Idea: binned, salvaged LM399's from taobao for forum members

[cellularmitosis]

Hello,

I have recently started dabbling in ordering salvaged electronic components (e.g. obsolete packages of precision components, etc) from taobao, including LM399's, and I had an idea of a service which might be of interest to forum members, particularly the volt-nuts on a budget who would like to track the volt at home.

Background:

If you try to track the drift of an LM399 by measuring it directly, you need to at least use a 6.5-digit meter, but even then, on the 10V range you are really just comparing your LM399 to the LM399 inside of the meter.  So what exactly are you measuring?

If you instead measure the difference between two LM399's, you can use the 1V range, which gives you better resolution, and also attenuates the perceived drift of your meter 10x relative to the drift of the LM399.

Even better, if you get lucky and have two LM399's which are less than 100mV apart, you get to use the 100mV range, with 100nV resolution and 100x attenuation.  Fantastic!

But that's just for 6.5-digit meter owners.  What about ballers on a budget, with 5.5-digit meters?

Well, some 5.5-digit meters (e.g. Keithley 195A, HP 3478A, etc) support a 20mV or 30mV scale with 100nV resolution.  This can allow you to track the drift of a few LM399's with results comparable to 6.5-digit meters, but for less money (the Keithley 195A can be found on ebay for under $100 if you are patient, and the 3478A as well if you are more patient).

The problem is, LM399's output voltage can vary quite a bit from unit to unit.

So, for a budget-minded volt-nut to track the volt, he needs access to cheap LM399's, and they need to be binned pretty tightly.

The LM399's are indeed cheaply available (about $3) on taobao, so to make this happen, we just need one person who can order a big enough batch to ensure tight enough binning.

The service I'm thinking about offering:

So, I had an idea to order batches of these LM399's, give them a quick measurement and bin them according to Vz, then sell bundles of them to forum members, binned according to the requirements of the lowest DCV range of their meter.

Would anyone be interested in this kind of service?

Edit: I was thinking bundles of 3 or 4, as you could keep 3 at home and track their drift while mailing a 4th to another forum member (to exchange calibration), and you'd have a reasonable guess at whether the 4th drifted when it was returned to you.  Users with just 3 refs could track them at home and wait for a cal-club ref to be mailed to them.

[TiN]

I'll be a fun breaker in this one.

I always felt like at one side LM399 are too elaborate for beginners (need to take care of heater, thermals, non-convenient Vz for ADC/DAC use, picky on noise). On other side they are too rigid and non-tweakable for hardcore voltnuts (no ability to play with heater setpoint, zener current, tempco tweaking) which make no much sense as LTZ is only 3-5 times of the price for way more fun. So, meh. 

Also 2nd-hand parts with taobao, bbrrrr.. who knows what blowtorch was used to remove them  .

[e61_phil]

If you instead measure the difference between two LM399's, you can use the 1V range, which gives you better resolution, and also attenuates the perceived drift of your meter 10x relative to the drift of the LM399.

Even better, if you get lucky and have two LM399's which are less than 100mV apart, you get to use the 100mV range, with 100nV resolution and 100x attenuation.  Fantastic!

A ppm is a ppm no matter to what voltage you scale it. There is no drift attenuation if you use the 1V range instead of 10V. In this case the drift of the LM399 in your meter becomes less relevant, but you change one LM399 (in the DMM) against another (the external comparasion) only. The only benefit I can see is the increased resolution.

[chickenHeadKnob]

I have to agree with TiN. The problem with reclaimed parts is we have seen them with dodgy new leads attached to them, for appearances  only. That can't be good. Also the whole idea that you save money with Taobao doesn't stand up to scrutiny when reasonable lifetime cost of ownership is tallied in. What I mean is that the real value  of a reference accrues after you have spent time monitoring it. That dwarfs the $3 or $9 original purchase price quickly if you value your own time/effort. Not to mention the other components in the BOM don't change. I think duplicating the Gellar SVT with perhaps some improvements in packaging would be of wider interest

[cellularmitosis]

If you instead measure the difference between two LM399's, you can use the 1V range, which gives you better resolution, and also attenuates the perceived drift of your meter 10x relative to the drift of the LM399.

