IXYS High Voltage current regulator

IIRC IXYS or whoever has at least one app note on the subject even a simian subspecies can understand...

These devices are interesting to me, in that they must use depletion-mode MOSFETs as the pass element. This gives them a free low-voltage power source to run a reference and opamp, to control the MOSFET's gate, and create a constant current.

But, one must ask, why not just use a depletion-mode MOSFET, such as offered by Supertex, to create the constant current, and avoid the integrated reference and opamp. One reason might be to obtain an improved current-source accuracy, but the IXCY' parts are spec'd with a huge nearly 50% part-part variation, plus another 15% tempco variation. It seems the extra integrated elements aren't providing much improvement over ordinary HV depletion-mode MOSFETs, especially when one consideres how much better-than-their-specs the depletion- mode MOSFETs are in practice. Maybe the wide-voltage-range current constancy is where the smart parts excel. But the fact that one can't buy these parts from distributors could be an additional troublesome factor.

Interesting idea: use a Supertex depletion fet to power an opamp and a reference in its own source circuit, which in turn actively regulates the current. That would make a precision, high-voltage, 2-terminal current regulator.

How's this for starters?

| | | | d +-------g | s | | | | | ---+--- | | | | | ldo |-------+ | | | | | |_______| R | | | | | | +--------+-----------+-----------

John

John, if you connect the gate to the ldo's output, you get a common circuit from the application notes that has super-high output resistance and accuracy commensurate with the quality and accuracy of the ldo and resistor you use.

John Perry

You typically don't get a very high -Vgs for the depletion-mode MOSFETs, whereas you need 1.2 volts for the LDO reference, plus the maximum dropout voltage, so I'd move the FET's gate to the LDO's output pin, like this:

high-voltage | input source | d g ----, s | | | ___|___ | Io = 50uA + 1.2/R | | | | LDO |----+--, |_______| | | R | | '-----------+-----------

There are troublesome issues: the required BFC output-stabilizing cap (not nice for a current source), the minimum load current spec, the rapidly-disappearing -Vgs at higher programmed currents, the ugly power dissipation with high input voltages, etc., but it's a cute idea.

Nice touch... gives the ldo more room to work and gets closer to Idss.

John

Sure. But if this stuff were easy, anybody could do it!

John

I must apologize. I can not discern the ASCII schematics. Could you post a pencil sketch on alt.binaries.pictures.radio please? or directly to me. Perhaps I can relate the differences for the future.

I am interested in this solution since IXYS emailed me saying their current regulators are out of production. I use these to replace 50A1 nitrogen filled regulator ballast tubes in Transoceanics.

Thanks, Paul P.

ASCII schematics are done in a fixed-width font. Ways to handle this are:

1) Set your newsreader for a monospaced font like Courier. 2) Cut & paste the text to something like Notepad. 3) Go to Google groups. Append &fwc=1 to the URL or click on **More options** then **Show original**. . . We don't much appreciate top-posters in this group either. Trim and bottom post here.

I see the Amperite ballasts generally work with more than 6V across them (can you confirm for the 50A1), so an ordinary MOSFET could be used with an ordinary LM317 3-terminal regulator to make a ballast-tube substitute, sort of like this:

positive ----+--/\\/\\--, | | d | n-chan g ----+ MOSFET s | | \\_|_ 7.5V ___|___ /_\\ | | | | LM317 |----+--, Io = 50uA + 1.2/R |_______| | | R | | '-----------+----- negative

Time to put stop playing around with theories, and actually make one of these high-voltage medium-power linear current-source beasts to see how they work.

