OK what do we have here:
- Increasing amplitude ... increases frequency and exhausts bandwidth.
- Z-diodes are too slow to protect gates (>15V in 1ns would need >15000V/µs VAS).
- External Diodes degrade sound, internals do not.
- Miller effect in followers.
- Mouser + Digi-Key sell fakes.
- The real performance ... cannot be measured.
- Everybody else misses the point or is not scientific.
Sorry but I think we have a troll. He is pulling your chains.
- Increasing amplitude ... increases frequency and exhausts bandwidth.
- Z-diodes are too slow to protect gates (>15V in 1ns would need >15000V/µs VAS).
- External Diodes degrade sound, internals do not.
- Miller effect in followers.
- Mouser + Digi-Key sell fakes.
- The real performance ... cannot be measured.
- Everybody else misses the point or is not scientific.
Sorry but I think we have a troll. He is pulling your chains.
Oh, man, you just tore my template into unuseful shreds...
"I have picked and used your old trannies" now show me very different meaning...
:facepalm:
Arrrgh...
There are different ways to discuss, more or less offensive or superficially argued.
Each designer take his choice and trying to cope with it.
Follower-style is much more easy way to obtain useful properties while common source OPS also can be very potent in skillful hands.
But keep in mind, you know first hand, skillful hands probably vote for follower-style.
Last edited:
@ctrlx, are you running the output stage Class AB? If so, your schematic No. 2 calls for trouble. It is a complementary feedback pair (or Sziklai) with unity gain and this usually is prone to short HF oscillation bursts when switching. I'd use CFP stages only when running class A all the time. Also, even your drivers in that configuration could leave Class A, something which isn't really a good thing.
When using BJT / MOS-FET emitter/source follower, two stages after VAS are OK.
When using only BJT's, 3 stages are better, otherwise you'll see a portion of the VAS high impedance at the output and have to correct for it by feedback.
Likewise, running output MOS-FET's off the high impedance VAS node like in the Maplin Amp doesn't seem the best thing, however interestingly enough some have found this to be better sounding than with an additional EF stage in those Maplin amps.
For reference, read also the Leach articles, and older posts from Hugh (AKSA) re CFP.
When using BJT / MOS-FET emitter/source follower, two stages after VAS are OK.
When using only BJT's, 3 stages are better, otherwise you'll see a portion of the VAS high impedance at the output and have to correct for it by feedback.
Likewise, running output MOS-FET's off the high impedance VAS node like in the Maplin Amp doesn't seem the best thing, however interestingly enough some have found this to be better sounding than with an additional EF stage in those Maplin amps.
For reference, read also the Leach articles, and older posts from Hugh (AKSA) re CFP.
Last edited:
i will be running the FETs at around 120mA.
i have not given up on cct2 yet, the osc i reported may be due to a construction error or bad part....
i have not given up on cct2 yet, the osc i reported may be due to a construction error or bad part....
Not really.
A transient denotes a rapid change. High currents obstruct fast changes and fast charging of capacitive loads.
A higher resistance implies an increase in mass, a decrease in velocity, amplitude and impedance, that is, damping. This is how electronic filters work. Damping is used to improve stability, reduce resonances and high speed transients. The base resistance accomplishes high frequency damping.
Consequently, a high current gain gives a narrow bandwidth.
Capacitance, charge, current, mass and weight are equivalent.
Maybe Holmer, in his extended knowledge acquired from credible sources, could come up with a better explanation.
Not.
Really.
Really not.
No need for credible sources, let's explain on the fingers.
Decade before fT mostly any follower acts as an impedance converter. Having low resistance at the input (being driven hard) it provides +6 dB rising slope at the output, so seen from the output it acts like inductance.
Having high impedance at the input (driven from current source like VAS output) paralleled with 3x to 5x Cob capacitance such a follower will provide you with a nice relatively flat response, so it converts capacitive -6 dB falling slope one order up and provides flat output impedance.
Now let's see on EF3 based on the relatively fast (comparatively fast) transistors.
Usual nominals of the VAS-to-OPS point are ~100-470 kOhm and 10-47 pF.
First follower stage will convert capacitive input impedance to resistive. Second will convert resistive impedace seen at the output of the first follower to the inductive. Third follower will convert inductive output resistance of the second follower to the inductive^2 residual impedance, which in case anyway provide you with an oscillation being loaded even at connecting wires capacitance.
Namely this is the main reason to decouple OPS from load with an inductor in most EF3-based amps.
There are two main possible ways to workaround.
First - switch off first follower stage with RC-filter like 100-220 Ohm + 100-220 pF between base-collector junction of the second stage.
Second - switch off last follower stage well before (by frequency) than instability region could occur. Namely - use very low fT devices with huge BE-capacitance and allow second follower to act with load directly at MHz's
What advantage do you expect from Circuit 2 vs Circuit 1?
Further decrease in distortion due to the output stage forming a local loop in the overall loop?
Sadly - no.
Such an OPS have intrinsically lowish input impedance and due to this for keeping relatively the same loopdepth will demand one additional stage.
Focus your attention on what happens at a time while one output shoulder MOS are going OFF. Keeped anyway ON BJT for a short time straightly interacts with load without being impowered by huge (but finite due to capacitance) MOS current gain. So achieved in 2-stage OPS impedance trasformation becomes are like 1-stage follower like beta ~100-150 times, which in case hugely loads VAS output and provides designer with a very bad high-freq distortion or short oscillation burst.
Moreover.
This provides you with a hugely variable OPS pole frequency position which not being properly accounted could provoke overall amplifier to become unstable!
Again!
Common-source or common-emitter stage is an intrinsic gain stage which have highish output impedance and needs to have huge global or local feedback for properly interact with speaker as load (which in case was designed to be well-damped by lowish impedance for proper work). So such an OPS straightly demands designer to be well-educated in terms of stabilizing control loop of the feedback of the whole amplifier.
Beware of such a stage!
Not being a "control-loop engineer" one could result with a bad sounding amplifier due to lowish achieved feedback deph which in case doesn't allow to compensate noted high-freq anomalies.
Again.
CCT2 demands for relatively high-freq and well-designed feedback loop, proper and hi-q PCB layout, accounting a tons of relatively bad high-freq aspects, BUT could result in a relatively better result if designed properly.
IMHO there are no argues for going common-source OPS now for audio applications. Too much efforts.
Denial of the reality in terms of others' experience is a bliss. Weaseling is not any better either.
Cheers!
P.S. Most of us are mere midgets standing on the shoulders of titans Like Gilbert, Widlar, Williams, Pease and so on.
Last edited:
an update.
cct2 is up and running, but it oscillates without a output zobel and unfortunately when pushed into clipping (with a load) the clipped elements have oscillations.
cct1 did not exhibit these behaviours.
cct2 is up and running, but it oscillates without a output zobel and unfortunately when pushed into clipping (with a load) the clipped elements have oscillations.
cct1 did not exhibit these behaviours.
The common emitter predrivers ought to run at a higher bias current and have a lower value of collector resistor than 1.5 Kohm, anyway, I`d rather use the 2N5551 at a low current than the MJE340.
I would increase the value of the gate stoppers at least to 470 ohm that need to be carbon composition or carbon film resistors to operate in the MHz region and also insert source resistors to reduce distortion.
I would increase the value of the gate stoppers at least to 470 ohm that need to be carbon composition or carbon film resistors to operate in the MHz region and also insert source resistors to reduce distortion.