How badly matched are your IXYS10M45S ?

[Will]

I bought a bunch of these to use as CCS.

Without prior screening for matches, I took 2 units and got it connected from a 270VDC,

1Kohm grid stopper and 1Kohm trimpot (2 legs shorted together) from G-K,

fed it to a Sovtek 6922.

The 6922 Cathode is Green LED wired to ground, developing about 2V on both sides of the cathode.

I tuned the trimpot so that anode voltage on both sides are 100VDC output, and found that the trimpot got pretty warm to touch after 5 mins. I guess for long term reliability I'd better replace it with a permanent resistor.

Now, once powered down, I measured the resistance of both the trimpot and got 272 ohms and 239 ohms respectively.

Is that how badly matched the 10M45S can be within a batch of purchase?

[Arnulf]

Just out of curiosity (since you're blaming the semiconductors): the two halves of your tube are perfectly matched ?

[Will]

That's a valid point and I haven't checked; just that my Sovtek 6922 are hardly used and I just took them out from my storage bins to test out. They are not NOS stuffs so the assumption of relatively good matching on both triodes are high.

[Arnulf]

NOS means "unused old stock" rather than "perfectly matched" And the former doesn't mean anything as far as matching is concerned - most dual tubes will be grossly mismatched (by a difference of 10+%), new, old, doesn't matter. Using fixed bias (or a LED) only makes this more apparent. Not that it matters much in audio where one section is generally in each channel and difference of 10-20% in audio volume is negligible ...

Unless your tubes were perfectly matched your testing method is invalid. If you want to match your current sources for whatever reason, use fixed resistors in your jig and leave the tubes out. If it was me I would be more concerned with getting similar THD and voltage swing rather than complicate my life by trying to match parameters that might ultimately be irrelevant

[kmtang]

It is important to measure the actual current through the 6922 tube. It can be measured through the cathode resistor, ie, V/R.

Then, you can tell if the current is right or there are difference in the tube. I usually use the same resistor for the 10M45 assuming the IC is more precise and stable.


Johnny

[sawreyrw]

The data sheet only gives one spec related to this. It says for Rset=300 ohms, the current will be between 7 and 15 mA with a typical value of 10 mA. Obviously, this is a very wide tolerance.

[kevinkr]

Why does anyone assume bogey parts bought in a batch are well matched? You'd need to make a jig and sort them for transconductance or current if you want a really close match. (or make the CCS adjustable)

An ideal CCS should deliver about the same current to the load regardless of the load's effective resistance once the voltage across the device is some level above its minimum compliance voltage. Ideally it should not and does not depend on the parameters of the active device in its load circuit. (This is far from an ideal device, but again above the minimum required voltage it should be pretty predictable...)

These appear to be basically high voltage mosfets, have you looked at the data sheet? You will see that the output current with a bogey resistor between source and gate can vary by -30% to +50% of design nominal..

Each device needs to be trimmed for the specific operating current you desire.

[kevinkr]

Originally Posted by Will

I bought a bunch of these to use as CCS.

Without prior screening for matches, I took 2 units and got it connected from a 270VDC,

1Kohm grid stopper and 1Kohm trimpot (2 legs shorted together) from G-K,

fed it to a Sovtek 6922.

The 6922 Cathode is Green LED wired to ground, developing about 2V on both sides of the cathode.

I tuned the trimpot so that anode voltage on both sides are 100VDC output, and found that the trimpot got pretty warm to touch after 5 mins. I guess for long term reliability I'd better replace it with a permanent resistor.

Now, once powered down, I measured the resistance of both the trimpot and got 272 ohms and 239 ohms respectively.

Is that how badly matched the 10M45S can be within a batch of purchase?

Note in this case you trimmed the current to result in 100V of plate voltage so in effect you do not know the actual operating current - you need to measure it. You will find the current in the two sections of your 6922 probably doesn't quite match due to differences in the transconductance of the two triodes. In addition add the uncertainty of the transconductance of the CCS itself and you have no idea what the current is.

You can make it easily settable by inserting a 1 ohm resistor in series with your bias leds or you can substitute matched precision resistors (or use the same resistor) for the 6922 and set for the voltage drop required for the desired current. Choose a resistor value that won't cause excessive dissipation in the CCS however. Were it me I would add the 1 ohm resistors which would make it easy to check over time for drift.

Use sufficiently large resistors in the CCS to prevent heating from causing drift in the set current. Adequate heatsinking of the devices themselves is a must.

[Michael Koster]

If you're curious, measure the exact voltage across each trimpot and knowing the R you can calculate the standing current in each side.

But I don't know how important Vgs matching could possibly be in ordinary CCS circuits. If you had a commercial build and needed a certain current range with a fixed value resistor, maybe you would need to know the spread and do some selection.

One thing I've noticed is that 10M45 seem to deviate from the published curve for I vs. Rset, needing smaller Rset values than indicated on the datasheet as the desired current I is increased.

PC mount trimpots are rated at about 1/4W dissipation, for the whole element. You are using 1/4 the resistance so need to derate to 1/4 power (about 0.06W), and your Rset at ~12mA should be dissipating about 0.03W. I use trimpots for Rset up to about 10-15mA and found that they are reliable. It's better to place the adjustment in the middle or towaerd the upper end of the full value for more dissipation headroom, e.g. for 12mA use a 500 ohm trimpot.

[TheGimp]

slight variations in the voltage across the green LEDs will also contribute to variations in bias. You may be (in part) compensating for this difference by changing your current source value to force the plate voltages to the same value.

Lots and lots of variables.