I've got a low power 801 SE amp that is resistor loaded and cap coupled to the OPT. Thinking of playing around with a CCS plate load. This will double or so the power (going from 0.5W to a massive 1W). I have some DN2540's but they are only rated to 400V and my supply is at about 625V. Can I use two FET's for the top of the cascode to divide the voltage load, something like this: Don't much mind the resistor values, I haven't done a calculation yet.
Attachments
What about to drive a high voltage N Mosfet with your fet?
You can place the mosfet over the fet, and thus cascode them.
It you can provide a fixed voltage (10V) to the Mosfet gate you can easily drive it with the DN2540, preserving the good fet characteristics while handling up to thousand volts.
I've used IRFBC40 mosfet, but you can use for example IRFBG20 or IRFBG30 (1000V).
This is a schematic that should work fine.
With this schematic the voltage across your jfet would never be more than 6-7 Volts.
Better again would be to use an external supply for the fixed voltage on the Mosfet gate; you can use a battery or, as I done, a well regulated low power supply (I used a spare 6V heater winding from y power trafo).
I never tried jfets (I always used Mosfet over Mosfet), but I think I would give them a try in the next future.
Best regards,
Giovanni
You can place the mosfet over the fet, and thus cascode them.
It you can provide a fixed voltage (10V) to the Mosfet gate you can easily drive it with the DN2540, preserving the good fet characteristics while handling up to thousand volts.
I've used IRFBC40 mosfet, but you can use for example IRFBG20 or IRFBG30 (1000V).
This is a schematic that should work fine.
With this schematic the voltage across your jfet would never be more than 6-7 Volts.
Better again would be to use an external supply for the fixed voltage on the Mosfet gate; you can use a battery or, as I done, a well regulated low power supply (I used a spare 6V heater winding from y power trafo).
I never tried jfets (I always used Mosfet over Mosfet), but I think I would give them a try in the next future.
Best regards,
Giovanni
The DN2540 is a depletion MOSFET. Biasing for depletion MOSFETs is similar to biasing for tubes and JFETs. AFAIK, JFETs are strictly small signal devices.
While genuine experts will have the last word, I believe max. performance from Constant Current Sources for plate loading is best achieved with "P" type semiconductors. OTOH, "N" type semiconductors are "best" in Constant Current Sinks for use with cathode followers and differential circuit blocks.
While genuine experts will have the last word, I believe max. performance from Constant Current Sources for plate loading is best achieved with "P" type semiconductors. OTOH, "N" type semiconductors are "best" in Constant Current Sinks for use with cathode followers and differential circuit blocks.
Have to disagree. I've gotten terrific performance from the DN2540 cascode as plate loads. I don't have Gary Pimm's measurement capability, but in theory, the sel-biased DN2540 cascode has better performance than a p-mode plate load since the p-mode will have a parasitic capacitance to ground via the collector-base (or drain-gate) capacitance of the device closest to the tube. Since that's usually the "big" device of the pair, that capacitance could be significant in some situations.
Channeling Brian Beck, I can also point out that this collector-base capacitance is modulated by the signal voltage. Now admittedly, I've never seen any evidence that this is a problem in my real-world measurements, but with high rp tubes and low currents, the effect could certainly show up.
Channeling Brian Beck, I can also point out that this collector-base capacitance is modulated by the signal voltage. Now admittedly, I've never seen any evidence that this is a problem in my real-world measurements, but with high rp tubes and low currents, the effect could certainly show up.
Good, no one gagged on the concept, so I take it as a sign that it is doable. Croccodillo's suggestion, as I understand it, is to replace the top two 2540's with a single mosfet, and bias that with a zener. Would there be any performance benefit/loss as compared to the two 2540's? Looking at parts I have in the box, though I could probably pick up some appropriate Mosfet's at the local surplus.
I also have some Lovaltech LU1014D. Suppose I could use those on the bottom?
Sheldon
I also have some Lovaltech LU1014D. Suppose I could use those on the bottom?
Sheldon
The only benefit in using a Mosfet in the upper side is the increased voltage range.
I never "heard" a depletion mosfet in audio circuit, so I cannot say if there are differences in sound.
What I can say is that if well designed a cascode mosfet CCS will sound very well.
My impression is that joining a dpletion mosfet togheter with a power mosfet one can realise a really good and powerful circuit.
Ciao,
Giovanni
I never "heard" a depletion mosfet in audio circuit, so I cannot say if there are differences in sound.
