New Keithley DMM6500 and now DAQ6510

[Brad O]

The 3706A is another instrument with dry circuit measurement, I think that's the only 3 we offer with dedicated dry circuit modes.  The DMM6500/DAQ6510 don't have the hardware required to do dry circuit testing, so that's a feature that can't be added in later (without a new model number).  It was looked at, but the team decided the cost for adding that one feature would've been too great.

As a compromise, is there anything which would prevent me from doing the following hack as a TSP script?
- Connect an external high stability 2 ohm in parallel with DUT (open would be 20mV on the 1 and 10 ohm ranges?)
- Measure the 2 ohm alone, store the value
- Deduce the DUT and display only the calculated value

One caveat is that pecision would become exponentially worse above 2 ohms,  but that's all I need anyway for small signal relay contacts.

In fact, that's exactly the type of thing TSP scripting was designed for!  For your test setup, I think that should work fine, I'm sure you know more about the measurement than I do.  Be sure to only use manual ranging and only ever switch in the dry resistance while the DMM is already measuring the 2 ohm resistor.

[nictinkers]

@nictinkers, thank you for the feedback!  I'm sorry you don't care for a lot of the design, but really, we do take all the feedback we can get to heart.

I rather hope you don't take it too close to heart - I know it can really suck to have something you worked hard on criticised.

Not exactly what you really want I think, which would require 2 full ADCs, but the DMM6500 can measure power with the voltage ratio function like the other DMMs in its class.

That's an interesting approach. I like that using an external shunt doesn't age the relays switching from voltage to current. Is there a way to fill a buffer of full ratio measurements at a high speed with a fixed sample frequency? (That is, like Digitise Voltage, but for full ratio.) Even if you couldn't hit 1 megasample/sec I'd assume you could get fast enough to get a low-distortion measurement for 50/60Hz signals. Then it would depend on whether the scripting language is then powerful enough to calculate AC power parameters from the buffer.

[Brad O]

Not exactly what you really want I think, which would require 2 full ADCs, but the DMM6500 can measure power with the voltage ratio function like the other DMMs in its class.

That's an interesting approach. I like that using an external shunt doesn't age the relays switching from voltage to current. Is there a way to fill a buffer of full ratio measurements at a high speed with a fixed sample frequency? (That is, like Digitise Voltage, but for full ratio.) Even if you couldn't hit 1 megasample/sec I'd assume you could get fast enough to get a low-distortion measurement for 50/60Hz signals. Then it would depend on whether the scripting language is then powerful enough to calculate AC power parameters from the buffer.

The Ratio function's sample rate is defined by NPLC or Aperture, so it has a minimum sample window of ~8 us like the rest of the box.  The maximum reading rate is slightly slower than that for buffer and trigger overhead though.  Just removing the delays in the script, setting a manual range, turning off auto delay, and setting the minimum NPLC, I measured noise at 200S/s, or ~5ms between readings.  I would suspect you could speed it up slightly if make a trigger model to control when readings happen, as I've said before, the trigger model has a dedicated processor.  I would have to dig deeper and talk to the firmware folks to find out what the limiting factor is here, it's not TSP though, I tried just reading the ratio function directly to defbuffer1 and it was about the same speed.  I suspect the ratio function just isn't optimized for speed. 

[attle123]

The 3706A is another instrument with dry circuit measurement, I think that's the only 3 we offer with dedicated dry circuit modes.  The DMM6500/DAQ6510 don't have the hardware required to do dry circuit testing, so that's a feature that can't be added in later (without a new model number).  It was looked at, but the team decided the cost for adding that one feature would've been too great.

Thank you for your answer.

As a compromise, is there anything which would prevent me from doing the following hack as a TSP script?
- Connect an external high stability 2 ohm in parallel with DUT (open would be 20mV on the 1 and 10 ohm ranges?)
- Measure the 2 ohm alone, store the value
- Deduce the DUT and display only the calculated value

One caveat is that pecision would become exponentially worse above 2 ohms,  but that's all I need anyway for small signal relay contacts.

Hi Brad, following up on this ratio topic, is there anyway to display the sense/Vs voltage onto the screen as well alongside the V_input. Like the keysight is able in the video (2:09)
https://youtu.be/Y6xnLkiUMn8?t=129
thanks!

[Brad O]

Hi Brad, following up on this ratio topic, is there anyway to display the sense/Vs voltage onto the screen as well alongside the V_input. Like the keysight is able in the video (2:09)
https://youtu.be/Y6xnLkiUMn8?t=129
thanks!