Paul says a 50A1 works at 50mA, so I used R = 25 ohms. I used an IRF640, which is a large-die 400-volt MOSFET, with Rth-JC = 1C/W, which is good for power-handling capability. I chose a 100k gate resistor, insuring more than 10uA of zener current for Vin-Vout of over 10 volts (note that zener diodes rated near 5.5 to 8 volts maintain their breakdown voltages down to amazingly-low currents).

positive ----+--/\\/\\--, 100k | | IRF640 d | n-chan g ----+ MOSFET s | | \\_|_ 7.5V ___|___ /_\\ 1n5236 | | | | LM317 |----+--, Io = 50uA + 1.25/R |_______| | | Rs 24.9 1% | | '-----------+----- negative

This setup started conducting with an Io above 20mA for Vin-Vout voltages above about 5V. This increased to roughly the 50mA design current for Vin-Vout > 6V.

I was testing with a 500-ohm load resistor, and found that I had to bypass the load with a 0.1uF capacitor, to prevent a small (1%) erratic relaxation oscillation with a roughly 300kHz rep rate. Further increase in the input voltage (Vin-Vout from 10 to 50V) caused the output current to be stiffly pegged at 50mA.

At 50V I noticed a little bit of output disturbance, and by 75V there was some RF oscillation underway, and the output current dropped to 46mA.

The propensity of high-voltage MOSFETs to oscillate when used in the linear mode above 50 to 100V is well known. I'll have to report back back later if I find a solution to this problem (or should I say, when).

The 300 KHz thing may be the zener itself; some do that at low current. You might try bypassing the zener, then adding a 47 ohm or so series gate resistor.

John

Le Mon, 27 Aug 2007 18:47:47 -0700, Winfield a écrit:

Even better. I've had some work for which I had to do some tests. Just checked a BZX85C 10V. At room temperature it roughly maintains its voltage down to 200pA (yep, that's pA) and starts falling badly at 100pA.

Here are some figures: Iz Vz

0.2nA 9.25V 0.5nA 9.73V 1nA 9.96V 10nA 10.27V 100nA 10.30V 1uA 10.31V 10uA 10.31V 100uA 10.31V 1m 10.33V

--
Thanks,
Fred.

Those numbers are unbelievable, that's one incredible zener.

How many do you contemplate using? Digikey is left holding the bag on

19,506 IXCP10M45S at $1.76/ea. Why don't you make a lifetime buy of quantity 3. As I understand the Transoceanics thing, the 50A1 can be replaced by a fixed power resistor most of the time, the tube was added on to the design later to provide regulation of a small low current bias voltage across installations and is not needed for any fixed installation line supply.

Le Tue, 28 Aug 2007 08:07:52 -0400, Fred Bloggs a écrit:

Oops, I got the reference wrong. It was a BZX55C10V (Philips).

From the tests I did some time ago, this seems to be quite general. I only tested 500mW zeners though :-)

Here's another one (BZX55C 15V), randomly picked from my drawers: Iz Vz

0.5nA 13.69V 1nA 13.71V 10nA 13.73V 100nA 13.76V 1uA 13.82V 10uA 13.89V 100uA 13.83V (not a typo) 1m 13.92V

--
Thanks,
Fred.

Fred,

I have been looking at the IXCP10M45. What throws me it the "10" meaning

10ma. I perused the spec sheets but could not ferret out if the chip is designed to handle the 50 ma plus of the application. I know some others use this chip in this application (Transoceanic). That leads me to a few conclusions:

1) I am not interpreting the spec sheets properly and I don't fully understand the implementation that I have seen.

2) Using a more common and available product for the application. That is what brought me here.

If the IXCP10M45 is appropriate I would make a bulk (not 19k) purchase and make the tube sub. As for why keep this ballast in place? It is desired of most collectors and hobbyists to modify the chassis as minimally as possible. I recommend to others to store the 50A1 and use a sub. That way the 50A1 keep on the shelf and not burn out. Then it can be said, "I have an original restored Transoceanic" by the owner.

Can you tell if the IXCP chip can handle the 50 ma at 120v rms mains (1/2 wave rectified)?

Here is the application (find the 50A1 on the incoming line cord):

I regret that I am not up on current component technologies. It is just where life takes you. So the help of this group's members is greatly appreciated.

