Calibratory D-105 DC Precision Voltage Reference Standard

[Sbampato12]

And if you take a good look in this thread, you will see the firsts attacks come from awesome14.

I agree, in the start, it was looking good, with good points on how to do, and how to evaluate this kind of circuits. But when the people from this forum started to try to get real tests, then the maker came and start to try make everyone to think the same way he does.
Using non-scientific things to explain the circuit, blah blah...

Then, after he show some attacks to whom disagree from him, the things come more comedy... From some pages there is nothing but funny posts, from one side or another.

No big deal. There is somethings that we don't really say to someone, but if he is concerned about this, stop to say something to us (like all that crap about how he get the circuit from heavens, or something about vampires, about the cardiologist who know so much of electronics, about death attempts, how all this are connected to some voltage reference circuit?)

More than a scam, it is a troll.

[TunerSandwich]

but hilarious, none the less.....I know I have gotten more than a few laughs out of this thread....if that makes me a bad person, oh well...

[gilbenl]

The 8-digit meter and the 732A belong to our calibration lab, which is owned by CalibratoryTM, LLC. I personally own a large part of CalibratoryTM, LLC, which also manufactures the voltage reference, D-105 10DC. But no man is an island! We all need other people to accomplish anything.

For the record, there's no LLC registered as Calibratory, or CalibratoryTM in the states of MN or WI. I also checked federal tax records and found nothing. This leads me to believe that The name is up for grabs

Anyhoo, we've beaten the piss out of a dead horse. We saw earlier that it actually performs relatively well and that is really a testament to the IC. Go buy the IC and some parts for $30 and you have a decent ref!

[paulie]

if that makes me a bad person, oh well...

IMO it does. Along with many others here for whom I had great respect until this came along. Even when early reviews indicated his product performed 10x better than it was supposed to the assumptions, misquotes, abuse of his posts and IP, and general blood lust continued.  I've seen this stuff elsewhere on the net but was surprised to find the infection spread here.

It was disappointing to see Eboy start going on about free energy, vampires, bible quotes, murder, etc but that has almost nothing to due with the despicable behavior shown here. At this point I wouldn't trust a single "independent" review.

[nctnico]

IMHO you have to admit that the build quality leaves much to be desired. Just look at the eval board from Linear for the LTZ1000. It is very hard to prove that the DC-105 offers the quoted performance. Perhaps it is adjusted properly to a certain level of accuracy and it has some temperature compensation (judging from how the copper foil looks it appears the thermal link is 'tuned')  but it seems to me that it does not get rid of long term drift and/or thermally induced offset voltages.

All of that is OK as long as the seller doesn't start to pull a whole load of BS from his ass. He could have sufficed to say that he doesn't want to disclose the inner workings. Some may accept that and others don't.

[timb]

He let the flood gates open paulie. At first we were poking fun at the build quality. That is no worse than what Dave does for a living.

Then we heard about Angels, mob hits and immortality. He brought it on himself.


Sent from my Smartphone

[timb]

My thinking is that there appears to still be a market for the voltage transfer service that Joe Geller had [which was more for transferring 10V from his 732B to the buyer, without costing US$500 each time].

It would take Joe about 8-hours in a small chamber he built to calibrate the TempCo of each SVR-T, and that combined with the high cost for a proper [low uncertainty] calibration of his 732B every year ended up being a zero-sum business, so he had to quit doing it.

I think this can still be done, but one would have to charge more for the reference itself, and more for the calibration in order for it to be a viable business model.

So, I've been designing a circuit similar to this, but using the AD587UQ.  No matter what I do, I end up with a circuit to adjust the TC of the AD587 and the time required to adjust the TempCo takes so long that it makes me think that I should just abandon the idea of temperature compensation of a monolithic reference IC altogether.  I think a better idea is to develop a low-cost voltage reference that has a [more or less] "very good" TempCo from the start, then provide a calibrated 10V using that.

