Calibratory D-105 DC Precision Voltage Reference Standard

[engineer_in_shorts]

When soldering the REF50xx chips, I put the .............

SNIP

Ref50xx is a bandgap device, what the point 

Heat still affects it? Just like all encapsulated silicon devices. The thermal shock creates rapid expansion and contraction leading to micro-fissures in the die. They start at the edges and spread inward. It can affect anything from the initial accuracy to long term drift.

Exactly the same as on a Zener based device. Why would you think it wouldn't?

(Incidentally, these cracks tend to appear when a wafer is diced. Laser "trimming" of resistors etc. before encapsulation can worsen them.)


Sent from my Smartphone

Bandgap devices are noisy, they don't really deserve the extra time. They are not suitable for this kind of 'working standard' reference work.

[engineer_in_shorts]

It has closed-chassis calibration controls. That means you don't need to open the chassis to calibrate the unit.

And how do you do that exactly?
Please be precise in your detailed explanation.
Also, please explain by what mechanism removing the rear case causes damage to the circuitry or calibration.
Once again, please be precise.

Dave not sure why you need this detail.  He is correct, you do not need to open it to calibrate it.
Calibration means comparison between measurements.
Clearly you can calibrate it without opening it 


When there is a discussion about calibration with a non-technical person, it helps for simplicity sake to use phrase "Calibration-Verification" and "Calibration-adjustment" to differentiate.

In the case of modern test equipment it's almost always the latter. Stop being pedantic, nobody likes that guy; don't be that guy.

No, it's the case FOR ALL test and measurement equipment, new or old, modern or antique. If I sent my instruments for 'calibration' and they adjusted it without clear instruction to do so, I would be pissed off.

It's not about being pedantic, it is explaining to people who do not understand basic metrology principles and terminology.

Any lab worth it's salt would say something like: "Do you want adjustment as well as verification? We would charge more for adjustment but we can do before and after calibration certificates."

[helius]

The phrase "closed-case calibration" is similarly nonambiguous. It always means, electronically adjustable by a diagnostic port. This is a standard usage throughout the industry.

[Terabyte2007]

I think the seller means by "closed-case calibration", is esentially, I whacked a couple of 10-turn pots inside and stuck them through the case! 

[splin]

no offense to the gentleman who makes the alternate ref mentioned here.....but why do people keep buying these things w/o seeing ANY specs or data.....

You are absolutely right. I did not give any spec's, but referred to the datasheet. And I did not make any promises, just relaying the datasheet. 

My intension has never bin to sell this reference. I made one for myself, and sold the leftovers. Min order was 10 PBC's.
No offense taken.

Generally surface mount REF IC's aren't as stable as their through hole counterparts.....

True again.
I aged all my references, and the datasheet shows clearly that the difference after 1000 hours is only 5 ppm/1000 hours.

I am thinking of making a second version, and perhaps I use a dip. I even consider putting the ref on a "island", to minimize the mechanical stress on the device.

No! It's 45 or 50ppm/1k hrs after the first 1k hrs (depending on package), not 5ppm! Did you actually bother to read TunerSandwich's post that you replied to? He helpfully showed you the actual drift specifications and a link to the current datasheet. Unfortunately I think you got caught out out by the headlined 5ppm drift claim in earlier versions of the REF5025 datasheet (but not stated in the actual specifications). It was almost certainly a mistake and corrected in later releases - see Reply #4 from this thread here:

https://www.eevblog.com/forum/projects/low-cost-voltage-references-long-term-stability/

You might also find this interesting about TI's approach to lifetime parameter specifications:

https://www.eevblog.com/forum/projects/semiconductor-ten-year-lifetime-parameter-shift-tests/

I'm with TunerSandwich and engineer_in_shorts - what is the point of saving a dollar or two when you are spending a significant amount of time building, testing and ageing a few boards?  - how much is your time worth? For example, consider the ADR4525 and the AD586 shown in the table below - you don't even need to age the latter two parts and would still have significantly better drift along with the lower noise and temperature drift benefits.

Of course it also depends on how much faith you put put in the published drift specifications as few people will have the time and resources to actually verify them.

There are other choices, such as hermetically sealed references, which may be better if hysterisis is important to you.

Reference	Cost	TC	Noise	Tol	Drift
REF5025	$6.00	3 (ppm/C)	7.50 (uV pk-pk)	.05%	125ppm (1k hrs)  45ppm (2nd 1k hrs)
ADR4525	$6.92	2 (ppm/C)	1.25 (uV pk-pk)	.02%	25ppm/sqrt(1k hours)
AD586M	$9.86	2 (ppm/C)	2.00 (uV pk-pk)	.05%	15ppm/sqrt(1k hours)

Note: Cost shown is from Digikey (25 off). The AD586 is 5V - the noise value shown is after dividing by 2 for comparison puposes.