Even better, if you get lucky and have two LM399's which are less than 100mV apart, you get to use the 100mV range, with 100nV resolution and 100x attenuation.  Fantastic!

A ppm is a ppm no matter to what voltage you scale it. There is no drift attenuation if you use the 1V range instead of 10V. In this case the drift of the LM399 in your meter becomes less relevant, but you change one LM399 (in the DMM) against another (the external comparasion) only. The only benefit I can see is the increased resolution.

Here is my understanding:

When on the 10V range, if the 399 in your meter drifts by a ppm, the measurement is now 10uV off.

When on the 100mV range, the same drift only results in 0.1uV of error.  So that’s 100x attenuation.

[Magnificent Bastard]

I would not trust parts from TaoBao.  For a few more $ you can get parts directly from Linear Tech (until the end of the month anyway when the web site will be taken down), and these have unquestionable pedigrees.  Also, it is instructive and valuable to take a fresh reference and monitor it as it burns in.  After viewing the many threads on this issue on this forum, I think an LM399 based reference has it's place, but only if it has a PWM divider (with filter and buffer of course).  The LM399 is no slouch-- it can have 1-2 ppm/a stability after a long (years) burn-in, and it is far less expensive than the LTZ1000 because no hyper-expensive support resistors are needed.  A well behaved LM399 based reference would be good enough to calibrate up to 6.5-digits (with a guard-band), and that fits the vast majority of hobbyist needs.  For extreme volt-nuts with 7.5+ digit meters, a LTZ1000 based reference is mandatory.

[cellularmitosis]

THanks for the feedback everyone.  I will however counter these viewpoints with “don’t let perfect be the enemy of good”.  These parts won’t be ideal as new parts, but there is still value in them.  I see a lot of “what if” focusing only on the potential down-sides.  Let’s think about the potential up-sides: a beginner gets their feet wet and maybe makes their mistakes / learns their lessons before they’ve invested much on expensive parts.

[Magnificent Bastard]

THanks for the feedback everyone.  I will however counter these viewpoints with “don’t let perfect be the enemy of good”.  These parts won’t be ideal as new parts, but there is still value in them.  I see a lot of “what if” focusing only on the potential down-sides.  Let’s think about the potential up-sides: a beginner gets their feet wet and maybe makes their mistakes / learns their lessons before they’ve invested much on expensive parts.

So, you have already made up your mind, and you are looking for "approvement" (a concatenation of agreement and approval)?

All I'm saying, is that if you were offering these on junkBay, I would probably not buy them.  Others might.  Whatever it is would probably be a damn sight better than a "Calibratory D-105" from @Awesome14...  But, I think this would be a big financial risk for you to take unless you get some pre-order commitments.  Let's say you plan on building 100 of these, at a parts cost of $10.  So, that's $1000 of your money.  Then add a butt-load of time to build them (unless it's a "kit"), and then a burn-in period (I'm guessing 1 year for this reference, but by no means less than 3 months), then calibrate all 100 of them to the best reference you have.  That's a lot of money and a lot of time.  What happens if you only sell 10-20 of them?  OTOH-- maybe all 100 will sell out in the first month-- but that's a big risk to take-- for what benefit again?

The real value of the Geller SVT was than you could get a cheap (but uncertified) transfer of the Volt-- the hardware that Volt was carried on is irrelevant, as long as it could transfer the Volt for 30 days or so with some level of confidence.

Sometimes people get so excited about their inventions and ideas that they forget to do the business analysis.  I'm not being an a$$hole here-- I'm trying to be your friend and to help you to see the whole picture.

Now, if you were going to do what Geller did, and get a Fluke 732A (or B), then get it calibrated at Fluke, then calibrate your references to THAT, well, that would be a totally different story and your gadgets would sell like hot-cakes!

No disrespect... Just sayin'...

[e61_phil]

If you instead measure the difference between two LM399's, you can use the 1V range, which gives you better resolution, and also attenuates the perceived drift of your meter 10x relative to the drift of the LM399.

Even better, if you get lucky and have two LM399's which are less than 100mV apart, you get to use the 100mV range, with 100nV resolution and 100x attenuation.  Fantastic!

A ppm is a ppm no matter to what voltage you scale it. There is no drift attenuation if you use the 1V range instead of 10V. In this case the drift of the LM399 in your meter becomes less relevant, but you change one LM399 (in the DMM) against another (the external comparasion) only. The only benefit I can see is the increased resolution.