Yeah, that's real easy.  So how the ratio function works is it will store the ratio as "the reading".  In a full buffer, it will also store the sense voltage in the "extra values" field.  You can get the input voltage from those two values of course.  Then you can just print those values as a string to the main screen.  Or, you can use them in the power calculation like the first script does.

The same engineer adapted his first script to replicate what the Keysight video looks like (change .txt to .tsp to use it, like normal). 

[MrFox]

Getting this DMM out of US customs is incredibly frustrating, is this paranoid crap a new thing?

[MikeP]

Yeah, that's real easy. 

 

 I think this tool has a lot of undisclosed features.

[thm_w]

Getting this DMM out of US customs is incredibly frustrating, is this paranoid crap a new thing?

Where did you buy from, from what I saw US vs CAD distributors the price wasn't far off.

[attle123]

thank you very much for the script, however there seem to be an issue with the readingbuffer, that every time the script is ran it seems to change the reading buffer to fill mode "once", even when i manually change the fill mode back to continuous on the dmm, so the script eventually stops itself. Is there anyway to change the fill mode to continuous in the script?. I noticed this aswell on the power measurement script aswell

Thanks!!

[drummerdimitri]

It doesn't have "dry circuit" voltage limits on the low ohm ranges, only the 7510 and the 2010 seem to have this option?
I was wondering if it could be added later in firmware, or in scripting, or whether it's an actual hardware limitation on the 6500.

The 3706A is another instrument with dry circuit measurement, I think that's the only 3 we offer with dedicated dry circuit modes.  The DMM6500/DAQ6510 don't have the hardware required to do dry circuit testing, so that's a feature that can't be added in later (without a new model number).  It was looked at, but the team decided the cost for adding that one feature would've been too great. 

@nictinkers, thank you for the feedback!  I'm sorry you don't care for a lot of the design, but really, we do take all the feedback we can get to heart. 

Also that the instrument doesn't measure voltage and current simultaneously. It would have been really nice to be able to measure power, power factor, VA, VAR, even at some much lower resolution and accuracy than the main ADC.

Not exactly what you really want I think, which would require 2 full ADCs, but the DMM6500 can measure power with the voltage ratio function like the other DMMs in its class.  One of our engineers wrote a small script to simplify the process a bit, the script and the full description are here: https://forum.tek.com/viewtopic.php?f=617&t=141154,
here's a video he's been putting together for it too:
https://youtu.be/oWTTrA20M0w

I'm glad there is an option to measure power.

My question however is what shunt ratings would I need to measure power as accurately as possible from 1 mA to say 300-500
A?

[MrFox]

Finally received it, the low resistance precision is pretty nice.

I just soldered three 14 awg in parallel on some machined banana plugs I had. With sense wires about an inch apart. So that is supposed to measure around 60 to 65 uohm I guess. It's just a test for noise, not accuracy.

The lowest resistance range is very noisy as expected, getting usable values that low requires slow integration and filtering. Averaging reaches the point of diminishing returns around 5 NPLC with 100x repeat filter. A hundred readings took a little over an hour. Doing 15 NLPC with 100x filter took 3 hours and provided very similar numbers. It's either the limit of the instrument or the limit of my crude methodology.

9.98 uohm pk-pk
2.26 uohm std dev
Average 61.3 uohms

That's kind of amazing....

[MrFox]

The interface is a lot of fun to use, there are still a few things that I hope will be adressed in a firmware update. Not a big deal, the yellow cursor have no reason to exist on the 6500 since there's no knob to control it (it seems to have been written more for the smus and the 7510) unless I didn't figure out how to move it around. It makes the behaviour of the enter key a bit random, it's whatever default was selected on the page.

The swipe to change pages can be risky to operate, I sometimes press one of the functions by mistake, which wipe out the currently running acquisition. I would love to cycle the pages with the home key or something, maybe an option to remove the function page completely since there's already a dedicated function button to popup that menu. Maybe a toggle to prevent interrupting the measurement? A dialog yes/no for anything which would clear the buffer? Something like that.  When running a trigger and count other than plain continuous, the content of the buffer is probably important.

[Kleinstein]

.....
The lowest resistance range is very noisy as expected, getting usable values that low requires slow integration and filtering. Averaging reaches the point of diminishing returns around 5 NPLC with 100x repeat filter. A hundred readings took a little over an hour. Doing 15 NLPC with 100x filter took 3 hours and provided very similar numbers. It's either the limit of the instrument or the limit of my crude methodology.

