Poweramp or Line Stage
Speaking of overkill. The Krill was using the MJF15030 & 31 in their top of the line preamp. But Krills line output stages have always been small poweramps...
It should work with the current mirrors. I have built a linestage with the current mirrors and after that a PNP differential stage as in the OPA627 and also a folded cascode stage with PFETs or NPN's. I later discarded the current morror as it did not bring any benefit sonically. Maybe it is all overkill and should we follow lumanauw's advice of deleting the current mirrors.
Speaking of overkill. The Krill was using the MJF15030 & 31 in their top of the line preamp. But Krills line output stages have always been small poweramps...
It should work with the current mirrors. I have built a linestage with the current mirrors and after that a PNP differential stage as in the OPA627 and also a folded cascode stage with PFETs or NPN's. I later discarded the current morror as it did not bring any benefit sonically. Maybe it is all overkill and should we follow lumanauw's advice of deleting the current mirrors.
adding some more confusion
...well the current mirrors will probably cause three things:
1. High and difficult to calculate bias of the VAS.
2. More gain than needed.
3. An additional pole in the Bode plot.
Especially 2 & 3 are good for oscillators.
So using resistors instead of current mirrors may
help to calm down, but without addidtional frequency compensation
you will not get lucky if you settle a closed loop gain of 1 as the circuit offer a hell a lot of gain in open loop condition. So you will have a massive feedback.
A simple and brutal force frequency compensation would be to slow down the VAS which typically is already the slowest part. You can make this pole dominant with some pf (may be 100pf) from basis to
collector of Q11 & Q12 each.
Just from stability this should be fine.
After this you should check for TIM distorsion.
OK, nearly nobody can measure them.
But a good method to check if you are running into issues with that
is to apply a high frequency signal at the input and check if the
differential input stage is still working linear or shows clipping (to be measured directly at the output of the differential stage) in closed
loop condition.
If you are dealing with normal signals like you might get them from
a CD player you should be fine if the input stage can handle
an input signal of 1V at 100kHz....
If your input stage shows clipping you can:
a)Decrease the gain of the differential input stage.
Or
b) Add a passive low pass in front of the amp to avoid high
slew rates at the input of the amp.
We could start a philosophic and emotional thread which solution
is sounding better... (I would prefer a)
)
An other way and much simpler would for sure the proposal
of mcp.
Good luck
Markus
...well the current mirrors will probably cause three things:
1. High and difficult to calculate bias of the VAS.
2. More gain than needed.
3. An additional pole in the Bode plot.
Especially 2 & 3 are good for oscillators.
So using resistors instead of current mirrors may
help to calm down, but without addidtional frequency compensation
you will not get lucky if you settle a closed loop gain of 1 as the circuit offer a hell a lot of gain in open loop condition. So you will have a massive feedback.
A simple and brutal force frequency compensation would be to slow down the VAS which typically is already the slowest part. You can make this pole dominant with some pf (may be 100pf) from basis to
collector of Q11 & Q12 each.
Just from stability this should be fine.
After this you should check for TIM distorsion.
OK, nearly nobody can measure them.
But a good method to check if you are running into issues with that
is to apply a high frequency signal at the input and check if the
differential input stage is still working linear or shows clipping (to be measured directly at the output of the differential stage) in closed
loop condition.
If you are dealing with normal signals like you might get them from
a CD player you should be fine if the input stage can handle
an input signal of 1V at 100kHz....
If your input stage shows clipping you can:
a)Decrease the gain of the differential input stage.
Or
b) Add a passive low pass in front of the amp to avoid high
slew rates at the input of the amp.
We could start a philosophic and emotional thread which solution
is sounding better... (I would prefer a)
) An other way and much simpler would for sure the proposal
of mcp.