I remember when I spoke with Bob Dobkin, Bob told me that the LM399 does not drift when it is off, and that the LM399A model *can* have a TempCo that is considerably less than the data-sheet value [he mentioned 0.1ppm/K].  He said that you could parallel several of these [he said at least 6, but I think 4 would be OK], and then only turn on the reference when you need to calibrate it, or to calibrate something else with it-- the rest of the time, leave it off.  He said you can get away with this because the LM399 has almost no hysteresis through power cycles.

So, why not make a "community voltage transfer device" based on one [or more] LM399A's?  If there is enough interest, I can do this, and sell them on eBay [with low cost follow-on calibrations as often as desired].  This would not need batteries, but could be powered by an external battery pack if desired.  Comments anyone?

I've actually got a partially finished board that sums the outputs of 6 LM399s. The analog section was designed to be ovenized at moderate temp (50c) to help keep the output consistent. (That could be abandoned though.)

I'd be interested in working on it with you.

Another thought, why can't the temperature compensation be automated? Stick it in a thermal chamber that's computer controlled and add a DAC or EEPOT for the trimming. Hook that up to your computer along with your meter and measure the offset over a temperature range or whatever.


Sent from my Smartphone

[paulie]

He let the flood gates open paulie. At first we were poking fun at the build quality. That is no worse than what Dave does for a living. Then we heard about Angels, mob hits and immortality. He brought it on himself.

Unfortunately true but not even close to an excuse for the behavior shown here. Anybody reviewing this thread will see the "he started it" line don't wash.

IMHO you have to admit that the build quality leaves much to be desired.

No, I don't. "Pretty" was never the number one item at the top of my list. Certainly not to the sacrifice of function and performance. IMO just another excuse for perpetrating the feeding frenzy.

It's obvious from all the threats to steal his name, design, or market that those individuals wish they had come up with something nearly as good. At least two of these guys have complained in the past about Chinese stealing ideas from good hard working westerners which makes the situation even more disgusting.

BTW claims it's simply mfg reference design appear lame. Let's see anybody succeed with that approach instead of coveting his innovations. I'm not an expert but do suspect it might actually be an improvement over the adored Geller product and this is just a case of "not invented here" syndrome.

http://en.wikipedia.org/wiki/Cult_of_personality

[EEVblog]

My thinking is that there appears to still be a market for the voltage transfer service that Joe Geller had [which was more for transferring 10V from his 732B to the buyer, without costing US$500 each time].

It's easy if anyone wants to and has the appropriate gear, you just buy the reference, stick in a box, and trim.
The reference on it's own is more than good enough to do this. No need for sprinkling fairy dust on it.

It would take Joe about 8-hours in a small chamber he built to calibrate the TempCo of each SVR-T, and that combined with the high cost for a proper [low uncertainty] calibration of his 732B every year ended up being a zero-sum business, so he had to quit doing it.
I think this can still be done, but one would have to charge more for the reference itself, and more for the calibration in order for it to be a viable business model.

It will never make you rich I suspect.
But no need for the chamber and labour, just rely on the reference, trim it, and ship it as a transfer standard. Spec it appropriately which will be more than good enough for most people.
Remember, the only thing that made the Geller reference so popular was the low cost.

[nctnico]

IMHO you have to admit that the build quality leaves much to be desired.

No, I don't. "Pretty" was never the number one item at the top of my list. Certainly not to the sacrifice of function and performance. IMO just another excuse for perpetrating the feeding frenzy.

I agree about a circuit not needing to be in a fancy casing or maybe the casing has some writing on it. I have bought some commercial products that way and I'm not complaining. And even in high end gear you'll find the occasional bodge wire. Again I'm not complaining.

BUT... When it comes to constructing a circuit which needs to be accurate at levels approaching single digit ppm there is a lot which can go wrong. Read about moisture and thermal EMF and their influence for starters. Lately I've been reading a bit about constructing circuits which work at ppm level accuracies/stability and in that arena there are lots of influences which can introduce a huge uncertainty in a circuit especially over a longer period of time.