Here is my understanding:

When on the 10V range, if the 399 in your meter drifts by a ppm, the measurement is now 10uV off.

When on the 100mV range, the same drift only results in 0.1uV of error.  So that’s 100x attenuation.

Yes, that is right for the LM399 in the DMM, but you simply replace one LM399 against another. Therefore, you still compare one LM399 against another. The only benefit is increased resolution.

[ramon]

2DW232 

[cellularmitosis]

So, you have already made up your mind, and you are looking for "approvement" (a concatenation of agreement and approval)?

Good points here, thanks.

Now, if you were going to do what Geller did, and get a Fluke 732A (or B), then get it calibrated at Fluke, then calibrate your references to THAT, well, that would be a totally different story and your gadgets would sell like hot-cakes!

Hmm, that's interesting.  I hadn't considered characterizing them over time.

Let's also ask that question: would people be interested in Vrefs which have been characterized over time, as a premium service?

[TiN]

Sometimes people get so excited about their inventions and ideas that they forget to do the business analysis.  I'm not being an a$$hole here-- I'm trying to be your friend and to help you to see the whole picture.

Now, if you were going to do what Geller did, and get a Fluke 732A (or B), then get it calibrated at Fluke, then calibrate your references to THAT, well, that would be a totally different story and your gadgets would sell like hot-cakes!

With 732 calibration in the mix and equipment time to provide verified uncertainty, even run of LTZ modules does not make business sense (even without duperpancy VPG resistors, but very reasonably priced PWWs from Edwin instead). However there is slight chance we all are wrong, as I was really surprised on interest of my latest 10pcs run, for which I did not even get fancy resistors delivered yet (still months away).

would people be interested in Vrefs which have been characterized over time, as a premium service?

I can only pretend I'm beginner nut from 7 year back, which just got his first K2001 repaired and in desperate for calibration reference, without 5700 or 600$ to spend on official calibration. I'd rather put 100$ extra on having properly tested and proven stable reference, than a 5$ board with 10$ LM399. Shipping alone of small packet from USA to Taiwan will cost more. However, that's not what I did back then, instead I started LTZ KX project, which ofc ended up in cost 30 times of the abovementioned 100$ for 5 boards...  YMMV.

[eurofox]

So, you have already made up your mind, and you are looking for "approvement" (a concatenation of agreement and approval)?


Let's also ask that question: would people be interested in Vrefs which have been characterized over time, as a premium service?

If quality is on the level or better than Geller Lab I think yes

[Kleinstein]

2DW232

The 2DW232 reference is different from the LM399. It is more comparable to an 1N821 or similar zener reference. It still needs additional circuit and individual adjustment of current to get a low TC. The long time stability is one of the big unknown properties - chances it is better than DIP parts, but there is still glue on the chip and it might need stable resistors more than the LM399 or LTZ1000. There seems to be a trade of between stability and noise: lower voltage compensated Zerners (e.g. 6.2-6.5 V) tend be lower noise while the higher voltage (6.8-7 V) ones can be lower resistance and thus less sensitive to the current setting resistors.

Due to the large scattering, binned (measured) 2DW232 refs could still be an interesting part  - just like the 1N829 maybe more specific. Not so much for absolute voltage, but more for suitable current for low TC and verified low noise.


LM399 accurately measured might make a limited sense, as they are likely a little more stable than the AD58x or similar ref boards one can get from china with a more or less trustworthy accuracy (there are also fake readings). With reused parts some long time test might be required to make sure to have a reasonable starting point - so the minimum would be something like a few days burn in. Due to the case and high normal operating temperature I would not expect that much damage during unsoldering.

[MisterDiodes]

Cell -

I'll be another wet blanket, but take it as constructive thoughts.  Yes, don't expect a 6 digit meter to give you any higher accuracy results at lower ranges when you really calculate your uncertainties.  The big "gotcha" about trying to measure 2 Vrefs back to back is you never know if they have both gone to driftytown for absolute value.  It happens.

