Need good current sink JFET or MOSFET?

[rlaury]

Hi Group:
I'm looking for a good current sink to use with a 6DJ8 long tail phase inv. I've run test using an NPN transistors, but would like to try a FET. I'm not sure if I see benefits between each. My main concern is saftey (relibility) linearity, and noise. Also what about some of the 3 terminal current sinks. Should I consider them also?
Anyone have some FET part numbers I should consider?
Any response apprecaited.

Thanks!

RonL

[bear]

JFETs will work fine.

The so called "current diodes" are merely jfets strapped to their IDSS value, selected from a given run for the specific value.

You can take any jfet and make a nicely adjustable current "sink".

Afaik, the only thing to worry about is that you don't exceed the JFET's breakdown voltage.

They are generally less noisy than most transistors... much less noisy than most any tube.

_-_-bear

[Miles Prower]

I'm looking for a good current sink to use with a 6DJ8 long tail phase inv. I've run test using an NPN transistors, but would like to try a FET. I'm not sure if I see benefits between each. My main concern is saftey (relibility) linearity, and noise. Also what about some of the 3 terminal current sinks. Should I consider them also?

For a differential phase splitter, best to stick with the BJTs. Since the CCS is in the tail, whatever it does is common mode, and therefore rejected by the amp's CMRR. Since BJTs are higher gain devices than JFETs, they make better current sources. The supplied current is steadier, and more importantly, the effective impedance higher. See This Design for how I implemented the same type of phase splitter. In testing this out, the voltage balance was excellent at an output of 40Vp-p. As for safety, simply use high voltage, small signal transistors, such as the MPSA42 (Fairchild) that I used which I got from Rat Shack in a 15 pack of transistors anonymous.

As for the IC current sources, can you find one that'll work at the design current? If so, then try 'em, but I'm not sure that these will provide any advantage (other than convenience) in this application.

"The so called "current diodes" are merely jfets strapped to their IDSS value, selected from a given run for the specific value"

I avoid those two-legged devils like the plague. Yuck!

[EC8010]

It's very much a "horses for courses" issue. If you're using 6DJ8, then you're probably wanting to pass somewhere between 4mA and 20mA. The most important question is, "How many volts can you afford to drop across your CCS?" If you have enough voltage, then a cascode of bipolar transistors will do very nicely and probably be cheaper than FETS. It will certainly free you from FET device tolerances.

[hacknet]

i was just curious, what is a good solution if my cathode voltage is low, say around 1v. is it still feasible to use a pair of BJTs?

[Tony]

afaik transistors can operate with zero collector volts, but you have to bias the base emmiter with 0.6 volts for it to conduct current. tight call....

[ray_moth]

A transistor operating at 0v collector would be switched hard ON and could not control the current. You need at least 15v to work with. If your LTP is direct-coupled to the plate of the preceding stage, that's easy but from what you say it is not like that.

Another way of getting a higher cathode-to-ground voltage is to raise the DC voltage of the 'grounded grid' of the LTP using a high resistance potential divider from B+. (e.g. 1Meg and 100k would give you 1/11 of B+). Ground the grid through a cap of 0.22uF or so. Connect the other grid to it through a 1Meg resistor and use a decoupling cap as input to the stage.

[rlaury]

Thanks everyone. Good food for thought.
I'm planning on using a -60 tail voltage from the bias supply.
This will insure there will never be 0 volts.