*Tricky* current measurement.

"Fred Bartoli" wrote in message news:45e46c61$0$7501$ snipped-for-privacy@news.free.fr...

Hello Fred,

This will only help a bit because these people make high current sensors:

Ultrastab 867-200I transducer.

200A full scale but errors and noise in the 10 ppm range .

If you wind more turns round it the sensitivity goes up but new errors may come in. If it worked with 100 turns it would be 2 A full scale so errors in the 20uA region which is 1% of 2mA which would be fine if you could afford

2A full scale current (only 8kW !! :-) With your 40mA at 4kV max you are looking at an equivalent voltage error of 2V which is about 50 times outside your budget.

However you may get some ideas from what they can achieve and the way they do it - you just need a 4A full scale device and you're there.

Good luck.

Michael Kellett

What's wrong with a current transformer? DC capacity?

Maybe you could mix a limited-bandwidth (less noise!) Hall effect signal with the transformer's signal output?

Tim

-- Deep Fryer: A very philosophical monk. Website @

Tim Williams a écrit :

A current transformer alone won't do. You need two, to compensate for the sensed wire parasitic current to shield/gnd. And then you still have to be very cautious: with a total 10p across the 100K load, you still need less than 10fF parasitics imbalance to achieve the 60dB CMRR. Can't go much over 10pF there because dV/dt are 20-50V/ns which is already 500mA. Still, 100pF requires less than 0.1pF imbalance, easing things a bit, but at a 5A switching transient cost.

Otherwise it's more or less what I was thinking about, but in a better way: if you just mux then you have some gain accuracy pb at crossover and most probably some non linearity due to the core hysteresis. Using FB to close a current compensating loop around the core takes care of most of this. But Hall sensors are... noisy. Just think: A 20mm dia toroid, mur=10K, will only give 8mT flux density at my

4kV/40mA peak current. But 1% of 60dB down is 100dB, so I want less than 80nT rms noise in the bandwidth in which the Hall sensor will take over, not accounting for the higher frequencies amplifier part. And CT, even shorted, have limited LF response so I can't make the Hall sensor branch crossover too low.

Oh, and I forgot to tell: 10MHz -3dB response, but that part is POC with CTs.

--
Thanks,
Fred.

Why not use an amplifier and ADC, all floating at the high voltage, and then clock out the output into a fibre-optic link. The ADC and fibre optic transmitter (and probably a microcontroller) would need some power but whether that is difficult depends on constraints of cost etc. (e.g. solar cells + bank of infrared LEDs can transfer a few milliwatts, or for more power a carefully constructed transformer.)

Chris

------------------

sounds like Massive Prong.

except for the above, I'm all out.

--

Bye.
   Jasen

ROTFLMAO!

Cheers Terry

Perhaps a comapany called "Ion Physics" has some CT type sensors that can do something close to what you want.

--
 JosephKK
 Gegen dummheit kampfen die Gotter Selbst, vergebens.  
  --Schiller

Some work at SLAC was done in that regard years ago; i forget even the order of magniture that they were able to achieve, but suffice to say they pushed existing technology a fair amount. Application? Measure the (pulsed) beam current.

The 16-bit measurement rate would be under, say, 200kHz? Why does the CMRR spec go to 10MHz, or are you saying that's an overall spec that can be met in part with a sharp signal filter?

My bet is that the spec is coming from a nice anti-cooperative working environment.

A (probably unexpectedly) sold a rough prototype, with as much carelessness designed in as possible (IOW a "demonstrator"), to B under the pressure of C (if you B still want to sell us in future, go have a look at what A do) without clear understanding of all the implications. Then D got a contract to ('just') industrialize the product. Now that A sold its s..t and that B almost don't understand how it works (I mean the fine important details), specs are flying over in an attempt to making all that falling in working order and also probably to cover one's ass as much as possible. And as A and B almost don't speak to each other anymore, and A refuses to speak to D, and B pretend to not being able to disclose some information, then having the real info is, hem, well hem... And this is taking interesting proportions.

So, despite that I think the sought CMRR is totally irrelevant we have to comply to what we are given. So yep, the CMRR has to be as asked, not a by product of some after processing.

For the pb I think I've now come up with a nice working solution but it probably won't see the light since B has now decided to engaged into the

2 x compensated dividers way and some other 'interesting' problems are pointing their nose.

--
Thanks,
Fred.

[ snip ]

Sheesh!