Range to Range Problem AC mV on HP 3458A

[Hawaii596]

I've had this problem for a while and ended up making the measurements using my Fluke 5790A.

I'm measuring around 20 - 30 Amps AC at 60 Hz using Empro shunt (50 mV / 50 A), which I characterized for AC and have a good charted value.  I'm taking AC mV readings using my nice Fluke 5440 cables (the nicer ones with spade lugs) and connecting guard at the HP 3458A end.  This is a high power AC analyzer that'll output 5 KW or so.

I set my source for about 10 Amps AC and get a very nice reading. with the Empro and HP 3458A.  I read about 10 mV on the 10 mV range.  I've tried a variety of ACBAND settings, SETACV settings, NPLC settings, LFILTER ON and OFF.  Not much difference in readings.

When I try taking my higher current readings (about 20 to 37 Amps = about 20 to 37 mV AC) the HP 3458A uses the 100 mV range.  I get quite a bit higher readings.  For example, if I try my 10 Amp (10 mV reading), on the 10 mV range I got about 10.001 mV.  Without changing anything, I manually upranged the HP 3458A, and the same output reads about 10.628 mV.  When I output about 27 Amps, I get readings anywhere from 30 to 33 Amps.  It is apparent that there is a range to range issue on the 3458A.  If I recall correctly, the input impedance spec on the 3458A is kind of sloppy.  I measured the input resistance (with a Fluke 187) on the 3458A.  I got about 0.88 Megohms on 10 mV range, and about 1.03 Megohms on the 100 mV range.

Anyone have any experience with this issue?  I know input capacitance and so forth are an issue with precision AC measurements.  I also seem to recall that HP never quite mastered some of those AC volts input impedance issues (an old Fluke 8506A might actually be better for such readings??).

Should I try the same readings with my Ballantine shunt box?  Would that isolate the impedance problem better?  This is a two day cal on a very high accuracy AC test system.

Rambling done.  Thanks folks for any thoughts.

[Bryan]

Totally wild speculation here but you are throwing out an alternating magnetic field or something with that AC running through the shunt, any chance the proximity of your measurement equipment to the shunt is causing the problem?

[Hawaii596]

Thanks for that input.

You may well be correct...  I'm sure there are some fields, and I've been considering that as part of my troubleshooting.  But.... What makes me THINK MAYBE that isn't the biggest issue is that I set everything up for 10 Amps (1mV/A shunt), read the 10 mV on the 10 mV range and get a nuts-on reading.  Then without changing a thing (leaving everything set up and running) I manually uprange the 3458A to read the 10 mV on the 100 mV range, and I get some way out reading like 11.628 (example).  So I would think the field interaction would be the same on each range.  I suppose it's possible that with such large fields, maybe some amp circuits on the 100 mV range are more susceptible than on the 10 mV range.

I finally was able to get some believable numbers after a few hours as follows:

(Refer to my original post for further details)

One small detail:  This involves some very large 5,000 Watt resistive load banks.  My 5 ohm load bank has 5 one ohm ceramic resistors about 20 inches long and 3 inch diameter (each) in a steel box about 18" x 24" x 24".  When I do full power tests it heats up to where it is burning hot to the touch.  So there are likely some very strong fields.  I also believe these resistors may be a little inductive.  Made by Milwaukee Resistor Corp. (if anyone's curious).

I played a lot with guarding and shielding.  I won't go into it in detail, but it involved tying various grounds together (single point gnd), grounding the load bank box, etc.  Finally, I connected from the chassis ground on the rear of the 3458A to both of the guard lead connectors  on the two ends of the 5440 lead set.  I did not connect to the 3458A guard connector, I did not connect to any other ground.  Then, I did a regimen for each measurement:
1. set 10 amps (read 10 mV) on 3458A (after a reset).  do LFILTER ON, SET ACBAND 10,3E6, SETACV SYNC on 3458A. 
2.  Read the 10 mV on 3458A.
3.  Do another reset on 3458A and manually uprange to 100 mV. 
4.  Do LFILTER ON, SET ACBAND 10,3E6, SETACV SYNC on 3458A.
5.  Read the 10 mV on the 100 mV range.
6.  Divide the 100 mV range reading by 10 mV range reading (to create a range to range cal factor).

I repeated this process about a half dozen times and got consistent results. 
Then I did my high wattage readings (about 130 Volts at about 35 Amps).   I got pretty good readings.  However, my formula made corrected readings a little low.  And uncorrected readings were a little high.

MY FINAL HYPOTHESIS:  I think it is a combination of range to range input impedance variation on the 3458A and field generation from the load and cabling.  I believe the reason my correction factor is a little off for the higher current readings is the amount of field generated at higher currents bigger and shifts the error.  If I have enough brain left, I think I'll also try moving the meter, cabling and other things around in a way to minimize field interaction (maybe locate 3458A away from the load).  I also seem to recall that DMM's such as Fluke 8506A were renowned for their excellent input impedance and accurate true RMS ability.  Matter of fact, the OEM procedure I am using (inspite of the many flaws - not a very good procedure) calls for use of HP 34401A for most things, but specifies Fluke 8506A for that one detail.  Too bad I don't have one (I wonder of the 8508A is as good in that regard??)