[EEVblog]

Well, if you have the chamber and all of the other equipment needed, then there is no reason why the TempCo compensation can't be totally automated [as well as the final voltage trim].  This could certainly be done "covers on" with the appropriate circuitry.
One of the reasons with using an AD587UQ in this application is cost.  One thing I thought of doing was add a small "jelly-bean" microcontroller, that measures the temperature and adjusts the output voltage in real-time.  This same MCU [through it's serial port] would have an EEPROM to store the TempCo constants and the final voltage adjustment value.

I just had the exact same thought, and yes, it would be the only way to go if you wanted to make a tempco adjusted reference at an affordable cost.
Just like they do for MCTCXO's. You build a custom jig with a dozen or more units and stick in a thermal chamber and automate the entire process. Should be able to make a few dozen units a day, or more if you stacked units in the chamber.

[paulie]

Lately I've been reading a bit about constructing circuits which work at ppm level accuracies/stability and in that arena there are lots of influences which can introduce a huge uncertainty in a circuit especially over a longer period of time.

No real experience in this realm but I've been following the LTZ1000 and LM399 threads since the start and everything you say sounds true. The problem is does anyone have a handle on what matters and how much. For example  endless arguments about the spiral holes cut in the PCB around the reference chip. Many said it was just a ritual holdover from an unrelated product. Lots of contradictory opinions about those but not so much in the way of hard evidence.

So the real question is are things like those perfboard holes and solder globs really important for the high performance as the designer claims or just defensive manic ramblings. It wouldn't be too hard to prove with proper experimental setup and I'd love to do that but lack the resources and expertise.

[IanB]

So the real question is are things like those perfboard holes and solder globs really important for the high performance as the designer claims or just defensive manic ramblings. It wouldn't be too hard to prove with proper experimental setup and I'd love to do that but lack the resources and expertise.

Every electronic component is affected at some level by mechanical stresses and soldering things to a PCB immediately clamps them tightly and transmits stresses through the leads. To avoid any issues arising from this source the best option would be point to point wiring with free floating connections.

PCBs were designed to automate production, not to make circuits look pretty.

[Terabyte2007]

Lately I've been reading a bit about constructing circuits which work at ppm level accuracies/stability and in that arena there are lots of influences which can introduce a huge uncertainty in a circuit especially over a longer period of time.

No real experience in this realm but I've been following the LTZ1000 and LM399 threads since the start and everything you say sounds true. The problem is does anyone have a handle on what matters and how much. For example  endless arguments about the spiral holes cut in the PCB around the reference chip. Many said it was just a ritual holdover from an unrelated product. Lots of contradictory opinions about those but not so much in the way of hard evidence.

So the real question is are things like those perfboard holes and solder globs really important for the high performance as the designer claims or just defensive manic ramblings. It wouldn't be too hard to prove with proper experimental setup and I'd love to do that but lack the resources and expertise.

I find it hard to believe that the perf board construction and globs of solder are necessary for proper operation. I know for a fact you can develop a very clean and tidy PCB with all the proper construction to accomplish the same thing or better. I really don't understand why the seller of these devices does not want to do this. Does he think that nobody would ever open this unit?

[nctnico]

So the real question is are things like those perfboard holes and solder globs really important for the high performance as the designer claims or just defensive manic ramblings. It wouldn't be too hard to prove with proper experimental setup and I'd love to do that but lack the resources and expertise.

As others pointed out the process of verifying long term stability is long and cumbersome. Only after a few months or even a year you can tell whether a certain construction is better or worse. So if you want to build or judge such a circuit you have to take a leap of faith and follow the path taken by people before you. Personally I have more confidence in the people at Linear Technology to know what they are doing (especially with the legacy of someone like Jim Williams to back their claims) than someone who put something together but is not able to produce numbers on long term stability.

[IanB]

very clean and tidy PCB with all the proper construction

This is mainly about aesthetics. Granted, many people take pride in their work and aspire to make their products look good, but one should not assume that looking good is necessary for performance. For instance, if you open up an IC package and look at the bond wires they might be very untidy. They may not be fixed in the exact center of the pads, and they may even have bends in them. But does this make you dismiss the IC as badly made?