Good luck
Markus
The circuit with dual current mirrors works but biasing is unpredictable and gain tends to be too high to obtain unity-gain stability
Some kind of servo is required in order to control the biasing of the complimentary VAS. I've used an optocoupler with its LED in series with the Vbe multiplier to sense VAS biasing
Frequency compensation is tricky. I've found useful to add a RC network between VAS base and its supply rail to introduce a pole/zero and roll-off gain quickly at 9dB/oct after the miller capacitor of the Vas takes effect
Some kind of servo is required in order to control the biasing of the complimentary VAS. I've used an optocoupler with its LED in series with the Vbe multiplier to sense VAS biasing
Frequency compensation is tricky. I've found useful to add a RC network between VAS base and its supply rail to introduce a pole/zero and roll-off gain quickly at 9dB/oct after the miller capacitor of the Vas takes effect
It seems to me this amp has a gain of 1, unintentionally I'm sure. However, it is even harder to make a voltage follower stable. The 10K resistor needs a resistor (say 499 ohm) to ground to give you voltage gain of 21, which is typical for audio amps.
The current mirrors are not the reason it is oscillating, they add confusion though. For now, take them out, just use matched resistors. To tame the oscillation, I've already mentioned starting points, adding VAS Miller compensation or feedforward compensation. Get that solved first.
There is a lot of gain in this circuit and layout could also cause oscillation -- in the real world, not on paper. So sloppy layout could be contributing to the problem too (don't mean to be critical, I found this out the hard way, by experience, and perhaps you are too).
Another minor point: you may find your current sources are noisy, since you reference the supply rails. Either use a fixed reference or add filter caps to the divider to quell the hum.
The current mirrors are not the reason it is oscillating, they add confusion though. For now, take them out, just use matched resistors. To tame the oscillation, I've already mentioned starting points, adding VAS Miller compensation or feedforward compensation. Get that solved first.
There is a lot of gain in this circuit and layout could also cause oscillation -- in the real world, not on paper. So sloppy layout could be contributing to the problem too (don't mean to be critical, I found this out the hard way, by experience, and perhaps you are too).
Another minor point: you may find your current sources are noisy, since you reference the supply rails. Either use a fixed reference or add filter caps to the divider to quell the hum.
Simple Schematic
Hi,
If we leave out the current mirrors and the output emittor followers (Q13 & Q14) we have in principle the schematic of the Gas Thaedra line amp.
http://www.gasaudio.com/pics/Thaedra2bipolarlineamp.gif
That was almost 30 years ago.! Nothing new under the sun.
Funny James B. does not apply the Lender circuit variation here, as he does in all his other designs of that time.
Hi,
If we leave out the current mirrors and the output emittor followers (Q13 & Q14) we have in principle the schematic of the Gas Thaedra line amp.
http://www.gasaudio.com/pics/Thaedra2bipolarlineamp.gif
That was almost 30 years ago.! Nothing new under the sun.
Funny James B. does not apply the Lender circuit variation here, as he does in all his other designs of that time.
I've built the real thing, not sims. This is the exact problem with Randy Slone's design. In the article itself (I have his book) he says it is the "possibly" the best design there is, where the input is full complementary differential+current mirrors, the VAS is pushpull+cascodes+quadruple miller cap, and the output stage is K1078 pairs. I'm attracted to the article and built the real thing. Just replace the current mirror with simple resistor, and ZAP!!, it works normally.
Right. It does not oscilate. But it only produces DC without any music(about 20Vdc, if I dont mistaken, I built it many years ago).
Maybe it will work in sims, but I dont believe it will work in real world, even if you have perfect matched transistors for all.
Again, can you define the voltage of base Q11 and Q12?
Right. It does not oscilate. But it only produces DC without any music(about 20Vdc, if I dont mistaken, I built it many years ago).
Maybe it will work in sims, but I dont believe it will work in real world, even if you have perfect matched transistors for all.
Again, can you define the voltage of base Q11 and Q12?
Hi Lumanau,
I can't follow your experience.
I used the full complemtary differential (without the
current mirros ) and a following complementary VAS several times.
It always was playing music....
.... had no issues with DC....
Must have been some particular draw back in
your specific case. Generally these fully complentary
designs can work fine.
Bye
Markus
I can't follow your experience.
I used the full complemtary differential (without the
current mirros ) and a following complementary VAS several times.
It always was playing music....
.... had no issues with DC....
Must have been some particular draw back in
your specific case. Generally these fully complentary
designs can work fine.
Bye
Markus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.