Dear colleagues,
In the style of the 20WPA thread, I am posting here an interpretation of the MF-A1 from the large neighboring forum for relentless discussion among experts.
Open fire!
The original should be well known, so I will refrain from unnecessary repetition of a well-known circuit.
In the hope of a lasting learning effect, because self-deception helps no one,
kindly
HBt.
In the style of the 20WPA thread, I am posting here an interpretation of the MF-A1 from the large neighboring forum for relentless discussion among experts.
Open fire!
The original should be well known, so I will refrain from unnecessary repetition of a well-known circuit.
In the hope of a lasting learning effect, because self-deception helps no one,
kindly
HBt.
Modifications, changes in comparison with the original:
The
If I scale the posted distortion spectra and compare them with my current measurements of the current original, the modification and incarnation of "the Babelfish A1 Zen Master" is unfortunately not a bit better.
Is it perhaps more reliable?
HBt.
The
- "level shifter and impedance converter" has been replaced by a HEX-FET
- two fixed-voltage regulators replaced two Z-diodes with their series resistors
- sensor resistance (let's just call it that for a moment) was reduced from 0.47Ohm (max) to 0.11Ohm
- negative feedback was significantly increased
- "slow and ancient power transistors" were replaced with modern and very fast sprinters !
If I scale the posted distortion spectra and compare them with my current measurements of the current original, the modification and incarnation of "the Babelfish A1 Zen Master" is unfortunately not a bit better.
Is it perhaps more reliable?
HBt.
One must never forget what the core of the A1 actually is, including two single-ended amplifiers working against each other ... dynamically they work in parallel, i.e. in push-pull.
If we are really honest, we have to admit that the circuit is "extremely strange". Basically one big design flaw! In terms of DC, you could even call it a ******-up voltage regulator.
Nevertheless, there is a splendid ingenuity in the whole construction.
This thing is crazy, isn't it?
Their TCR will be far lower than any of the active devices. It would be far more important to group (or thermally couple) the active devices.
I designed and built a JLH-like HA using two dual-transistor packages for thermal tracking but then stupidly wired them up wrong. Touching one of them with a finger caused the output offset to change by quite a bit -- 100's of mV. I initially thought the circuit was oscillating but the offset only slowly recovered so I suspected it really was a thermal-match problem. Using a q-tip that had freeze-spray on it confirmed it was all thermal.
Distorsion of this schematic should be huge, beside MFA1 has never been a good design, and never will be,
it s well below other such symmetrical diferentials that were published roughly 40 years ago.
it s well below other such symmetrical diferentials that were published roughly 40 years ago.
Wahab and EdGr are absolutely right!
Let's look at the core and note that a meaningful output stage (EF2 or EF3) is completely missing. The present two-stage VAS will bring more problems (as a power amplifier) than solutions.
This idea can only work as a class A representative with really large quiescent currents. At the moment, a quiescent current of more than 2.25Adc would have to flow permanently in order to be able to supply the propagated 20Wrms to static 8Ohm loads.
Unfortunately, this crazy idea does not work without an active counterpart to replace the passive 30 Ohm resistor.
Without the impedance-converting level shifter between the IP and the VAS-OP stage, the principle outlined above with the specified resistance values will not work anyway.
To be continued ...
Let's look at the core and note that a meaningful output stage (EF2 or EF3) is completely missing. The present two-stage VAS will bring more problems (as a power amplifier) than solutions.
This idea can only work as a class A representative with really large quiescent currents. At the moment, a quiescent current of more than 2.25Adc would have to flow permanently in order to be able to supply the propagated 20Wrms to static 8Ohm loads.
Unfortunately, this crazy idea does not work without an active counterpart to replace the passive 30 Ohm resistor.
Without the impedance-converting level shifter between the IP and the VAS-OP stage, the principle outlined above with the specified resistance values will not work anyway.
To be continued ...
A deliberately created DC offset!
If we were now to short-circuit the quasi-voltage regulator with the static internal resistance of 0.47Ohm and an original /source voltage of 376mV, i.e. connect the output to the zero-volt potential, we would have a largely constant current flow of 0.8Adc.
