New Keithley DMM6500 and now DAQ6510

[macaba]

I cannot say I am disappointed but perhaps I was expecting a little better than this

What in particular could have been better? The peak to peak noise looks good to me.

I did some testing with a voltage reference (VRE102CA) that I've confirmed to have no popcorn noise and a much lower 0.1-10Hz noise than a typical LM399.

2 images attached.

5 NPLC, no filter, AZ off

Looks good, there appears to be a hint of popcorn noise from the onboard LM399. (Edit: popcorn doesn't show up in further captures, probably just a random occurrence)

5 NPLC, no filter, AZ on

While it's well within the expectations of a 6.5 digit meter, I would say there's a slight bug here as shown by the periodic waveform. Looking at the CSV sample data, it appears that the time for each sample is 2x 5 PLC + 2.09ms (202.09ms). I presume that 2.09ms is the total mux settling time for the AZ mode, but I'd say the duration should ideally be 2 PLC, 1 PLC before each phase. Just a guess though, the cause could be something else. Does the DMM6500 synchronize to actual AC frequency? (Edit: I would guess it might; time between samples in non-AZ mode corresponds to 50.016 Hz)

I did find a software bug where if you set up a waveform trigger, use it, then clear all the triggers, you still can't change the range to 'Auto' (gives an error about being in analog trigger mode, despite the triggerflow model being empty).

[E-Design]

Does the DMM6500 synchronize to actual AC frequency? (Edit: I would guess it might; time between samples in non-AZ mode corresponds to 50.016 Hz)...

Yes it does.

[Vgkid]

I cannot say I am disappointed but perhaps I was expecting a little better than this

also the resistance value never quite settles, the last two digits (only) keeps moving around that average

Thats rather good. Your results are not unexpected. On the voltage, you are looking at microvolt level input voltages. Rather hard to measure with good stability without specialized equiptment. IE a nanovoltmeter, or microvoltmeter. On my 34420a you can watch it increment nanovolts as the machine warms up.
With the resistance you have good results. When it comes to measuring ohms that low, you need to force in more current, with a high gain(running a pulsed dc current to cancel out the thermal offsets), or inject an ac signal, also with a high gain.

[Kleinstein]

In the non AZ mode, there can be additional popcorn noise from the ADC or amplifier part, not just the reference. The non AZ result looks quite good. For use with the high speed digitizing mode good performance also without AZ is kind of needed.
Part of this could be the configuration of the input stage in the modern Keithley meters (from K2000 on): they use a chopper stabilize buffer (amplifier in some cases) at the input and the auto-zero mode only comes after this. This can also explain the slight offset even in AZ mode. It needs an adjustment for the input part from time to time.
I am not even so sure the jumps are true popcorn: they look a little aligned to a 3 second grid, a little like the spikes in the AZ mode.

In contrast most HP/Keysight meters use the AZ mode all the way to the front and have no extra chopper stabilization at the input. So there the non AZ mode also adds extra LF noise from the input, but also reduces the higher frequency noise of the input - so the difference between AZ and non AZ mode is larger there.  Both solutions have there pros and cons.

The periodic background in the AZ mode looks a little odd, though still at a low level. I don't know for the DMM6500, but with other Keithley meters the AZ mode is or seems to be more than just the simple difference between an input and zero reading. There is also some measurement of the gain and zero readings seem to be averaged over some time. The period of some 3 seconds may fit the cycle time there. A reference measurement instead of a zero may effect the next signal reading and thus cause a periodic glitch.

One may call this a design decision or in some sense also call it a software bug: it helps a little with the shorter time scale noise and measuring the ADC gain (at least done with the old meters and K2001 and likely also DMM7510) helps with a stable gain. However it also comes at a price: it adds to noise on a longer time scale (e.g. 100 PLC). I don't know if it's a problem for the DMM6500, but it is a weak point with the DMM7510 and 2002.  Ideally the user would have the choice between a simple AZ mode (lower noise at 100 PLC) or the current mode (slightly lower noise at 1 PLC).

[macaba]

Thanks Kleinstein. having built a successful integrating ADC that you helped me with earlier this year, that all made sense to me!

I agree that it would be ideal to have more choice over the low level workings but I suspect fixing the egregious software bugs would deliver more value at this point in time.

analogRF - I've attached my results of your experiment. By this point, the meter has been switched on for around 20 hours, perhaps you might want to repeat in a few days.

[analogRF]

Thanks Kleinstein. having built a successful integrating ADC that you helped me with earlier this year, that all made sense to me!

