Datron 1271--How to access "extra digits"?

[View[+]Finder]

With thanks to "Re: Excel adding digits when logging with VBA script" RIP

Back in the h/p 41c calculator days, those "extras" were called guard digits and were there to protect the quality of the calculations. "CalcNuts' back then would try to access them. The modern ADC's have more digits than are displayed, e.g. KY 6500 6.5 digit average calculation has an extra digit. NI DMM's offer a lot more.

Are these digits meaningful? Well, if the first extra is worth a dollar, the next is worth a penny, maybe. If a manufacturer made a cheaper model by throttling the  more expensive one, some folks would look for a hack. But digging through the numbers past the "sixth place of decimals" is unlikely to yield a 9.5 digit voltmeter. Still, it's good to check rounding errors and lack of randomness.

I digress. The Datron 1271 provides a command {RDG?} to access the last triggered reading when accessed over GPIB. I use python ad suspect that I am not getting the most out of measurements.

The blue dots are from NI4071 7.5 digits and the orange from the Datron 1271 at 8.5 digits. The orange marching band might result from use of a filter or the ADC. All advice appreciated . . .

[unseenninja]

> All advice appreciated . . .

Exactly what advice are you seeking? All I see here is an axis label free graph. I have no idea what you are measuring, what the time period is and what the temperature was during measurement.

Your Datron 1271 will give you 8.5 digits of resolution. The digits are all meaningful, but there is measurement noise.

[Mickle T.]

While digging around the Datron 1271 I discovered a few secret commands, including YEUK and YEUK?, which provide access to the 8Kb buffer with raw ADC counter data. For example, each 8.5-digit measurement loads 32 integer triplets (32x16 PLC) into the buffer, which are then filtered and averaged in the firmware.
Google it, it's been on the forum before.

[View[+]Finder]

> All advice appreciated . . .
Exactly what advice are you seeking? All I see here is an axis label free graph. I have no idea what you are measuring, what the time period is and what the temperature was during measurement.

The graphic presentation should have been sufficient, however vague the context, or so I thought. The orange Datron voltage markers appear to fall on straight lines; the blue NI4071 markers appear random. Why is this so? My guess is that it is 'normal' and related to the ADC's in the two voltage measurements. I have seen this before in the context of measuring an 8.5 digit reference with a 6.5 digit meter. The YEUK? command may provide a better explanation. Also there might be averaging in the measurements by the Datron.

[Kleinstein]

The reading from the 1271 fall on straight lines / a limited number of values. This indicates either an actually limited resolution (quantization limit) of the ADC or rounding before sending the data. When comming from formating the data for the interface the line should match digit levels. If coming from the ADC the possible levels usually don't excatly line up with the digit steps due to a scaling factor. One may have to look at a slightly larger range and the actual data in ascii format to see this.

Even with the DA1271 and the visible lines there is still significant noise - at least in the measurement shown. With that noise level there is not much lost from the quantization. Additional digits could have some meaning but the gain from including them is limited. The way the NI ADC is made (more SD like with 2nd or even higher order)  one often gets very small quantization steps, not visible in the noise. Not much gained much from these, but why not send them if the come essentially for free.

[unseenninja]

> All advice appreciated . . .
Exactly what advice are you seeking? All I see here is an axis label free graph. I have no idea what you are measuring, what the time period is and what the temperature was during measurement.

The graphic presentation should have been sufficient, however vague the context, or so I thought. The orange Datron voltage markers appear to fall on straight lines; the blue NI4071 markers appear random. Why is this so? My guess is that it is 'normal' and related to the ADC's in the two voltage measurements. I have seen this before in the context of measuring an 8.5 digit reference with a 6.5 digit meter. The YEUK? command may provide a better explanation. Also there might be averaging in the measurements by the Datron.

More information, like how you collected the readings, how the graph was created and scaled, how the 1271 was configured and all the other information relevant to your problem gives other people the chance to make a reasoned and useful post in answer to your questions.

I suspect there is either something wrong with how you collected the data or configured the meter, or that there's something wrong with your meter.
Here's a short extract from a log where my 1271 was measuring an LTZ1000 voltage reference with the full 8.5 digit resolution. I see no missing digits and no reduction in resolution.

2024-07-05 12:51:23,27.24,48.64,1001.01,+07.1137100E+00
2024-07-05 12:51:39,27.24,48.64,1001.01,+07.1137098E+00
2024-07-05 12:51:55,27.24,48.64,1001.01,+07.1137096E+00
2024-07-05 12:52:11,27.24,48.64,1001.01,+07.1137096E+00
2024-07-05 12:52:27,27.26,48.68,1001.02,+07.1137092E+00
2024-07-05 12:52:43,27.26,48.68,1001.02,+07.1137095E+00
2024-07-05 12:52:59,27.26,48.68,1001.02,+07.1137094E+00
2024-07-05 12:53:15,27.26,48.75,1001.00,+07.1137093E+00
2024-07-05 12:53:31,27.26,48.75,1001.00,+07.1137097E+00
2024-07-05 12:53:47,27.26,48.75,1001.00,+07.1137093E+00
2024-07-05 12:54:03,27.26,48.75,1001.00,+07.1137098E+00
2024-07-05 12:54:19,27.26,48.74,1001.00,+07.1137094E+00
2024-07-05 12:54:35,27.26,48.74,1001.00,+07.1137091E+00

As Kleinstein says, an example of the raw data would have been much more useful to see where the problem comes from.

[View[+]Finder]

As requested, I have attached an excerpt from the data file that was used for the graphic posted earlier in this thread. The graphic presentation tells the story better than a table of numbers for some, others want the raw data. The column "LastDigits" in the attached file captures the repeated values in the digits past the sixth place. The Datron 1271 ADC at 8.5 digits is at its limit of meaningful information. Recall the number-free graphic: the NI4071 DMM observations are not lined-up because their ADC includes more digits. Not to suggest that there is relevant information, but rather to confirm that there is noise in the measurements.

I'll leave the question of "YEUK?" for another day.