This is a sort of clone of my 50W project (wich still isn`t finished due to an extremely empty wallet and other bad excuses).
I just wan`t your opinion, the input is balanced, also the differential input impedance is equal...but I had to implement a DC-servo to surpress a DC-offset wich appeared at the output (for some *strange* reason, or not strange -because of my limited design experience).
I`ve tried lots of different topologies, and this one seemed to have the lowest distortion figures, though the S/N is a bit worse than in the 50W design....but again this simulation numbers.
I probably rely too much on the simulator...
The output power is 180W class AB (100W class A), 8 ohm load.
The outputs are the same as in my 50W design...actually most of the circuit is rather similar, with a few changes...
It`s that balancing issue I`m worried about, mostly because I had to implement a DC-servo.
Damn that picture was blurry!
(I`m also working on a sym-schematic with hawksford cascoded VA-stage and MosFET output`s, but that`s for now. -thanks for the hawksford papers, Andy C)
I just wan`t your opinion, the input is balanced, also the differential input impedance is equal...but I had to implement a DC-servo to surpress a DC-offset wich appeared at the output (for some *strange* reason, or not strange -because of my limited design experience).
I`ve tried lots of different topologies, and this one seemed to have the lowest distortion figures, though the S/N is a bit worse than in the 50W design....but again this simulation numbers.
I probably rely too much on the simulator...
The output power is 180W class AB (100W class A), 8 ohm load.
The outputs are the same as in my 50W design...actually most of the circuit is rather similar, with a few changes...
It`s that balancing issue I`m worried about, mostly because I had to implement a DC-servo.
Damn that picture was blurry!
(I`m also working on a sym-schematic with hawksford cascoded VA-stage and MosFET output`s, but that`s for now. -thanks for the hawksford papers, Andy C)
Attachments
Crazy biest
Hi Panzer!
You love tough designs right?
As far as I got through this crazy biest, it looks like
both tails of the differential stage should run at a similar
current. Comming from this the basis currents of Q4 and Q5
will be nearly the same. If we now look to Q5, we find a DC path
with low resistance to ground.
The DC path for Q4 has a much higher resistance.
If you run it without DC-servo, then I would expect that the output will show some positive offset in order to deliver enough DC through the feedback resistor.
For proper DC levels without DC servo, I would propose to increase
R43 to the same value as the feedback resistor R41.
And to decrease R40 to the same value as R44.
For which reason did you choose it so unsymmetrical?
Cheers
Markus
Hi Panzer!
You love tough designs right?
As far as I got through this crazy biest, it looks like
both tails of the differential stage should run at a similar
current. Comming from this the basis currents of Q4 and Q5
will be nearly the same. If we now look to Q5, we find a DC path
with low resistance to ground.
The DC path for Q4 has a much higher resistance.
If you run it without DC-servo, then I would expect that the output will show some positive offset in order to deliver enough DC through the feedback resistor.
For proper DC levels without DC servo, I would propose to increase
R43 to the same value as the feedback resistor R41.
And to decrease R40 to the same value as R44.
For which reason did you choose it so unsymmetrical?
Cheers
Markus
This isn`t the type of answer I was expecting, why bother?
I`ve changed the resistor values, the DC-offset is now 40mV, and the input resistance is 1k, a bit low, but my preamp will handle it without any problems...
The reason I chose non-symmetrical is because...well I like it. For the same reasons Per Anders likes symmetrical I guess.
1K input resistance??
1k input resitance?
From my proposal I would have expected around 3.5K Ohms.
Or did you keep it unsymmetrical and simply lowered all values in order
to avoid high DC-drops across the resistors?
Well from theory I would prefer symmetrical values, but in fact
the smashing criteria is to listen to the amp and then decide what
you prefer!
Good luck
Markus
1k input resitance?
From my proposal I would have expected around 3.5K Ohms.
Or did you keep it unsymmetrical and simply lowered all values in order
to avoid high DC-drops across the resistors?
Well from theory I would prefer symmetrical values, but in fact
the smashing criteria is to listen to the amp and then decide what
you prefer!
Good luck
Markus
If I understand this right, the only frequency compensation you're using is the lead compensation cap in the feedback loop? If this is true, the bandwidth must be very large. How large of a capacitive load can it drive while still maintaining stability?
