hi thanh !
So the amp works again ?
How does it sound ?
Offsetvoltage, do you mean DC-offset at output ?
Want to show the schematic ?
Openloopgain of 150, so you gave up on low thd ?
Last night i "destroyed" somehow my symamp, i wanted to try a
modification, but didn't work. So i reverted the modification, but still
not functional...
Seems that one of the 18 small bjts silently blew.
And yes, the low maxrating of jfets is annoying, and they are very
expensive ! I tried cascoding the inputstage, but it got unstable...
The sj74 costs 37 times more than a bc556b.
Mike
So the amp works again ?
How does it sound ?
Offsetvoltage, do you mean DC-offset at output ?
Want to show the schematic ?
Openloopgain of 150, so you gave up on low thd ?
Last night i "destroyed" somehow my symamp, i wanted to try a
modification, but didn't work. So i reverted the modification, but still
not functional...
Seems that one of the 18 small bjts silently blew.And yes, the low maxrating of jfets is annoying, and they are very
expensive ! I tried cascoding the inputstage, but it got unstable...
The sj74 costs 37 times more than a bc556b.
Mike
hi lumanauw !
I finished prototype for the folded cascode circuit. But somehow it
absolutely doesn't behave like in sims, it oscillates...
It plays music undistorted, so it can't be heavy miswiring. I've spent
5 hours adjusting caps, but nothing stopped oscillation. I will analyze
the prototype this evening, hopefully finding some bad wiring.
In sims the circuit never oscillates due to wrong caps, that's a bit
confusing and dissapointing, even worse is the fact that it does not
oscillate when loaded with a simple 8ohm resistor...
With some caps-settings the oscillation even got audible !
Mike
I finished prototype for the folded cascode circuit. But somehow it
absolutely doesn't behave like in sims, it oscillates...
It plays music undistorted, so it can't be heavy miswiring. I've spent
5 hours adjusting caps, but nothing stopped oscillation. I will analyze
the prototype this evening, hopefully finding some bad wiring.
In sims the circuit never oscillates due to wrong caps, that's a bit
confusing and dissapointing, even worse is the fact that it does not
oscillate when loaded with a simple 8ohm resistor...
With some caps-settings the oscillation even got audible !
Mike
Try to connect some caps (33pF) from each side of the collector of the folded cascode transistor to ground. And put additionally two 33pF caps from the output of both folded collectors to the negative input of the long tail input. Simulating is sometimes different from real behavior. The folded cascode is not so easy as ist seems...
Hi bocka !
The 33pF from each collector to ground i already have. I also added
10pF millercaps to the cascodes. These helped a little, at least it
stopped oscillating into pure resistive load.
I wanted to avoid the direct feedback from vas, as it reduces feedback
from output to null for higher freqs. Maybe i can use a cap + res ?
But i will try, as long as the amp oscillates it's useless...
Until now sims predicted quite good stability of an amp. What drives
me crazy is the fact that a normal vas-circuit is stable with much less
caps. If i add that much caps, speed of the circuit becomes a joke.
So i hope there is some miswiring / bad layout. I never had a that
much instable circuit before. I had more complex circuits beeing stable
without any cap... Is this a local oscillation to fight ?
Did you use symetrical or asymetrical topology for your folded cascode ?
Mike
The 33pF from each collector to ground i already have. I also added
10pF millercaps to the cascodes. These helped a little, at least it
stopped oscillating into pure resistive load.
I wanted to avoid the direct feedback from vas, as it reduces feedback
from output to null for higher freqs. Maybe i can use a cap + res ?
But i will try, as long as the amp oscillates it's useless...
Until now sims predicted quite good stability of an amp. What drives
me crazy is the fact that a normal vas-circuit is stable with much less
caps. If i add that much caps, speed of the circuit becomes a joke.
So i hope there is some miswiring / bad layout. I never had a that
much instable circuit before. I had more complex circuits beeing stable
without any cap... Is this a local oscillation to fight ?
Did you use symetrical or asymetrical topology for your folded cascode ?
Mike
Hi MikeB,
it's not a joke. Sim is only as accucate as the models you're using. I spend weeks to get the folded cascode stable. If you need more speed you should increase the current of the long tail pair.
