#2500 Post : Teardown and study of Fluke 792A AC/DC XFER Standard [56K warning!]

[TiN]

Not very long ago I've started hobbyist first thread about Fluke 5700A series calibrator. It's time to continue this sick practice to tear apart expensive metrology equipment, don't you agree? Calibrator in that thread living well since the overhaul last year and was used to calibrate number of 3458A's, few Datron 1281's and even my second 3458A back in January'17.

With access to 732A/B and 1ohm/10K SR104 resistance standards it's not hard to calibrate 5700A to it's 24hr spec. But how about AC voltage and current calibration? Calibration manual calls out for Fluke 792A AC/DC transfer box. Well, some patient hunting and here we go now..



Intro

The Fluke 792A is an very high accuracy AC/DC Thermal Transfer Standard, designed to meet the most demanding ac traceability requirements. Using the patented Fluke RMS sensor and thin film range resistors, the 792A offers an extraordinary transfer accuracy, with total uncertainties of as low as ±10 ppm (±5 ppm from the NIST). The 792A also provides a wide voltage range of 2 mV to 220V (1000V with external range resistor), and a wide frequency range of 10 Hz to 1 MHz.

The Fluke solid-state RMS sensor also provides the 792A with remarkable temperature stability and fast settling time. Now you can be ready to make measurements in 30 seconds instead of 30 minutes. To simplify the transfer process, the 792A's 2V output permits you to use a high resolution digital multi meter rather than a null meter. The Fluke 792A is designed to support the calibration of the most accurate ac instruments in your standards lab workload, including calibrators like the Fluke 5700A, voltmeters like the Fluke 8506A, voltage amplifiers and other ac/dc transfer standards

Fluke 792A is main standard used to test and characterize multifunction calibrators, such as Fluke 5700A/5720A/5730A and AC voltage measurement standards, such as Fluke 5790A, Fluke 5790B and Wavetek 4920.

Manuals and documentation

Fluke 792A Instruction manual

Exterior and user interface



Standard consists of two boxes, one being battery block/power supply (on the left) and actual transfer standard (on the right). If you buy it from Fluke for serious amount of money, you also get fancy transfer switch to toggle input connection for standard and external 1000V range resistor, to measure high voltages. Transfer box itself can only measure up to 220VAC/VDC. In this ebay deal we got no 1000V range resistor, so will to hunt that rare beast separately...

Teardown and design evaluation

Well, like we say it here, "Don't turn it on, take it APAAAAART!" . Let's start.



Shield under the shield under the shield. Like those Matreyska dolls of metrology



Don't breathe...





Look at the construction. Wonder how many engineering hours was spend on just chassis alone.



Some componentry around:



Attention to details, zip-tie hooks to get board out if required. Steel wool in the chamfered corner to ensure good RF contact with shield box, to make sure no electrons escape!



Some of the areas of PCB are exposed of soldermask, to have more stable frequency response over the tracks.



Plenty of filtering and LC circuits for power.







According to date codes on components, this specific 792A was manufactured in the middle of 1996, some 20 years ago.

Range switch assembly

To maintain specification and long term stability, Fluke 792A RMS sensor and hermetic range resistors contain Beryllium Oxide ceramics. While normal use of 792A is safe, BeO dust or fumes from damage to these components HIGHLY TOXIC and breathing them can result in serious injury or worse. Beryllium is IARC Group 1/EPA Class A EPA carcinogen and exposure can cause Chronic Beryllium Disease, an often fatal lung disease. Never disassemble, grind, lap, fire or chemically clean on any ceramic parts inside 792A.

We don't plan any of that, phew...









Isn't this most beautiful range switch assembly you ever saw?







There is even spot for few PTF's

Bit of wiper action:





Sensor PCBA





Core of the instrument, Fluke's thermal solid state sensor hybrid:



And few overview shots:







Sensor test

The key component of Fluke 792A performance is custom solid-state thermal transfer RMS sensor. It's not impossible that this RMS sensor is failed, and that can quickly make unit's complete repair very expensive.