I agree that it would be ideal to have more choice over the low level workings but I suspect fixing the egregious software bugs would deliver more value at this point in time.

analogRF - I've attached my results of your experiment. By this point, the meter has been switched on for around 20 hours, perhaps you might want to repeat in a few days.

yeah this is more like what I expected...I'll try it again for several hours this time but it seemed it was pretty stabilized when I posted my results

probably the environment around is also important, isn't it? or the AC power line? but my lab is not noisy though...

[KedasProbe]

Because the DMM6500 is sensitive to overload
Usually one of the first remarks of new users even if they don't know that is the reason.

I thought lets measure it and see how small it can be to trigger an overload. (and compare it with my DM3068)

I did a few measurements and there is a very big difference that is clearly not due to input capacitance or BW differences.

Test method:
Set the DMM to 1PLC DCV 100mV and sent very short square pulses (<10ns rise time) up to 200mV to it, 1 every 10ms so: 2 square pulses every 50Hz power cycle. (The coax was terminated with 50 ohm before the DMM input) (DG4102 used as source, burst mode)

DMM6500:
When the overshoot was increased the required pulse width to trigger overload was lower (seems logical)
Example measurement:
170mV square (70mV above max range) of 4.05µs triggers the overload  (Pulse Duty Cycle = 0.04%)
130mV square (30mV above max range) of 9.1µs triggers the overload

DM3068:
Doing the same on the DM3068 (200mV range) was a wast of time I basically had to increase the pulse until the average value was too high: 80% DC of 300mV square pulse or 8ms high 2ms low * 2 in 1PLC  (100Hz)

So we are talking in the range of a factor 1000 difference between both DMM.
Since we are comparing 'µs' spikes while doing an 1PLC DCV measurement 'ms'
I'm assuming that the fast digitize ADC or another very fast ADC is keeping track of the min. max. input.

Maybe E-Design can elaborate on this design difference?

Not sure if this is a blessing or a curse when doing 1PLC measurements. I would think just average those spikes out (analog), but maybe you want to know they are there although you decided not to use the fast ADC.

[macaba]

Thanks for revealing those results KedasProbe... it is a bit of a curse I'd say. A switchable LPF on the input would have gone far here. As it is, I'll probably make a small LPF PCB in a form factor similar to the 4 wire short PCBs.

Edit: I felt it important to clarify to E-Design that there is no ill intent in this post, the design is a work of art in the vast majority of it.

[E-Design]

We have heard this feedback before about the sensitivity to the overloads. Thanks for the comments!

Design for it either way has trade-offs and this graphical digitizing DMM leans more towards faster input response vs only capturing DC levels. Some users want to see as true of signal as possible, others only DC.  Unfortunately, a switchable built-in input filter wasn't in the cards for cost (and calibration / specs  / test) reasons. Its easy to add one externally if needed. If you find your signal has noise on it,  you can try to go up one range and likely still have acceptable / usable readings.

[E-Design]

I could check into finding some typical test data if its useful to you?

Yes please, it's useful to me by virtue of educational value, and it's probably useful for others to add to DMM comparison charts. I may do my own INL testing (mostly for the educational value to learn procedure, etc) so the data will give me a broad benchmark for what the output from my testing should look like.

Typical linearity of a 65xx instrument. Of the ones I looked at, this one has a very common shape. The data is represented as error / allowable error.

[exe]

Typical linearity of a 65xx instrument. Of the ones I looked at, this one has a very common shape. The data is represented as error / allowable error.

Wow, that's by far better than most ADCs I've heard of. As I understand the max non-linearity is 1ppm, and is typically below 0.5ppm. Do you have similar data for dmm7500? I wonder how the two compare to each other.

[Kleinstein]

The linearity data look indeed quite good. The file has a few odd labels, saying ppm where the data are µV, but the graphs seem to be OK again ,giving µV.

If the data where recorded in just one run, so all 10 readings for each voltage done in a row, the curve may show additional effects from reference drift and possibly popcorn noise from the DMMs reference.  So not all the error shown may be INL.

On the other hand the data are separate for the positive and negative side - so there may be some glitch near zero that is not caught in the test.  A problem here is that even a Fluke 5700 calibrator as source may show some errors when switching sign - so it makes sense to separate the curve in two.

[E-Design]

Thanks for revealing those results KedasProbe... it is a bit of a curse I'd say. A switchable LPF on the input would have gone far here. As it is, I'll probably make a small LPF PCB in a form factor similar to the 4 wire short PCBs.

Edit: I felt it important to clarify to E-Design that there is no ill intent in this post, the design is a work of art in the vast majority of it.