[timb]

I generally find the REF50xx series fairly consistent from one part to the next. The REF5010's I've been tracking all have an almost identical drift curve. I've also consistently gotten much better than 125ppm/1khr, closer to 25ppm. They're easy to obtain and work with, so I usually use them on boards that have ADC's and DAC's on them.

(It's also an inexpensive way to make a very accurate 1kV reference. Chain 100 of them together.)


Sent from my Tablet

[JohnnyBerg]

I aged all my references, and the datasheet shows clearly that the difference after 1000 hours is only 5 ppm/1000 hours.

No! It's 45 or 50ppm/1k hrs after the first 1k hrs (depending on package), not 5ppm! Did you actually bother to read TunerSandwich's post that you replied to?

Read again .. I stated that the difference between the so8 and the dip version is only is only 5 ppm/1000 hours after 1000 hours.

I'm with TunerSandwich and engineer_in_shorts - what is the point of saving a dollar or two when you are spending a significant amount of time building, testing and ageing a few boards?  - how much is your time worth?

Excuse me, for wasting my time. Sorry, I will not do that again.

There are other choices, such as hermetically sealed references, which may be better if hysterisis is important to you.

Sure there are. But you have to start somewhere. I found the REF5025A (the A is important) a very good device. Of course paying more money, will get you a better device. The ADM586M is $10. How about a LTZ1000?

The thing is, when engeneering there is always a trade off between specs - result - price and invested time. My goal was a device that gave rock solid reference, to verify and test multimeters up to 6.5 digits. The REF5025A for fills that spec.

[timb]

The A is the lesser model though...


Sent from my Tablet

[JohnnyBerg]

The A is the lesser model though...

It is not about who has the biggest ..
It is about which device gets the job done.



Has been so for a week or 2 .. rock solid.

[TunerSandwich]

34401A specs *after 1 hr warmup

[JohnnyBerg]

@TunerSandwich: I did say nothing about its absolute value.  Neither did I make a statement about the temperature in my lab.
What is the point you are trying to make?

[timb]

The A is the lesser model though...

It is not about who has the biggest ..
It is about which device gets the job done.



Has been so for a week or 2 .. rock solid.

Yeah, that's very true! Sorry, I've seen some people posting in another Vref thread thinking the A model was the higher grade. TI's binning scheme is confusing as hell. (Non-Suffixed parts are the highest grade; A suffix are the lowest.)

I wasn't sure if you were trying to say the part was better or not, but reading back I see now. My bad!

I can totally believe those results as I've seen it too on a 7.5 digit meter.


Sent from my Smartphone

[alhoop]

I generally find the REF50xx series fairly consistent from one part to the next. The REF5010's I've been tracking all have an almost identical drift curve. I've also consistently gotten much better than 125ppm/1khr, closer to 25ppm. They're easy to obtain and work with, so I usually use them on boards that have ADC's and DAC's on them.

(It's also an inexpensive way to make a very accurate 1kV reference. Chain 100 of them together.)


Sent from my Tablet

Tim:
do you know what 1kv is?
Al

[timb]

...Uh, it's o...one thousand volts...isn't it?


Sent from my Tablet

[timb]

The Mars-Energo lab used 10,000 REF5010's, connected serially, to create a 100kV reference. Initially they did a test with 100 chips to create a 1000V reference.

Source: http://www.ti.com/lit/an/sbaa203/sbaa203.pdf


Sent from my Tablet

[JohnnyBerg]

The Mars-Energo lab used 10,000 REF5010's, connected serially, to create a 100kV reference. Initially they did a test with 100 chips to create a 1000V reference.

Source: http://www.ti.com/lit/an/sbaa203/sbaa203.pdf

Wauw .. not only  the voltage, but also the money 

I like the REF50XX. It has proven to be very stable.

I have a some other reference IC's coming. LT1236, LT1021. For my B revision
It will come in dip, and so8 and will have the reference IC's placed on an island, as shown by Lineair in the video. I am designing the current source (perhaps a Howland Current pump), with respect to ground, so one does only have to change the "plus" lead when switching from voltage to current.

I also tried the make a "poor mans reference", as  nearly all the money goes into the reference IC. I did some tests with a TL431B. That failed miserably. Also I tried a LM4040. That was better then the TL431B, but still not good enough for the "poor mans reference" 

[paulie]

LOL. Looks like "poor man's" is the new "el-cheapo" or "ghetto". Back in the old days it meant for a man that was actually poor. In other words lowest cost solution. Lately it seems to mean expensive but not outrageously so. Or maybe only a little outrageously so. For example:

So, some projects that have been "stewing" in my lab are:

• Poor-man's primary voltage transfer standard [based on LTZ1000(A)]
• Poor-man's voltage transfer device
• Poor-man's multi-junction thermal converter for AC/DC transfers
• Poor-man's resistance standards [1R, 10R, 100R, 1K, 10K, 100K, 1M, 10M, 100M, 1G]
• Poor-man's resistance synthesizer [that uses the above resistors to get any value]
• Poor-man's direct current comparator bridge for low resistance measurement
• Poor-man's electrometer for high resistance measurement
• Poor-man's calibrator for DCV, ACV, DCI, ACI
• Poor-man's automatic LCR bridge
• (and a few others that are still in the idea stage)

Probably most of those end up costing hundreds if not thousands. I've built many references based on 4040a and tl431 and also 2 cent zeners. Preliminary tracking (few days) show more than adequate for "poor man's" meter (4-5 digit) calibration or reference for most electronics experiments. Or maybe 6 1/2 - 7 1/2 digit lab meters costing only few thousand are considered "poor man's"?

In my world it means something can be built for pennies or a dollar or two at most. Let's see you build a "good enough" voltage reference for 25 cents. You know, like with 4040a or tl431. Wait... tl431 runs only 5 cents plus penny for the resistor!

ps. I figure if we are gonna hijack this guys thread I might as well get in on it.

[blackdog]

Hi,

If you want to see what is possible with a TL431, look at this topic from me, on this Dutch site.

You can use Google translate 
http://www.circuitsonline.net/forum/view/116156/1/tl431

Afther about 40 hours the stability was about +-3 PPM, not bad for a 431A
I dit not test the reference for about 1000 hours, because its used in a power supply with a 10 turn potmeter.
Long-term stability is not so interesting in this application.

Kind regarts,
Blackdog

[TunerSandwich]

time + resources to characterize these "el cheapo" components doesn't allow for a "poor mans" anything.....unless your time is worthless

[blackdog]

Hi TunerSandwich,

This all depends on your point of view...
time + resources to characterize these "el cheapo" components doesn't allow for a "poor mans" anything.....unless your time is worthless

I spent time on low cost references in building them in ovens.
The out come, not wasted time, but a lot of knowledge on references and ovens.
I never thought that a TL431, LT1234, LT1021 or even a LTZ1000 I use, will be better than a Fluke 732B.
I simply don't have the resources to characterize them.
I have i different point of view, not money is important, but knowledge < now i sound like Yoda 

Just my 2 cents 

Kind regarts,
Blackdog

[paulie]

time + resources to characterize these "el cheapo" components doesn't allow for a "poor mans" anything.....unless your time is worthless

Why am I not surprised to hear you say that? In reality a poor man's time is worth very little due to low wage or maybe even no job at all. Fortunately not either for me but I do enjoy accomplishing the most with the least. A wonderful challenge.  Unlikely most on this forum have any capacity to understand this philosophy. Maybe one or two others if that, including dannyf who initiated the el-cheapo/ghetto craze.

FYI I'm about half way through a 1000hr characterization of true "poor man's" references. 16 baking in a temperature controlled oven necessary because of wild temperature extremes in my shop. Not including the oven but counting 5 digit meter, datalogger, and references total cost less than $7. As mentioned initial results might be promising for most casual hobbyists but obviously not of any interest to the volt-nerd crowd.

ps. Dog, thanks for the link. I'm really liking this tl431. Even cheaper back-to-back zeners also interesting with almost zero TC by balancing 5.6v to 6.2v parts.

[TunerSandwich]

time + resources to characterize these "el cheapo" components doesn't allow for a "poor mans" anything.....unless your time is worthless

Why am I not surprised to hear you say that? In reality a poor man's time is worth very little due to low wage or maybe even no job at all. Fortunately not either for me but I do enjoy accomplishing the most with the least. A wonderful challenge.  Unlikely most on this forum have any capacity to understand this philosophy. Maybe one or two others if that, including dannyf who initiated the el-cheapo/ghetto craze.

FYI I'm about half way through a 1000hr characterization of true "poor man's" references. 16 baking in a temperature controlled oven necessary because of wild temperature extremes in my shop. Not including the oven but counting 5 digit meter, datalogger, and references total cost less than $7. As mentioned initial results might be promising for most casual hobbyists but obviously not of any interest to the volt-nerd crowd.

ps. Dog, thanks for the link. I'm really liking this tl431. Even cheaper back-to-back zeners also interesting with almost zero TC by balancing 5.6v to 6.2v parts.

A poor man can't afford 2 3458A's + a few 732B's......

show me how you can amortize the cost of equipment + resources into a sub $100 MSRP, and come out ahead?  You can't....

If you are doing this for purely academic purposes....then you are once again not a poor man, because you have no financial gain...and only spend resources.....so again not an excercise for a poor man.....the insight you gain and it's value is 100% personal and only equitable based on your own needs....

"Unlikely most on this forum have any capacity to understand this philosophy"

that is ridiculous.....and an elitist attitude.....maybe YOU don't have the capacity to understand basic economics?

http://www.merriam-webster.com/dictionary/poor

[TunerSandwich]

Hi TunerSandwich,

This all depends on your point of view...
time + resources to characterize these "el cheapo" components doesn't allow for a "poor mans" anything.....unless your time is worthless

I spent time on low cost references in building them in ovens.
The out come, not wasted time, but a lot of knowledge on references and ovens.
I never thought that a TL431, LT1234, LT1021 or even a LTZ1000 I use, will be better than a Fluke 732B.
I simply don't have the resources to characterize them.
I have i different point of view, not money is important, but knowledge < now i sound like Yoda 

Just my 2 cents 

Kind regarts,
Blackdog

you still have to buy points of reference to even know what you have produced.....they aren't "cheap".....when you talk about word such as "poor mans" you are describing economic value.....intellectual value has no value to others....it's only basis for equity is your own assigned value....but that value is not spoken in terms of "dollars"....so "poor" or "expensive" can only be related to things other than monetary condition.....in that regard everything you learn from has value, but to implement that value towards others, one has to spend an economic measure of value to 1. disseminate the information 2. market the resource

If you follow NONE of the above procedure, you wind up with Awesome14's "reference"....which has little to no academic value, and little to no economic value....it's just crap (technical term)

comparing anything anyone here has done to a 732B is absurd, because it isn't just a DiY reference....it has a built in PSU and battery back-up....those have orders of magnitude more value than a PCB with a ref IC soldered in place.....for example...let's say you are trying to characterize performance of one of these DiY jobs and the power goes out?  Or you have to transport it....you are relying on an un-characterized external power source to keep your data relevant.....a wall wart or 9 volt battery is NOT anywhere close to what the Fluke 732B is doing....it can seamlessly switch power sources, and give a predicted sDEV while doing so.....that is what you are paying for....a warranted set of claims from a trusted manufacturer......how could anyone here have anywhere near the level of industry trust of a company like Fluke?  I see no real data from anyone on ANY of these devices.....a static screenshot of something hooked up to a DMM is worthless....show me provable long term drift figures, under a variety of conditions....I am frankly shocked anyone would compare their DiY standard to a Fluke 732B.....apples and oranges

[IanB]

Suppose I make a bunch of homebrew references and compare them against each other. If they drift apart over time I have a problem, they are not very predictable and I can't do much with them. But if they all stick together, then either they are very stable, or they are all drifting at similar rates. Either outcome is good. If they are all drifting at similar rates, that rate is predictable and can be compensated. Making something predictable at low cost is good in terms of a transfer standard, especially if the predictability survives power down during shipping. Just because there are whizz-bang industry devices doesn't mean there is no room for the amateur to make a contribution.

[TunerSandwich]

Suppose I make a bunch of homebrew references and compare them against each other. If they drift apart over time I have a problem, they are not very predictable and I can't do much with them. But if they all stick together, then either they are very stable, or they are all drifting at similar rates. Either outcome is good. If they are all drifting at similar rates, that rate is predictable and can be compensated. Making something predictable at low cost is good in terms of a transfer standard, especially if the predictability survives power down during shipping. Just because there are whizz-bang industry devices doesn't mean there is no room for the amateur to make a contribution.

you still have to take the time and buy the resources to gather those benchmarks.....I am all for DiY and lower cost solutions, but I don't think people take into account the actual investment....

even if you did it on borrowed gear and your mom pays the rent and electricity....it still costs someone something.....and those DMM's and baseline ref aren't "cheap"

I have a sizable collection of DiY "standards", and they were all "fun learning experiences", but they weren't cost effective and I would never compare them to a Fluke 732B with a NIST traceable cal cert.....nor do I think any industry with a large consumer liability would trust them either.....if for any reason insurance purposes....if some whizz bang widget takes someones arm off and you can't document or prove you followed generally accepted industry practices.....yikes, I am not brave enough to employ my DiY ref in any critical application.....so again they become hobby devices, and since a hobby generates no revenue...they aren't "poor mans" devices....in fact they are shockingly expensive, because I can't apply them to anything that has financial gain.....

personal knowledge is a great thing....and I am all for experimentation etc....but when terms that classify economic standing are assigned, then it all goes to hell....