I had built up a few cascode CCS's for a gas tube shunt regulator, running at 82mA. I have some pcb's used for a standard heatsinked design which has
worked well at lower currents. What I found, at least with this latest batch of 10M45 IC's, is that the Vgs was very low, around 0.5V for 82mA. The net
result was that the bottom DN2540 was unable to provide this low a voltage, and the 10M45 would not reach the desired 82mA target. So the problem is
that the bottom device was saturated due to insufficient Vds.
This got me thinking about how to provide more Vgs to the top devices. More voltage is also better for the bottom device, which decreases in output
capacitance as the voltage increases.
I have attached a possible design with paralleled upper IC's. Running them in parallel now requires less current, which equates to higher Vgs. It is
true that they may not share current equally since you force the same Vgs, but it will certainly be improved over trying to run only one. The two
devices can even share the same heatsink without isolation pads. Particular people could even screen for equal transfer characteristics at the desired
operating point. Yes, you could just parallel two separate cascodes, but this takes four devices instead of three, along with separate adjustments. The
DN2540 can easily run over 100mA, given enough Vds.
I have used up my IC's, so will have to order more in order to test it out, but it seems to me this will work, even three devices.
Thoughts?
worked well at lower currents. What I found, at least with this latest batch of 10M45 IC's, is that the Vgs was very low, around 0.5V for 82mA. The net
result was that the bottom DN2540 was unable to provide this low a voltage, and the 10M45 would not reach the desired 82mA target. So the problem is
that the bottom device was saturated due to insufficient Vds.
This got me thinking about how to provide more Vgs to the top devices. More voltage is also better for the bottom device, which decreases in output
capacitance as the voltage increases.
I have attached a possible design with paralleled upper IC's. Running them in parallel now requires less current, which equates to higher Vgs. It is
true that they may not share current equally since you force the same Vgs, but it will certainly be improved over trying to run only one. The two
devices can even share the same heatsink without isolation pads. Particular people could even screen for equal transfer characteristics at the desired
operating point. Yes, you could just parallel two separate cascodes, but this takes four devices instead of three, along with separate adjustments. The
DN2540 can easily run over 100mA, given enough Vds.
I have used up my IC's, so will have to order more in order to test it out, but it seems to me this will work, even three devices.
Thoughts?
Oh yeah, forgot to mention the most important part: I placed a Bybee in series with the bottom device, and the sound became more 'liquid'.
😉
😉
Why Zeners instead of reverse-biased "normal" diodes? And you might want to put them on the "far" side of the gate stoppers.
Just to protect the gates in both directions- ESD, transients, static conditions, whatever might put them at risk. When used in power supplies I have had these CCS's schmoke before; ever since installing the zeners I haven't had any problems.
I could put them on the other side; not sure it would make much difference either way unless there is concern about oscillation due to zener leakage effects?
I could put them on the other side; not sure it would make much difference either way unless there is concern about oscillation due to zener leakage effects?
I'll move them in the next pcb revision.
So I assume the method with one lower device driving two, three, or four uppers will work, up to the IDSS of the lower device.
So I assume the method with one lower device driving two, three, or four uppers will work, up to the IDSS of the lower device.
Update after a few months:
Finally got around to building and testing the scheme. Works well, to over 100 mA, when the single upper device could only get to maybe 70 mA.
This is good news, as the non-cascoded version would let through just a tad of 120 Hz ripple that I was able to measure. This new cascode should clean up that last bit for a clean PS.
Finally got around to building and testing the scheme. Works well, to over 100 mA, when the single upper device could only get to maybe 70 mA.
This is good news, as the non-cascoded version would let through just a tad of 120 Hz ripple that I was able to measure. This new cascode should clean up that last bit for a clean PS.
- Status
- Not open for further replies.