# Safely Measuring HV DC

1. Sep 1, 2009

### Jdo300

Hi Everyone,

This thread is a modified continuation of my "Digital Voltage Meter circuit " thread I started a long time ago here:

But with a few changes. In this case, I have a large DC capacitor that I want to monitor the voltage on. This capacitor can hold up to 1000VDC across it and I am trying to come up with a simple and safe way to measure the voltage with a microcontroller for processing purposes.

The sensor circuit I want to design has to have a bandwidth of about 50-100kHz and be isolated so that if something goes wrong, my whole board won't go up in smoke! Below I have attached a diagram of how I *think* it could be done. I wanted to get some opinions on this to see if anyone here has any ideas on how this could be done more simply (if even possible).

I did find a 'voltage transducer module' here: http://search.digikey.com/scripts/dksearch/dksus.dll?Detail?name=398-1019-ND which seems like it would work perfectly for my application, but the price made me want to cry :yuck:

It would be great if there was another canned solution out there that was much more affordable, but assuming that there isn't anyone have any other ideas?

Thanks,
Jason O

File size:
10 KB
Views:
72
2. Sep 1, 2009

### Staff: Mentor

You would need to use 500V resistors to just put two in series to stand off 1kV -- more practically, you put a lot of real resistors in series to add up to the total resistance, and each resistor is still within its voltage specifications. Otherwise you get arc-throughs or arc-arounds.

A pretty inexpensive and simple solution is to use a HV probe as your safe monitor:

http://img.directindustry.com/pdf/r...e-probe-meter-differential-probe-50951_1b.jpg (the middle one)

They do cost $70-$100, though, so I'm guessing you want to do something less expensive.

3. Sep 1, 2009

### Staff: Mentor

Also, I don't think that transducer is necessarily for measuring high voltages. It looks to be intended to measure regular voltages, and transmit that measurement info across a boundary between an AC Mains domain and an SELV domain (if I'm reading the datasheet right).

4. Sep 1, 2009

### Jdo300

Hi berkeman,

Yeah good point there about the resistors. I did find a 10MOhm resistor on Digikey that could handle up to 3.6kV across it (though I'm not sure I believe them just from seeing the datasheet).

I also found yet another problem that I didn't see originally. Even if this circuit works ok and doesn't arc out on me at high voltages, the resolution would be very bad in the low voltage range. I'm planning on using a 12-bit ADC to convert the votlage value to a digital output that I can clock across the isolation barrier but I'm concerned about the output voltage. With less than 20V on the cap, the output voltage of the divider would be less than 100mV. Under normal conditions, this would be fine but I'm a bit concerned about noise issues since this sensor circuit would be located on the analog side of my board where it could be subjected to noisy switching transients.

I have heard of sensors using a 4-20mA current loop to help mitigate the issue, but any traces coming from the resistor divider to the op-amp and ADC probably wouldn't be longer than a couple of inches at most so I'm wondering if I'm just over thinking this... What do you think?

- Jason O

5. Sep 1, 2009

### Staff: Mentor

Yeah, high impedances like that can cause capacitive noise pickup issues. Why do you need the high bandwidth? Look at those resistances by themselves, and you'll see that with a straight resistor divider, you are not going to get much bandwidth.

If you want a big divider and good bandwidth, you need to use the oscilloscope construction trick, with lead caps around the input resistor(s). For good signal integrity, you will need some way to tune the capacitors (for 1-up circuits, you could just solder caps in parallel to trim their values).

Beyond that, I'd be inclined to put a uC on the HV side of the iso barrier, and just communicate with it via serial comm through standard optos. That's easier than having to try to get ADC clock, data, strobes, etc., across the barrier.

6. Sep 1, 2009

### Jdo300

Hi berkeman,

I'm basically using the voltage sensor circuit so that I can regulate the output voltage on the capacitor. It's basically like a HV flyback converter circuit only the capacitor output can fluctuate fast (maybe 50-100kHz is way overrating it). I mentioned the high bandwidth so that I could get an accurate reading from the capacitor. Basically, it will be periodically discharged into a load circuit and I need to monitor the ramp up curve on the cap and keep it within a specified voltage.

What kind of bandwidth can one typically get out of just a straight resistor network? Also, in case you're wondering, I need to be able to accurately monitor the output voltage on the cap so that I can keep the votlage below a digitally preset value.

- Jason O

7. Sep 1, 2009

### waht

The resistor network won't give you much bandwidth. How about using a small toroid step down transformer to monitor the AC transients, and a resistor network to monitor the DC voltage?

8. Sep 1, 2009

### Jdo300

Hi Waht,

Sounds like an interesting idea, though I think that it would be impractical for lower voltage measurements. To handle the bandwidth issue, I'm leaning more towards the scope probe type setup. But then again, how low is low for the resistors? Low Hz? 100's of Hz? I inevitably will need to filter the input voltage somewhat, so, depending on the response time of the resistors, I may be able to get away with not using AC coupling of any kind. After giving this some further thought, I may only need an input bandwidth of about 10kHz max but I won't know for sure until I do some bench testing in that regard.

- Jason O

9. Sep 2, 2009

### Mike_In_Plano

This isn't so bad. You need a large value, high voltage resistor, but Ohmite makes them, and Mouser stocks them.

As for the frequency response, that's easily handled as well. Place properly valued capacitors across each resistor in the divider. For example, given a simple resistor divider comprised of R1 and R2, where R1 is attached to high voltage, R2 is attached to ground, and R1 is connected to R2.

Place C1 in parallel with R1 and C2 in parallel with R2. As long as the following equation is met:

C1/C2 = R2/R1

the response will be flat. The only difficulty is that C1 tends to be a low value and the slight parasitic capacitance across R1 can throw the circuit off. Then, you have to add a little capacitance to C2 to get the response flat.

You might try R1=5Meg, C1=270pF, R2=20.5K, C2=68nF for about 4 volts FS.

I used to tune mine with a regular signal generator (square wave) and scope. The signal loss is huge, but still manageable.

Best Luck

- Mike

10. Sep 2, 2009

### Mike_In_Plano

One other thing, kinda important, C1 should be rated for 2Kv or more ;)

11. Sep 2, 2009

### waht

Last edited by a moderator: May 4, 2017