Keysight's new 34465A (6.5 digit) and 34470A (7.5 digit) bench multimeters

[Alex Nikitin]

I left both meters (34465A and 3458A) to run overnight measuring the LTZ1000 reference voltage. The Keysight meter did drift about -3ppm, while LTZ1000+HP3458A combination did show a drift of less than 0.7ppm. However ACAL on Keysight did get it back at <1ppm difference from the HP. It looks like a regular ACAL is a necessity here unless the drift would stabilize after a while. By the way, when I've powered the 34465A for the first time and let it run for an hour, it was about +35ppm off at 10K before I've done ACAL. Here are two graphs from the overnight logs (same scale, 1ppm/div). The Keysight has about 4-5 times larger p-p noise.

Cheers

Alex

[Dr. Frank]

Received the 34465A today. After ACAL it agrees within 1ppm with our HP3458A Opt002 meter (calibrated in January) on my self-made 10K reference and on 7.16V from LTZ1000 . Quite impressive for both meters to agree that nicely.

Cheers

Alex

Hello Alex,
congrats on your new 34465A.
Your 10V / 10k check is to be expected, and proves that it arrived in undisturbed condition.

Check the Test Report of your instrument.. it has been adjusted and verified to < 1ppm uncertainty in DCV and Ohm modes.
As you probably received it within about 3 weeks, and it had not been powered on in between, practically no drift occurred.
So you might take these readings as an additional calibration verification of your 3458A, about +/- 1ppm after ACAL.´

The difference in noise levels of the 34465A vs. the 3458A arise directly from the LM399H vs. their internal  LTZ1000A references.

Your 0.7ppm drift of the 3458A vs. the external LTZ1000 is quite high, maybe your room temperature changes too much; <= 0.2°C change is required.

Frank

[Alex Nikitin]

Received the 34465A today. After ACAL it agrees within 1ppm with our HP3458A Opt002 meter (calibrated in January) on my self-made 10K reference and on 7.16V from LTZ1000 . Quite impressive for both meters to agree that nicely.

Cheers

Alex

Hello Alex,
congrats on your new 34465A.
Your 10V / 10k check is to be expected, and proves that it arrived in undisturbed condition.

Check the Test Report of your instrument.. it has been adjusted and verified to < 1ppm uncertainty in DCV and Ohm modes.
As you probably received it within about 3 weeks, and it had not been powered on in between, practically no drift occurred.
So you might take these readings as an additional calibration verification of your 3458A, about +/- 1ppm after ACAL.´

The difference in noise levels of the 34465A vs. the 3458A arise directly from the LM399H vs. their internal  LTZ1000A references.

Your 0.7ppm drift of the 3458A vs. the external LTZ1000 is quite high, maybe your room temperature changes too much; <= 0.2°C change is required.

Frank

Hi Frank,

Thank you!

1) The calibration date on the 34465A is 30th of June, so over 90 days ago .

2) Yes, according to the calibration data the unit was calibrated to less than 1ppm error on 10V DC and 10K.

3) Yes, the lab is not temperature controlled, with at least 3-4C temperature variation overnight. Also my LTZ1000 reference is on an open board with only a foam cup. I will move it into a proper case soon  . It is probably sufficient for my current needs though.

Cheers

Alex

[Dr. Frank]

..

1) The calibration date on the 34465A is 30th of June, so over 90 days ago .

..
Cheers

Alex

The DCV drift is dependent on the instruments uptime. If it's not used so often, it may as well have its initial uncertainty.
Mine is from about June 2015, I think, and it's still within about 1ppm.

HighVoltage's 34470A, which is running 24/365, showed a drift of +5ppm only, after about 10 months.
This is also typical for the LTZ1000A reference, at 95°C.

The Ohm ranges, as their components are running on RT, won't drift as much, and the 34470A was also found to be within 1ppm for 10kOhm.

So your observation is an indicator, that your instrument is still in very good condition; 3 months of operation is nothing.

Ah yes, you should try to get the 2MB option for free.. I also got it this year, despite mine was an elder one.

Frank

[Alex Nikitin]

Hi Frank,

The DCV drift is dependent on the instruments uptime. If it's not used so often, it may as well have its initial uncertainty.
Mine is from about June 2015, I think, and it's still within about 1ppm.

HighVoltage's 34470A, which is running 24/365, showed a drift of +5ppm only, after about 10 months.
This is also typical for the LTZ1000A reference, at 95°C.

Thank you - I thought that the 34465A has LM399? In my case this unit will be running 24/7, same as the HP3458A. At home I have two K2015 which are also running 24/7 for last four years (and were used for many years before I bought them). LM399s in both Keithley meter are very stable, drifting not more than a couple of ppm per year (if that - I will have a chance to check soon).

