# High voltage CC current source design idea

#### Janis59

Joined Aug 21, 2017
1,256
Whole day sitting on the task I got in stupor. Probably You may give a hint in what direction to dig further....
The problem is of need for 100 pieces identical 50 mA CC (constant current) stabilizers, having common ground and + in the regulator side. Working voltage 580V(!)

First my idea was Wilson current mirror, and my Spice model works with that brilliantly, BUT.... if Farnell has no ANY 600V candidate for p-channel mosfet that’s bad sign, and Octopart has only few for ... 18 USD piece, absolute nonsense. While n-channel is plethora for 24 cents a piece. So, seems working with n-channel is mandatory.

Then I turned the circ in all possible ways but only result is common plus and separate grounds what is absolute not permitted. Logically, moving the load from drain to source kills the stabilizing functionality. Adding any emitter follower kills the stabilizing effect by great part too.

So I guess the regulator with current sense element in series with load would be best choice, so forget the Wilson. But even then I cannot to see any other choice as to put load in the source instead of drain - as soon it is drain the common minus is lost what was so much needed (it is actually galvanic bath very body).

So, what may be SIMPLE design ideas for such 600V 50 mA current source?
Probably try the sense resistor upper than load in the source, serially, and TL431 between gate and sense resistors bottom regulates all?? I wonder the www is full of similar circuits on npn bjt, but nowhere I was able to find anyone had used such made on mosfet. Why?

Last edited:

#### #12

Joined Nov 30, 2010
18,217
I don't really understand your question and I don't remember where I saw this. I think it was in a vertical deflection circuit. The idea is to use a low voltage transistor to control a high voltage transistor so you get some kind of goodness of the LV transistor and the voltage capabilities of the HV transistor.

#### Attachments

• 5.6 KB Views: 21

#### Hymie

Joined Mar 30, 2018
945
Could you not use a circuit as simple as this, with appropriate resistor & zener values to achieve the required 50mA. Transistors are available with ratings well over 600V, in theory the transistor could dissipate 29W if the output is a 0V – with additional impedance in the circuit, the maximum transistor power dissipation could be reduced.

#### Bordodynov

Joined May 20, 2015
2,753
See

Joined Mar 10, 2018
4,057
Something like attached. You have to calc Pdiss for Rs and Zener.

Regards, Dana.

#### Attachments

• 12 KB Views: 13

#### Hymie

Joined Mar 30, 2018
945
Something like attached. You have to calc Pdiss for Rs and Zener.

Regards, Dana.
Yup, but Q1 may need to be a darlington transistor (as in Bordodynov’s circuit) if the gain of Q1 is not adequate when the emitter voltage is close to the 580V rail.

#### -live wire-

Joined Dec 22, 2017
912
If you want efficiency, you could make it a SMPS. It switches a transistor (an IGBT is best for this application) rapidly and the output gets filtered. There is some kind of feedback (hall, shunt, or sense) that tells it to adjust the duty cycle.

So here is my idea. It would require you to step it down to low voltage DC for certain control parts of it. It would use a lot of common ICs. You need an IC that generates a precise, high-frequent PWM (150-500kHz) based on a high/low input. The input would be your feedback. It must increase a counter when it is low, decrease when it is high. The counter would control the duty cycle of the output PWM. These are commonly used in DC-DC buck converters. You would have two of these controlling IGBTs that control two series filters. This would give you more precision so that you will actually get close to the desired 50mA. The first one would be the "coarse adj." and the second one would be the "fine adj."

So now what do you use as feedback? Well, you could just use a high-value shunt (10s of ohms) in series with the output. But this alone is a TERRIBLE idea. Let's say you accidentally short it. Or even apply a low resistance load. Now you have 100s of volts across it! This will completely destroy you measurement stuff. So, you need proper protection. Use a clamping zener diode and another series resistor. This way you will NEVER get more than a certain voltage across it. You will want a 50 ohm shunt with a 20V zener and a 500 ohm series resistor. This will ensure the voltage across the shunt never exceeds 20V, protecting your measuring circuitry. This will also allow you to use a zener that is rated for 100v instead of 600V. Alternatively, you could use a hall effect sensor, which would give you more isolation. BUT the accuracy would be pretty bad for such low currents and with all the high voltage EMI. So it is not too recommended.

So then V=IR, meaning that V=50ohms*.050A. That means 2.5V for 50 mA. That means 5V for 100mA, etc. This should go into a comparator (rated more than 20V on the input) that is high if greater than 2.5V, low if not. This should go to both your PWM ICs. Or, have that go to the second one. Then possibly have a Schmidt trigger or something that is high if greater than, say, 4V and low if greater than 2.6V.

Hardware wise, use a high voltage (rated 700V+) IGBT with a parallel capacitor and series inductor with protection. The cap should be at least 20-75uf and rated at least 650V. The inductor should probably be in the uH to mH range. Again, everything should be rated for at least 650V to keep it safe. Or, you can even use an inductor and diode like this if you want higher voltages in the Kv range. Just use this configuration or something similar.

But unless you really need those higher voltages, that is not recommended. It would cause the component values to have to be increased greatly. Increasing cost and creating other problems.

So then have another identical stage (switch, inductor, cap) controlled by the second IC. You will have a common gnd for HV and low voltage for control circuitry, so be very careful!