I've used a sym amp design. The folded cascode is a linear high speed amp and it needs special know-how to make it stable. The problem is that you using low speed (high-power) output transistors with a rf design. I've build several folded cascoded amps and after all they are stable, extraordonary stable. You need a rf path from the output of the folded cascode to the long tail input. Maybe my values are a little bit large but the are a starting point.
it's not a joke. Sim is only as accucate as the models you're using. I spend weeks to get the folded cascode stable. If you need more speed you should increase the current of the long tail pair.
I've used a sym amp design. The folded cascode is a linear high speed amp and it needs special know-how to make it stable. The problem is that you using low speed (high-power) output transistors with a rf design. I've build several folded cascoded amps and after all they are stable, extraordonary stable. You need a rf path from the output of the folded cascode to the long tail input. Maybe my values are a little bit large but the are a starting point.
Yes I'm shure you build a microwave oven 
I never used a testboard. Maybe keeping the tracks short will help. When I started with this topology I build a line amp first. After some experience I switched over to a power amp. A Zobel network on the output of the EF also helps sometimes to prevent oscillation.
I never used a testboard. Maybe keeping the tracks short will help. When I started with this topology I build a line amp first. After some experience I switched over to a power amp. A Zobel network on the output of the EF also helps sometimes to prevent oscillation.
Mike, what is the schematic for your folded-cascode power amp?
I don't know how you are referencing the bases of the folded-cascocde stage, but it may be that you need to install a small resistor (47 ohms for example) in series with the base of each folded cascode device (and don't add any subsequent capacitance to ground or the power rails). Or, should you be referencing the bases with a voltage source and you have added capacitors across the voltage source to reduce noise, you may need to eliminate the capacitors. It is possible to get local oscillations with a cascode when the drive impedance into the base is too low, and the problem occurs more frequently when the output of the cascode is feeding into a high-impedance node.
Also, the traces feeding directly into the cascode device bases (after whatever base stopper resistor you install) should be kept short.
A folded cascode does have different quirks from a more typical multiple gain-stage topology. But once you learn what it likes and doesn't like, it tends to be _more_ stable than a multi gain-stage topology, not less (particularly when global feedback is applied).
Recently I have been working on a power amp with a floating-reference folded-cascode stage, a Sziklai output stage, 0.1 ohm output resistors, no output inductor/resistor network (an output Zobel network was retained), and a good amount of global NFB. It has been stable with any load (including capacitive) that I have tested it with.
best of luck, jonathan carr
I don't know how you are referencing the bases of the folded-cascocde stage, but it may be that you need to install a small resistor (47 ohms for example) in series with the base of each folded cascode device (and don't add any subsequent capacitance to ground or the power rails). Or, should you be referencing the bases with a voltage source and you have added capacitors across the voltage source to reduce noise, you may need to eliminate the capacitors. It is possible to get local oscillations with a cascode when the drive impedance into the base is too low, and the problem occurs more frequently when the output of the cascode is feeding into a high-impedance node.
Also, the traces feeding directly into the cascode device bases (after whatever base stopper resistor you install) should be kept short.
A folded cascode does have different quirks from a more typical multiple gain-stage topology. But once you learn what it likes and doesn't like, it tends to be _more_ stable than a multi gain-stage topology, not less (particularly when global feedback is applied).
Recently I have been working on a power amp with a floating-reference folded-cascode stage, a Sziklai output stage, 0.1 ohm output resistors, no output inductor/resistor network (an output Zobel network was retained), and a good amount of global NFB. It has been stable with any load (including capacitive) that I have tested it with.
best of luck, jonathan carr
Hi jonathan !
A nearly identical schematic is 1 or 2 pages back.
I use 3 diodes in series for the Vref, loaded with 20k to the other
side. I have a 500ohm baseresistor, is this too much ?
And yes, i have 1nF in paralell to the diodes. (installed afterwards)
I added 10pF millercaps to the cascode (C-B), these helped a little.
The traces to the cascodebase are very short, a few mm.
The fact that the 10pF helped, shows to me that its likely a local
oscillation at the cascode.
And yes, the "vas" has high impedance. (At output of cascode)
Okay, solder iron heated up, i will try some mods...
Mike
A nearly identical schematic is 1 or 2 pages back.
I use 3 diodes in series for the Vref, loaded with 20k to the other
side. I have a 500ohm baseresistor, is this too much ?