What I can say is that if well designed a cascode mosfet CCS will sound very well.
My impression is that joining a dpletion mosfet togheter with a power mosfet one can realise a really good and powerful circuit.
Ciao,
Giovanni
I'm not even sure that's the case. Gary's site is down at the moment, but he showed dynamic impedance versus frequency for several different CCS configurations connected as plate loads. Best was his IRF820-based unit. Second best was the FAR simpler self-biased DN2540 cascode, and it was some ridiculously high number.
There's another alternative: connect the drain of a DN2540 to the cathode of a triode, and then connect the triode gate to the DN2540 source.
Then, bias and use the DN2540 in the usual way.
Of course, the triode plate should be connected to B+.
In other words, replace the upper mosfet with a servo-triode (sorry for my bad english, I hope the idea is clear).
This would preserve the good from the DN2540 and add a tube philosophy the the whole CCS.
I tried it with a KT88 placed over a N-channel mosfet (IRF640), fixed biased with 5V and with a source resistor of about 33 Ohm (about 100mA). It worked fine (at least, for my ears).
Ciao,
Giovanni
Then, bias and use the DN2540 in the usual way.
Of course, the triode plate should be connected to B+.
In other words, replace the upper mosfet with a servo-triode (sorry for my bad english, I hope the idea is clear).
This would preserve the good from the DN2540 and add a tube philosophy the the whole CCS.
I tried it with a KT88 placed over a N-channel mosfet (IRF640), fixed biased with 5V and with a source resistor of about 33 Ohm (about 100mA). It worked fine (at least, for my ears).
Ciao,
Giovanni
Jeb-D. said:
It's not the power, it's the voltage limitations I'm concerned with. I think that the 10M45 is has a rating of 450V, and I have a B+ of 625V. The CCS wouldn't actually see a full 625V, but it could see to 450V or more on transients. No sense courting failure.
No one has scoffed at the original proposal, so I assume that it would address the issue. But, I'm always open to better ideas, and more importantly, the discussion helps me (and I hope, others) learn more about the topic. And of course, it's always nice to have good options.
Sheldon
mach1 said:
Busy, busy...
I haven't yet seen a single-transistor CCS design that would decisively eliminate the voltage-variable capacitance (whether Ccb or Cdg), at least on paper. If a single transistor is used, at least it should be biased with very high Vds (or Vce) to reduce the absolute value of this capacitance and to reduce its percent variability per volt. As SY alludes, when used with a low-rp tube, a single transistor chosen for low C and biased with high voltage may be quite acceptable to many ears, because the variable phase shift is reduced within the audio band. But it seems to me that to dramatically reduce capacitance, compound devices like cascodes or more elaborate solutions like some of Pimm’s would be needed. Pimm showed several CCS designs with equivalent shunt capacitance pushed down to well below a picofarad.
I haven’t tried the 10M45 yet. I assume that it is a compound device, perhaps even using internal gain and feedback loops. George (Tubelab) likes them. But the 10M45 data sheet is silent on its AC characteristics.
Fascinating.
I have searched for a while on low capacitance BJT's and my best solution was to use the MJE 3439 (current sink, not source). It is rated for 10 pF maximum at 10V, 1 MHz. Cascoded with an MPSA18, rated 3pF collector to base. I thought I was getting optimal devices.
The other dilemma I see is whether or not to limit the voltage across the sink transistor, dissipating the heat in a wirewound resistor instead. One argument is that the resistor somewhat shields the amplifier circuit from the Ccb capacitance. The other argument is that a higher voltage across the BJT will lower the effective capacitance.
It would appear that the capacitance is lowered by about 50% between a devices effective operating range, possibly as low as 25%.
Has anyone compared the effect Vce has on AC impedance?
I have searched for a while on low capacitance BJT's and my best solution was to use the MJE 3439 (current sink, not source). It is rated for 10 pF maximum at 10V, 1 MHz. Cascoded with an MPSA18, rated 3pF collector to base. I thought I was getting optimal devices.
The other dilemma I see is whether or not to limit the voltage across the sink transistor, dissipating the heat in a wirewound resistor instead. One argument is that the resistor somewhat shields the amplifier circuit from the Ccb capacitance. The other argument is that a higher voltage across the BJT will lower the effective capacitance.
It would appear that the capacitance is lowered by about 50% between a devices effective operating range, possibly as low as 25%.
Has anyone compared the effect Vce has on AC impedance?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.