[robrenz]

There are ways to shape leads to greatly minimize the stresses transferred to the component if short lead lengths are not necessary but I rarely see any examples on the forum.

[Terabyte2007]

very clean and tidy PCB with all the proper construction

This is mainly about aesthetics. Granted, many people take pride in their work and aspire to make their products look good, but one should not assume that looking good is necessary for performance. For instance, if you open up an IC package and look at the bond wires they might be very untidy. They may not be fixed in the exact center of the pads, and they may even have bends in them. But does this make you dismiss the IC as badly made?

I understand what you are saying, but at the same time it speaks prototype. Sure, you can make a product stable and work with prototype materials but do you really want to take it to market that way? I just feel it can be taken to the next level, that's all.

[rx8pilot]

PCBs were designed to automate production, not to make circuits look pretty.

PCB's were also designed to get better performance by being bale to arrange components in ideal positions while isolating delicate areas with ground planes etc. Try making a high performance SMPS on a perfboard or high-speed digital circuits. The PCB is a part of the circuit performance, sometimes critical and sometimes less so.

[Jay_Diddy_B]

Hi group,

I think if I was building a reference I would follow the design shown by the forum member 'Blackdog' in this thread:

https://www.eevblog.com/forum/testgear/6-5-digit-dmm-keysight-34461a-or-rigol-dm3068-or/msg604327/#msg604327

Here is the schematic:



I propose the following changes:

1) Replace the LM10 with a LT3082. This will have lower noise performance. The noise of the reference is not gained up in this single resistor set topology.

2) I would use LT5400-1 Quad matched resistor network for the critical resistors R12 and R17. The LT5400 is very economical way of buy the tempco matched resistors.



What I like about this design

1) It is flexible, the circuit will work with just one LT1021-5 references. More can be added if the need arises. Using multiple references reduces the noise by SQRT n.

2) Blackbox has a note that he built the circuit in a 42^oC bimbox oven. This could be optional. The circuit will behave reasonably well without the oven.

3) Using multiple reference will also reduce the tempco

To get the best performance, use the metal can versions of the LT1021-5

Blackdog also included battery back-up to keep the references powered continually.

I should add:

This is about building a very good reference at reasonable price. It is not about building a reference to challenge the fluke 732a/b at any price using expensive parts.

Regards,

Jay_Diddy_B

[richiem]

I mentioned on the LTZ1000 thread that I was interested to see what Awesome14's D-105 could do so I was buying one. It came yesterday. In the interest of more light and less heat, here is what I've found so far. The results are not metrology lab measurements -- they are a guide only, but I've provided enough info to make the results useful. One note -- I have thought for a while that my HP 3458A is reading 10VDC about 6ppm high, although it is in spec....

Calibratory D-105 10V standard evaluation, started on 2-7-15
Instruments used in evaluation:
— HP 3458A, calibration checked on 7 Aug. 2014, by Keysight in Loveleand — no adjustments made, found to be completely in spec; this is the standard model — 8ppm ref board. Used to monitor 10V standard outputs. 3458A powered on for last 6 days; internal temp 36.5 °C. ACAL 1 (DC) run at 4:30PM. NPLC=60, NDIG=8
— HP 3456A, cal’d against HP 3458A. Used for monitoring ambient temps with Pt100 RTD.
— Heath 2718 Linear Triple Bench Supply — each 20V supply powering one standard..
— Geller SVR (not the temp comp unit), purchased in 2007, used for approximately 12 hours total since then up to this time. On for the last 8 hours (included in total of 12), since 8AM with 15VDC power. Geller’s 2007 certificate says 10.000 00V, Tolerance 5ppm/°C, 15ppm/1k hrs. Cal’d at 23°C ambient.

D-105 unboxed in lab at 4PM. Connected to bench supply at 14.5VDC, initial ambient temp at case 23.0 °C.  Ambient measured immediately adjacent to and in between the two standards, all sitting on top of an HP 8904A that was powered off — makes a handy stand with some thermal mass.. D-105 output started a few hundred microvolts below 10V, quickly rose to 10.000 05VDC. Calibratory’s certificate says 10.000 0014V, cal’d at 20°C ambient.