I would agree with TiN and others here.  As a business model selling garbage parts is probably not the best plan - I've seen some of the guys that work for us try to by parts like '399's on the garbage sites...and even if you receive the parts there is usually a reason they are so cheap.  Usually because they are really drifty, out of spec or just non-operational - and they were probably removed from non-operational equipment in the first place.  Removed literally with a blowtorch (TiN is not making a joke).  These are the same sellers that will be selling bags of "recycled" parts in China for $1 or $5 per -bag-.  So even at $2 or $3 each you're paying way, way too much for those junk Vrefs.

Just get new parts that are guaranteed.  If something goes wrong with the batch you'll be covered for replacements.

Follow the rule: If it seems too good to be true - it is not true.  If you're known to be selling products built with junk parts that's not going sit well with many buyers, even at a low price point.

If you look around and at past history, Vref makers usually release a product and see a surge of interest initially - and then sales slow down dramatically and they are selling just a few units here and there.  It's just not something people really need all the time, and not a lot of folks are willing to spend a lot of money on products they aren't going to be using all the time.

I know you're aiming for the hobby market - but as a business owner I would have no need for anything suited for a 6 digit meter since all critical measures on done on 8 digit meters, and we keep cal'd 732's on hand keeping those verified.  And that's what our customers expect us to have on hand to measure to their specs.  Even if you're building nice LTZ based 10V Vrefs it would be hard to break into the 732a/b market because those expensive Flukes are the defacto standard and are well proven over time.  We would never waste time on buying an LM399 Vref made from junk parts - why even bother when a good '399 is < $10 anyway?.  Even then a '399 wouldn't fit our needs - but other companies might want that IF they were built with good parts.

It's kind of a very small & narrow market.  There could be a small business in building some economy Vrefs; not sure you'd see any real profit.  The rule of thumb is:  Figure out your  complete cost of production of your gadget (Parts, boards, labor time, testing, burn-in, characterization time, shipping, etc.) then multiply that by about 3 to get an approx retail value.  If you can't sell them for at least that amount it's probably not going to be a good business model.  It all depends on volume and competition of course but that's just a typical starting number.

Just do research on other attempts at selling VRefs at a lower price point and decide what you want to risk.  You're not the first or last one with this idea, so learn from others.

[zhtoor]

hello,

reportedly some versions of valhalla 2720GS implemented some kind of online monitoring
of upto 8 LM399's (1N827's also) and maintained online drift data by comparing each
of the references with either their means or within themselves, and the system could
automatically take out the offending (excessive drifts) references out of the average.

if some board like that could be produced using LM399's and some cheap microcontroller / adc
combination and using binned LM399's for relatively tight grouping, thereby using a cheaper
measurement of relative drift.

such a board could be pretty useful, i am not so sure how difficult it would be to implement.

keep up the good work Mr. Pepas !

regards.

-zia

[mycroft]

As a newbie I see this as a valid proposal. Not as a product but as a learning tool. As such the burn-in is optional.

THanks for the feedback everyone.  I will however counter these viewpoints with “don’t let perfect be the enemy of good”.  These parts won’t be ideal as new parts, but there is still value in them.  I see a lot of “what if” focusing only on the potential down-sides.  Let’s think about the potential up-sides: a beginner gets their feet wet and maybe makes their mistakes / learns their lessons before they’ve invested much on expensive parts.

[cellularmitosis]

Thanks for the replies, all.  I have a handful of these salvaged parts already, so perhaps I should start be characterizing them and seeing if they are any good.

[Cerebus]

So, you have already made up your mind, and you are looking for "approvement" (a concatenation of agreement and approval)?

All I'm saying, is that if you were offering these on junkBay, I would probably not buy them.  Others might.  Whatever it is would probably be a damn sight better than a "Calibratory D-105" from @Awesome14...  But, I think this would be a big financial risk for you to take unless you get some pre-order commitments.  Let's say you plan on building 100 of these, at a parts cost of $10.  So, that's $1000 of your money.  Then add a butt-load of time to build them (unless it's a "kit"), and then a burn-in period (I'm guessing 1 year for this reference, but by no means less than 3 months), then calibrate all 100 of them to the best reference you have.  That's a lot of money and a lot of time.  What happens if you only sell 10-20 of them?  OTOH-- maybe all 100 will sell out in the first month-- but that's a big risk to take-- for what benefit again?

