Agree on this one. I towards faster devices. However, BD139/40 doesn’t spec its fT. I only use those when there is no other to-126 parts available.
I would insert resitors of about 1k to 1k5 in the collectors of the T13 and T15 predrivers to protect the VAS trnsistor from ovedrive.
By doing so, the “bjt early affect” will creep in and lower your gain.
Alternatively add a diode on top of the miller cap to prevent overdrive. Below is the example of internal JRC4558 opamp.
Yes, but with low velue of this resistor effect is very small and VAS transitor is protected and not burned from very high current and with lower distortion than using diode.
Not really, I investigated this years ago. The best way is using a saturated constant current source to load the collector. The 2nd best is putting 1n4148 on miller cap. The last one is the resistor method.
The other is explicit current limiting for the VAS transistor (Re current sensing). Which also protects against excessive VAS current if output SOA protection is employed.
I haven't seen the diode version before, thanks for the education.
Like @dadod I would expect the distortion to be worse, but I haven't tested it.
I have tested the resistor and it is simple and works well.
Do you have comparison data?
Best wishes
David
Edited after the time window.
But it doesn't seem to fit your criteria, a traditional "Blameless"/power op-amp circuit would be more simple, less parts means probably more reliable, easier to build for the same reason.
Possibly more (loop) stable, based on your tests your amp is OK but 62° is usually considered a bit suspect.
But comp-sym input has its attraction, even if this is a use case where I wouldn't use it, so I am interested to see your amp.
The issue is the VAS current stability, well known to be a problem for the comp-sym with current mirrors.
I looked at this (after it was recommended be Erno Borbely IIRC) but it is not intuitive to me how it behaves when the CE input has feedback from the Miller capacitor while the cascode connection doesn't.
I think this is the core of the comments by @jxdking. It may be balanced at DC but quite one sided at audio frequencies.
Also the Q5 collector is not tied to a stable potential, which seems to sacrifice performance.
This is probably the price of VAS current stability in your circuit, but it seems a heavy price.
Do you have simulation data on these issues?
I did some sims of a comp-sym input with a common mode control loop for the VAS current that looked really nice, but they are LTSpice, so not much use to you.
Best wishes
David
Many people are attracted to the complementary-symmetrical input because it looks nice on the schematic, and I admit I am attracted too.
But it doesn't seem to fit your criteria, a traditional "Blameless"/power op-amp circuit would be more simple, less parts means probably more reliable, easier to build for the same reason.
Possibly more (loop) stable, based on your tests your amp is OK but 62° is usually considered a bit suspect.
But comp-sym input has its attraction, even if this is a use case where I wouldn't use it, so I am interested to see your amp.
The issue is the VAS current stability, well known to be a problem for the comp-sym with current mirrors.
Q5 also acts as a a follower to drive Q6 as a Common Emitter as a conventional VAS.
I looked at this (after it was recommended be Erno Borbely IIRC) but it is not intuitive to me how it behaves when the CE input has feedback from the Miller capacitor while the cascode connection doesn't.
I think this is the core of the comments by @jxdking. It may be balanced at DC but quite one sided at audio frequencies.
Also the Q5 collector is not tied to a stable potential, which seems to sacrifice performance.
This is probably the price of VAS current stability in your circuit, but it seems a heavy price.
Do you have simulation data on these issues?
I did some sims of a comp-sym input with a common mode control loop for the VAS current that looked really nice, but they are LTSpice, so not much use to you.
Best wishes
David
Thanks, nice to read the thread but it doesn't address my question about the Miller capacitor diode vs the resistor in the VAS collector as alternate methods to limit the VAS overdrive, and which has lower distortion. Do you have any data for this?
Best wishes
David
No one seems to have noticed that this should cause easily detectable overshoot (of about 7% but it is quite sensitive to the exact PM) but your scope pictures show no trace of overshoot.
It seems unlikely that the PM is better than the simulation, unmodeled effects practically always make it worse, and there always are unmodeled effects, like inevitable upper frequency poles and perhaps RHP zeroes
If your simulation is accurate then there must be some limitation in your measurements.
Either the instrumentation doesn't show the overshoot, oscilloscope set-up (too slow sweep or too coarse quantization?) or the square wave source doesn't expose it (too slow rise time?).
Or perhaps the slew rate limitation of the amplifier suppresses the overshoot?
Anyone have ideas?
Best wishes
David
It depends on how he probed the loop gain. If he just simply used Vin_neg/(Vin_pos-Vin_neg), that would see worse margin than that in real world.
As he stated, there is no coil and it is stable with capacitive load. That is pretty stable.
Why always worse? I always use Tian probe to avoid this sort of issue, so I have never had to think about it.
I assumed PMA would use an accurate probe too but you may be correct, perhaps he didn't.
And yes, it looks pretty stable, but the failure to show expected overshoot makes me a little concerned that there may be other indicators that don't show up.
The Uni-T scope is pretty slow and low resolution for instance.
This should be adequate to show the main loop stability but could miss fast parasitic oscillation or low level oscillation.
Best wishes
David
Last edited:
I am just guessing here. The ltp has no degeneration. Thus the impedance looking from the base of ltp is pretty low, only couples of K. Makes it slightly towards current feedback. If you only probe the voltage at the negative input, you miss some feedback that is through the current.
We all know that simulation is not reality.
For the many suggestions posted here, I am very looking forward to some of those proposals being put to the test,
i.e. being built, measured, and auditioned by their respective inventors.
Best regards,
Patrick
For the many suggestions posted here, I am very looking forward to some of those proposals being put to the test,
i.e. being built, measured, and auditioned by their respective inventors.
Best regards,
Patrick
Patrick, I think it will never happen. Armchair design is a discipline of its own. No need to initiate PCB production, no need to buy components, no need to make soldering, no need to make real testing and experience possible failures for the reason that stability or thermal issues were overlooked in the simulation.
Maybe then they should open their own thread, instead of hijacking yours.
😊
Patrick
😊
Patrick
Who are the "they" who should open their own thread?
For example, surely not @dadod, who has built, tested, and discussed in DIYaudio for everyone's enjoyment and education, many, many excellent pieces of work.
And has, in fact simulated, built, tested and explained the improvement he recommended.
Whereas I have only asked questions about the circuit, it's simulation and measurement because I am interested in it, as I already said.
If that is to "hijack" the thread then I have misunderstood the purpose, I assumed some discussion was why the OP opened it in the first place.
Are questions about the VAS drive in the OP's circuit a "hijack" of the thread? because I never received any answers and I'm still interested.
Best wishes
David
Last edited: