I was trying to resurrect some old CCS's boards from long ago for a new headphone amp I'm building, but IRF820's are not too available, or that is to say, the versions available today are quite different in performance to ones from 10+ years ago. So far I've gotten a pretty effective oscillator going in the 600Khz-1Mhz range pretty reliably. They're also pretty hard to troubleshoot because of all the high node impedances that aren't really probe-able even with a good differential without affecting the stability of the circuit.
I've tried a few mosfets to varying success, though all are unstable to various degrees. The closest I've gotten to something working is the FQPF2N80, though I had to increase the upper mosfet gate stopper to 15K and even then at higher voltages and lower currents it still oscillates. I've gone back and tested some of the ones I built 10+ years ago using some FQPF2N60C's, and found though they seemed stable at the time, reducing the current to the 3ma-10ma ranges I need, they also quickly become unstable.
If anyone has had luck with current production/availability mosfets and has a working gate stopper value for them, please share. In the meantime I'm spinning some new test PCB's that have extremely tight loop distances, and one version with mostly SMD parts to see if they show any improvement. The next MOSFET I have my sights on testing is the Vishay IRFIBF20G. I also might explore replacing some of the lower current CCS's (3ma) with simple cascode depletion ones for my project, but I would like to find working enhancement mosfets for these circuits as well because I had some plans for voltage regulators and/or the SS pentode operation modes as well.
Some relevant posts on the topic by Rod Coleman and Gary Pimm:
https://www.diyaudio.com/community/...-bias-ccs-500khz-ringing.318407/#post-5332345
https://www.diyaudio.com/community/...-regulator-projects.82641/page-5#post-1734963
Wayback link to Pimm's original page on the self bias CCS:
https://web.archive.org/web/20110514215841/http://pimmlabs.com/web/self_bias.htm
I've tried a few mosfets to varying success, though all are unstable to various degrees. The closest I've gotten to something working is the FQPF2N80, though I had to increase the upper mosfet gate stopper to 15K and even then at higher voltages and lower currents it still oscillates. I've gone back and tested some of the ones I built 10+ years ago using some FQPF2N60C's, and found though they seemed stable at the time, reducing the current to the 3ma-10ma ranges I need, they also quickly become unstable.
If anyone has had luck with current production/availability mosfets and has a working gate stopper value for them, please share. In the meantime I'm spinning some new test PCB's that have extremely tight loop distances, and one version with mostly SMD parts to see if they show any improvement. The next MOSFET I have my sights on testing is the Vishay IRFIBF20G. I also might explore replacing some of the lower current CCS's (3ma) with simple cascode depletion ones for my project, but I would like to find working enhancement mosfets for these circuits as well because I had some plans for voltage regulators and/or the SS pentode operation modes as well.
Some relevant posts on the topic by Rod Coleman and Gary Pimm:
https://www.diyaudio.com/community/...-bias-ccs-500khz-ringing.318407/#post-5332345
https://www.diyaudio.com/community/...-regulator-projects.82641/page-5#post-1734963
Wayback link to Pimm's original page on the self bias CCS:
https://web.archive.org/web/20110514215841/http://pimmlabs.com/web/self_bias.htm
What are the voltage and currents you are targeting?
It's true that older MOSFET'S was better than new. I made a pushpull of IRF510 for a linear output amplifier from an old President McKinley CB rig modified (by me) to ham 10mts band. And they performed very well (we are talking of SSB @ 28MHz 😳🙄😯).
It's true that older MOSFET'S was better than new. I made a pushpull of IRF510 for a linear output amplifier from an old President McKinley CB rig modified (by me) to ham 10mts band. And they performed very well (we are talking of SSB @ 28MHz 😳🙄😯).
When I've first brought them up on my bench supply at 80v, generally most have been stable, except for a few extremely fast mosfets. However, when moving to a real B+ supply at 350v, almost all start oscillating. So the Vds on the upper mosfet seems to play a role there. Also the older circuits I used were higher current 24ma and up, and seemed more stable. I'm experimenting with some lower current tubes and the circuit seems much less stable at low currents of 10ma or below.
In this test case I just have the mosfet ccs sitting above a 3k load resistor, that's it. And even the tiniest bit of capacitance added, like putting just one end of a handheld dmm probe between the ccs and resistor stabilizes it. It's so quickly stabilized by probing that the first clue something was wrong was my load resistor being incinerated. When the circuit oscillates, the bias charges up and current skyrockets, but every time I measured it it stopped. Finally using an analog panel meter I could watch the current shoot up when I switch in/out a small film capacitor across the load.
In this test case I just have the mosfet ccs sitting above a 3k load resistor, that's it. And even the tiniest bit of capacitance added, like putting just one end of a handheld dmm probe between the ccs and resistor stabilizes it. It's so quickly stabilized by probing that the first clue something was wrong was my load resistor being incinerated. When the circuit oscillates, the bias charges up and current skyrockets, but every time I measured it it stopped. Finally using an analog panel meter I could watch the current shoot up when I switch in/out a small film capacitor across the load.
Perhaps it is not a problem strictly of the MOSFET's themselves. Being them capacitances, the output conductance of the drivers play a crucial role. How are you driving them? In SMPS's it it well known that leaving a gate open circuit or driven from a high impedance circuit is synonnimus of hard problems. Not only with MOSFET, also with IGBTs.
The positive supply of this CCS feeds two circuits, so careful decoupling is needed - maybe 47uF electrolytic Panasonic M (NO low impedance types!). Try also with 1 ohm series resistor. The ground return of the cap should be copper tape or other wide path, back to the tube ground point.
But it should be noted that this circuit is much more difficult to get right, compared to a depletion fet cascode.
The 100uF cap risks being a feedback loop from output to the voltage setting node, which can easily turn nasty. Then, the 4.7k in the upper gate is a sure sign of marginal stability. Stoppers should never be set this high: the real problem should be solved.
Much better performance can be had from IXTP10N1000 (upper FET) and DN2540 lower. 100R stoppers are sufficient for these. Or simply get Ale's mu follower as an anode load - optimised for the purpose.
But it should be noted that this circuit is much more difficult to get right, compared to a depletion fet cascode.
The 100uF cap risks being a feedback loop from output to the voltage setting node, which can easily turn nasty. Then, the 4.7k in the upper gate is a sure sign of marginal stability. Stoppers should never be set this high: the real problem should be solved.
Much better performance can be had from IXTP10N1000 (upper FET) and DN2540 lower. 100R stoppers are sufficient for these. Or simply get Ale's mu follower as an anode load - optimised for the purpose.
Osvaldo, the link to the circuit is in the first post in the wayback machine link.
Rod,
Thanks for the input. I've re-read some other comments of yours about this circuit and other CCS's stability and this definitely seems like a trickier one. Especially given I'm using mosfets extremely close to the original specs and still having a variety of problems. Ultimately I might try some more with layouts where I can easily put everything extremely close together on one PCB (psu cap, ccs, tube, parafeed cap, output trans, everything...), but as sort of a 'drop in' CCS load for an anode, it's hard to recommend. Especially since oscillation seems to quickly drive up the bias network and drive max current into whatever is underneath it, be it just a resistor or maybe an expensive NOS tube.
I'll keep reading about Ale's mu-follower a bit, as well as Pete Millet just put up another fairly classic cascode depletion ccs. I was always suspicious of how little voltage they put across the lower fet in those designs, but I guess the newer IXT units especially with jfets underneath seem to have fairly reasonable op points. And my guess would be that trying to bias up that upper cascode FET to provide more Vds for the lower one is precisely where things start to get trickier with stability.
Rod,
Thanks for the input. I've re-read some other comments of yours about this circuit and other CCS's stability and this definitely seems like a trickier one. Especially given I'm using mosfets extremely close to the original specs and still having a variety of problems. Ultimately I might try some more with layouts where I can easily put everything extremely close together on one PCB (psu cap, ccs, tube, parafeed cap, output trans, everything...), but as sort of a 'drop in' CCS load for an anode, it's hard to recommend. Especially since oscillation seems to quickly drive up the bias network and drive max current into whatever is underneath it, be it just a resistor or maybe an expensive NOS tube.
I'll keep reading about Ale's mu-follower a bit, as well as Pete Millet just put up another fairly classic cascode depletion ccs. I was always suspicious of how little voltage they put across the lower fet in those designs, but I guess the newer IXT units especially with jfets underneath seem to have fairly reasonable op points. And my guess would be that trying to bias up that upper cascode FET to provide more Vds for the lower one is precisely where things start to get trickier with stability.