Products > Test Equipment
Keysight's new 34465A (6.5 digit) and 34470A (7.5 digit) bench multimeters
<< < (60/145) > >>
Dr. Frank:
Finally, here's my story, up to now..

I got my 34465A end of June, GPIB and DIG options.
Very nice instrument, compared to my 34401A from 1993.. especially the comfortable graphical interface and all these direct access / display features. Big win ! Also see other post about 2 very mighty digitizing measurements, I already did with this instrument (DCV and DCI).
 
Now to the bugs I very soon discovered.
The calibration date had been 3 weeks earlier, 2nd June, with very freshly calibrated Fluke 5720A.
As the instrument already had been produced in January, Keysights calibration policy required a 2nd calibration run before delivery, as 3 months had passed in between (No orange sticker any more!)

Normal calibration is exactly 76, would have been 152 for my instrument, but it came with 222!
Very odd already.

I always check new instruments on receipt.
Have to say, that I own a 3458A, and a Fluke 5442A, which both are uncertain to about 1ppm in DCV, on all DCV ranges .. otherwise I wouldn't have had the chance to find these bugs..

OK, I checked 100mV, 1V, 10V DCV. All three ranges were in accordance to my equipment within < 1ppm!
With each instrument, there's a "Certificate Of Calibration" and a verification Test Report on the backside, which contains the read-back of some of the just-before calibrated modes and ranges.

These DCV ranges were also spot-on during verification, as expected, i.e. less than 1ppm.
So were 10KOhm and other modes.

2nd odd thing now was, that the divide-by-hundred HV ranges, i.e. 100V, 500V and 1kV were all off by 6 .. 14ppm.
(Again, the 3458A and the 5442A both do exact 10V => 100V transfers , < 1ppm uncertainty, by auto-calibration!!)

That already caught my eye, as such an instrument should not drift so far off, in 3 weeks time, and without any possible voltage / power burden  in between, when all the lower DCV ranges were spot-on..
The 100:1 divider should also be stable, at least it should not drift by 16ppm.. that would bust the 90d limit as it would bust the 1yr. limit.

Anyhow, I now re-calibrated the HV ranges only, by applying 100V and 500V, as described in the manual.
After storing the cal constants, the instrument displayed exactly 100.00000V, 500.0000V and 1000.0000V, within 1ppm (the 34465A displays 7 digits, if using statistics math function), as expected, and as already checked during verification at Keysight.

Temperature is constant to a few tenths of °C in my lab, during short term measurements, by the way.

As soon as I triggered the ACAL function, the very same calibration offset as before was introduced, i.e +9ppm for 100V, +16ppm for 1kV range. I repeated that several times, but very obviously , the ACAL does not work correctly in the HV ranges. It irreversibly busts the calibration.
Specification is at least violated, as ACAL claims to improve the T.C. from 5ppm/K to 1ppm/K.
Also the 90d limit will be violated, if additional drifts occur.

Changing the inner temperature lead to even more crazy ACAL readings.
I strongly assume a firmware error . The exchange instrument, I received  a few days ago, is also spot on in the low DCV ranges, but off by around 10ppm on the HV ranges. (Already FW 2.09)

The 2nd bug is accessible to everybody, and confirmed already on both 34465A, and also on a 34470A.
(All TueVolt DMMs share the same FW 2.08 or 2.09)

If you apply a constant voltage on the HV ranges, like 100V on the 100V range, change the impedance setting from AUTO to 10MOhm.
The instrument is definitely fixed in the 10M divider mode, so there is absolutely no change in the hardware setting.
Anyhow, the reading will increase by +3ppm, reversibly.
Only explanation is a firmware bug also.

3rd bug is a mechanical one.
The lever of the front / rear switch is much too flexible. The plastics is too soft or too thin, so it bends by engaging the Shadow switch.
In my case, releasing the switch to front input again, did not feel fine, and the switch often was hooked.
The occurrence of this error depends on the spring force of the switch, I think.
Typical mechanical design fault.

Anybody else might also check these faults on his own instrument, even though the first fault is only evident to 34465A/34470A owners, and who were capable of checking 100V and 1kV calibration very stably / precisely.

I expect an analysis from Keysight, and hope for confirmation and repair of the assumed faulty firmware.

I have to say, that the reception of this assumed fault by Keysight is very good (starting with kind assistance by Mike Kawasaki), and the support extremely alert, as I already got another new instrument, and they wish to analyse the "defect" instrument in their development in Malaysia.

Frank

 
saturation:
Thanks Dr Frank, this issue would be difficult to catch without a higher accuracy reference. 