The limited effect of more filtering / longer averaging is a kind of weak point with quite some Keithley meters. If think they should work on this, as it is very likely fixable in software and it effects not just the 6500.  This seems to be an old problem (AFAIR I saw the odd low frequency noise already on an old 196 or 199) so I would no expect a fast solution. My hope would be more on finally finding out when developing a new 8 digit meter.

The DAQ6500 is more like a meter for fast measurements and only specified for 6 digits. So the relatively poor noise with 10-60 seconds averaging is not that significant for this type of meter. It's more the higher grade 7510 where this really hurts. The good thing is, chances are it gets better once averaging is much longer (e.g. 5 minutes).

The very low resistance means measuring very low voltage and there thermal EMF effects at the "resistor" can also limit the accuracy. It might help to cover the resistor to protect it from air drafts. Thermal EMF combined with temperature fluctuations is one of the limiting factors for long time averaging. This is inside the meter and also at the DUT.  For the copper resistor this could also be just a fluctuation of the resistance due to temperature changes.

[MikeP]

The limited effect of more filtering / longer averaging is a kind of weak point with quite some Keithley meters.

I did a comparison of "fast" (filter 10, NPLC 5, readings 10) and "slow" (filter 100, NPLC 12, readings up to 160) measurements for 10mΩ +/-0,01%. In a series of ten "fast" measurements, the maximum deviation is about 200 ppm.

[MrFox]

I think the integration rate (NPLC setting) is altering the slope rate of the ADC? I mean it's not averaging multiple samples during thst time, otherwise 100x 1 NPLC would give the same result as 10x 10 NPLC and it doesn't. As a reference the 1 ohm range is insanely low voltage considering the current source is 10mA. So the noise in that range warrants more samples at a faster rate.

I made 3 different reading at the same 100 readings, 100 total NPLC per reading, I repeated the test 5 times to get a good sampling. Room is electronically controlled to 22C. Same 60uohm copper wire. The value shown is the std deviation, in uohms, of each iteration. The test was done 1,2,3,1,2,3 etc... to avoid corellated external factors.

Filter 7x at 15 NPLC:
7.85
9.73
8.75
9.02
9.63

Filter 10x at 10 NPLC:
7.59
7.86
9.04
7..35
7.99

Filter 100x at 1 NPLC:
3.75
3.48
3.23
3.45
3.92

I wish they enabled a 1000x repeat filter for a 1 NLPC reading, which would probably yield one least significant digit in the 1 ohm range. All other ranges are ultra stable and don't really need this. That 1 ohm range is already crazy impressive but it can still be improved in software even more.

[Brad O]

Getting this DMM out of US customs is incredibly frustrating, is this paranoid crap a new thing?

Occasionally customs gets freaked out by test equipment.    Nothing new as far as I'm aware.

thank you very much for the script, however there seem to be an issue with the readingbuffer, that every time the script is ran it seems to change the reading buffer to fill mode "once", even when i manually change the fill mode back to continuous on the dmm, so the script eventually stops itself. Is there anyway to change the fill mode to continuous in the script?. I noticed this aswell on the power measurement script aswell

Thanks!!

Yeah, by default buffers are fill-once.  Add these commands after the buffers have been created (so around line 15):

Code: [Select]

readingBuffer.fillmode = buffer.FILL_CONTINUOUS
powerBuffer.fillmode = buffer.FILL_CONTINUOUSManually changing them should work too, but you have to change both buffers.  The fillmode command in on page 742/15-41 of the reference manual.
EDIT: oh, you also need to change a couple other things: the for loop needs to become a "while true do" and you need to add "i = 1" before the loop and "i = i + 1" at the bottom of the loop since we're removing the iterable for loop.

The lowest resistance range is very noisy as expected, getting usable values that low requires slow integration and filtering. Averaging reaches the point of diminishing returns around 5 NPLC with 100x repeat filter. A hundred readings took a little over an hour. Doing 15 NLPC with 100x filter took 3 hours and provided very similar numbers. It's either the limit of the instrument or the limit of my crude methodology.

There are diminishing returns for higher NPLC for most signals, and in fact higher NPLC may be less accurate.  Page 203/5-58 in the reference manual discusses this a little bit.  At higher NPLC, the measurement interval starts to cover a time frame where the DMM's internal drift starts to matter.  Unless you expect a lot of power line noise, I would recommend 1-5 NPLC and averaging. 

The interface is a lot of fun to use, there are still a few things that I hope will be adressed in a firmware update. Not a big deal, the yellow cursor have no reason to exist on the 6500 since there's no knob to control it (it seems to have been written more for the smus and the 7510) unless I didn't figure out how to move it around. It makes the behaviour of the enter key a bit random, it's whatever default was selected on the page.