Also, with the paralleled PNP common emitter amps in the VAS (U14 and U25 I think?) and no emitter degeneration, what happens if two transistors that are otherwise identical, but with Is values that differ by, say, 20% are used? Does the DC current mismatch have any undesirable effects?
One thing I found with common base VAS transistors is when the amp clips, surprisingly large base currents are needed. Without the 47uF capacitors it would easily be enough to turn off the zeners. What happens with low-frequency clipping, say 20 Hz? Do the zeners turn off?
With the wide bandwidth, does the high-frequency grounding of the common base VAS amp affect stability? From what I've seen, electrolytic caps in the neighborhood of 47 uF have around 20-30 nH of inductance, and an ESR of about .25 Ohm. This gives a series resonant frequency of about 150 kHz or so. Will this inductive impedance seen by the base affect the open loop gain/phase in the MHz region?
Regarding the Hawksford cascode, there was a thread started by peufeu in which it was discovered that both it and the "super pair" had a tendency to oscillate (or at least have frequency response peaking or time domain ringing) in the neighborhood of 50-100 MHz. This was verified for the super pair by both peufeu and myself in simulation, and by peufeu on the bench. This seems to be the result of sampling the base current and feeding it back to the emitter. I also observed that the Hawksford cascode had rather poor clipping behavior. There's apparently no way to incorporate a Baker clamp into the Hawksford cascode due to the biasing technique it uses. For a power amp I'm working on, I ended up just using a plain cascode with a Baker clamp and an emitter follower to sink the current of the Baker clamp diode when it turns on.
That's all I can think of at the moment.
Also, with the paralleled PNP common emitter amps in the VAS (U14 and U25 I think?) and no emitter degeneration, what happens if two transistors that are otherwise identical, but with Is values that differ by, say, 20% are used? Does the DC current mismatch have any undesirable effects?
One thing I found with common base VAS transistors is when the amp clips, surprisingly large base currents are needed. Without the 47uF capacitors it would easily be enough to turn off the zeners. What happens with low-frequency clipping, say 20 Hz? Do the zeners turn off?
With the wide bandwidth, does the high-frequency grounding of the common base VAS amp affect stability? From what I've seen, electrolytic caps in the neighborhood of 47 uF have around 20-30 nH of inductance, and an ESR of about .25 Ohm. This gives a series resonant frequency of about 150 kHz or so. Will this inductive impedance seen by the base affect the open loop gain/phase in the MHz region?
Regarding the Hawksford cascode, there was a thread started by peufeu in which it was discovered that both it and the "super pair" had a tendency to oscillate (or at least have frequency response peaking or time domain ringing) in the neighborhood of 50-100 MHz. This was verified for the super pair by both peufeu and myself in simulation, and by peufeu on the bench. This seems to be the result of sampling the base current and feeding it back to the emitter. I also observed that the Hawksford cascode had rather poor clipping behavior. There's apparently no way to incorporate a Baker clamp into the Hawksford cascode due to the biasing technique it uses. For a power amp I'm working on, I ended up just using a plain cascode with a Baker clamp and an emitter follower to sink the current of the Baker clamp diode when it turns on.
That's all I can think of at the moment.
Hawksford cascode
What is the idea with Hawksford cascode?
When talking about the Hawksford cascode I can imagine that can oscillate, the Cbc is over collector and emitter of the input transistor an reminds of a typical HF oscillator circuit, I have for long time ago bulit one for around 100 MHz.
I wonder if it could be solved by using a small base resistor to the U21-22, that would decrease somewhat the inherent capacitor effect on the input transistor Q4-5....
Another thing that could help is to keep the input transistor collector resistor very close as well as keeping U21-22 close too to keep copperboard strips and component leggs inductances at a minimum.
What is the idea with Hawksford cascode?
When talking about the Hawksford cascode I can imagine that can oscillate, the Cbc is over collector and emitter of the input transistor an reminds of a typical HF oscillator circuit, I have for long time ago bulit one for around 100 MHz.
I wonder if it could be solved by using a small base resistor to the U21-22, that would decrease somewhat the inherent capacitor effect on the input transistor Q4-5....
Another thing that could help is to keep the input transistor collector resistor very close as well as keeping U21-22 close too to keep copperboard strips and component leggs inductances at a minimum.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.