All about Keithley DMM7510. Bugs and features, recipes, advice, notes.

[MegaVolt]

Analog

Relay

* Secondary measurement without relay clicks - https://www.eevblog.com/forum/testgear/all-about-keithley-dmm7510-bugs-and-features-recipes-advice-notes/msg2180195/#msg2180195
* Additional information about the relay           - https://www.eevblog.com/forum/testgear/new-keithley-dmm6500/msg2022019/#msg2022019

DCV

* An input of 10 MΩ in the range of 0.1 V and 1 V creates an offset voltage of 5-6 μV and additional noise. Do not use this input impedance for accurate measurement. No solution. - https://forum.tek.com/viewtopic.php?f=363&t=141241

DCI

* Measuring current from sources with low internal resistance gives a big error. - https://forum.tek.com/viewtopic.php?f=363&t=141262

Digi V

* The input frequency band does not depend on the sampling frequency and depends only on the range. - https://www.eevblog.com/forum/testgear/all-about-keithley-dmm7510-bugs-and-features-recipes-advice-notes/msg2710538/#msg2710538
* At high digitizing speeds (> 200kSps), noise in the 0.1V range is very different from white noise. No solution. - https://www.eevblog.com/forum/testgear/all-about-keithley-dmm7510-bugs-and-features-recipes-advice-notes/msg2609100/#msg2609100

Temperature

* Does not support measurements for Pt1000. No solution.
* When measured from a thermocouple, it does not measure the temperature of the cold junction. Solution - https://www.eevblog.com/forum/testgear/all-about-keithley-dmm7510-bugs-and-features-recipes-advice-notes/msg2233422/#msg2233422

Digital

Digi V; Digi A

* Parameter dmm.APERTURE_AUTO does not work as described in the documentation. https://www.eevblog.com/forum/testgear/all-about-keithley-dmm7510-bugs-and-features-recipes-advice-notes/msg2648655/#msg2648655
* The position of each sample on the timeline has a virtual error due to the transfer of the internal timeline to the external. https://www.eevblog.com/forum/testgear/new-keithley-dmm6500/msg2140273/#msg2140273

Programming

* Low data reading speed via remote interface (v1.7.0). I could not get more than 800 KB/s (<7% 100MB Ethernet). This is about 50 KS/s (8 bytes of data and 8 bytes of time for 1 count). The processor inside the device allows you to work at higher speeds. No solution.
* The warm-up status on the remote interface can only be known once. The program launched the second time will not be able to get this status. No solution. - https://forum.tek.com/viewtopic.php?f=363&t=138795
* NTFS flash drive is not supported, only FAT32 works. No solution. - https://forum.tek.com/viewtopic.php?f=363&t=140287

Firmware bug`s fix history

1.7.0

Improved remote control of the device:
1. Data reading speed increased 3 times to 50 KSamples per second. (When reading a large buffer, the reading speed at the beginning is greater than the reading speed at the end.)
2. Reading data digitized at high speeds (1 MHz) is performed without a preliminary long delay.
3. Virtal front panel works much better. If you return to the device using the http://IP_Device/front_panel.html link, you can open a window without a menu. It looks great on a smartphone ))

Other:
4. The measurement block (trigger.BLOCK_MEASURE_DIGITIZE) now has the same name for the Measure and Digitize modes. (Just as it is done in the DMM6500).
5. The labels on the vertical axis on the graphs have become much better readable.

1.6.7

[MegaVolt]

Brad O I will be very grateful for the answers about 7510.

1. DataSheet promises 140,000 measurements per second via LAN (measurement + time). So far I was able to get a maximum of 15,000. What can I configure to get the reading speed stated in the datasheet? Binary mode is on.

2. How can you remotely know if the warm_up period has passed or not? Is there any flag or status for this?

3. I managed to get the error of measuring time beyond 20ns during the first 2s   It seems to me that these outliers occur at random times and are not related to the first 2s. I also saw peak to peak emissions on the order of 900ns on a sample of 8MS.