The quiescent current of the original MF-A1. Unfortunately, we have to create a virtual short circuit (with our active opponent) at the output-node.
R2 without Ib (i.e. equal to zero) is exactly 1.5MOhm. Unfortunately, we never know Ib and Ib is not constant either ...
If we were now to short-circuit the quasi-voltage regulator with the static internal resistance of 0.47Ohm and an original /source voltage of 376mV, i.e. connect the output to the zero-volt potential, we would have a largely constant current flow of 0.8Adc.
The quiescent current of the original MF-A1. Unfortunately, we have to create a virtual short circuit (with our active opponent) at the output-node.
R2 without Ib (i.e. equal to zero) is exactly 1.5MOhm. Unfortunately, we never know Ib and Ib is not constant either ...
Last edited:
If we are not very careful, the output stage of the (two-stage, total) VAS will conduct beyond all intended limits.
Danger!
Our DC offset control circuit (R1 & R2 - loop) must function immediately under all circumstances.
Zen-Masters 1. Mod:
Nothing will happen here on its own, nothing at first! Unfortunately, this modification does not change the actual crux of the entire circuit, i.e. the two-stage VAS concept remains untouched - with all its disadvantages.
Break ...
☕
Nothing will happen here on its own, nothing at first! Unfortunately, this modification does not change the actual crux of the entire circuit, i.e. the two-stage VAS concept remains untouched - with all its disadvantages.
Break ...
☕
We should close and open the ac nfb loop.
OLG is approx +47dB, in closed loop it is +24dB as voltage-gain. Can the counterpart, the second SE amplifier so to speak, improve the situation?
OLG is approx +47dB, in closed loop it is +24dB as voltage-gain. Can the counterpart, the second SE amplifier so to speak, improve the situation?
Some loop gain plots would be useful to determine stability margins.
It is a very idiosyncratic circuit for sure. But, if you've ever listened to one, and especially with classical music, you’ll know why people raved about the sound. Small class A amps have a magic of their own when paired with a sympathetic load.
It is a very idiosyncratic circuit for sure. But, if you've ever listened to one, and especially with classical music, you’ll know why people raved about the sound. Small class A amps have a magic of their own when paired with a sympathetic load.
Let's compare the scenario with the modification proposed by the "Zen Master" !
CLG approx +19.2dB and OLG approx +43dB
43 - 19.2 = 23.8
vs.
47 - 24 = 23 !
There is no significant improvement whatsoever - only the impedance converter (and level shifter) is fundamentally different. It is single-stage (as a source follower) and not self-conducting.
However, the longitudinal element is now (still to be replaced by) a 30MHz ft BJT instead of the usual <3MHz boys - that scores a few points.
Now we should take a closer look at h11e, h12e, h21e and h22e - but one thing is already certain, the use of a HEX-FET promises no advantages in terms of THD response and distortion spectrum. What we need is a much larger OLG!
HBt.
CLG approx +19.2dB and OLG approx +43dB
43 - 19.2 = 23.8
vs.
47 - 24 = 23 !
There is no significant improvement whatsoever - only the impedance converter (and level shifter) is fundamentally different. It is single-stage (as a source follower) and not self-conducting.
However, the longitudinal element is now (still to be replaced by) a 30MHz ft BJT instead of the usual <3MHz boys - that scores a few points.
Now we should take a closer look at h11e, h12e, h21e and h22e - but one thing is already certain, the use of a HEX-FET promises no advantages in terms of THD response and distortion spectrum. What we need is a much larger OLG!
HBt.
Without the 101µF ac-short cap ---> OLG approx +56dB and fh' is about 22kHz.
But "Kudos for Tim de Paravincini" ...
But "Kudos for Tim de Paravincini" ...
absolute !
This helps and at the same time represents an improvement -> if we can now still steer reasonably /swing between the rails ?! Fortunately, what we still need now is the mirror image, the opponent!
The ludicrous one, thinks now:
"I'll go straight for the IGBT pair GT20-D101/D201"! ... cumbb's ingenious water cooler is soon to be used as the cooler.
;-)
"I'll go straight for the IGBT pair GT20-D101/D201"! ... cumbb's ingenious water cooler is soon to be used as the cooler.
;-)