And yes, i have 1nF in paralell to the diodes. (installed afterwards)
I added 10pF millercaps to the cascode (C-B), these helped a little.
The traces to the cascodebase are very short, a few mm.
The fact that the 10pF helped, shows to me that its likely a local
oscillation at the cascode.
And yes, the "vas" has high impedance. (At output of cascode)
Okay, solder iron heated up, i will try some mods...
Mike
Mike, is your schematic the one included with Post 184?
First thing - do add a Zobel network to the output. This is _not_ a unique recommendation specific for a folded cascode. I would recommend doing this for multistage topologies as well.
You shouldn't need as much as 500 ohms for the base resistor. Something between 47~100 ohms should do the job. You _really_ don't want capacitance directly on the cascode base, so maybe you could try lifting the base (only) ir Q16/Q17 from the PCB and soldering a small 1/8W stopper resistor from the biasing diode network directly to the base.
If the base resistor is isolating cascode devices Q16/Q17 from the 1nF cap, there should be no problem. You may want to change C5 into a series RC network from the base of Q7 to ground, and I would try adding a similar RC network right at the base of Q6.
I would also add base stoppers to some of the BJTs that comprise the triple Darlingtons. I'd definitely add a 4.7 ohm stopper resistor to the base of Q15/Q9.
I would also try adding stopper resistors to the bases of Q7/Q6. Experiment with the values - maybe start with something like 220 ohms, and if that works (with reactive loads), gradually reduce the value until problems appear again. OTOH, adding base resistance to Q7/Q6 can have an effect on any RC network that is added to these same bases - I'd recommend using your ears as well as the scope.
Also, you may want to try lifting C1 from ground and reconnecting it from the bases of Q10/Q13 to the bases of Q11/Q12. Increasing the value may also be appropriate. I would try 220pF, but OTOH, I normally use cascoded JFET inputs in preference to BJTs, so I don't know how this would work in your case.
hope this give you some fruitful avenues to pursue, jonathan carr
First thing - do add a Zobel network to the output. This is _not_ a unique recommendation specific for a folded cascode. I would recommend doing this for multistage topologies as well.
You shouldn't need as much as 500 ohms for the base resistor. Something between 47~100 ohms should do the job. You _really_ don't want capacitance directly on the cascode base, so maybe you could try lifting the base (only) ir Q16/Q17 from the PCB and soldering a small 1/8W stopper resistor from the biasing diode network directly to the base.
If the base resistor is isolating cascode devices Q16/Q17 from the 1nF cap, there should be no problem. You may want to change C5 into a series RC network from the base of Q7 to ground, and I would try adding a similar RC network right at the base of Q6.
I would also add base stoppers to some of the BJTs that comprise the triple Darlingtons. I'd definitely add a 4.7 ohm stopper resistor to the base of Q15/Q9.
I would also try adding stopper resistors to the bases of Q7/Q6. Experiment with the values - maybe start with something like 220 ohms, and if that works (with reactive loads), gradually reduce the value until problems appear again. OTOH, adding base resistance to Q7/Q6 can have an effect on any RC network that is added to these same bases - I'd recommend using your ears as well as the scope.
Also, you may want to try lifting C1 from ground and reconnecting it from the bases of Q10/Q13 to the bases of Q11/Q12. Increasing the value may also be appropriate. I would try 220pF, but OTOH, I normally use cascoded JFET inputs in preference to BJTs, so I don't know how this would work in your case.
hope this give you some fruitful avenues to pursue, jonathan carr
Hi MikeB,
I took another look on your schematic. I build something very similar and it
does not oscillate. As I see nearly everything I recommended has been done, I
overviewed it. The only thing I did the other way round was using a CFP instead
of a triple EF. I'd recommend to use larger R18 - R21 as well as a Zobel network
as I've stated before and also Jonathan Jarr does. I never needed base stopper
resistors, but I also never used caps over my referencing diodes. For these
I used LEDs, cheap a good choice to maintain thermal stability.
I'm sure this amp is stable. Did you blocked the supply with 100nF caps or
similar as close as possible to the ground? If it also oscillates give a PCB
a chance or rebuild your amp using tracks as short as possible.
I took another look on your schematic. I build something very similar and it
does not oscillate. As I see nearly everything I recommended has been done, I
overviewed it. The only thing I did the other way round was using a CFP instead
of a triple EF. I'd recommend to use larger R18 - R21 as well as a Zobel network
as I've stated before and also Jonathan Jarr does. I never needed base stopper
resistors, but I also never used caps over my referencing diodes. For these
I used LEDs, cheap a good choice to maintain thermal stability.