The real value of the Geller SVT was than you could get a cheap (but uncertified) transfer of the Volt-- the hardware that Volt was carried on is irrelevant, as long as it could transfer the Volt for 30 days or so with some level of confidence.

Sometimes people get so excited about their inventions and ideas that they forget to do the business analysis.  I'm not being an a$$hole here-- I'm trying to be your friend and to help you to see the whole picture.

Now, if you were going to do what Geller did, and get a Fluke 732A (or B), then get it calibrated at Fluke, then calibrate your references to THAT, well, that would be a totally different story and your gadgets would sell like hot-cakes!

No disrespect... Just sayin'...

You seem to have invented some board, gadget or kit that cellularmitosis is planning to sell. The problem with your argument is that he made no such suggestion. He's suggesting getting naked salvaged LM399s and collecting them into packs which have Vz in the same 100mV band, that is all.

[Echo88]

Regarding measuring voltage references in opposition to use smaller voltage ranges on the used DMM:
Which supply-design is best suited to supply multiple voltage references in regard to supply-voltage noise and isolation to the primary side? Are there proven isolated ultra-low-noise-DCDC-converter-topologies or should one just use modules based on transformer/diode-bridge/low-noise-linear-regulator? Since ive not been able to find small transformers with a shielding winding im concerned about HF-stuff not may couple in from primary to secondary.

[Svgeesus]

For a few more $ you can get parts directly from Linear Tech (until the end of the month anyway when the web site will be taken down), and these have unquestionable pedigrees.

Wait, what! I'm getting to like ordering from Linear.com, especially since Mouser barely carries Linear. What did I miss? (Sorry to derail, a pointer would be fine).

[cellularmitosis]

Regarding measuring voltage references in opposition to use smaller voltage ranges on the used DMM:
Which supply-design is best suited to supply multiple voltage references in regard to supply-voltage noise and isolation to the primary side? Are there proven isolated ultra-low-noise-DCDC-converter-topologies or should one just use modules based on transformer/diode-bridge/low-noise-linear-regulator? Since ive not been able to find small transformers with a shielding winding im concerned about HF-stuff not may couple in from primary to secondary.

I was thinking about this just recently.  If you don’t need them to be continuously isolated, you could use relays to take it off of the common supply bus, at which point you might float it on a large capacitor while taking a measurement.

[Echo88]

That actually sounds like a nice idea. If heard that the Fluke Super Thermometer uses such a concept during ADC sampling: https://us.flukecal.com/literature/articles-and-education/temperature-calibration/papers-articles/recent-advances-resistance -> https://patents.google.com/patent/US20100308772

But using this topology you cant make continuous measurements with two voltage references in opposition, only sequentially.

[Cerebus]

I was thinking about this just recently.  If you don’t need them to be continuously isolated, you could use relays to take it off of the common supply bus, at which point you might float it on a large capacitor while taking a measurement.

The obvious choice is not a capacitor, but a secondary cell with as low internal resistance as possible. Many years back NIST did a paper on noise from batteries and the TLDR conclusion was that the lowest noise power source was a battery with the lowest possible internal resistance. This was pre-NiMH cells, so they didn't have a figures for those; but NiCd (70pV/rtHz@1kHz) and (from a different source) Lead-Acid batteries came out well.

[Echo88]

Throw a super-low-noise-regulator like the LT3042 at the battery and the battery related noise doesnt matter anymore. But in this case batteries are superior when it comes to available isolated measurement time indeed.

[Cerebus]

Throw a super-low-noise-regulator like the LT3042 at the battery and the battery related noise doesnt matter anymore. But in this case batteries are superior when it comes to available isolated measurement time indeed.

Go back and look at the noise spectral density for the NiCd battery 70pV/rtHz, if you add a regulator you'll add noise, much much more noise than from the battery. The best figure for the LT3042 is 2nV/rtHz, that's about 30 times noisier than the bare battery. You do not want to add a regulator.

[Andreas]

. You do not want to add a regulator.

Hello,

Usually PSRR is limited and the references have a much larger noise than the batteries.
At the LM399 you have also to stabilize the heater voltage to get a stable output.
So you really want a regulator near the LM399.

With 90nV/sqrt(Hz) for the LM399 I would not worry to spend a low noise regulator.
Even the LT1763 with 20nV/sqrt(Hz) is below that figure.