The swipe to change pages can be risky to operate, I sometimes press one of the functions by mistake, which wipe out the currently running acquisition. I would love to cycle the pages with the home key or something, maybe an option to remove the function page completely since there's already a dedicated function button to popup that menu. Maybe a toggle to prevent interrupting the measurement? A dialog yes/no for anything which would clear the buffer? Something like that.  When running a trigger and count other than plain continuous, the content of the buffer is probably important.

Interesting idea for the enter key to move between swipes...  To remove the Functions swipe screen, you can use same command I gave to PTR_1275 earlier:

Code: [Select]

display.delete(display.SCREEN_FUNCTIONS_SWIPE)to bring it back:

Code: [Select]

display.create(display.SCREEN_HOME, display.OBJ_SWIPE, display.SCREEN_FUNCTIONS_SWIPE)You could add these to autoexec.tsp to have the swipe removed on start up.
In fact, here's the list of the swipe screens if anyone else wants to remove one of them. You could even remove all of them! If for some reason you wanted that...

Code: [Select]

display.SCREEN_FUNCTIONS_SWIPE
display.SCREEN_SETTINGS_SWIPE
display.SCREEN_SECONDARY_SWIPE
display.SCREEN_GRAPH_SWIPE
display.SCREEN_STATS_SWIPE

The very low resistance means measuring very low voltage and there thermal EMF effects at the "resistor" can also limit the accuracy. It might help to cover the resistor to protect it from air drafts. Thermal EMF combined with temperature fluctuations is one of the limiting factors for long time averaging. This is inside the meter and also at the DUT.  For the copper resistor this could also be just a fluctuation of the resistance due to temperature changes.

Also make sure the covering is a Faraday cage, stray electrical signals become a problem at the 1 ohm range.  Really the 4-wire 1 ohm range is approaching the volt-nut territory so you'd have to start looking a the physical limitations of your measurement setup like you say.

I wish they enabled a 1000x repeat filter for a 1 NLPC reading, which would probably yield one least significant digit in the 1 ohm range. All other ranges are ultra stable and don't really need this. That 1 ohm range is already crazy impressive but it can still be improved in software even more.

Improving the physical measurement setup would have a much larger impact than more averaging if you're talking about going from 100 to 1000 readings.  Software can only do so much if you have too much noise.  Even our electrometer has a 100 reading limit on averaging. 

[MrFox]

.....
The lowest resistance range is very noisy as expected, getting usable values that low requires slow integration and filtering. Averaging reaches the point of diminishing returns around 5 NPLC with 100x repeat filter. A hundred readings took a little over an hour. Doing 15 NLPC with 100x filter took 3 hours and provided very similar numbers. It's either the limit of the instrument or the limit of my crude methodology.

The limited effect of more filtering / longer averaging is a kind of weak point with quite some Keithley meters. If think they should work on this, as it is very likely fixable in software and it effects not just the 6500.  This seems to be an old problem (AFAIR I saw the odd low frequency noise already on an old 196 or 199) so I would no expect a fast solution. My hope would be more on finally finding out when developing a new 8 digit meter.

The DAQ6500 is more like a meter for fast measurements and only specified for 6 digits. So the relatively poor noise with 10-60 seconds averaging is not that significant for this type of meter. It's more the higher grade 7510 where this really hurts. The good thing is, chances are it gets better once averaging is much longer (e.g. 5 minutes).

The very low resistance means measuring very low voltage and there thermal EMF effects at the "resistor" can also limit the accuracy. It might help to cover the resistor to protect it from air drafts. Thermal EMF combined with temperature fluctuations is one of the limiting factors for long time averaging. This is inside the meter and also at the DUT.  For the copper resistor this could also be just a fluctuation of the resistance due to temperature changes.

Ok thanks I just saw the discussion about this in the 7500 thread, I'll be following it there. The details are a bit above my degree of knowledge.

[MikeP]

 I performed another experiment - the measurement of 1 milliohm. The settings: NPLC 0.1 - 1 - 3 - 5 - 12, the filter 100 or OFF, the measurement time of 1 or 10 minutes. About 60 measurements. I was very surprised by the results - a smaller error at the NPLC 1, the filter 100 time measurement 10 min. Much more accurately than with NPLC 12.

 The user manual has the following information: At these rates (lowest noise region in graph), DMM6500 will make corrections for its own internal drift and...

 How is autocorrection related to aperture? How it work? And why is this only possible in a narrow range?

[Brad O]

The user manual has the following information: At these rates (lowest noise region in graph), DMM6500 will make corrections for its own internal drift and...