I'm sure this amp is stable. Did you blocked the supply with 100nF caps or
similar as close as possible to the ground? If it also oscillates give a PCB
a chance or rebuild your amp using tracks as short as possible.
Hi, Mike,
Don't give up. You cannot built a power amp with 1 way work (I never experience that). Always need touch here and there. The first I encountered usually is stability/oscilation problem, like you do now.
But now you have learn important thing. SIM is not real world.
These are my opinion:
1. For voltage reference (Like R1-R2, V24-V26) I always use standing current more than 4-5mA. It gives more stability to the voltage reference. Don't use small bias for these. Especially if you use Zeners, they do not like small bias. LEDs are better.
2. C5 and C4. Maybe you got these from SIM. Usually there is needed only 1 small pf cap, that is from base of Q7 to base of Q11-Q12. It is for compensating NPN vs PNP output stage behavior. You put C4 in the wrong place (in PNP), making the need of bigger value of C5.
3. You should try replacing R10-R11 with CCS (not just plain R). It gives far better folded cascode work.
Don't give up. You cannot built a power amp with 1 way work (I never experience that). Always need touch here and there. The first I encountered usually is stability/oscilation problem, like you do now.
But now you have learn important thing. SIM is not real world.
These are my opinion:
1. For voltage reference (Like R1-R2, V24-V26) I always use standing current more than 4-5mA. It gives more stability to the voltage reference. Don't use small bias for these. Especially if you use Zeners, they do not like small bias. LEDs are better.
2. C5 and C4. Maybe you got these from SIM. Usually there is needed only 1 small pf cap, that is from base of Q7 to base of Q11-Q12. It is for compensating NPN vs PNP output stage behavior. You put C4 in the wrong place (in PNP), making the need of bigger value of C5.
3. You should try replacing R10-R11 with CCS (not just plain R). It gives far better folded cascode work.
Hi !
I stopped oscillation by simply adding a RC from emitter of cascode
to rail. Like in my previous amps: Oscillating ? RC always fixes it !
to lumanauw: I planned the ccs for the Vref, but not much place left
on my testboard... But by connecting both vrefs and avoiding going
to ground, the vref is at least always symetrical.
I wanted to use leds, but i had to use what i've found in my box...
A bit too early to talk about the sounding, but it sounds different.
It's missing the smoothness of my other symamp, but finetuning
should fix this. The trebles, the best i've ever heard ! Not a single
sign of the weakness from my previous amp, they just play, no sign
of any exaggeration or distortion. No singerssss any more !
One strange experience: The music "flows", it's never too loud, never
not loud enough, it just plays, hard to describe !
And, it has the same tight bass like my other symamp, so this seems
to come from the combination of symetrical topology and the outputbjts,
the mjl3281a/1302a.
Mike
I stopped oscillation by simply adding a RC from emitter of cascode
to rail. Like in my previous amps: Oscillating ? RC always fixes it !
to lumanauw: I planned the ccs for the Vref, but not much place left
on my testboard... But by connecting both vrefs and avoiding going
to ground, the vref is at least always symetrical.
I wanted to use leds, but i had to use what i've found in my box...
A bit too early to talk about the sounding, but it sounds different.
It's missing the smoothness of my other symamp, but finetuning
should fix this. The trebles, the best i've ever heard ! Not a single
sign of the weakness from my previous amp, they just play, no sign
of any exaggeration or distortion. No singerssss any more !
One strange experience: The music "flows", it's never too loud, never
not loud enough, it just plays, hard to describe !
And, it has the same tight bass like my other symamp, so this seems
to come from the combination of symetrical topology and the outputbjts,
the mjl3281a/1302a.
Mike
Mike B. I'm sure happy that i posted my folded cascode 2 weeks before you posted yours in post #184 of this thread Otherwise i might accidently have copied your topology
http://www.diyaudio.com/forums/showthread.php?s=&threadid=40065&perpage=10&pagenumber=4
post #39
Otherwise i might accidently have copied your topology http://www.diyaudio.com/forums/showthread.php?s=&threadid=40065&perpage=10&pagenumber=4
post #39
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.