The Ohm ranges, as their components are running on RT, won't drift as much, and the 34470A was also found to be within 1ppm for 10kOhm.

So your observation is an indicator, that your instrument is still in very good condition; 3 months of operation is nothing.

3 month of sitting on a shelf at Farnell in this case!

Ah yes, you should try to get the 2MB option for free.. I also got it this year, despite mine was an elder one.

Frank

Yes, I've ordered the extended memory option as soon as I've received the meter!

Cheers

Alex

[TiN]

34470A has LTZ1000, 34465A - LM399 .

I'd leave meter running for at least a week, better two and then repeat this test just to make sure. Same procedure apply for 3458A as well, long cold instruments "lose" their stability and need to run awhile to get back to normal stability. Few cases over EEVBlog nuts already confirmed this.

[Dr. Frank]

...

 Same procedure apply for 3458A as well, long cold instruments "lose" their stability and need to run awhile to get back to normal stability. Few cases over EEVBlog nuts already confirmed this.

Does your statement refer to the regular LTZ1000A references,@95°C oven temperature?
AN-18 might apply generally to these references.

Mine is pimped to about 65°C, as you know, and it does not really show hysteresis, maybe about < 0.2ppm, even after a longer period of idle state.

The other two LTZ1000 @45°C, as already stated, should show no hysteresis at all, as they are 25°C above RT, only.

Frank

[TiN]

Does your statement refer to the regular LTZ1000A references,@95°C oven temperature?

Hm, you definately right. I referenced to SN18 for standard A9, and did not tested on pimped ones. My box runs 24/7 since repair done.

I might have chance to test this theory in neareast future though   

[Alex Nikitin]

Well, the 34465A did drift about 3.5ppm down again over the weekend and returned to <1ppm difference with the HP3458A after ACAL. Here are the graphs. It also looks like a large part of this change is a temperature related drift. I am logging the voltage again just to see if the measured value would go up now when the lab temperature is increasing.

Cheers

Alex

[HighVoltage]

I am logging the voltage again just to see if the measured value would go up now when the lab temperature is increasing.
Cheers
Alex

Nice data on your 34465A and comparison to the 3458A.
Can you also log the temperature during this next test?

[Alex Nikitin]

Can you also log the temperature during this next test?

Unfortunately, no. I will check the temperature occasionally though.

Cheers

Alex

[Alex Nikitin]

OK, the tempco of the 34465A appears to be below 1ppm/C, as the temperature rise from 24C to 25.5C only increased the readings by 1ppm or so, not corrected by ACAL and with the rise of only 0.2ppm measured by HP3458A. Not conclusive  . I'll compare the meters again in a while, running 24x7 for now.

Cheers

Alex

[Alex Nikitin]

It looks like the 34465A is stabilizing. Here is the data from an overnight run. The lab temperature varied by about 3C. ACAL after this run resulted in about +1ppm change on the 34465A and no change on the HP3458A. For that run I've synchronized the meters, by triggering the HP3458A from the 34465A, measuring every 10 seconds at NPLC100 . The vertical scale is the same as before, 1ppm/div

Cheers

Alex

[TiN]

You can poke 3458A with TEMP? command and it will return temperature of it's A1 PCBA for you.
Don't do it each sample though, as it uses relay to switch DCV to internal TEMP signal and then switches back. I usually log TEMP? reading each 50th sample on my runs.

[Dr. Frank]

All three devices are good for stability on the order of +/- 0.2ppm/ 24h.
So you might zoom your measurement to that resolution.
The 3458A should at least show no continuous drift in one direction (on that order), to prove that its resistor array inside U180 is fine.
The temperature change is a problem, so that also should be measured in parallel, you can use TEMP? for the 3458A as well.
(TEMP? tracks RT)

Anyhow, nice measurements under difficult environmental conditions.

Frank

[Alex Nikitin]

Thank you. One interesting observation - a repeated ACAL on the 34465A returns a somewhat different measured values for LTZ1000, with around 1-1.5ppm variations (easily visible on the averaged 7.5 digit value). My suspicion is that variation is due to the LM399 noise, which is about 1ppm p-p.

Cheers

Alex

[Dr. Frank]

Thank you. One interesting observation - a repeated ACAL on the 34465A returns a somewhat different measured values for LTZ1000, with around 1-1.5ppm variations (easily visible on the averaged 7.5 digit value). My suspicion is that variation is due to the LM399 noise, which is about 1ppm p-p.

Cheers

Alex

That may be the root cause.
I also see some variations between consecutive ACALs.

