The floating approach works for the resistors too, although one doesn't get the benefits of high impedance. Still a PITA.
Seems to work best that way to me also... except I use 40Khz for the -3db point OL.
What about you, JC, what do you find as the minimum OL bandwidth works best for you?
THx-RNMarsh
Some topoliges, like a diff amp driving another diff amp, if the second amp is biased high, at 25 mA or better, can provide full power OL bandwidth of better than 100 kHz, however with both first and second stage gain limiting.
H/K routinely does something like 80 kHz, working backwards, CL response is better than 250 kHz, and global NFB factor is 3:1, or 12 dB.
H/K routinely does something like 80 kHz, working backwards, CL response is better than 250 kHz, and global NFB factor is 3:1, or 12 dB.
I'm with you, Christophe, but I wonder why stop at 10 kHz, why not go even higher?
At first glance, the somwhat obvious answer is that with higher gain, you can use more GNFB to reduce distortions, but please also conside this: if you reduce the open loop gain with local degeneration and let the VAS work with 20 mA or more, your inherent open loop distortion will also be reduced. Also, of your VAS is working into fixed loads like 10 Kohm or more, it will maintain its performance over a wider bandwidth than otherwise.
This also depends on what you take for your least permitted THD factor in conjunction with the GNFB you plan to use.
The only obscured downside to this approach is that since your GNFB factor is reduced, to get rid of crossober distortions you will likely have to use between 100 and 150 mA of bias per output device. This is no big deal to achieve, but in case of several outpout device pairs this can tax your heat sinks. Even if you have say 125 mA per device, at just 51V, with three device pairss you'll have a normal dissipation of 38 Watts with zero signal.
Payback is that you will also have less thermal shifting once the amp heats up.
big a$$ JFET
Not sure why I couldn't find it before, but now Mouser is carrying it!
InterFET Corp. Page Selector - IF3601 datasheet
They are expensive.
1: $19.16
10: $16.11
100: $14.83
250: $13.51
500: $12.63
1,000: $11.59
Not sure why I couldn't find it before, but now Mouser is carrying it!
InterFET Corp. Page Selector - IF3601 datasheet
They are expensive.
1: $19.16
10: $16.11
100: $14.83
250: $13.51
500: $12.63
1,000: $11.59
Last edited:
Yes, maybe, but if you achieve that, there is no way you'll pass a current through R5: if everything is perfectly floating, you'll just bias the transistors, no more: Kirchhoff's Current Law has no provision for compromises.
Even if you magically surmount this hurdle, you will defeat the cascode action of the OP transistors
Nope. It works fine. Kirchoff is not being asked to compromise anything. The transistors still steer current in and out of R5, and the voltages are still anchored by Q2 and Q3 (note that the bias on their gates is not provided by the floating I source).
I'll clarify with a schematic in a bit.
Last edited:
That is new for Mouser to have them. They are tough to work with if you self bias them the noise performance is lost and the input capacitance is of course large. Fortunately we stocked up on Toshiba parts. Maybe someday they will get used for something.
> Maybe someday they will get used for something.
Something like this :
http://users.cosylab.com/~msekoranja/tmp/04447683.pdf
Patrick
Something like this :
http://users.cosylab.com/~msekoranja/tmp/04447683.pdf
Patrick
Think of the current source resistors as being open circuits for signal frequencies.
It's a cascode, but instead of connecting to the supply through a load resistor,
the signal path is folded to ground with an opposite polarity "upper" device.
Last edited:
Yeah that capacitance is huge. I fear I will be gone before a lot of my cherished parts are used for anything
Another thing I did and mentioned somewhere was to stack a bunch of BF862 parts and interconnect them via a wire on the gate leads, which also serves as a heatsink. To my surprise a simple amplifier using the composite was not oscillating, even without any lumped inductance in the gate. And hey---it doesn't take up much extra board space---just height
Yeah I remember that paper. I would have done something about that input capacitance
Don't think so, but let's wait for the clarification of bcarso. One thing is certain: without a return path from supplies (CCS?) to ground, there is no way to drive the output, you can just bias the transistors.
- Status
- Not open for further replies.