An externally hosted image should be here but it was not working when we last tested it.
All BJTs 2N3904/6... or monolithic pairs if you're made of money. Mirrors should probably have balancing emitter resistors (not shown) if discrete.
With a couple compensation components, it seems to do a good job in simulation. Whaddya think?
Tim
Clever circuit, Tim,
How linear are the current mirrors?
The two floating power supplies would be a PITA, but doable.
Any distortion figures, profiles?
Hugh
How linear are the current mirrors?
The two floating power supplies would be a PITA, but doable.
Any distortion figures, profiles?
Hugh
Linearity seems to correspond to the input differential... i.e. terrible without feedback (or predistortion diodes). Not unusual to expect.
Assuming the simulator's AC Analysis knows what it's doing, I found gain ca. 96dB and fT = 11MHz, with something like -240 degrees phase shift at fT. Some feedback and compensation cleans that up for a reasonable gain-of-20 monoblock, but it's still prone to oscillation when the two stages (high and low side) are conducting.
Because it's current mode, the voltage waveforms can look weird, especially around zero crossing, where crossover distortion is easy to make.
I'm still not entirely sure how I want to handle the input stage. A differential current steering doesn't allow big current peaks, so you have to bias it way into class A to get much output. An op-amp working into "nothing" (use supply rail current to drive the current amps) is one option.
Tim
Assuming the simulator's AC Analysis knows what it's doing, I found gain ca. 96dB and fT = 11MHz, with something like -240 degrees phase shift at fT. Some feedback and compensation cleans that up for a reasonable gain-of-20 monoblock, but it's still prone to oscillation when the two stages (high and low side) are conducting.
Because it's current mode, the voltage waveforms can look weird, especially around zero crossing, where crossover distortion is easy to make.
I'm still not entirely sure how I want to handle the input stage. A differential current steering doesn't allow big current peaks, so you have to bias it way into class A to get much output. An op-amp working into "nothing" (use supply rail current to drive the current amps) is one option.
Tim
Fixed it.
Basically, switched sides of the current regulating diffamp things, which eliminates all those biasing mirrors and makes the stages inverting = inherently stable.
Without feedback, the transient response is to die for. Of course, the unloaded gain of 120dB(!) isn't very useful, even if it's -3dB at only 600Hz (again, assuming the AC analysis isn't lying to me). The same high frequency pole is still present, unity gain out at 96MHz or something like that, with too much phase shift to wrap feedback around blindly.
This amplifier appears to have some very interesting properties, I just hope someone can come up with a way to compensate it. Any takers?
Tim
Basically, switched sides of the current regulating diffamp things, which eliminates all those biasing mirrors and makes the stages inverting = inherently stable.
Without feedback, the transient response is to die for. Of course, the unloaded gain of 120dB(!) isn't very useful, even if it's -3dB at only 600Hz (again, assuming the AC analysis isn't lying to me). The same high frequency pole is still present, unity gain out at 96MHz or something like that, with too much phase shift to wrap feedback around blindly.
An externally hosted image should be here but it was not working when we last tested it.
This amplifier appears to have some very interesting properties, I just hope someone can come up with a way to compensate it. Any takers?
Tim
I might be out of my league but wouldnt the combination of small emitter resistors, and supermatched piggybacked SMD transistors do the job?
Emitter resistors add local NFB, so the gain will come down, but the -3dB point will go up, about like adding global NFB as it is. Not sure what that'll do to the VHF pole, might make it better.
Supermatched pairs aren't a big deal here, the current mirrors aren't operatings over a wide range and the three diff pairs are inside feedback loops. The input diff should be well matched, but the other two aren't critical.
The bigger problem is compensation. This thing is useless until gain can be brought down to a reasonable level without oscillation.
Tim
Supermatched pairs aren't a big deal here, the current mirrors aren't operatings over a wide range and the three diff pairs are inside feedback loops. The input diff should be well matched, but the other two aren't critical.
The bigger problem is compensation. This thing is useless until gain can be brought down to a reasonable level without oscillation.
Tim
- Status
- Not open for further replies.