4. How can I reset the device without turning off the power? I notice that if the device stopped working as needed (does not respond via the network or something strange responds), then you can fix it only by turning off the power. It is very uncomfortable. We must wait again for an hour and a half. Resetting from the touchscreen does not help: (

[MegaVolt]

I see a clear periodicity. Every 5s is a new failed time.

Does this mean that data is selected at the wrong moment in time? Or is it a timeline problem and you can ignore them?

[MegaVolt]

Current and voltage measurement without relay clicks. This is possible if you select secondary measures in accordance with the tables below:

[MegaVolt]

When measured from a thermocouple, it does not measure the temperature of the cold junction. See the first picture.

A possible solution to this problem is indicated in the second picture.

[MegaVolt]

New bug`s: The problem of changing the size of the buffer.

UPD: To repeat the bug for the newly turned-on device, you need to repeat the script twice. But between operations you need to disconnect from the device and reconnect. The red arrow indicates the location of the reconnection.

UPD2: The maximum size that can be obtained after the appearance of the bug is only 7997000.

[MegaVolt]

New bugs DMM7510 (features):
And there is nothing to complain about. Branded short Keythley 6820.

[MegaVolt]

If you have access to the Keythley DMM7510 please check to see if you have the same bugs.
I'll be very grateful.

[MegaVolt]

The device is mocking me. I can’t understand how it works.

I was trying to figure out how to control the aperture for DigiV. To do this, I applied a rectangular signal with a frequency of 400 kHz from the generator (high level [2.5V] 90% of the time; low level [-2.5V] 10%).

And he turned on the digitization with the same frequency of 400 kHz. (frequencies are not synchronized). The period is 2.5 μs. This means that the aperture can be 1 or 2 μs. And Auto should correspond to 2 μs.

And I was hoping to see two types of graphs.

BUT!!!! I see 3 types of graphs. The graph when installing Auto Aperture is not similar to the 2 μs graph .... It is rather similar to the graph when the samples go through 1.25 μs. But this is only an assumption. In fact, I do not know how it is done in the device.

Though disassemble this miracle box and become an oscilloscope on the DAC

[MegaVolt]

Measured the cutoff frequency in DigiV mode.
The cutoff frequency does not depend on the sampling frequency .

For ranges: 0.1 V; 1 V; 10 V cut-off frequency: 600 MHz
See the picture.

For the 100V range, the cutoff frequency is 7.5 kHz.

For 1000V, the cutoff frequency is 17.5 kHz.

[Kleinstein]

The quite limited frequency response in the 100 V and 1000 V ranges is a bit disappointing. Especially the low cut of for 100 V is odd, as there is the same divider and just the amplifier in x 10 mode instead of x 1. This corresponds to the 1 V and 10 V ranges, that seem to be fine to much higher frequency.

For 1 MSPS sampling the Nyquist limit is at 500 kHz - so there should be large drop in this range. If looking at more than just the pure RMS or peak to peak values, there should be signs of aliasing. So how does the waveform look for something like a 550 kHz signal, when measured at 1 MSPS.

It looks like there is no anti aliasing filter, not even a fixed one (e.g. 400 kHz) for 1 MSPS sampling.

[MegaVolt]

The quite limited frequency response in the 100 V and 1000 V ranges is a bit disappointing. Especially the low cut of for 100 V is odd, as there is the same divider and just the amplifier in x 10 mode instead of x 1. This corresponds to the 1 V and 10 V ranges, that seem to be fine to much higher frequency.

If you look at the noise in the specification, then the 100V range has a higher noise level. Is the band reduced to reduce noise?

For 1 MSPS sampling the Nyquist limit is at 500 kHz - so there should be large drop in this range. If looking at more than just the pure RMS or peak to peak values, there should be signs of aliasing. So how does the waveform look for something like a 550 kHz signal, when measured at 1 MSPS.

Yes, all crossings across the Nyquist border have a drop. This is clearly visible if you change the frequency linearly and not in steps.