I post later if my hypothesis ends up proving true.

[mdbuike]

O.K. a few things...

The three lowest ranges on the 3458A are the worst spec'd for input impedance (+/- 15% plus some other stuff).

The 3458A is not a true RMS voltmeter, it samples points on the wave shape, which makes it unsuitable for anything but sine waves (courtesy of an Agilent engineer..I'll post his email from work)

Finally, a suggestion from the old PG506 days..put a precision 1.1 Mohm resistor accross the input terminals to improve the input impedance.

I also have an excell spreadsheet with the accuracies for the ranges, you can plug in the numbers, and it will tell you the accuracies (also at work..but tomorrow's Friday, I'll be there).

Good luck with your adventures

Mike

[dallanta]

Although the 3458 is handy, its accuracies in ac are complete crap. I also always use nplc100 on any function if I am forced to use the 3458.  There is too much room for error with this instrument.  If I remember correctly, dcv is the only function that is really useful

[scottbp]

My 2 cents:

1. The HP3458A AC section is as accurate as a shotgun, and to think I used to calibrate Fluke 5500As with it, but I always characterized it reading for reading with a Fluke 5700A.

2. Now we have a Fluke 8508A and a Fluke 5790A. The 8508A has an internal AC thermal transfer standard of the same type used in the 5790A, but not quite as tight as the 5790A.

3. You're using plain old Empro "bar" shunts?? I've never had any success with AC measurements with those, they're only good for 0.25% accuracy. Yeah, they might be OK for the AC output of a welding machine, but for higher accuracy AC measurements, such as calibrating transconductance amplifiers, we have a Ohm-Labs CS-50 50 amp and CS-100 100 amp shunts, which are 0.01% accurate from DC to 100 Hz (0.1% to 1 kHz) and are designed to have very low inductance. We also have a Valhalla Scientific 2575A shunt set that is 0.1% from DC to 10 kHz.

4. For very high current devices, such as big honkin' circuit breaker test sets, we use a JAMB Industries amplifier aided current transformer; ours goes up to 1000 amps at 0.01% accuracy to 1 kHz, but they're available to 5000 amps.

No, I'm not an employee for the companies mentioned above, nor do I get paid to endorse anything, they're just the ones I've had success with.

[WestCoastCal]

Thank god for automation! To be able to use the specs If you don"t have the overlay configuration keys set up as short cuts remember to:
PRESET, NPLC 100, MATHNULL, ±1 °C of Acal ±5 °C of Tcal, . . . Then after you get that all set up, don't forget to add (for DCV) 12 ppm · (Vin/1000)E2 for inputs greater than 100 V. . . Oh and by the way these are 99. 73 % confidence level--don't forget to convert for calibrating a 99. 45 % unit under test, yadda yadda yadda

[mdbuike]

Hawaii, PM me and I'll send you a copy of the spread sheet.  Covers a lot, and if nothing else, clears up the 4:1 TAR

Mike

[RichieRich]

I don't think your issue could have anything to do with range imput impedance because you are working with such a low output impedance from the shunt that your meter input impedance will not cause any problems.  If you were working in the reverse situation say with a high voltage probe (very high output impedance), the input impedance of your meter will be significant. 

I wonder if you may have some harmonic or spurious signals that are being picked up in the higher voltage range that aren't being seen on the 10mV range.  It might be instructive if you filtered for your fundimental.  I wouldn't be surprised to see the range to range difference disapear then. 

[Hawaii596]

Wow.  Some interesting and thought-provoking points above.  I'll try to reply to some of them..

I couldn't agree more that the 3458A AC isn't worth much. 

Regarding the frequency response of my shunt, I would have to discount that as I got right on the money readings on the 10 mV range.  I pre-calibrated with the Ballantine 1620A/Fluke 5700A at 10 Amps and charted DC as well as 60 Hz response.  It was pretty tight at 60 Hz.  My 100A/50mV shunt was 0.5005mV/A at DC and 0.5019mV/A at 60 Hz.  I was using the 50A/50mV shunt for the 37A and it had similar response.  When I tested the unit being cal'd at 10A @ 60Hz, I got pretty tight readings (I believe I calculated 10.0014A).  I did try using my Ballantine 1625A shunt box (similar to the Valhalla), and actually got worse results.

Because things are so tight at 10A and everything goes out the window at 37A (using a 5KW resistor bank), I am leaning toward thinking it is a field issue.  One of the replies suggested something in the higher ACV range on the 3458A may be picking up more field transmission.  That seems pretty believable. 

Next time I do one of these systems, I'm going to experiment with physically isolating the load bank more from the meter and everything else.  If you imagine the big ceramic wirewound resistors with exposed metal conductor fins; the resistors in my loads have "wavy" fins (i.e.: inductive??). 

All I can say is that system pretty well wore me out.  It's gone for this time around.