 How is autocorrection related to aperture? How it work? And why is this only possible in a narrow range?

Most every DMM will do some delays/corrections/reference checks between each or every-so-many readings.  Most of these are controlled by the autozero settings but there are some settling delays and reference checks that can't be changed because they're necessary to get a reading at all.

The instrument can't do those corrections and checks while the measurement is actually happening.  So when you set a longer aperture/NPLC, the end of the measurement is more likely to have drifted away from the reference than the beginning of the measurement.  The solution to this is often to use a smaller aperture (but no smaller than 1), even when you want a stable reading, and do some averaging.  This isn't always true though since a stable DMM will not drift very much at all. 

[Kleinstein]

The ADC and input amplifier might show some 1/f type noise. The longer the individual integration, the more of this 1/f noise will come through. For this reason there is often an upper limit at some 10 PLC and even if one chooses a longer integration like 100 PLC this would internally be done by averaging shorter conversions. In this sense the 1 PLC mode might be the best, but there can be some downsides too, as limited INL and also more time lost to switching the source. The 4 Wire Ohms usually switches between Sense_Lo and Sense_Hi.

The Ohms mode may be special, but in voltage ranges the Keithley meters tend to use some averaging on the zero reading of the AZ mode. This can slightly reduce the higher frequency noise, especially seen over short times and thus can be a good thing there. However there is a downside to this: averaging the zero readings interferes with 1/f noise suppression from alternating zero and signal reading. If too much averaging is used the 1/f noise will come through and this can cause poor performance for longer times.
The PLC setting can also have an effect on this hidden zero reading averaging.

The PLC setting also effects the suppression of mains hum. Depending on the timing an actual line frequency this can be better at some settings. While some prefer a long simple conversion, there is also an advantage with the average over the right number of 1 PLC readings.

[MikeP]

 Gentlemen, thank you very much for the explanation. There are so many things that are not obvious to me. Moreover, the user manual recommends: Measuring DCV with high accuracy ...Set the integration rate to 10 power line cycles (PLCs)... (page 5-3)
 Another question is how often and when does auto-zeroing occur? Sometimes I get strange results. Perhaps there is a connection with AZ.

[Kleinstein]

The 10 PLC mode can have some advantage with linearity over 1 PLC - this at least is true for many older DMMs. With good amplifiers with low 1/f noise levels the 1 PLC mode may not have that much noise advantage.

The usual auto zero mode is alternating between one conversion with the actual input and one conversion with a zero input. There might be cases where something like 2 times signal and 1 zero reading might be used. Some meters also used a sequence line zero, signal and reference. However I don't expect this for the 6500, more for slower reading DMMs with higher priority on accuracy than speed.
Depending on the meter there can be some averaging with the zero readings - my guess is not much averaging in the Keysight meters and a little too much with the Keithley meters.

Another factor can be synchronization with the mains frequency. Sometimes there is a separate setting for line synchronous mode.

The instructions to older meters usually have a more detailed description on how exactly AZ mode works. The newer instructions often leave out these details - maybe because they are not at the high end anymore. Sometimes it still matters how the measurement is actually done, e.g. if the input signal is not constant. One would not need all the details from the old theory of operation section, but a little more information would be good.

[admiralmaggie]

Quick question: How is your experience with continuity test on your DMM6500? I read earlier that there is a delay of some sort...

[sstepane]

Quick question: How is your experience with continuity test on your DMM6500? I read earlier that there is a delay of some sort...

There's no delay on continuity detection, but the sound is at least 1 second long - no matter how short was the connection. Which is a bit unusual, but, IMO even a little better solution - so you'll not miss it.

[MrFox]

(cross-posting this here from the 7500/2450 bugs discusion)

Resizing buffers can sometimes cause memory fragmentation or memory leak.

Steps to reproduce on the 6500 (I have firmware 1.0.02a):
1. Reboot (power off, power on)
2. Go to Reading Buffers
--> The bug will appears if you do NOT touch the Buffer selector button and leave it at defbuffer1
3. Change capacity to 1M
4. Set buffer to defbuffer2
5. Change capacity to 10
--> Here the bug we see is that the buffer selection goes back to defbuffer1, but the capacity button is the value from defbuffer2
6. Set buffer to defbuffer1 (reselect it even if already selected)
7. Change capacity to 10
--> From this point, 1M is missing, it's not possible to either resize or create a buffer larger than 6M (7M is the max on the 6500). Until the next reboot.

In the mean time, it's simple to avoid the bug, always reselect the buffer before changing capacity.