Thanks to you everyone will benefit.  Its good to hear Keysight is very responsive to questions, I had the same experience with the 1272a DMM bug when it was first introduced.

I found one range off spec on the 1272a but the Keysight calibration/adjustment method allows you to adjust one range only unlike Fluke DMMs, you need to calibrate the whole meter to do just one range.  I also had a 1252a with the same issue, but a different range.  I presume these were calibration errors only as the values have held true ever since they were adjusted years ago. 

Always good to break in new instruments and do a performance test!




--- Quote from: Dr. Frank on August 09, 2015, 07:40:28 pm ---Finally, here's my story, up to now..
...
Anybody else might also check these faults on his own instrument, even though the first fault is only evident to 34465A/34470A owners, and who were capable of checking 100V and 1kV calibration very stably / precisely.

I expect an analysis from Keysight, and hope for confirmation and repair of the assumed faulty firmware.

I have to say, that the reception of this assumed fault by Keysight is very good (starting with kind assistance by Mike Kawasaki), and the support extremely alert, as I already got another new instrument, and they wish to analyse the "defect" instrument in their development in Malaysia.

Frank


--- End quote ---
LabSpokane:
Keysight surfs the EEVBlog and will likely ping Dr. Frank in short order today.  :)
Dr. Frank:
Hi,
I recently reported errors with my new 34465A, which I discussed intensively with the Keysight DMM support engineer, Jim Durr.

In the meantime, Keysight exchanged my instrument for analysis. The new instrument showed similar irregularities.

Jim Durr already confirmed, that the +3ppm shift between impedance AUTO and High setting is a real defect in the firmware, affecting all TrueVolt DMMs.

About the 2nd assumed error, caused by the ACAL function, I finally had a phone conference today, with Scott Stever, previous PM of the 3458A, and now also involved in the 34465A/470A development.

He explained to me the characteristics of the ACAL function.
(It's clear that he could not go into details, how this is realized in hardware.)

This function is able to make small corrections to the gain constants of volt and ohm modes, obviously by re-measuring the ratio factors.

It works very well in the 10V range, here the process is "precise", as he told me.

For the the HV range, as there is this big 100:1 attenuation, the function has some limitations in the repeatability of the ACAL function.
One of the limitations is caused by the noise level.
Obviously this ratio is determined in the fab (or during calibration) at the 100V level, output 1V, whereas during ACAL, the output is in the 100mV range or smaller, causing higher noise and lower repeatability of ACAL.

So the repeatability for the HV ACAL function is about 5..10ppm only, which is well inside the 90d specification limits, and also safely included in the 24h specification.

Keysight generally estimates the operation of the ACAL function being at around 1/5 of the corresponding 90d specification.

I also made a partial recalibration of the 100V and the 1kV ranges only, which is not applicable, as the ACAL relies on the complete calibration procedure, in the correct order.

Therefore, this observed irregularity is not an error, but a sign of limitation of the ACAL function.

It also can not be compared at all with the "real" ACAL of the 3458A, which relies on the ultra linear A/D converter (0,02ppm INL), and also is capable of calibrating all modes and ranges. So, my expectations maybe were too high, especially being experienced with the 3458A.


Scott Stever assured me, that the 34465A / 470A perform much better than their predecessors, and that the ACAL feature greatly improves the T.C. in all ranges, so that in summary, the specification is superior over the older instruments.

Anyhow, ACAL is not needed on daily basis, but only when the room temperature changes by more than about 5°C.


What I have seen so far, after nearly 3 months, the instrument is still spot on in the 10V and lower ranges, and also the deviation in the 100V / 1kV range against the Test Report of about 8ppm is constant.. each parameter is still well below the 24h limits.
The ACAL will always bring the reading to 1ppm of the initial ones, so it's performing much better than Scott Stever explained to me. Same goes for the Ohm ranges, which are still at 1ppm accuracy.


In retrospective, the people at Keysight really have done a great job on my issue.
They really have taken care of my concerns, even when I insisted on explanations about the ACAL peculiarity.
They gave a great support, including a new instrument!

A new FW for the +3ppm failure will be released in the next days / weeks, it's currently undergoing their tests.

Frank
saturation:
Wow, very insightful and most helpful, Dr. Frank.  Thanks for all the work to bring this to light, a 65a or 70a is one of my next purchases so this was good to know.



--- Quote from: Dr. Frank on October 12, 2015, 07:03:58 pm ---Hi,
I recently reported errors with my new 34465A, which I discussed intensively with the Keysight DMM support engineer, Jim Durr.
...A new FW for the +3ppm failure will be released in the next days / weeks, it's currently undergoing their tests.

Frank


--- End quote ---
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod