The Current Cost of Chasing PPM's (Fluke 732b price list)

[MisterDiodes]

I know a lot of people here on EEVblog will know this already, but this might be interesting for others new to Voltage Reference design.

Since I just helped a client setup a critical test procedure and required equipment list on a critical semiconductor test process, I thought I post here current prices from Fluke for 732b voltage reference units.

The method below would apply to a reasonably profitable business or very serious hobbyist, of which we have a few on EEVBlog.

Background:
For most people, if you're a student of LTZ1000 circuits, one of the more economical solutions is having a bank of 732b's and possibly a good 7 decade Kelvin-Varley Divider and a battery-operated null meter (Keithley 155 / Fluke 845ab / similar) will give good measurement results down into the sub 5ppm range, and sometimes even into the sub 3ppm range - but anytime you're going after about 2ppm or better gets exponentially harder.  The advantage to NOT using a 3458a DMM here is that you can get an absolute voltage measure while running the entire test system completely on battery power - and we'll even switch over to battery powered DC lab lighting for attempting an accurate low-ppm measure.  Any switching power supply in the vicinity can sometimes cause a noise / drift problem with critical measures.  Also realize that a LTZ1000 - especially the non-A version - can pick up any vibration or board stress in the vicinity of the part.  So really keep an eye on the test environment and shielding is key. 

Generally, single 3458a will not give you a full story on how an LTZ1000/a circuit is performing for drift.  It will tell you how your circuit is drifting compared to the Vref in the 3458a, but you won't have a lot of low-ppm confidence in the absolute voltage value.  Basically if your working with an LTZ circuit and 3458a, you're comparing two of the same reference circuit.  Depending on where your 3458a is very recently calibrated, you can get good ballpark results in the ppm range, but even a 3458a is not considered a true transfer standard for voltage.  If you have several 3458a's that can help - the late Jim Williams would measure into the low PPM range with usually 3ea or 5ea 3458a's to increase measurement confidence.  Good, properly kept, aged and calibrated 3458a's can typically have drift rates as good or better than 3458a's with -002 or -HFL options, so we rarely worry about that unless you need that low drift performance on an infant "new" unit.

Comparing a Voltage Reference to at least a few 732b's is one accepted way to get an accurate measure of of an absolute voltage in the lab to better than 5ppm.  The 732b will have a similar zener diode inside, but it's drift characteristic is presumably (not always) a bit different than an LTZ circuit, and more than one 732b on hand will reduce uncertainty.  A 732a that is well aged and working properly (not easy to find on the used market) can even be more stable than a 732b, but be aware there are quite a few drifty 'A units in circulation now, and most seller's haven't got a clue of what they are selling.  The 732a / 732b does NOT like to be powered off for long periods of time - if they've been sitting in a hot / cold warehouse for years they can take months to stabilize when powered up - or never stabilize.  A 732a/b that has a drifty zener or oven assembly is basically useless since the critical zener / transistor and surrounding PWW / laser trimmed resistors are a carefully matched set, and Fluke isn't going to ever hand those replacements out, especially for 732a.  Sometimes a 732a can be repaired with the simpler things like caps in the power supply, new batteries, etc.  But sometimes they are 100% useless, so beware.

SO:  If you're into LTZ circuits as a profitable business, probably the most direct, time-saving way is to buy a new or off-lease 732b units from Fluke, with a warranty.  Even on the new 732b's we've seen are having to be returned to Fluke for repair, so getting a warranty on these units is a valuable time & money-saver.   Being acquired by Danaher has done nothing for Fluke quality, at least from what we're seeing.

PRICES:
Current prices quote direct from Fluke as of December 2016 (there will be a price increase for January but I wasn't able to get that in time for this essay).  Prices in USD$, do not include shipping:

732b - Basic unit, shipped to you cold (power off), no calibration: $8,500
732b/h - Basic unit shipped to you hot (shipped on battery power) with basic calibration certificate (not accredited): $9,525
732b/c - Basic unit shipped to you hot, with basic calibration certificate and a 90 day drift rate characterization: $10,705

Basic calibration for 732a or 732b runs about $550.  If you want accredited Z-540 documentation that'll run around $1,632.  That will get you within +-2ppm when your unit is shipped.  You can pay for tighter tolerance calibrations as well if price is no object.

Most ordinary cal labs are not equipped to calibrate (and perhaps adjust) a 732b to low ppm uncertainty.  Fluke is.  You typically will never find a cal'd unit on eBay or any used equipment auction.

HEAD'S UP: If your 732b goes without power for long enough for the batteries to die (about 72 hrs with good batts), then your calibration certificate is immediately invalid with Fluke.  It will need to go back in to get a new cal certificate at that point.  Keep them powered up and have a good lab backup power plan.

NOTE that a 734A is really just a metal frame for 4ea. independent 732b's of your choice, and some built-in power sockets at the back.  The idea here is that to comply with NIST guidelines you always have at least 3ea independent voltage references (with no common point of failure) to inter-compare references, and the 4th unit is for shipping into the field for adjusting  field equipment OR for shipping back to Fluke for calibration.  When the 4th unit returns from calibration or field test it is immediately compared against the other 3 units on a 734A bank to make sure it is still in-spec and no damage occurred.

Price for 734A is basically price of the 4ea 732b's of your choice plus another $1250.  So you're looking at around $40k for some 732b's and maybe another $10k or so for each 3458a for a well-equipped lab that maintains a good and accurate "Volt" with low uncertainty.

Of course when you talk to Fluke sales they may offer you a deal on refurbished lease-returns or loaner units.

So a 734A set of 4ea. 732b's can get you in the ballpark for some accurate low PPM measures, and that's about what it takes if you're really checking high-end (LTZ and similar) voltage reference performance accurately.

Some interesting reading - How many references do you need, and how often do you need to calibrate to maintain a 0.3ppm lab "Reference Volt" (this would allow to to measure accurately to say about 1ppm at any time with confidence)?

http://download.flukecal.com/pub/literature/aug96pp2.pdf

I know there are other methods to maintain an accurate reference lab Volt, but the method above is probably the more popular and relatively economical way.  If you have an unlimited budget, the other way would be to maintain an Josephson Junction Array, but that is generally much more expensive.  You can call these guys for a current price quote:

http://www.hypres.com/products/voltage-standard/

 

[Vgkid]

Thanks for this, it is always neat to read about what is involved in maintaining the upper echelons of voltage references.

[retrolefty]

Volt-nuts!

 I really wonder what the total U.S. market is for such a set-up? How many publically listed manufactures order/need this level of standard? Does Fluke publish sales volume by product?

[MisterDiodes]

Volt-nuts!

 I really wonder what the total U.S. market is for such a set-up? How many publically listed manufactures order/need this level of standard? Does Fluke publish sales volume by product?

You'd be surprised. I don't know the actual number but at semiconductor fabs plants they will typically have a calibration room where the "good" references are kept - and you'll see quit a few 734b sets and racks of 3458a's on the production line test jigs.  For one outfit I was in where they do very high reliability laser diodes (where every diode produced has a serial number) -  they cycle the production test jig 3458a's into the cal room often (at least every 30 or 45 days) to get a touch up - because these devices being produced come with a 30yr warranty, and so this company are absolute sticklers for detail, and if anything is drifting it is pulled out of service right away.  Plus there is that Dept. of Defense traceability requirement as well for every piece of equipment used to build a component.

For these places, they don't worry much  about 1yr drift on an LTZ circuit, they are more concerned with the reliability of 30 or 90 day drift and the low noise you get with several LTZ's running in parallel on a production line test apparatus.

In this case the Fluke standards are kept in cal often with Fluke mobile cal service, and both Keysight and Fluke have their own offices in-plant to do detailed inspections and repairs, and many pieces of equipment are kept in stock as hot spares.  If a test engineer or production manager needs a 3458a or whatever, they can go get one of several already warmed up off the shelf and put it in service right away - the production line has to run 24/7.  So an application like that will keep quite a few 732b's busy as reference checks and as artifact sources for other gear.

So you know, on the price list above, there is about a 1 to 2 week wait on the basic 732b's, the 732b/c has about a 6 week wait.  Which sounds like they have several in the pipeline at any time, which means Fluke probably has a fairly steady supply of orders.


 

[VintageNut]

I visited a cal lab last year that has 4 x 732B and has had this group of references for many years. The drift is about 0.03 ppm/month in the up direction.

[MisterDiodes]

That sounds about right.  732's tend to drift up slowly, LTZ's tend to go down over time.

[CalMachine]

Great Thread!!! I'm actually in the process of doing just this type of thing here at ENI Labs, getting our DC/Low Frequency Voltage program up to par.  We have 1 732B that we've had for about 5 years or so.  It has been cal'd by PSNA for 4 years and then last year it was sent to Process Instruments (0.5 µV/V Uncertainty at that point).  We just purchased a Datron 4910 for fairly cheap and we are in the works of obtaining another one!  That will give us 7 seperate 10V Cells with 3 points of failure, so we're almost there.  Some of the best scopes I've seen have got like 0.2 or 0.3 µV/V @ 10V.  So you're talking like 2-3 µV uncertainty in their measurement... so roughly 10:1 TUR.  You would be surprised at the amount of accredited labs that can actually make a *decent* enough accredited measurement of a 732b @ around 2:1 TUR.

In general, you would almost never want to adjust a 732B, or any 10V DC standard for that matter, you have had for awhile.  If you have history of data on a unit, you can do linear regressions to predict drift and get your uncertainties fairly low. 

I wish we could get a 734A... Those things are sexy.  I'm super amped about getting a Datron 4910, though.  4 Cells in 1 is amazing and the average of the 4 have half the stability of a 732b

Another thing...  I actually got a specialized quote from Fluke for Calibrating a 4910 directly against their JJA...  $12k       While I was on the phone with Jeff Glust about the special request, we got to chatting about Fluke vs Datrons and just general precision metrology mumbo jumbo.  Well.. he offered to dig up the birth certificate of our 732B, which was pretty sweet!  So I can track my 732B from the factory through its entire life until now. 

That sounds about right.  732's tend to drift up slowly, LTZ's tend to go down over time.

Mine is drifting down 

[Pipelie]

That sounds about right.  732's tend to drift up slowly, LTZ's tend to go down over time.

agree, more specifically,  the SZA263's tend to drift up slowly,   LTFLU and LTZ's tend to go down over time.

[Pipelie]

[/quote]
Mine is drifting down 
[/quote]

include the drift of the boost-up resistor pair.
@CalMachine

[CalMachine]

Mine is drifting down 
[/quote]

include the drift of the boost-up resistor pair.
@CalMachine
[/quote]

What do you mean, include the drift of the boost-up resistor pair?

[TiN]

Minor correction, there is no Fluke 734B, only 734A which takes 4x732B

CalMachine
As we discovered earlier and covered here, there are multiple versions of 732B. Older ones with Motorola SZA, newer with LTFLU-1 refamp chips. Hence drift difference .

You can also opt for NIST JJA calibration, but since 4910 has four references, result pricing is about same. Our local Taiwan NMI quoted me 500$ for one DCV point vs their JJA measurement.
So perhaps ask around other top level callabs to get the spread. Or get own JJA, then you can charge others for paper patch with 9-digit numbers written down  .

[CalMachine]

Minor correction, there is no Fluke 734B, only 734A which takes 4x732B

CalMachine
As we discovered earlier and covered here, there are multiple versions of 732B. Older ones with Motorola SZA, newer with LTFLU-1 refamp chips. Hence drift difference .

You can also opt for NIST JJA calibration, but since 4910 has four references, result pricing is about same. Our local Taiwan NMI quoted me 500$ for one DCV point vs their JJA measurement.
So perhaps ask around other top level callabs to get the spread. Or get own JJA, then you can charge others for paper patch with 9-digit numbers written down  .

Yeah for sure, you are definitely right about the differing drifts.   $500 bucks for a cal directly against a JJA is a pretty damn good deal IMO.  I'm sure NIST is probably more expensive than Fluke and Keysight is probably along the same scale as Fluke.  There are a few other labs around that I know of with JJA, but they don't do outside commercial calibrations.   My plan is to have a JJA here at ENI within roughly 5 years of now.

[Pipelie]

[/quote]

What do you mean, include the drift of the boost-up resistor pair?
[/quote]
the resistor pair i said is R401/R402.
@CalMachine

[CalMachine]

the resistor pair i said is R401/R402.
@CalMachine

How do you propose I characterize the drift of these resistors that reside in my 732B, without deconstruction and voiding the cal?  And for what particular reason?

holy crap that serial number order graph is crazy...

[HighVoltage]

MisterDiodes
What a great thread and explanations.
After much research over the last few month, I have come to about the same summary about 732A/B

So far I can only afford the older 731B, but I have 4 of them constantly powered up.
I am hoping to get lucky one day to find a used 732B for an acceptable price in good condition.

Which one is more sensitive to have not been powered up for month or years?
The 732A or the 732B?
The problem is that we never know what we are getting, if we buy it used.

[quarks]

very good and helpfull information
thanks for sharing

[MisterDiodes]

Minor correction, there is no Fluke 734B, only 734A which takes 4x732B

You are correct, my typo on original post - I went back and made the corrections in case someone is reading this thread later on.  The actual prices I double-checked and are correct according to my quotes from Fluke.  As of December 2016... I'm sure there will be a price adjustment for 2017.  The price also depends on who the buyer is and how much equipment is purchased every year, etc.  They also have occasional demo or off-lease returns etc.

For what it's worth, you could just have 4ea 732b on a shelf and that would be equivalent to 734a - but the 734a frame is nice if you have a rack shelf or similar situation.  Plus the 732b's slip in and lock in place, so it give a more organized, substantial feel and look to the whole setup.

[MisterDiodes]

MisterDiodes
What a great thread and explanations.

Which one is more sensitive to have not been powered up for month or years?
The 732A or the 732B?
The problem is that we never know what we are getting, if we buy it used.

Therein lies the problem if you buy a used one - you won't know for a while if you have a drifter.  I tried contacting this seller several times to see if his 732A's were stored powered up or not (I assume probably not).

http://www.ebay.com/itm/Fluke-732A-Reference-STANDARD-/302176388032

If they are what I think they are I think these came from University of Washington or Washington State University, and maybe they were possibly all part of a big Fluke donation made to both institutions years ago.  See the 731b the same seller has and you'll see the trail of stickers.  Notice we can't see all sides of the 732A.  They were at a surplus auction before (I think), and I'm sure they've been powered off for a long time.  I asked seller if he'd be willing to let me  drive over there to do a quick check of some of these at his site with a smaller cal'd 6.5 digit DMM, and then I could report back here for you guys - just to verify to see if outputs are anything remotely close to 10V - but I never heard back.  I'm wondering if he's playing games about never really telling us if they are working or not to get out the 30-day refund policy.  Sort of like the eBay scam phrase "Have no means to test further" which gets the seller out of all sorts of refund responsibilities.

I could also be that he's still gone for holidays, or maybe doesn't check eBay messages often.  I'll let you know if I hear anything.

Of course you wouldn't know how bad the drift problem is for months after power up, especially if they've been off for a long time.  I know neither the 732A or B likes to be off for long periods, and I know both can be rockin' and rollin' for months after power up if they were powered down for years, but sometimes they settle down within a year.  Sometimes not - and then you have an expensive doorstop.

If you think that the 732B's were an improvement (with LTFU,  laser trimmed resistor packs and smaller oven design) vs the 732A's with PWW resistors and earlier Moto reference... well here's Fluke's own data.  Take a look at page 7 at the typical drift rate per year.  The 732A specs are very conservative, and we've seen them stay put to within a ppm for years.  If it were me the 732b is still easier for shipping and it does settle down pretty well with time - but expect a "new" 732b will be on the move for at least several calibration cycles - this is a case where older can be better.   Notice how both 732A/ 732B types -tend- to drift up slightly, but that doesn't apply to -every- unit.  I think the very newest units might tend to drift down (Newest LTFU manufacture) but all the ones I've seen at least (that are at least 10 yrs old) tend to go up slightly every year:

http://download.flukecal.com/pub/literature/Predictability_Of_Solid_State_Zener_References.pdf

A working 732A works for us because we can easily drive it to Fluke for cal, and we'll use either A or B type as long as they continue to be stable.  Each user's situation will be different.

[ap]

From the Fluke predictability data PDF (referenced above) a few things can be derived (fig. 9 and 10):
- The likelyness of the 732A to be low-drift (<1ppm/a) is higher than that of the 732B.
- quite a number of 732B tested did not meet the annual drift rate spec (2ppm/a), none of the 732B exceeded their spec (3ppm/a)
- using a 2ppm/a limit for both A and B, still way more 732B exceed this than 732A.
- when you buy a brand new 732B, it does not necessarily meet the 2ppm/a spec.

When I back then saw this document, my first thought was: not really a good advertisement for the 732B; product marketing may not have seen it when published 
So, unless you have a broken 732A unit that is just not working, this reference tends to be the more stable standard.  This is in line with what many people from call labs, having using those over years, have told me.

I am operating a 732A, a 732B, and a 4910. The 732B is pretty new, so no history yet, but the 732A is well below 0.5ppm/a drift. The 4910 too. That said, the current 732A is not my first unit, bought 3 overall, and at least one did strange things (jumped during warm-up; not usable). The second one was ok, just not as stable /about 1ppm/a) as the one I have now. So my yield, if you will, in buying used 732As, was 67%.
 
The advantage of the 4910 is, it is specified to 1ppm/a. So starting from scratch, with no history available, the 4910 gives a better drift rate per year, if you do your tolerance prediction. Next best is 732B. For the 732A (or any other), to establish a history for a specific item, DAkkS (the German cal. accreditation body) requires at least 3 data points to establish a history. So this takes time.