Yes it can be a curse sometimes..   

 if you do make an input filter....  I would suggest using film capacitors for any caps (polypropylene) and not use too large of series input resistor to preserve the noise floor. Also consider not only the differential (HI-LO) filter but also a common-mode filter to earth as well. By balancing the impedance from HI-EARTH and LO-EARTH you will be able to filter out common mode noise sources which will improve your measurements if you have floating setups. The input capacitance from HI-Earth and LO-Earth is naturally unbalanced, so you can equalize the impedance by your filter component choices.

If you are doing low resistance measurements, filtering the SLO and SHI pathways likely wont help much because  the noise on the 1 Ohm range is dominated by the input protection resistors (internal) just FYI...

[E-Design]

The linearity data look indeed quite good. The file has a few odd labels, saying ppm where the data are µV, but the graphs seem to be OK again ,giving µV.

If the data where recorded in just one run, so all 10 readings for each voltage done in a row, the curve may show additional effects from reference drift and possibly popcorn noise from the DMMs reference.  So not all the error shown may be INL.

On the other hand the data are separate for the positive and negative side - so there may be some glitch near zero that is not caught in the test.  A problem here is that even a Fluke 5700 calibrator as source may show some errors when switching sign - so it makes sense to separate the curve in two.

This example isnt the best nor the worst I have seen but fairly typical.

Yes the Flukes do glitch and there is extra care taken to avoid those switching / ranging scenarios.

Additionally, sometimes we get just a random flyer reading from noise or who knows what (unattended) and it causes a very bad data point so we routinely sweep through it again looking for systematic errors related to the linearity and not random events. It can take  a long time to cover the whole range adequately so these tests have been automated.

[macaba]

Does anyone know what the tiny vertical lines on the chart trace are supposed to indicate?

[exe]

Does anyone know what the tiny vertical lines on the chart trace are supposed to indicate?

Do you still get them with leads disconnected or shorted? If not, then they come from environment, i.e., noise . If they still present... well, may be it also comes from environment, or may be not.

[Kleinstein]

The thin lines look a little like glitches, however they are suspicious equal in size, so there may be some internal, possibly software reason behind them.  They are quite small in in many normal applications would not be noticed. 

[The Bootloader]

I recently received a brand new DMM6500. I made a super basic unboxing video with my first impressions on the unit :



It's a great instrument, I love it. I hope this video will help others trying to decide if they want one or not.

[Mike G]

Hi, I have had my 6500 for nearly 2 years now and have a question please.
It has always been a bit twitchy, blue screens, locking up etc but I persevered as I hoped firmware updates would resolve these issues as many others seemed to be experiencing similar problems.
I upgraded to 1.7.3c a month or so ago and this seemed to solve the prolems.  I was really having trouble with the meter just halting readings while in continuous mode , meaning I would return to the lab to find no readings had been taken
However, I still have one issue remaining which is a real nuisance. If the meter does lock up (I accept that while writing scripts it may get into a muddle sometimes)  and a power off reset is performed it often hangs on the first Keithley logo screen. The only way I have found to escape this is to turn off the meter (the power button still functions) and have to remove the mains power for a while (about 30 minutes usually does it but on rare occasions may need longer) this is a major problem which the latest firmware has not helped with.  Any advice or similar experiences would be interesting to hear about.
A quick thank you to E-Design for offering inside support and help, it was because of Brad's input way back that made me choose this meter.
Mike

[E-Design]

Hi, I have had my 6500 for nearly 2 years now and have a question please.
It has always been a bit twitchy, blue screens, locking up etc but I persevered as I hoped firmware updates would resolve these issues as many others seemed to be experiencing similar problems.
I upgraded to 1.7.3c a month or so ago and this seemed to solve the prolems.  I was really having trouble with the meter just halting readings while in continuous mode , meaning I would return to the lab to find no readings had been taken
However, I still have one issue remaining which is a real nuisance. If the meter does lock up (I accept that while writing scripts it may get into a muddle sometimes)  and a power off reset is performed it often hangs on the first Keithley logo screen. The only way I have found to escape this is to turn off the meter (the power button still functions) and have to remove the mains power for a while (about 30 minutes usually does it but on rare occasions may need longer) this is a major problem which the latest firmware has not helped with.  Any advice or similar experiences would be interesting to hear about.
A quick thank you to E-Design for offering inside support and help, it was because of Brad's input way back that made me choose this meter.
Mike

Yes, fortunately the recent firmware releases have cleaned up a lot of bugs. However, I am sorry to hear you can still experience a lockup. Needing to leave the meter off for 30 minutes or more to resolve seems odd and makes me think it might be related to the meter being warmed up.

