1000V power supply - advise needed

[AlexResch]

I need to make a ( slightly over ) DC 1000V power supply to calibrate my multimeters. I intend to divide this voltage by 20 and compare with 5 REF102's. Would a voltage doubler do ? Must it be regulated ? etc etc
I would appreciate any suggestions on how to go about this, maybe some schematics.

[Wolfgang]

Hi Alex,

- how accurate does it need to be ?
- What is the output current it should provide ?
- How much ripple and noise is OK ?

Regards
  Wolfgang

[AlexResch]

Hi Alex,

- how accurate does it need to be ?
- What is the output current it should provide ?
- How much ripple and noise is OK ?

Regards
  Wolfgang

Well, it should be accurate enough to calibrate my Protek 608 - 50,000 counts. The Protek has 10MO input impedance - I suppose a couple of mA from the power supply will suffice (?). Ripple and noise I could not say - I suppose people here who know about these things will suggest reasonable values.

[Wolfgang]

... lets see. 50000 Counts means 20ppm of precision.

This means:

- an ultra stable reference (LM399 at least)
- divider resistors should be 10ppm parts, and you need quite some of them due to the voltage range
- Ripple should be below 20mV. Ambitious, you probably need a preregulator/filter, but can be done.
- Output current of 1mA is enough. Your meter will draw 100uA.

If you are serious about the precision, I would buy a Keysight 34465A. Its probably cheaper than to build a calibrator for a 250€ meter that costs a lot more than the meter itself.

Are you serious about 50000 counts *precision* ? Or did you mean *resolution*, and accuracy is a lot less ?

[AlexResch]

... lets see. 50000 Counts means 20ppm of precision.

This means:

- an ultra stable reference (LM399 at least)
- divider resistors should be 10ppm parts, and you need quite some of them due to the voltage range
- Ripple should be below 20mV. Ambitious, you probably need a preregulator/filter, but can be done.
- Output current of 1mA is enough. Your meter will draw 100uA.

If you are serious about the precision, I would buy a Keysight 34465A. Its probably cheaper than to build a calibrator for a 250€ meter that costs a lot more than the meter itself.

Are you serious about 50000 counts *precision* ? Or did you mean *resolution*, and accuracy is a lot less ?

Let me try :

1. I don't quite understand what 50k counts mean. What I would like to achieve is accurate readings to 0.01V up to 2.5V , 0.1V up to around 50V and 1-2V up to around 500V. These are my needs based on the kind of tinkering I usually do . I am not knowledgeable enough to translate those in proper techno-speak.

2. I ordered a couple of '399s. I have 11 REF102's which I could use in series and get 100V with reasonable margin of error ( for my modest needs ). Of course, I will use what you guys advise.

3. I can buy Vishay resistors. Also, I have a HUGE qty of NOS pre WWII resistors - maybe those are well "aged" ?

4. No idea how to build the power supply ! I could probably construct a regulated 40V supply and then use some sort of "converter" to raise the voltage to 1000V , rectify and filter   Really no idea how to do this.

5. Yes, buying a proper multimeter is an option and could do it. But for me this is rather a learning experiment i.e. I am not however discounting buying a good voltmeter - I've seen quite a few on eBay and will ask a couple of questions here in a day or two.

[eliocor]

Suggestion: take a look at this very interesting application note:http://www.analog.com/media/en/technical-documentation/application-notes/AN118fb.pdf

[Wolfgang]

OK, I answer item per item.

1) 50000 counts mean that your meter can display 50000 different values (resolution), NOT to be confused with accuracy. Your meter has 500ppm (parts per million), which is an error of 0.5V in 1000Volts.

2) LM399 is good. There is no need to stack them, because a (precise) voltage divider can bring the HV down to measurable values (several volts) used by the error amplifier

3) Vishay precision resistor are great. Pre-WWII resistors certainly cannot compare with the Vishays.