So overall, depends also on what you want to do (and budget of course). My call on this, if you have time to establish a history, or just want a stable voltage and can recalibrate it frequently, and also in view of price, used 732As are hard to beat.
Second best, from a cost/drift perspective, 4910. Hard to get. A used 732B, imo, is the last choice. Offers 'only' 2ppm/a spec data, is relatively expensive, and is statistically (and in the real word, if I believe all these cal lab guys I talked too; and I have no reason not to) not as stable as a 732A.

Few comments to the 732A:
when switched off, and getting cold, it may reverse its drift direction after new turn-on, so your drift history is completely gone. So make sure your battery is ok. It is known that the 732A has a potential issue with battery charging, many seem to overcharge them, so the battery should be replaced every year. I bought a spare battery compartment, so I just remove the old one, insert the new one, with the batteries already installed, and thats it. Worth the money.

[Pipelie]

@CalMachine
sorry, my English is not good, but  I'm working on it.
what I'm trying to say is,
we  known the  zener and resistor pair (R401/R402)  are the primary cause of drift.
since we don't known the drift rate of the zener .....  and we characterize the drift of these resistors  without deconstruction and voiding the cal.
but I'm plan to remove the divider resistor of 1.018V in my third unit, and connect   "7V"  to the binding posts of 1.018V.
after that, I can keep tracking the drift rate of  resistor and Zener.

[CalMachine]

@CalMachine
sorry, my English is not good, but  I'm working on it.
what I'm trying to say is,
we  known the  zener and resistor pair (R401/R402)  are the primary cause of drift.
since we don't known the drift rate of the zener .....  and we characterize the drift of these resistors  without deconstruction and voiding the cal.
but I'm plan to remove the divider resistor of 1.018V in my third unit, and connect   "7V"  to the binding posts of 1.018V.
after that, I can keep tracking the drift rate of  resistor and Zener.

That is a very good idea!!  I might actually see if I can do that with our 732B.  As you said, that would allow us to characterize the zener reference and the output voltage, and therefor derive the drift of the resistor pair!  I love it 

[Echo88]

I thought the Datron 4910/11/12 use LTZ1000-references? I was always under the impression that LTFLU/SZA-references including the matched resistors were more stable than LTZ1000. Maybe Datron used better selected LTZ1000?

[CalMachine]

I thought the Datron 4910/11/12 use LTZ1000-references? I was always under the impression that LTFLU/SZA-references including the matched resistors were more stable than LTZ1000. Maybe Datron used better selected LTZ1000?

Well the thing about the stated Datron 4910 specs, is that the 1.0 ppm/yr spec is for the average of the 4 10V cells.  So it's not 1 cell vs 1 cell, as you would think.

[Pipelie]

@CalMachine

[/quote]

 I might actually see if I can do that with our 732B. 
[/quote]
@zlymex did this years ago, He modify all his 732A and 732B to get the 6.5V out  for characterize the zener reference.
here is the article Relate to this modification.
https://www.eevblog.com/forum/metrology/teardown-voltage-standards/
more detail
http://www.gfjl.org/forum.php?mod=viewthread&tid=8431&highlight=732B

[MisterDiodes]

Yes, the alloy-junction devices are old, almost primitive tech (process first used in the 50's for transistors, diodes, etc) - and hard to get decent yields and repeatability compared to more modern processes.  Not  used much anymore.  Once you do get an actual working device they work well though and with low noise.

The bigger issue with 732A's is the physical size and very expensive to ship, even cold - from a production standpoint the B's are much easier to send for cal, batteries last long enough for most FedEx shipments, and they will settle down after a while.  It takes years or a decade but they do get there eventually.  Some of those shipping requirements are impossible with a 732A, depending on location.

If you can handle keeping a 732A in cal, it can be a very good reference.  Even if a lab knows a lot of time they really don't -have- to get a well-aged reference shipped back for cal, a lot of times cal is required no matter what for customer audits and legal reasons.

For instance:  If you're building something for Dept. of Defense or Aerospace, on critical components and systems they want to see all equipment used to production  build and test is kept in specified calibration, at specified intervals.  They don't care if a 732A hasn't moved a ppm over the last several years, they want to see current calibration history, cross check tests and documentation.  And so on.

That's why Boeing just maintains their own JJ Array - now they can tend to their own references easier and on shorter intervals  - and they will measure a 732a/b for you (starting at around USD $800 ) but they are not in the business of adjustment - which is something different entirely.  For Calibration documents -and- Adjustment you send your 732A/B reference to Fluke.

And that brings up another Cost of Chasing PPM's:  The annual expense of keeping everything up-to-date on calibrations.