Even though it may be intermittent, Is it the same script running when you experience it?
Do you know the date of manufacture (or purchase) ? There could be other possibilities if it is always a hang at the boot screen.

[Mr. Scram]

I recently received a brand new DMM6500. I made a super basic unboxing video with my first impressions on the unit :



It's a great instrument, I love it. I hope this video will help others trying to decide if they want one or not.

I'm surprised you're so positive about the fan noise. It's easily the most annoyingly instrument in my lab. It's reasonably loud but above all it whines. I do have a comparatively early model so maybe the fan was upgraded since. Don't get me wrong, it's a great meter. I just feel the fan is a letdown.

[Mike G]

Hi, I have had my 6500 for nearly 2 years now and have a question please.
It has always been a bit twitchy, blue screens, locking up etc but I persevered as I hoped firmware updates would resolve these issues as many others seemed to be experiencing similar problems.
I upgraded to 1.7.3c a month or so ago and this seemed to solve the prolems.  I was really having trouble with the meter just halting readings while in continuous mode , meaning I would return to the lab to find no readings had been taken
However, I still have one issue remaining which is a real nuisance. If the meter does lock up (I accept that while writing scripts it may get into a muddle sometimes)  and a power off reset is performed it often hangs on the first Keithley logo screen. The only way I have found to escape this is to turn off the meter (the power button still functions) and have to remove the mains power for a while (about 30 minutes usually does it but on rare occasions may need longer) this is a major problem which the latest firmware has not helped with.  Any advice or similar experiences would be interesting to hear about.
A quick thank you to E-Design for offering inside support and help, it was because of Brad's input way back that made me choose this meter.
Mike

Yes, fortunately the recent firmware releases have cleaned up a lot of bugs. However, I am sorry to hear you can still experience a lockup. Needing to leave the meter off for 30 minutes or more to resolve seems odd and makes me think it might be related to the meter being warmed up.

Even though it may be intermittent, Is it the same script running when you experience it?
Do you know the date of manufacture (or purchase) ? There could be other possibilities if it is always a hang at the boot screen.

Hi E-Design, thanks for responding.
I bought it from Farnell UK in early April 2019 S.No 04415206
There is no consistency regarding when it is going to freeze, or for what reason. All I can say is that, apart from script lockups, it has the habit of occasionally just stopping the continuous measurement usually sometime during the first 30 minutes after power on in the morning, sometimes different button presses and random combinations will allow it to start but not able to discern any definite procedure. This is not the serious problem, the main gripe is when a power off reset is made it often hangs on the opening Keithley screen before even the relay clicks.  I am a repair tech so am more than happy to investigate if you can offer any information. Leaving it in the off condition but with power applied doesn't work, must be unplugged from the power.  One possible clue is that usually  it has just crashed when I perform the reset, if I turn it off then on from working state I don't seem to have the boot up problem
Thanks for any help you can offer.  Mike

[KaneTW]

I recently received a brand new DMM6500. I made a super basic unboxing video with my first impressions on the unit :



It's a great instrument, I love it. I hope this video will help others trying to decide if they want one or not.

I'm surprised you're so positive about the fan noise. It's easily the most annoyingly instrument in my lab. It's reasonably loud but above all it whines. I do have a comparatively early model so maybe the fan was upgraded since. Don't get me wrong, it's a great meter. I just feel the fan is a letdown.

I bought mine very recently and while the fan's kinda loud the most annoying thing is the transformer noise, especially when it's off. Have to power off the entire strip so it doesn't annoy me when not in use.

[Mr. Scram]

I bought mine very recently and while the fan's kinda loud the most annoying thing is the transformer noise, especially when it's off. Have to power off the entire strip so it doesn't annoy me when not in use.

Mine has the same issue. Transformer noise can be heard from across the room.

[KedasProbe]

I bought mine very recently and while the fan's kinda loud the most annoying thing is the transformer noise, especially when it's off. Have to power off the entire strip so it doesn't annoy me when not in use.

Mine has the same issue. Transformer noise can be heard from across the room.

They cheaped out on that important part, production may have already changed that test to '= True' or increased the tol. to high because they had to much 'problems'.
I would have gladly paid 10Euro more for one without so much hum, I also kill the power strip if not in use.
Just so you know an active noise canceling headset is also fixing it (I tested it)
But I think al these big vibrations in the unit can't be that good on the long term.

Maybe next time they should provide a DC power input at the back so you can do it right yourself. (It's Cheaper!!)