Both the Geller SVR and the D-105 outputs are connected with 24ga. solid copper insulated wires (old-school telephone extension wire) to gold flashed dual banana plugs to ease connections to the 3458A. The wires are soldered to the Geller SVR and clamped in the gold-flashed bindind posts of the D-105.

Elapsed time     Ambient   Geller SVR     D-105
for D-105      °C
0.6 hr               23.5           10.000 060     10.000 058
3.4 hr      22.9            10.000 071     10.000 068
5.3 hr      23.6            10.000 060     10.000 066
6.4 hr      23.8            10.000 053     10.000 068
Overnight break in measurements with lab cool-down to thermostat set 66°F, reheat in AM to 71°F; standards left powered on.
20.1 hr           22.4      10.000 092     10.000 078
22.5 hr           22.5      10.000 088     10.000 078
Moved to warmer location on top of HP 3456A, standards and RTD grouped as before.
22.8 hr           28.9      10.000 017     10.000 047
23 hr              29.3          10.000 002     10.000 042
24 hr              29.5        9.999 997     10.000 043
Standards powered down. Will begin from room ambient start in 18 hours or so.

At this point, a simple "box" of drift v. temp shows that average drift for the D-105 is approximately 0.5ppm/°C; the Geller SVR average drift is approximately 1.4ppm/°C (pretty good).

[TunerSandwich]

none of these results are "astounding".....

If someone is inclined to tie up their equipment for LONG term testing.....then ok....+/- 2 ppm would be "somewhat impressive"....

Keep in mind the REF102C has an "enhanced" stability rating, after "168hrs stabilization period".....

A two day test shows very little....as I can get a piss poor MAX6350, on a piss poor perfboard layout....in NO enclosure, to keep +/- 1.5 ppm on my home bench DM3068....across 2 days....and that ain't no 3458A (nor even close)....

If you guys want transfer standards....www.voltagestandard.com makes some fairly decent units.....with some realistic quotations on "specs"....

At some point, when doing this kind of design, you have to take a step back and set a "realistic expectation".....

I have played with parallel REF102C, as shown in Ti's applications schematics.....it does help noise.....but again I am not sure at what point you try to make something "better" than stated in the datasheet, and then at what point you start going off into volt-nut land and wind up with a wholly impractical design....which you can't profit from.....

At some point you will spend as much as one would just buying a JJA primary standard.....which you will need to verify your own little standard against, at some point anyhow.....

When you get down to these kinds of numbers....then every last little detail matters.....you might even consider "shock mounting" the components, to avoid harmonic resonances from cosmic rays impacting the poor little ref....and other such nonsense....which obviously Awesome14 has done....with his "heatpipes".....it just doesn't make sense, from an practical standpoint....that one should spend such vast resources and effort to try and build such a thing.....I mean what is it's real practical use?

At some point, like I said above, you will have to have it sent out for comparison against a primary standard....and that will cost more than you will ever net back from such a thing.....

I've been down this road, and gotten to the point of diminishing returns......no sweat....we all go a bit nuts sometimes......it's a lot of fun, but I don't think you would ever be able to provide such devices, on the open market and turn a profit.....

I would be fun to have a community effort to design an easy DiY EEVblog voltage standard.....NOW that sounds like fun, because we could all measure our "identical" standards against eachother....and log the results....which would actually be a HUGELY useful set of data to take academic indicators from.....imagine a loose comparison of a few dozen identical standards, against all the varying conditions each individual "lab" has.....that sounds like a  nice statistical effort and might actually lead to some discovery, for those that don't have any experience in this field and want to learn....and even those who think they know alrady

[rx8pilot]

At this point, for me, it does not matter so much for a couple of reasons.

The first is that a voltage standard is a critical piece of equipment that I should be able to rely on as a true reference. It's too much for me to trust a dead bug/perf board constructed device. I feel that it's long term performance would be hard to predict. If I have such critical work that I need standards in-house, it needs to be something that goes beyond a science fair experiment. The functionality should not be mysterious.

