Calibratory D-105 DC Precision Voltage Reference Standard

[lowimpedance]

While there is some talk on the aspect of resistors I have scanned and attached a copy of an ESI catalogue page on their precision resistors construction and specs etc.
For your reference only.

[AlfBaz]

On the picture, 06:00 is 6 minutes.

ok thanks

As far as the number of digits, the 34461A returns the full floating point number for the measurement over SCPI (i.e. ADC + calibration function), so it has plenty of digits and there is quite a bit more effective resolution than the 6.5 digits on the display.

That's what I find confusing, specs stipulate +-(0.0015+0.0004) on the 10V scale for 24hrs. Doesn't that equate to roughly +-190uV rendering the nanovolt bits as mathematical artefacts?
 

[LaurentR]

That's what I find confusing, specs stipulate +-(0.0015+0.0004) on the 10V scale for 24hrs. Doesn't that equate to roughly +-190uV rendering the nanovolt bits as mathematical artefacts?

Not arguing with that, just clarifying where the extra resolution comes from. Clearly, the stability shown is way beyond the 24hr accuracy.

I guess the argument can be made that if it looks stable within a couple of uV over that period of time (that's fractions of 1 ppm), it probably is unless the drift of the DMM miraculously compensates the drift of the DUT. If there was drift, you certainly couldn't say if it's the DMM or the DUT.

Of course, without knowing what the actual ADC resolution is (which I am not sure has been discussed), this could all be, as you said, averaging artifacts.

[Dr. Frank]

On the picture, 06:00 is 6 minutes.

As far as the number of digits, the 34461A returns the full floating point number for the measurement over SCPI (i.e. ADC + calibration function), so it has plenty of digits and there is quite a bit more effective resolution than the 6.5 digits on the display.
IIRC, the 34401A returns what is on the display, so no gravy.

Hi,
That's not correct!
The 34401A also returns one more digit for each single measurement, i.e. 7 digits, over the bus .
Also the statistics function delivers 7 digits, which are really useable when averaged over an appropriate time.


Generally, ALL HP bench meters deliver one or two digits more over the bus, than displayed .
BenchVue also displays this additional resolution.

The 3458A outputs up to 9 digits, and at least 8 digits also for functions which were displayed to 7 or 6 digits only.

All other 6 1/2 and 7 1/2 bench instruments up to now were very similar, or exact copies, concerning the A/D topology, including the reference.
That can be seen also by the Multi Slope III or IV designation, where obviously mostly software related improvements were made.. if you study the schematics in contrast to that.

So the length of the raw A/D output words was always very similar, something like 24..25 Bits  (>30 bits for the 3458A.)

I also think, that even the new 34461 reused the A/D from the 34410A, only the user interface, graphical and statistical functions and interfacing was greatly improved.

And I'm curious about the 34465A, 34470A, if further improvements on reference and A/D were made, or if they only reused again.

Indicator for latter assumption is, that the A/D linearity still is on the order of 1ppm only, like the old 34401A, although 7 1/2  digits would urgently require something like 0.1ppm, which is only possible by the A/D in the 3458A (which has down to 0.02ppm linearity).

Well, if you would like to judge about the stability of the additional digits of the 6 digits instruments, you would best make a longer series of measurements on a known stable reference, and then calculate the Allan stability deviation, like we've done in a recent thread.:

https://www.eevblog.com/forum/projects/project-kx-diy-calibrator-reference-sourcemeter/msg592440/#msg592440

It will probably turn out, that NPC 100 does not give useful results, but at about 12 seconds averaging time, this additional digit can be used.

Frank

[timb]

Hey guys, I'm writing up the OP and gathering links for a general "Voltage Reference & Standards Megathread" right now. If you've got any useful links (to other threads or websites/documents/projects) post them here (or pm them to me). The idea behind the thread is as a general post to talk shop regarding that elusive last ppm. It'll also help keep track of relevant information that's currently spread over several threads. (It's not designed to replace any of the specific threads, just a catch all for things outside their scope.)

I'll be posting it tonight, assuming I've got internet after the foot of snow that just came out of nowhere last night.


Sent from my Smartphone

[rx8pilot]

Digi-Key sent an email about "new releases" including the Linear LS8 voltage references. I have not poured over the data sheets but it looks like a humidity insensitve ceramic package. Not sure if they are anything special.

http://cds.linear.com/docs/en/datasheet/6655fe.pdf

[mrflibble]

Hey guys, I'm writing up the OP and gathering links for a general "Voltage Reference & Standards Megathread" right now.

...

It'll also help keep track of relevant information that's currently spread over several threads.

Great idea. The various threads have some very useful information in it, but as you indicate it's spread across a rather large number of posts. So one place to condense "the good bits" sound like a plan.

[JohnnyBerg]

Digi-Key sent an email about "new releases" including the Linear LS8 voltage references. I have not poured over the data sheets but it looks like a humidity insensitve ceramic package. Not sure if they are anything special.

http://cds.linear.com/docs/en/datasheet/6655fe.pdf

The C version is nothing special, but the B version looks promising. 2ppm/K. max. and 1ppm/K. typ. High initial accuracy of 0.025%. €11/piece at Farnell.

@timb: great idea!

[paulie]

Appending a question to this thread, since it seems to have entered more enlightened territory:

Evidently, not... quite... yet...  I think I am going to stop trying to help people that are not nice to me.  ('nuff said...)

Hey diligent, if that was in reference to my "cheap" trick I apologize. It was intended as a joke and I thought, considering your avatar, you might appreciate such. It seemed funny to me that you took off on that word after my use of it. I Hope you may also notice my many complements on your knowledge and experience. Sorry again if you took my humor the wrong way. In any case I appreciated the help given so far in that other thread and will follow your advice to start an "el cheapo" thread as soon as my test results get organized.

*sigh* ... i feel like DIY-ing a LM399 in memory of leonard nimoy

[LaurentR]

...
Generally, ALL HP bench meters deliver one or two digits more over the bus, than displayed .
BenchVue also displays this additional resolution.
...

Thanks Dr Frank. That prompted me to look for more information on the ADC and I stumbled on this HP journal issue with 6 articles on the ins and outs of the 3458A with a lot of very interesting content (very relevant to this thread). I assume it's been posted at some point before, but I thought I'd use the occasion to repost it:
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf
Also found this 1990 paper that has useful general content of DMM performance verification:
http://cp.literature.agilent.com/litweb/pdf/5991-1266EN.pdf

[Terabyte2007]

...
Generally, ALL HP bench meters deliver one or two digits more over the bus, than displayed .
BenchVue also displays this additional resolution.
...

Thanks Dr Frank. That prompted me to look for more information on the ADC and I stumbled on this HP journal issue with 6 articles on the ins and outs of the 3458A with a lot of very interesting content (very relevant to this thread). I assume it's been posted at some point before, but I thought I'd use the occasion to repost it:
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf
Also found this 1990 paper that has useful general content of DMM performance verification:
http://cp.literature.agilent.com/litweb/pdf/5991-1266EN.pdf

Thanks for posting the links to these white papers. These should be a good read!

[HighVoltage]

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf

This HP Journal is amazing.
Dated from 1989 and full of technical stuff on the 3458A.
I did not care too much of the firmware development explanations, but the rest is just really good information, including a list of the people who developed the 3458A.

Also interesting, that this 3458A was developed in the 80's and is still state of the art.
Those were probably the prime years of HP.

[Awesome14]

Update on the Calibratory D-105 DC. A 288-hour and running trend-plot shows minimal temperature drift (+-0.5ppm, deltaT=6 degrees C), any long-term drift--if ity exists--is masked by temperature drift, no drift due to humidity changes (delta RH% 17), and the device was chosen as a worst-case-scenario device. It is still steady at 10.00000VDC, and rises to 10.00001VDC once in a while. Even when the device was functioning below its temperature range, it still remained within the +-2ppm spec.

The plot trigger is 2 seconds, perpetual. I've set aside two 6-1/2 digit meters to run and monitor the plot for 45 days. 

My observations thus far are encouraging--better than predicted! 

[Terabyte2007]

Where is the graph data plot? I don't see an attachment.

[timb]

288 hours isn't long term. Temperature drift is still drift, so it ultimately counts towards the accuracy of the system as a whole. This is why those figures are given in ppm/C. For long term temperature-independent results, the unit needs to be placed in a small chamber that can maintain absolute temperature (25c, 40c, whatever). Lastly, if you've only got 6.5 digit meters, don't bother. At that voltage the 10uV decade is noise anyway and accuracy to 100uV is nothing special. (Power Designs was doing 10uV accuracy in their benchtop power supplies in 1964 with a Zener and a few resistors).


Sent from my Tablet

[Awesome14]

288 hours isn't long term. Temperature drift is still drift, so it ultimately counts towards the accuracy of the system as a whole. This is why those figures are given in ppm/C. For long term temperature-independent results, the unit needs to be placed in a small chamber that can maintain absolute temperature (25c, 40c, whatever). Lastly, if you've only got 6.5 digit meters, don't bother. At that voltage the 10uV decade is noise anyway and accuracy to 100uV is nothing special. (Power Designs was doing 10uV accuracy in their benchtop power supplies in 1964 with a Zener and a few resistors).


Sent from my Tablet

I'm just reporting what I have so far. If you don't like it that's really not my doing. But I'm impressed with the performance thus far.  You should be too. As I said the test is "and running". I can't speed up the passage of time. But you can still criticize me for not doing that. I don't know what drives you. But those who contradict everything another man says, and who criticize everything another man does, are the sand in the intricate mechanism of civilization that makes cooperation among men impossible.

[Awesome14]

Where is the graph data plot? I don't see an attachment.

I have to photograph the meters. Stay tuned. 

[Terabyte2007]

Where is the graph data plot? I don't see an attachment.

I have to photograph the meters. Stay tuned.

A photo of the meters? I think you're going to need long-term plot data and other statistics to back up your claims. Coming back to this thread with a photo of the meters and the same argument, well, I don't think your going to win over the crowd with that. I'm just sayin' 

[Terabyte2007]

@Awsome14, Also if you read my post earlier about my testing of the D105 it is marginal to do test's with a 6.5 digit DMM. If you know anything about TAR (Test Accuracy Ratio) then you would know that to test the D105 accurately to the claimed precision you would need a TAR rating of 4 or better. A 6.5 digit DMM barely makes a rating of 1. At a minimum, you would need a 7.5 preferably an 8.5 digit DMM to get a rating of 4 or better and that is only going to happen at a cal lab. I am not trying to burst your bubble, but testing to the claimed precision with a 6.5 digit DMM is going to be marginal at best.

[paulie]

LOL!

[Awesome14]

@Awsome14, Also if you read my post earlier about my testing of the D105 it is marginal to do test's with a 6.5 digit DMM. If you know anything about TAR (Test Accuracy Ratio) then you would know that to test the D105 accurately to the claimed precision you would need a TAR rating of 4 or better. A 6.5 digit DMM barely makes a rating of 1. At a minimum, you would need a 7.5 preferably an 8.5 digit DMM to get a rating of 4 or better and that is only going to happen at a cal lab. I am not trying to burst your bubble, but testing to the claimed precision with a 6.5 digit DMM is going to be marginal at best.

We're not dealing with absolutes, only degrees of certainty. In the trend plot, the measurement uses a seventh digit of precision, because the trigger is set to 2 seconds, with 50,000 samples/trigger. So, since the meter can use a statistical mean of 50,000 readings per trigger, it gains a legitimate digit of accuracy. Correction: the meter uses 1000 samples per trigger. I have the benefit of checking our meters at least once per month. That adds certainty to the meter's readings. I also record the historical data, so I have drift figures. That increases certainty still more. The total range of the trend plot is  10VDC < +-2ppm. And, we have a 3458A to confirm the research findings.

You can do R&D with a 6-1/2 digit and confirm the findings using an 8-12 digit. Using an 8-1/2 digit for everything takes too long! The meter in lower position in the photo is within 0.5ppm of our 732A, and its internal noise 0.1-0.2uVDC. You might say, "There's no way I could know that!" But there is! Just because something is unknown to one person does not mean it is unknown to everyone! I can test the internal noise of any DM meter, in five minutes!

Everyone who has tested the device intact has agreed that it works as claimed. I guess because of my personal beliefs or idiosyncrasies some people on this board reflexively question my abilities, intent, and credibility. I assure you I take good care to be correct in my claims. You simply have to imagine the catastrophic and immediate end that would come to an enterprise selling voltage standards that don't work as specified.

I've been doing this for years, and I haven't had a single complaint regarding the claims made for my devices. That speaks much louder than a handful of amateurs spouting canned jargon in a desperate plea to disparage me and my product. Let's see someone prove me wrong, in the real world! Two unsolicited, outside sources with 8-1/2 digit meters confirmed my own findings. What more do you want? Sheesh!

So I live in a world that's fun. Big deal! I like being entertained by everything! In my world, I have everything within the three-dimensional reality + many other valuable things. I am never bored. My mind ceaselessly creates. I love creating things. It makes me God's playmate. We create together. It's a lot of fun! I like having fun! If anyone is to fault me for it, they are not having enough fun. 

[Terabyte2007]

@Awsome14, Also if you read my post earlier about my testing of the D105 it is marginal to do test's with a 6.5 digit DMM. If you know anything about TAR (Test Accuracy Ratio) then you would know that to test the D105 accurately to the claimed precision you would need a TAR rating of 4 or better. A 6.5 digit DMM barely makes a rating of 1. At a minimum, you would need a 7.5 preferably an 8.5 digit DMM to get a rating of 4 or better and that is only going to happen at a cal lab. I am not trying to burst your bubble, but testing to the claimed precision with a 6.5 digit DMM is going to be marginal at best.

We're not dealing with absolutes, only degrees of certainty. In the trend plot, the measurement uses a seventh digit of precision, because the trigger is set to 2 seconds, with 50,000 samples/trigger. So, since the meter can use a statistical mean of 50,000 readings per trigger, it gains a legitimate digit of accuracy. I have the benefit of checking our meters at least once per month. That adds certainty to the meter's readings. I also record the historical data, so I have drift figures. That increases certainty still more. The total range of the trend plot is < 10VDC +-2ppm. And, we have a 3458 to confirm the research findings.

You can do R&D with a 6-1/2 digit and confirm the findings using an 8-12 digit. Using an 8-1/2 digit for everything takes too long! The meter in lower position in the photo is within 0.5ppm of our 732A, and its internal noise 0.1-0.2uVDC. You might say, "There's no way I could know that!" But there is! Just because something is unknown to one person does not mean it is unknown to everyone! I can test the internal noise of any DM meter, in five minutes!

Everyone who has tested the device intact has agreed that it works as claimed. I guess because of my personal beliefs or idiosyncrasies some people on this board reflexively question my abilities, intent, and credibility. I assure you I take good care to be correct in my claims. You simply have to imagine the catastrophic and immediate end that would come to an enterprise selling voltage standards that don't work as specified.

I've been doing this for years, and I haven't had a single complaint regarding the claims made for my devices. That speaks much louder than a handful of amateurs spouting canned jargon in a desperate plea to disparage me and my product. Let's see someone prove me wrong, in the real world! Two unsolicited, outside sources with 8-1/2 digit meters confirmed my own findings. What more do you want? Sheesh!

So I live in a world that's fun. Big deal! I like being entertained by everything! In my world, I have everything within the three-dimensional reality + many other valuable things. I am never bored. My mind ceaselessly creates. I love creating things. It makes me God's playmate. We create together. It's a lot of fun! I like having fun! If anyone is to fault me for it, they are not having enough fun. 

Sigh! 

Honestly, I don't have time to argue. I will say this, Simply post the data and trend plots you speak of! I can care less how quickly you can determine the internal noise of a DMM or any of the other repetitive rhetorical BS that you keep filling this thread with. "Canned Jargon"? Have you read your eBay page lately? I think you should tread lightly my friend, I was not attacking you!

[paulie]

LOL!

[mrflibble]

Pure comedy.

Have you read your eBay page lately?

Well darnit, now you made me read it. From the item description on ebay:

What this device is: It is a revolutionary voltage standard at an affordable price. If you want to get your lab synced to the SI second, our 0.0002% standard provides one possible way to do it.

Synced to the SI second, how quaint.

Ah screw it, I'll just quote it here for the benefit of the readers. It has too many good bits in it to just let it slide past. XD

The item is a 10.000000VDC Precision Voltage Reference Standard. It accepts 13.0VDC to 35.0VDC INPUT. The OUTPUT is 10.000000 VDC ±0.0002% (±2ppm) at 14.5VDC (+-1.0VDC) INPUT. The other unit we have listed is a 10.000000VDC +-0.0006% (+-6ppm) unit. The absolute accuracy, added temperature-control, and the price, are the only differences between this unit and  our novice product, the +-0.0006% voltage standard. The +-0.0002% standard (the one in this description) has (4) four nonmagnetic, gold-flashed, color-coded banana-sockets w/binding posts for INPUT and OUTPUT connections. The internal connections are soldered or bolted. The regulator IC is socketed for low hysteresis and to prevent damaging the IC by soldering.  The mainboard is a low-hysteresis design, which prevents output drift due to mechanical stresses on the regulator IC and other internal components. We hand-pick the precision voltage reference ICs we incorporate into the units—TI REF102C—and integrate it with (2) low-pass filters for ripple-suppression, shielding against induced thermoelectric effect, and passive temperature-compensation. The +-0.0002% standard can be  power-cycled without affecting the absolute accuracy (+-2ppm). Among the photos is a laboratory null voltage-offset setup. The reading is: 0.0092mVDC (9.2uVDC; 0.92ppm). That is the maximum difference between the 732A 10.000000VDC OUTPUT and the 10.000000VDC OUTPUT of the +-0.0002% standard at its maximum actual error. The null setup is as follows: Fluke 732A OUTPUT (-)ground(-) to OUTPUT (-)ground(-) on the +-0.0002% standard. Power Supply 14.5 Volt DC (+) and (-) to INPUT of the +-0.0002% standard; 10.0VDC (+) OUTPUT (+)  of Fluke 732A to DMM (-) low (-); 10VDC  (+) OUTPUT (+) of +-0.0002% standard to DMM (+) high (+) . All units are nulled to a 732A or 732B Precision Voltage Reference. The +-0.0002% standard consumes roughly 18mW, so it may also be powered by (2) 9V alkaline batteries wired in series. The device settles in roughly five minutes. Maximum output load is 10.0 mA. The unit is durable and maintenance-free. It measures approximately 1.5" x 2.25" x 1.75". Calibration is available any time. We customize calibration parameters on request, i.e. input voltage, calibration temperature, output load, etc.. Test equipment and interconnect cords are not included. Questions, use the link below. Thank you for viewing our eBay.com listing! The following information may be of use to certain buyers. What this device is not: the device is not designed for use in mission-critical applications. What this device is: It is a revolutionary voltage standard at an affordable price. If you want to get your lab synced to the SI second, our 0.0002% standard provides one possible way to do it. The device does not require that it be powered on continuously. It will endure power cycling and retain its absolute accuracy of +-2ppm. Therefore, the device is rated for drift per 1,000 hours of use. Due to demand, we have begun long-term drift testing over 100% power-up conditions. See the graphs in the photos. We have also recently conducted a test to determine the effects of relative humidity on the output voltage. Initial results demonstrate a 17% RH range between 25% and 42% RH has no immediate effect on output voltage.  Based on tests done with each unit, and also a number of samples from outside sources, we have estimated the actual absolute error of our +-0.0002% product to be within +-2ppm, with 95% certainty. In many instances it is desirable to ship our unit (shipping weight = 4 oz.; 112 gm.) from place to place in a powered-down state than to ship a Fluke 732B in a powered-up state, because the accuracy required for a particular application may be less than the 732B provides, and shipping costs are much less. Based on preliminary findings we will guarantee by traceable certificate, +-0.0002% absolute accuracy within the specified operating environment conditions, at >=10 MOhm load, between 0 and 60 days; +-3.5ppm for 61-120 days; and 4.5ppm for 121-180 days. Nothing else will get the volt in the door at a lower cost. The +-0.0002% (+-2ppm) voltage standard was released January 2015. The only data we had acquired at that time was drift figures in ppm/1,000-hours-of-use format, and temperature drift ppm/degreeC. Presently, we are doing research on drift per %RH, and drift per multiple 1,000 hours of use. However, we have years of data for our +-0.0006% (+-6ppm) voltage standard, which uses the same regulator IC. Therefore, we can make certain assumptions that hold true for any device using the same IC. I am not an amateur. This device was created by University graduates outstanding in their fields! I rigorously test, and confirm my findings, before any public claims are made regarding our products. We have 100% positive feedback. We've sold hundreds of voltage standards on eBay! They have been tested on 8-1/2-digit meters by at least five separate, outside sources, which have confirmed our own claims.

[mrflibble]

Curious. Another thing I just noticed on that ebay page:

External Calibration Controls:    Yes

Are those the external calibration controls inside the box that shall not be opened, or some other external calibration controls?