And usually as calibrator the 0.1 ... 10 Hz noise is the more interesting since broadband noise is integrated away by the DMM.

Noise is also what limits resolution if you use of a 100 mV range instead of the 10V range of the instrument.
And of course every voltage range has its individual divider/amplification resistors in the DMM.
Usually the native (3 or 10V range) is the most stable. The stability in the 20mV or 100mV range may differ.
So practically you do not get a factor 100 improvement by this. Its more like a factor of 10.
And of course it does not help if all references drift in the same direction. (Which is very likely).

It was a big surprise for me when I did the first re-calibration of my LTZ#1 and LTZ#2.
I told our calibration department that I have a difference drift of 12uV between both.
And LTZ#2 has a zero drift against my best ADC at that time.
But the calibrator showed that both had drifted down. One by 8 the other by 19 uV.
So the ADC had drifted also nearly 3 ppm.

Edit: if you have a 34401A it might be better to use the ratio mode for relative measurements.
I really should do some stability measurements.

with best regard

Andreas

[Magnificent Bastard]

For a few more $ you can get parts directly from Linear Tech (until the end of the month anyway when the web site will be taken down), and these have unquestionable pedigrees.

Wait, what! I'm getting to like ordering from Linear.com, especially since Mouser barely carries Linear. What did I miss? (Sorry to derail, a pointer would be fine).

Analog Devices bought Linear Technology, and they are going to get rid of the (excellent) Linear Technology site-- and you will be redirected to the "new and improved" (and really crappy) Analog Devices website.  I do not know what will happen to the Linear Technology store, as Analog Devices has their own, and I think it is going to be a nightmare until everything gets integrated.  I suspect that this is just the beginning-- Analog Devices says that they are "going to leave Linear Tech alone", and I say BULL$HIT-- they are going to screw it all up-- you will see.  If you need to order something from Linear Tech you better get it done fast as they are closing everything by the end of this month.  So, pretty soon, when you are looking for data on Linear Tech devices, you will need to (often) "click here to see more" (which is for smart-phones-- engineers use big screen workstations, so this is some kind of bull$hit to please executives).

[3roomlab]

So, for a budget-minded volt-nut to track the volt, he needs access to cheap LM399's, and they need to be binned pretty tightly.

what about ALEX's JVR ?
https://www.eevblog.com/forum/metrology/building-your-own-voltage-reference-the-jvr/?all
buying a new JFET and new resistors eliminates the uncertainty of used 399s. and the extra budget can be shifted to the op-amps and better resistors (maybe PTF56 is already sufficient for this, but i dont know, ALEX n others will need to chime in, ANDREAS? any thoughts?).

it seem like this JVR doesnt care about temp drifts and has no heaters (very low power consumption). likely an added oven will make it even better? 

i feel like suggesting JVR because, it looks like a one up DIY jellybean project that is cheaper (likely better?) than using a used 399 from a scrapyard source. why not just leave the 399 in the scrapyard? i think is worth re-using a 399, when it is not removed from its PCB, like a battered old fluke.

[Andreas]

i feel like suggesting JVR because, it looks like a one up DIY jellybean project that is cheaper (likely better?)

Hello,

I think that Alex has spent very much time in adjusting his reference.
You will also need some equipment to do this.
So I would not call it "cheap" when you count in your time.

A naked LM399 reference (without the 10V transfer) is the cheapest project (price/performance ratio) that you can have.
You need only a voltage regulator 2 100nF caps and a pull up resistor.
Of course you can use a bootstrap cirquit, but I fear the most of us will not find any difference with their equipment.
Wait half a year until it stabilises and then you have only to find a possibility to calibrate it.
At best you build 3 or 4 of them so you can determine (and sort out) the one with largest drift.
You then have a robust working horse which you can use for many other things.

If you need stable 10V then it gets complicated.
You either need a temperature stabilized housing (like ap with his DVR1 device)
(which is clever when you want to sell the device)
or much time to characterize + adjust the device.
(which is fun but no one wants to pay your time).

with best regards

Andreas

[cellularmitosis]

A bit off topic, but I jumped on the 'bay and got lucky.  Found an old Keithley scanner full of low EMF cards.  Wether I end up selling Vrefs or not, time to start recording data!