The second would be that such a bizarre rubbish of conversation subtracts any confidence at all even if I was willing to overlook the construction.

I don't know much about design of super-precise voltage standards and the unique testing required to verify the performance. I do know that when the end result must be VERY EXACT, everything that goes into it must be EXACT. Obviously that is not really the case here. Is it unique and special? yes, but not in a good way. If the box and the seller are VERY difficult to believe, why would I use this device as a REFERENCE STANDARD?

As for the community project idea, I will pitch in CNC machined cases. They can have a lot of thermal mass making them easy to temp control. If temp temp and humidity are really factors to consider, I can easily make a nitrogen purged case that can be temp controlled to something above typical ambient so that external temp/humidy will not have a chance to impact the circuit. Not sure if that will help, but I am hearing a lot about various tempco challenges and humidity. It seems that isolating a good circuit in a small box from almost all outside influences should off the best chance of making something with god like specs.

[blackdog]

Hi Jay_Diddy_B,

The 8V powersupply build around the LM10 verry low noise!
There is a lot of filtering going on.
The schematic of my 8V powersuppply kicks LT3082 but on noise ;-)

The 10K resistors are Vishay presision type, somting like S102.
No problem, if u want to use the LT5400.
If wil change this week, the LT1012 for a LTC2057.

P1 at the moment is a 10 turn spectrol 534 model potmeter.
But for the best stability use a Vishay special series trimpot in de heathed box, but they are expensive about 15$

Kind regarts,
Blackdog (not Blackbox *grin*)

[TunerSandwich]

One suggestion, in building a "community standard" is to keep it simple....as most of us that have 3458A or similar, can't afford to tie up those devices for "long term" stability testing....so perhaps designing standards that simply meet the potential of "run of the mill" 61/2 digit DMM's is the way to go....the insight gained from broad use is MUCH more important than insight gained in ONE specific controlled environment.....

The value of data across a diversity of unique scenarios is in fact VERY valuable.....especially in determining WHY or WHY NOT something meets "spec"....

At least that has been my experience.....I deal with a lot of precision measurements and this has been invaluable experience to me. 

Step one is in standardizing a set of benchmarks for qualification, based on a cross section of "average" DMM drifts.....I mean what good is a "2ppm" standard, if you can't keep your DMM stable past 15ppm.....it's useless at that point....

The ideal way to do this would be to build "identical" transfer standards...and then have them all calibrated from a SINGLE primary reference.....send them out and see what they do in the real world....then try to figure out why they did or did not do this or that.....

Getting down to sub ppm drifts isn't entirely practical or useful, for anything other than academic purposes and speculation....I know I personally don't mind tying up a DMM to do this with....

In fact i have been designing a standard that has a built in 24 bit ADC, and associated display....and has a built in wi-fi signal out to the network....that way I can run it indefinitely, not tie up any gear, and gain insight on how and why things happen or don't happen....I had posted a block diagram of this a while back in another thread.....

IMO the value in a standard isn't it's absolute resolution....but it's absolute drift over time.....I would love to have a dedicated, temperature controlled chamber...shielded, that i could run 24/7 and log the data from....and 5 years down the line have some real answers on what long term accuracy means....

I have been working on this project off and on for the last few months....and I actually chose the REF102C as a starting point....because it's easy to implement and a decent value.....

Another, possibly "better" scenario...is to build a discreet standard....that way folks can test what each individual component does, and how manipulating them changes outcomes....

Again we have to segment what is practical day to day, and what we do to gain academic insights....

Obviously I am throwing a lot of opinions out there, but opinions usually lead to facts.....if those facts support the initial opinions, is what engineering is all about....

The idea behind the wi-fi output...is a simple way to send data back to a centralized database, and be able to "check-in" on the standard remotely....or set alarms for large changes etc....imagine a "community standard" where each participants device checks in on a regular basis...and the database does a comparative analysis across ALL devices....and updates a master drift figure....

So for example 100 standards in a specific region drifted a total of XX over a given period of time.....then analyze WHY....

Another key here is battery backup....a device going offline basically destroys ANY time put into it before it lost power....