Paul P.

To calm the MOSFET I added a 620-ohm gate resistor, and this worked fine. The output current stays at exactly 50mA over the voltage-drop range of 10 to 80 volts. At 80V the MOSFET is dissipating 4W, which is close to the limit with the small clip-on heatsink I've been using.

However, above 86 volts the output current drops by 3% and even more at higher voltages. An oscillation can be seen at the MOSFET-source and LM317-input node. It's about 1V at 2.5MHz, which is apparently enough to cause the LM317 reference voltage to drop a bit.

Adding a 10nF cap across the LM317 stopped this new oscillation. Now the observed output current doesn't change all the way up to a 200V drop, which I tried for only a very short time, due to an anticipated MOSFET overheating above 80 volts.

Here's the final quiet seemingly-robust circuit:

100k positive ----+---/\\/\\---, 50mA Current Source | | with a 5 to 200-volt IRF640 d 620 | compliance range, if n-chan g --/\\/\\-+ the MOSFET heatsink MOSFET s | is adequate. _______| \\_|_ | ___|___ /_\\ 1n5236 7.5V _|_ | | | --- | LM317 |------+ Io = 50uA + 1.25/R 0.01 | |_______| | |_______| Rs 24.9 1% | | '----------+--+--- negative _|_ 0.1uF --- | gnd

It does lose its appeal as the parts count goes up.

The first number on your tubes indicates filament voltage, they are all in series: 3V4, 1U4, 1L6, 1U4, and 1U5. This totals to 9V ( where "1" is understood to mean 1.5V and "3" for 3.0V, which is inline with various web page descriptions of the filament circuit operation voltage. For best longevity with only a slight degradation of receiver sensitivity, most users keep the total filament drop in the 7.2-7.5V or about 80%, and this gels with original manufacturer filament life data. The 3V4 is rated absolute maximum filament of 100mA with the two halfs in parallel, the Transoceanic puts them in series, so 50mA would be a maximum limit for the Transoceanic circuit. The "50" in 50A1 almost certainly means

50mA regulation. The "A" means A-series. Looking at Amperite AB-51, the first page shows a graph of the typical A-series thresholding at 10V and regulating up to 70V before the current deviates outside the +/-1% regulation window, with a fairly steep rise. The same operating voltage vs life applies to these devices too, so it is clear the reason they are all blown out is because Zenith was running them at too high a voltage. The T600 chassis schematic shows the 50A1 in series with 700 ohms to the filament string. This puts 105-700x0.05-9=60V across the ballast or 85% at 117VAC. At the more usual 125 VAC this drop maybe more like 68V, uncomfortably close to the maximum, and shortened life, say 2000 hours vs 10,000. In addition to replacing the 50A1, it would be a good idea to mount a 1W 7.5V zener across the entire tube filament string. This diode will survive quite nicely and is cheap compared to replacing those tubes. So now all you need is a 50mA constant current, not necessary for spectacular regulation, with operating voltage up to 100V. Because the regulator is on the output filter side of the rectifier, it will not require any transient protection. Expected power dissipation will be something like (115-35)x0.05=4W, which means use a 10W clip-on heatsink for maybe 27oC temperature rise on the outside. The IXCP10M45 can be operated like so, derived from figures 2 and 4 of the IXCP10M45 datasheet Don't think about using this circuit without the 7.5V 1W zener, breadboard it and wire the output into B(-)Bus as an initial test, verifying that the voltage across the 51 ohm resistor in around 2.5V, let it cook a bit under the heat, only then connect the output into the filaments: View in a fixed-width font such as Courier.

. . . . . . from R27 and C41A . | . | . | . ------ . | A | . | | . .--|G | IXCP10M45 . | | | . | | K | . | ------ . | | . | [51] . | | . '------+ . | . | . | . to R33 and C41B . . . . .

Best of luck to you, I know that radio, it's a good one.