every time power amplifiers were discussed is inaccurate.
The actual Lin configuration (APAD6 p700) is 2-stage. Shunt feedback is used so a single transistor can operate as both feedback-subtractor and VAS. The output stage is unity-gain quasi-complementary, and as I understand it. that was what was new about it. The Blameless amplifier is 3-stage, which allows it to use series feedback; this is much superior, not least because the amplifier doesn't go mad if you disconnect the input.
I'll sort out that web page when I get a moment. Thanks for drawing it to my attention.
Last edited:
Member
Joined 2009
Paid Member
The only symmetric design that I currently like is unusual - the OPS is the VAS. It's a push-pull VAS but the VAS is the output device. It's a super-clean sounding amplifier.
I'd call it a 2 stage design even though it uses an EF buffer between IPS and OPS.
http://www.diyaudio.com/forums/solid-state/239418-tgm7-amplifier-based-greg-ball-ska.html
I'd call it a 2 stage design even though it uses an EF buffer between IPS and OPS.
http://www.diyaudio.com/forums/solid-state/239418-tgm7-amplifier-based-greg-ball-ska.html
Bigun,
There are ways to control vas current with the pp vas. Using a small signal diode or diode connected bjt, it negates the temperature sensitivity if it is placed close to the vas bjt. I use this in my cfa, though it is not the vssa design, and have found once it's powered up it stabilizes quickly and maintains solid bias current. The only downfall is the expense of a little open loop gain.
Colin
There are ways to control vas current with the pp vas. Using a small signal diode or diode connected bjt, it negates the temperature sensitivity if it is placed close to the vas bjt. I use this in my cfa, though it is not the vssa design, and have found once it's powered up it stabilizes quickly and maintains solid bias current. The only downfall is the expense of a little open loop gain.
Colin
That's interesting, because looking at the schematics, a push-pull VAS is about the only thing the AU217 and AU317 have in common. (apart from the EF output stage) The AU217 is a 3-stage design with PNP BJT input devices. The AU317 is a 4-stage design with JFET input devices. The compensation schemes look somewhat different, but it's hard to tell without doing a full analysis.
So I am being asked to believe that the superior qualities of a push-pull VAS shine through despite these differences? I don't think I can do that.
Even 100R may be too high for the best results; all I can really say is that the degradation with 50R appears to be negligible.
It's an interesting point- perhaps driving an amplifier from 0R would show differences. However that would mean using an opamp buffer, and the opamp would need to be carefully chosen for very low noise and distortion. An LM4562 might do it.
More generously-funded research needed!
No it's not easy. You are trying to keep identical the junction temperatures (not the case temperatures) in two devices with different dissipations.
Why deliberately get yourself into a position where you have to attempt that?
I have the 5th ed so in no hurry to buy the 6th
but I can't tell form recent comments whether you are talking about complementary symmetry diff driving push-pull VAS or some other pairing, or truly trying to abstract the push-pull VAS
for instance Cordell's MOSFET Amplifier uses a "balanced" current mirror diff pair driving a "folded diff pair" VAS
CordellAudio.com - A MOSFET Power Amplifier with Error Correction
does Bob's "balanced current mirror" on the input diff pair, cascoded 2nd diff pair then overcome your matching or power/Vcb balance objections?
but I can't tell form recent comments whether you are talking about complementary symmetry diff driving push-pull VAS or some other pairing, or truly trying to abstract the push-pull VAS
for instance Cordell's MOSFET Amplifier uses a "balanced" current mirror diff pair driving a "folded diff pair" VAS
CordellAudio.com - A MOSFET Power Amplifier with Error Correction
does Bob's "balanced current mirror" on the input diff pair, cascoded 2nd diff pair then overcome your matching or power/Vcb balance objections?
Last edited:
I seem to recall that the transition from bandgap to avalanche is a voltage dependent thing. Something around 1.8 volts?
I do not recall it exactly, it was in '83 that I larned that..
jn
Well, far be it from me to frogmarch you down the bookshop, but the fact is that my investigations into the various forms of push-pull VAS are only in the 6th edition. They take up 28 closely-reasoned pages, and there's a limit to how effectively they can be summarised in a discussion like this.
I also have schematics for both amps and know the differences between the two. Amps do sound different but my point was that push-pull VAS did not spoil anything subjectively in any of them. Both sound very good.
- Status
- Not open for further replies.