The ACAL mechanism of the 34465A/470A is not as precise as the one of the 3458A, as Scott Stever last year explained, in a discussion about that feature (it also cannot do a full re-calibration, of course).
He estimated the accuracy to be about 5..10ppm only.

So it might as well arise from the basic ACAL-principle.

After correction of the FW, as of vers. 2.14, the mechanism improved, and I estimate / observe the precision to about +/- 1ppm for 10V and 1V, and about +/- 2ppm for 100V, 1kV and 100mV.

I did not investigate on how it behaves under bigger environmental temperature excursions.

Frank

[Alex Nikitin]

The ACAL mechanism of the 34465A/470A is not as precise as the one of the 3458A, as Scott Stever last year explained, in a discussion about that feature (it also cannot do a full re-calibration, of course).

So it might as well arise from the basic ACAL-principle.

The ACAL on 34465A is also very quick, which means it is based on a spot reference value and not on an averaged one. I would probably be happier with a somewhat slower but more consitent ACAL. It is still accurate to 1-2ppm though, which for that meter class (and price) is very good.

Cheers

Alex

[Dr. Frank]

The ACAL mechanism of the 34465A/470A is not as precise as the one of the 3458A, as Scott Stever last year explained, in a discussion about that feature (it also cannot do a full re-calibration, of course).

So it might as well arise from the basic ACAL-principle.

The ACAL on 34465A is also very quick, which means it is based on a spot reference value and not on an averaged one. I would probably be happier with a somewhat slower but more consitent ACAL. It is still accurate to 1-2ppm though, which for that meter class (and price) is very good.

Cheers

Alex

Well, we currently are digging 10 times below the specification limits.. and it's intended to compensate for temperature driven drifts on the order of tens of ppm. It cannot and should not be compared with the 3458As ACAL performance.
That's what Scott Stever also wanted to stress.

Even the 34470A, promoted as 7,5 digits class (!!)DMM, does not benefit accordingly, though it has the less noisy reference inside.
But it works better than suggested by Keysight.. I did not figure out how they realized it, and they didn't want to explain in detail.

Comparing with the possibilities of the other ACAL methods (superlinear A/D and D/A converters, Hamon type, precision transformer), I suggest, that the ACAL in the 465/470A can't be as precise , for physical reasons, even at longer integration times.

[Alex Nikitin]

My first (sort of) disappointment with the 34465A . On resistance (both 2-wire and 4-wire) it is noisy. If you do ACAL, the next few readings usually come with a decent accuracy, however when I've tried to see how stable a 10K resistor is with temperature (as our HP3458A is currently with Keysight UK for calibration), the noise-like variations on the meter readings were so large that any resistor changes with about 3-4 degrees C room temperature variations are completely invisible. Here is a graph with a 2 hour run using my LTC450C 10K "standard" (measured by HP3458 about a week ago and by Keysight 34465A today) . Overnight variations were even larger - about 25ppm p-p. Is this a typical behaviour for the 34465A ?

Cheers

Alex

[TiN]

Since measurement in comparison were taken different time, perhaps in case of 34465 some EMI/RFI coupling to resistor?

[HighVoltage]

The noise seems to be high.
What I have noticed with my 34470A on resistance measurements is noise injected from nearby WiFi, or wireless phones, so I make sure to take these kind of measurements far away from any known noise and test the place, before I start.
 

[Alex Nikitin]

As far as I can tell, it is not EMI/RFI, the variations were present even overnight, with doors closed and nobody/ nothing in the lab to interfere. As a side note - if I do an ACAL during one of the far out variations (as you may see on the graph some large shifts are long lived), than the value returns to accurate, with ACAL correcting up to 10 ppm. Below the overnight (18 hours) run using a different 10K W/W resistor (UPW25). When I've looked at it this morning it was not a pleasant sight, so I've re-run the measurements using the LT450C - with the same result  .

Cheers

Alex

[saturnin]

@Alex

I am just repairing one of my DMMs (an ohms stability issue). As part of tests I have made, I measured short-term stability of my 10K standard (two Fluke 19.995K 0.03% resistors in parallel). The measurement was performed by healthy K2000. Its 24hour spec is 26 ppm (34465A has 25 ppm). Therefore, I would expect 34465A should perform very similar to my K2000 in 2h interval. I have PC, WiFi, and smartphone within 1.5m from my DMM and yet the measured variation of 10K resistance is no more than 2ppm (max. temp change was 0.4°C)...

[Alex Nikitin]

It looks like our 34465A is failing the noise spec on 10K. I'll talk to Keysight on Monday. It would be nice if somebody with the 34465A would repeat this resistance test (10K, 100NPLC, 10 sec logging interval).

Cheers

Alex