Holidays are over, had to get back to work on his own stuff. If specific questions he would probably respond sooner. Otherwise, we are all waiting patiently for a build and testing to start
Last edited:
Holidays are over here, had to get back to work on his own stuff. If there are specific questions he would probably respond. Otherwise, we are all waiting patiently for a build and testing to start
Bloody work.
Hahahaha
One minor change to regulators:
I already have a bunch of heatsinks, but they're drilled with the TO-220 hole height, not TO-126.
So I've swapped the MJE340/350 for BD243C/BD244C.
They simulate a bit poorer on phase margin (62º -> 58º), but a bit better on dropout voltage (63V -> 62V) and attenuation (98dB -> 99dB).
I already have a bunch of heatsinks, but they're drilled with the TO-220 hole height, not TO-126.
So I've swapped the MJE340/350 for BD243C/BD244C.
They simulate a bit poorer on phase margin (62º -> 58º), but a bit better on dropout voltage (63V -> 62V) and attenuation (98dB -> 99dB).
Mechanical bits or the electrical bits? Here's what I have for the latter:
I'm using monoblock chassis from China this time:
4315F Aluminum Power Amplifier Chassis Class A amplifier Enclosure HiFi DIY Case | eBay
Note that it's a bit bigger than it looks: those heatsinks are 70mm instead of the more common 40mm.
Cheers,
Jeff.
I'm using monoblock chassis from China this time:
4315F Aluminum Power Amplifier Chassis Class A amplifier Enclosure HiFi DIY Case | eBay
Note that it's a bit bigger than it looks: those heatsinks are 70mm instead of the more common 40mm.
Cheers,
Jeff.
I was playing around with the KISS front-end (formerly Beefcake) and discovered that some frequencies (such as 140KHz) throw it into minor oscillation. Evidently, the circuit is just that little bit simpler than it should be.
There's a big spike in the gain curve in the MHz region which appears to be some interaction between the phase lead comp and the input filter. I tried using only one or the other of them, but it took too much capacitance to get reasonable margins. Just a big input filter killed the frequency response, and just a lot of phase lead comp killed the slew rate.
I tried 2-pole comp both with and without the phase lead and input filters, but all it did was add another spike.
I tried Miller comp, again both with and without phase lead and input filters, but none of them gave satisfactory margins without killing the slew rate.
Inclusive Miller comp was interesting: I could get the amp compensated and the slew rate up, but the 20kTHD still suffered. Not sure what that was all about.
Transitional Miller comp was uniformly worse than inclusive Miller comp.
A little bit of phase lead comp in the IPS worked a treat with the existing input filter and global phase lead comp.
I didn't try Miller input comp because I was tired of running sims at this point, and I had one that worked.
So here it is with the input filter, IPS phase lead comp, and global phase lead comp:
(At one point I was getting somewhere with just the phase lead comp by driving the IPS much harder, but I ran out of room. I wonder if this is why Nelson paralleled 2SJ109s in the front-end of the J2?)
Cheers,
Jeff.
There's a big spike in the gain curve in the MHz region which appears to be some interaction between the phase lead comp and the input filter. I tried using only one or the other of them, but it took too much capacitance to get reasonable margins. Just a big input filter killed the frequency response, and just a lot of phase lead comp killed the slew rate.
I tried 2-pole comp both with and without the phase lead and input filters, but all it did was add another spike.
I tried Miller comp, again both with and without phase lead and input filters, but none of them gave satisfactory margins without killing the slew rate.
Inclusive Miller comp was interesting: I could get the amp compensated and the slew rate up, but the 20kTHD still suffered. Not sure what that was all about.
Transitional Miller comp was uniformly worse than inclusive Miller comp.
A little bit of phase lead comp in the IPS worked a treat with the existing input filter and global phase lead comp.
I didn't try Miller input comp because I was tired of running sims at this point, and I had one that worked.
So here it is with the input filter, IPS phase lead comp, and global phase lead comp:
(At one point I was getting somewhere with just the phase lead comp by driving the IPS much harder, but I ran out of room. I wonder if this is why Nelson paralleled 2SJ109s in the front-end of the J2?)
Cheers,
Jeff.
I don’t think that was the reason, both front ends are biased with a ccs of the same value regardless of number of devices.
Maybe it was just to reduce the impedance at input and the feedback loop, by increasing the effective capacitance of the stage, and/or improve noise performance further.
If you use 22k/220k at input and feedback loop it may take care of your problems, or 10k/220k depending on gain requirements
Or maybe increase the value of C3.
Maybe use base stoppers on Q3 and Q4 and maybe also Q7 and Q8.
Maybe it was just to reduce the impedance at input and the feedback loop, by increasing the effective capacitance of the stage, and/or improve noise performance further.
If you use 22k/220k at input and feedback loop it may take care of your problems, or 10k/220k depending on gain requirements
Or maybe increase the value of C3.
Maybe use base stoppers on Q3 and Q4 and maybe also Q7 and Q8.
Last edited:
Increasing the resistance in the voltage dividers pushes the spike down, and adding the base stoppers reduces the height of the spike.
The spike is now well below unity gain, but oddly the oscillation gets worse, going from the odd hiccough to constant fuzz.
If I inject my AC signal at the output of the LTP and measure the amplifier output there's a few very sharp corners in the curve that aren't changed by the voltage dividers, and get worse with the base stoppers. The spike isn't at the frequency where I catch it in oscillation, though, and the phase never goes over 180º, so I'm not sure how to interpret that....
The spike is now well below unity gain, but oddly the oscillation gets worse, going from the odd hiccough to constant fuzz.
If I inject my AC signal at the output of the LTP and measure the amplifier output there's a few very sharp corners in the curve that aren't changed by the voltage dividers, and get worse with the base stoppers. The spike isn't at the frequency where I catch it in oscillation, though, and the phase never goes over 180º, so I'm not sure how to interpret that....
Tried resistor CCS (which produces some really cool oscillation, slowly building, then stopping, then repeating), and the 1ohm degeneration on the VAS MOSFET (and both together). Back to SPICE saying it shouldn't oscillate, but having it oscillate anyway.
Interestingly, all of these schemes can be fixed by re-enabling the LTP phase comp.
Interestingly, all of these schemes can be fixed by re-enabling the LTP phase comp.