And all that is higher than the Nyquist frequency is spectrum overlay. The ADC has a very good sample/hold device with sampling times of the order of 100 ns.

It looks like there is no anti aliasing filter, not even a fixed one (e.g. 400 kHz) for 1 MSPS sampling.

Unfortunately it is so. As Bred said, the user must make an external good filter himself if he wants to watch the signal in a limited band. If this is not done, then all the noise from the Nyquist frequency to the MHz pair will fall into the signal ((

[MegaVolt]

Updated the header:
Added by:
- improvements to the new firmware;
- added new bugs and features;

[MegaVolt]

The new screen of death.
The device after it did not react to anything. I had to reboot. Unfortunately, I can’t say how to repeat it. I switched between DCV and Digi V modes, used a large buffer and played with the Count parameter.

[E-Design]

The quite limited frequency response in the 100 V and 1000 V ranges is a bit disappointing. Especially the low cut of for 100 V is odd, as there is the same divider and just the amplifier in x 10 mode instead of x 1. This corresponds to the 1 V and 10 V ranges, that seem to be fine to much higher frequency.

For 1 MSPS sampling the Nyquist limit is at 500 kHz - so there should be large drop in this range. If looking at more than just the pure RMS or peak to peak values, there should be signs of aliasing. So how does the waveform look for something like a 550 kHz signal, when measured at 1 MSPS.

It looks like there is no anti aliasing filter, not even a fixed one (e.g. 400 kHz) for 1 MSPS sampling.

Yes the 100V and 1kV ranges have bandwidth limited by the huge 10Meg divider. Internal changes when going from 100V and 1kV can make those differ a bit as well. The divider wasn't AC balanced because most of the user base doesn't do high voltage at high frequency generally.

I can confirm there is not a specific anti-alias filter in hardware for the digitizer. The instrument was released as primarily a DMM where mostly band limited signals are used. This was the marketplace assumption.

[MegaVolt]

Yes the 100V and 1kV ranges have bandwidth limited by the huge 10Meg divider. Internal changes when going from 100V and 1kV can make those differ a bit as well. The divider wasn't AC balanced because most of the user base doesn't do high voltage at high frequency generally.

The 100V range behaves very strangely.
For example, I see a reaction from a change in input resistance.
At 10 MΩ - 750 μV
At Auto - 65 uV

At the same time (Auto) it can be seen that emissions appear on the noise track

Is there any difference between 10 MΩ and Auto?
If you select a range of 1000V there is no such effect.

[E-Design]

Yes the 100V and 1kV ranges have bandwidth limited by the huge 10Meg divider. Internal changes when going from 100V and 1kV can make those differ a bit as well. The divider wasn't AC balanced because most of the user base doesn't do high voltage at high frequency generally.

The 100V range behaves very strangely.
For example, I see a reaction from a change in input resistance.
At 10 MΩ - 750 μV
At Auto - 65 uV

At the same time (Auto) it can be seen that emissions appear on the noise track

Is there any difference between 10 MΩ and Auto?
If you select a range of 1000V there is no such effect.

Your comment I believe is referring to low frequency (DC) noise and not the AC bandwidth .

There is an internal signal path difference based on selection 10M or Auto.  So yes.
For 1000V, there should be same effect but it might not be apparent due to the higher noise (because of higher gain)

[MegaVolt]

There is an internal signal path difference based on selection 10M or Auto.

Can you reveal the details? What is the difference? What is the input impedance in Auto mode? Does the device have the ability to work with an input impedance of 1 GΩ for the ranges 100V and 1000V

[MegaVolt]

New firmware 1.7.0. Trying to write a large contiguous buffer (8 000 000 samples) at a time:
Settings:
Digi V + [Freq=1MHz] + [Aperture=1mks] + [count = 8,000,000]



When time-critical processes are not implemented on the hardware, but are given to programmers, it’s a disaster   When the product is more than 5 years old. But he still cannot do a simple operation - write a buffer of a given size with a given frequency. But on the screen you can run the watch and other toys.