4) You could make a power supply from a small 1:1 isolation transformer (Conrad has some) with 20VA, plus a voltage quadrupler. This gives about 1400V unregulated.
    A preregulator plus a final precision regulator make 1000V from that.
   I am in the process of building a 1kV PSU (not completed). The preregulator is here:

https://electronicprojectsforfun.wordpress.com/power-supplies/high-voltage-lab-power-supplies/a-1kv-50ma-linear-power-supply/

Much fun, but play safe !!
  Wolfgang

5) Its always better to learn. You will need a good multimeter to calibrate your new reference powers supply, however.

[jeroen79]

Is it necessary to build a power supply just for calibrating your meter?

If you put several lower voltage power supplies in series you can get the desired high voltage.
Maybe borrow a bunch from friends, school, local hacker space, etc.

Then you use a voltage divider, voltage reference and null detector to trim the power supplies to a known value.
This doesn't have to be 1000.000 Volts, anything close to it will be fine as long as you know what it is once trimmed.

Then you measure it with the meter you want to calibrate.
Write down the indicated value besides the known value and you have a calibration.

Also keep in mind that this will only tell you how your meter compares to your reference and other meters you calibrated against your reference.
If your reference or voltage dividers are not what you think they are then you still don't know how your equipment compares to the rest of the world.

[AlexResch]

Thank you guys !!!

[AlexResch]

Is it necessary to build a power supply just for calibrating your meter?

If you put several lower voltage power supplies in series you can get the desired high voltage.
Maybe borrow a bunch from friends, school, local hacker space, etc.

Then you use a voltage divider, voltage reference and null detector to trim the power supplies to a known value.
This doesn't have to be 1000.000 Volts, anything close to it will be fine as long as you know what it is once trimmed.

Then you measure it with the meter you want to calibrate.
Write down the indicated value besides the known value and you have a calibration.

Also keep in mind that this will only tell you how your meter compares to your reference and other meters you calibrated against your reference.
If your reference or voltage dividers are not what you think they are then you still don't know how your equipment compares to the rest of the world.

The Protek 608 is calibrated by giving to it 1/10/100/1000 V , amps, ohms etc and then pressing the Enter key. I don't see how I can go around having a 1000V supply. Initially I thought of putting together a chain of REF102's which if the 2.5mV tolerance is to be believed are more than enough for my needs. But 100 of those is a bit much and too close to actually buying a new meter. . But I could do that with 10 of them and then I could calibrate a 1000V power supply through a 9:1 divider against that. Or at least I hopw it would work....

[matseng]

I might be a bit dense, but why not just calibrate the DMM against two or three REF's so you got a good reading at about 10 volts.  Then take a pile of maybe 15  fresh 9-volt batteries and measure the "exact" voltage on them individually and then sum all the readings.  Stack the batteries up in series and measure on the higher voltage range on the dmm.  If you're only aiming for about a volt of so of error in the higher ranges then I imagine that this would be ok - just make sure the batteries have a stable room temperature and don't handle them too much.

This sounds much easier and cheaper than 1000 volt PSUs and precision resistors.  The requirements seems very modest to me...

[AlexResch]

I might be a bit dense, but why not just calibrate the DMM against two or three REF's so you got a good reading at about 10 volts.  Then take a pile of maybe 15  fresh 9-volt batteries and measure the "exact" voltage on them individually and then sum all the readings.  Stack the batteries up in series and measure on the higher voltage range on the dmm.  If you're only aiming for about a volt of so of error in the higher ranges then I imagine that this would be ok - just make sure the batteries have a stable room temperature and don't handle them too much.

This sounds much easier and cheaper than 1000 volt PSUs and precision resistors.  The requirements seems very modest to me...

For reasons which escape me, 9V batteries are VERY expensive in RSA - aprox $4 each. 15 batteries will make around 135V. That's middle of the road for me - I have decent 300V power supplies in valve equipment and can chain REF102's to around 110V.  To calibrate the 608 on the 1000V scale I need 1000V - no way to go around that one. I speculate that by comparing 1/10 of 1000V with 10 REF102's in series and adjusting accordingly I can get 0.1% for a couple of seconds. Enough time to press the Enter key on the 608.

