EZ Sink: Depletion-mode MOSFETs and the LM334
Posted: Fri Oct 19, 2012 5:56 am
I'm working on a new 300B amp which will use a DC-coupled cathode follower driver. The benefits of a constant current sink here are substantial. The higher resistance will let me swing more voltage before my driver starts to draw grid current and really starts distorting. A CCS lets me adjust the current easily without changing the DC-coupled value from my voltage gain stage to feed a hungry 300B as needed. And of course, there is added linearity.
Depletion-mode MOSFETs like the IXYS 10M45S and Supertex DN2540 have really been boon for DIYers the last ten years or so. They provide a simpler active plate load versus BJT based CCSs, as they self bias with one resistor. Walt Jung published two landmark articles in 2007 in AudioXpress magazine. With their high voltage rating and more than adequate performance, these have been used in many tube DIY designs very effectively.
The main problem I have with using these in a design - particularly a DIY design - is that these need a trimmer pot to adjust the current. The trimmer needs to be adjusted to ensure minimum current on power up, and then the user then needs to adjust the current with a multimeter with the unit powered up (add one more resistor at this point - maybe a 100 ohm for sensing). FETs can die pretty easily plus you have exposed HV on the drain tab when you are fiddling with this adjustment.
The simple solution is to cascode with a LM334Z. The LM334Z gives us precision control at currents 10mA and less, and just needs 1V to 'turn on' (though at 10mA, I think we can swag this to be closer to 1.2V). The biasing region for the DN2540is about 1.8V to 2.2V (for all the currents 10mA and less). The biasing region for the 10M45Sis about 2.9V to 3.5V (for all the currents 10mA and less). We can even skip the "gate stopper" resistor, as the gate will be directly connected to ground. Since the LM334Z controls the current, you can co back and forth between these two different depletion-mode devices without re-biasing. So with three parts, we have about a perfect sink.
Here's my tests at 9mA comparing AC impedance. Note that the Supertex has better high frequency performance as recorded by Mr Jung in 2007.
IXYS 10M45S (300 ohm current set resistor)
20Hz 4.5M
1kHz 7.2M
20kHz 460K
100kHz 95K
IXYS 10M45S cascoded with LM334Z (8.2 ohm current set resistor)
20Hz 60M
1kHz 21M
20kHz 1.15M
100kHz 250K
Supertex DN2540 cascoded with LM334Z (8.2 ohm current set resistor)
20Hz 160M
1kHz 12M
20kHz 1.9M
100kHz 1.75M
Research Links
http://waltjung.org/PDFs/Sources_101_P1.pdf
http://waltjung.org/PDFs/Sources_101_P2.pdf
http://www.diyaudio.com/forums/tubes-va ... value.html
Shannon
Depletion-mode MOSFETs like the IXYS 10M45S and Supertex DN2540 have really been boon for DIYers the last ten years or so. They provide a simpler active plate load versus BJT based CCSs, as they self bias with one resistor. Walt Jung published two landmark articles in 2007 in AudioXpress magazine. With their high voltage rating and more than adequate performance, these have been used in many tube DIY designs very effectively.
The main problem I have with using these in a design - particularly a DIY design - is that these need a trimmer pot to adjust the current. The trimmer needs to be adjusted to ensure minimum current on power up, and then the user then needs to adjust the current with a multimeter with the unit powered up (add one more resistor at this point - maybe a 100 ohm for sensing). FETs can die pretty easily plus you have exposed HV on the drain tab when you are fiddling with this adjustment.
The simple solution is to cascode with a LM334Z. The LM334Z gives us precision control at currents 10mA and less, and just needs 1V to 'turn on' (though at 10mA, I think we can swag this to be closer to 1.2V). The biasing region for the DN2540is about 1.8V to 2.2V (for all the currents 10mA and less). The biasing region for the 10M45Sis about 2.9V to 3.5V (for all the currents 10mA and less). We can even skip the "gate stopper" resistor, as the gate will be directly connected to ground. Since the LM334Z controls the current, you can co back and forth between these two different depletion-mode devices without re-biasing. So with three parts, we have about a perfect sink.
Here's my tests at 9mA comparing AC impedance. Note that the Supertex has better high frequency performance as recorded by Mr Jung in 2007.
IXYS 10M45S (300 ohm current set resistor)
20Hz 4.5M
1kHz 7.2M
20kHz 460K
100kHz 95K
IXYS 10M45S cascoded with LM334Z (8.2 ohm current set resistor)
20Hz 60M
1kHz 21M
20kHz 1.15M
100kHz 250K
Supertex DN2540 cascoded with LM334Z (8.2 ohm current set resistor)
20Hz 160M
1kHz 12M
20kHz 1.9M
100kHz 1.75M
Research Links
http://waltjung.org/PDFs/Sources_101_P1.pdf
http://waltjung.org/PDFs/Sources_101_P2.pdf
http://www.diyaudio.com/forums/tubes-va ... value.html
Shannon