This is just a rough outline and could probably be improved on. But it should get you a decent efficiency (75-90%) and be pretty accurate. The parts should not cost you more than ~$50 total. Thread Starter #### Janis59 Joined Aug 21, 2017 1,256 RE: #12 Thanks for reply, but reality of the process below is that frequent shortcomes (load shorted), long idles (ie disconnected load) as well as half-loaded states are having place. If the first two means only stress of voltage or overcurrent for F(T) small needle of time, then last means heavy stress for the thermal power. 50 mA*500V=25W. 100x25W=2,5kW in the place where area and cubature is enormously hardly limited. RE: Hymie This is exactly what I thought be possible. Only my idea is to put instead of Zener the TL431 (more accurate somehow), so the first free time I shall make a Spice model to see how it works. Only instead of npn bjt I would use a n-ch mosfet because it are available for 24 US cents at 65W 650V what is unimaginably for any bjt. RE: LiveWire Thank Yo for care but igbt is bit slowish for the task where physical process beyound is sparcles series, so the timing may be the heavy MHz if not hundreds of MHz. So the MOSFETs are more suitable in this meaning. And more, how far I understand the Octopart stands absolutely over any of Elfa, Farnell, Mouser, etc etc companies if question is about rarities or expensive things. However for simple, typical everyday purchasing anything what is coming from America is damn unhandy. For example, last year I bought some very rare and expensive transistors APT6038, what would cost me at least few thousands if to obtain in Europe, so I bought it somewhere at America at 5 USD piece of about 100 pieces. The result - I got the DHL paper that they stay at Riga contor for each day costing me 50 Eur, but they are declining to make any custom clearing. I had done that clearing hundreds of times, but in this case it is prohibited by law, that must be a accredited custom declarant. What have that licence - dhl, but they deny to do it. So few days while I found an another firm, week with a cheek while they cleared, and logically 21% VAT tax was imminent, so together I spent near the 1000 Eur for just clearing this mess out. And I bet it would cost the rather similar money to clear out even the 3-cent purchase. Why I tell this? Because Farnell sends in the purchases cleared of taxes, but never the Mouser. One is seem curious to me, I never had such problem with China purchases. Last time we bought a machinery worth of 65 000 Eur from them in weight of 7 tons, and even that they managed going over boarder as cost-free sample. RE: Danadak Thanks... I never was using Nexperia.... time to start it to get the taste. I always used a Farnell as its search engine is order of magnitude better as other, and if there is no results then Octopart what has everything but searcher is bit skew-made. Often one may demand a p-ch and get the bunch of offers what opening the datasheet is n-ch instead, or ask a 600V and get the 60V, or written the price 0,1 USD but is 100 USD etc etc. RE: LiveWire and SMPS Undoubted the SMPS is a best choice in near all life cases. But it is seriously more complicated, so if I must produce some 100 such pieces there is difference if I have stupid simplicity or not. Other what I am aware is sparkling determined fast changes between zero load, shorted load and normal load conditions, what may happen that driver may not understand what is that the heck. Thus, I think the extreme efficiency is not the goal per se, however thermal flux still is. When I shall mount the first sample and shall experiment a bit in the real galvanic bath let the sparcles jumps in and off, then I shall be able to comment on real thermal flux, this moment I had only quess about maximum maximorum 25W to channel. Actually, that is microsparkle technology of rare-metal oxidizing Thread Starter #### Janis59 Joined Aug 21, 2017 1,256 RE: Live Wire <<But it should get you a decent efficiency (75-90%) and be pretty accurate. The parts should not cost you more than ~$50 total>>

Wow, You are joking. I bought a few years ago those bit unstable but anyway usable MC34063 tablets for about 5 USD per 200 pieces!!. And cheap Cheeneeze SMPS units already soldered costs between 10 cents and 50 cents depending on whats inside. For 23 USD I purchased some 20 pieces of smps of 27V and 25Amp. Wonder but them are rather stable and still after of year exploitation all are alive - but I am using em for 100 MHz generator feeding so their life is very dangerous. Thus I cant say that all what is China made is garbage, but for 50 USD.... thats a minor village there at them may be fed for years and years long

#### Janis59

Joined Aug 21, 2017
1,256
RE: Bordodyunov
Its exactly what I meant by TL431 only instead of npn the n-channel fet. However, the solution with Wilson in left upper corner is eliminable, its may be substituted by resistor. At such high voltages it is stable enough but simpler. KISS rule - keep all things stupidly simple.

#### Bordodynov

Joined May 20, 2015
2,753
RE: Bordodyunov
Its exactly what I meant by TL431 only instead of npn the n-channel fet. However, the solution with Wilson in left upper corner is eliminable, its may be substituted by resistor. At such high voltages it is stable enough but simpler. KISS rule - keep all things stupidly simple.
I calculated that the load of this power source is very different. There may be one LED, or maybe as I have shown - 150 white LEDs. I agree, for one LED is enough and a resistor. But then one powerful resistor is enough!

#### waker

Joined Feb 21, 2018
12
Excellent and proper Idea

#### DickCappels

Joined Aug 21, 2008
6,941
A completely analog solution is only reasonable if your can afford to dissipate about 30 watts. Switching solutions don't have that problem -at least not that much of a problem.

#### Janis59

Joined Aug 21, 2017
1,256
RE: DickCapells
Yes, as more am I meditating on this problem as more I am afraid of the thermal flux. But I have no experience over high voltage SIMPLE smps. I have made a much the full H and half H bridges, have made a simple low voltage circuits, but then how to make the driver IC will not burn? Optos? Maybe. Actually, if there is some rather simple CC circuitry, I would evaluate the con and pro.

#### DickCappels

Joined Aug 21, 2008
6,941
There are some very simple, cheap, and efficient LED driver ICs for that current range. The ones I saw when looking a few years ago were for use between 1000 and 240 VAC. There might be such parts for the kind of voltage you are concerned with by now.

I guess you could look at adapting some of the single chip power supply solutions that were designed for lower voltages to use your higher voltage.

https://ac-dc.power.com/products/