[matseng]

Ah, so it's for an autocal procedure - that changes things ;-)   Is tere any way of doing the 1000v range manually?

[AlexResch]

Ah, so it's for an autocal procedure - that changes things ;-)   Is tere any way of doing the 1000v range manually?

Nope. I wish it were.

[jeroen79]

The Protek 608 is calibrated by giving to it 1/10/100/1000 V , amps, ohms etc and then pressing the Enter key. I don't see how I can go around having a 1000V supply.

It can adjust itself.

Then the same method would apply except that you do need to trim the power supplies as close to 1000V as you need.
But you still won't have to build something just for an auto adjustment once in a while.

Can you skip the 1000V step of the procedure?
Then you can also do a manual calibration to see if the meter is off.
If it is off too much to your standards then you record the calibration into a conversion table.
(indicated value -> actual value)

[AlexResch]

The Protek 608 is calibrated by giving to it 1/10/100/1000 V , amps, ohms etc and then pressing the Enter key. I don't see how I can go around having a 1000V supply.

It can adjust itself.

Then the same method would apply except that you do need to trim the power supplies as close to 1000V as you need.
But you still won't have to build something just for an auto adjustment once in a while.

Can you skip the 1000V step of the procedure?
Then you can also do a manual calibration to see if the meter is off.
If it is off too much to your standards then you record the calibration into a conversion table.
(indicated value -> actual value)

I'm afraid I don't understand.

I can skip calibration steps - no problem there. But w/o 1000V supply I can not calibrate the 1000V range.  The meter is off for sure - but I can't say how much w/o a 1000V power supply.
I'll read the manual again - quite possible I don't understand something.

[tkamiya]

Another thing is....  is your meter really capable of handling 1KV?  Many meters have range up to 1KV but many are not made to measure that high safely.  Assuming it is, do you even need that?  AND... at that high voltage, do you have any application to need any sort of precision?

I play with tube based Ham Radio equipment.  They have 850V or so in final cage.  So I do have a need.  But I also know if it's 830V or 870V, it won't really make any difference.  To measure this level of voltage, I'd turn the equipment off, bleed capacitor, hook it up, check, hands off, and turn on the power.  It's kind of whole new world once you pass few hundred volts.

If you are *just* calibrating your meter, you'd only need a small power supply capable of supplying just few milliamps.  But regulating such would require good knowledge on how to safely handle high voltages. 

If I were you, I wouldn't bother.  If I really need that, I'd get it done by a cal-house. 

Assume you made it, how would you verify it's 1000V exactly?  Do you have a calibrated anything that can measure that?

[AlexResch]

1. Another thing is....  is your meter really capable of handling 1KV?  Many meters have range up to 1KV but many are not made to measure that high safely. 

2. Assuming it is, do you even need that? 

3. AND... at that high voltage, do you have any application to need any sort of precision?

4. I play with tube based Ham Radio equipment.  They have 850V or so in final cage.  So I do have a need.  But I also know if it's 830V or 870V, it won't really make any difference. 

5. To measure this level of voltage, I'd turn the equipment off, bleed capacitor, hook it up, check, hands off, and turn on the power.  It's kind of whole new world once you pass few hundred volts.

6. If you are *just* calibrating your meter, you'd only need a small power supply capable of supplying just few milliamps.  But regulating such would require good knowledge on how to safely handle high voltages. 

7. If I were you, I wouldn't bother.  If I really need that, I'd get it done by a cal-house. 

8. Assume you made it, how would you verify it's 1000V exactly?  Do you have a calibrated anything that can measure that?

1. Well, that's what the manufacturer says.

2. Yes. Why are you asking ?

3. Absolutely

4. It depends. Sometimes does. sometimes doesn't.

5. I understand. It's not my approach though.

6. No, one needs quite a bit more than a few milliamps.

7. This is somehow patronizing. You are not me and because you did not ask you have no idea why I need to be pretty sure on the actual voltage.

8. I don't need "exactly" - whatever that means. I explained in the first post how. Did you actually read it ?

[TiN]

8. I don't need "exactly" - whatever that means. I explained in the first post how. Did you actually read it ?

First post does not explain how you going to calibrate 1000V source. Division and comparing to REF still need calibrated meter to determine exact ratio of your divider , including voltage and temperature coefficients of it, to estimate the total error of 1000V output. Not even mentioning low voltage reference stability and accuracy, which also must be calibrated. So while it all can be done, based on questions and attitude in this post, sounds like in need of own research quite a bit, before any simple advice can help.

Why don't you start with something more simple, like 20VDC source first. Once that is done and verified, go up to 100V, and repeat same validation. Then perhaps 200-300V... Then you don't need much advice to build your 1kV box.

[vindoline]

You have the classic "chicken and egg" problem here. If you build a 1000V source, how do you calibrate it? With your meter? By far the easiest and simplest solution is to send your meter in for calibration on a regular basis to develop historical confidence in its accuracy and drift rate... No, I don't do that either!

I don't know what your use case or budget are, but I would consider picking up a used Power Designs 3K10B Precision power supply on eBay. I have several I've picked up over the years for under $100 each. Like all the Power Designs precision supplies, these are amazingly accurate and stable. You can set any voltage between 10 and 3000 to the nearest 0.1v!

It will source up to 3000 volts at 10 mA. That's enough to ruin your day, so you need to be careful with it.

[Neilm]

8. I don't need "exactly" - whatever that means. I explained in the first post how. Did you actually read it ?

First post does not explain how you going to calibrate 1000V source. Division and comparing to REF still need calibrated meter to determine exact ratio of your divider , including voltage and temperature coefficients of it, to estimate the total error of 1000V output. Not even mentioning low voltage reference stability and accuracy, which also must be calibrated. So while it all can be done, based on questions and attitude in this post, sounds like in need of own research quite a bit, before any simple advice can help.

One thing not mentioned is the input impedance of the meter used on the divider. This will have a significant effect on the divider ratio. Also, some meters have switchable impedences depending on the sselected voltage range - this causes havok if you calibrate the divider using a different voltage range to the one you actually use.

Also, the accuracy of the REF102 will mean you will have 0.25V error on you 1kV. Is that acceptable?

[David Hess]

Use as little capacitance on the output as feasible which should not be a problem for a calibration source.  I know this from a personally painful experience.

- divider resistors should be 10ppm parts, and you need quite some of them due to the voltage range

Do not forget the voltage coefficient of resistance which can be the largest error term at high voltages.  To reduce or eliminate this, use the same resistor and value of resistor on both legs of the voltage divider so every resistor has the same voltage across it.  This will be inconvenient at high divider ratios but trying to calibrate this source of error out is even more inconvenient.

This also equalizes the power dissipation in each resistor limiting the error from the temperature coefficient of resistance.  Mount the resistors so that they maintain the same temperature as much as possible. For example, use constant lead lengths.

Given the accuracy requirements, I would buy a lot of resistors and grade them for value, voltage coefficient of resistance, and temperature coefficient so these errors cancel out as much as possible.

[Wolfgang]

... another idea would be to buy a prefabricated laser-aligned voltage divider resistor array from Vishay or Caddock.
If you can get one for a kV, they are hard to beat, but not cheap.

[David Hess]

... another idea would be to buy a prefabricated laser-aligned voltage divider resistor array from Vishay or Caddock.
If you can get one for a kV, they are hard to beat, but not cheap.

The common inexpensive decade divider is good to 1kV and 0.1% but if I was calibrating a multimeter which likely uses something similar, I would want better.

[Wolfgang]

... you can go one order of magnitude better, like here:

https://www.rhopointcomponents.com/components/resistors/networks-and-voltage-dividers/caddock-usvd-series

they have 0,01% in ratio. Not so cheap anymore, however.