So one of reasons why we get everything apart to this level, is to actually check if RMS sensor good or not. Procedure of this simple test listed in instruction manual in troubleshooting section. It calls to measure resistance between pins 6 and 7 of U111 sensor hybrid assembly. If it's no longer between 368 Ω and 432 Ω then U111 sensor is damaged and must be replaced.



Our measurements with portable Agilent DMM shown sensor resistance 392 Ohm, very well within expected good range. Big "Phew!" moment here!

Front filter PCB















Standard DC output, guard and ground terminals are connected to high quality copper 5-way binding posts, very similar to ones used in Fluke 57xx series calibrators and other Fluke standards. Similar connectors can be bought from Low Thermal for about $25/piece.



Troubleshooting and repair

Before we test transfer box, need to fix PSU first.

Model 792A Power pack repair

In condition as received power pack does not show any signs of life, and does not turn on with AC power. Simple check procedure hinted from instruction manual.
Initial cover removal only shows more shields inside. Fluke 792A build like a famous Matreska dolls, box inside a box inside a box.



Shielded box in center contain four 6V sealed batteries, which are likely dead after 20 years. Let's see.



This can be done either by measuring battery pack output at J303 connector (should measure around 12V pin 1-2 and 12V pin 3-4). Or simply remove the cover and just to check each battery terminal voltages.







Well, no comments required, all four batteries are completely dead. Let's remove then and get access to power supply PCBA. This is done by removing front chassis side.



After temporary change with two good 12V battery, output voltage was measured at expected +/-11.3 V levels. However relay K30x which connect charger to the battery got very hot quick. Powering unit with AC made relays click very fast for a second and that was it.



At least we know at this point that output voltage regulator is correctly working, which bring some confidence that even if transfer 792A unit have problem, it's not came from power supply. So focus on troubleshooting now is on charge/detection circuitry. Fluke supplies schematics of 792A to our benefit (Thank you, Fluke!) so that makes troubleshooting bit easier.



Both negative and positive regulators are symmetrical, and can be used to compare each side for faster troubleshooting.





After investigation and few days of troubleshooting, next parts were replaced. Relays were replaced just in case.



* Make sure fuses F301, F302 are OK. Voltage difference between fuse terminals should equal zero. OK..
* Measure supply voltage on U307 at pins 14-7. Should be around ~12VDC.  OK..
* If OPERATE indicators briefly goes on at power on, there is likely a short. Check resistance between test points TP302 (+11V), TP304 (-11V) and TP305 (Common). Automatic shutdown also activated in case of too low battery voltage (can be checked at J303 connector).  OK..
* If battery charger is faulty, check mains fuse and voltage setting.  OK..
* Use scope to measure U301 pin 2 and pin 3 with respect to TP308 (Charger COMMON). Should see half-sine waves about 24V[~pk-pk~].  OK..
* Logic supply voltage should be around +22VDC between cathode of CR309 and TP308 test point. OK..

As a result, Power pack block now working as expected. Larger batteries were used for test, while replacement already on their way.



Both BT1 and BT2 battery now charging as expected, reaching ~13.2 and ~13.3 VDC after 16+ hours of charge.



Output voltages are within few millivolts positive and negative, well in spec. Now order some LEMO connectors, make a cable (straight 1:1 coonection, no worry) and we are ready.

Initial tests

Input of 792A has an internal protection circuitry that guards against accidental overvoltage. However, the instrument may be damaged if the PSU turned on while voltage is applied to the Type-N input connector. Always turn the power switch ON BEFORE applying a signal to the input connector.

All range resistors were checked good, so time to connect 792A to Fluke 5700A, some random 3458A laying around and see some transfers.



2.2VAC 1kHz applied to 792A input (left) and  2.2VDC applied to 792A input (right)


It works, yay!

As usual full worklog is maintained and updated on xDevs.com site.

[Vgkid]

I will read this, while eating breakfast.

[Dr. Frank]

Really nice catch!
What did you pay for this unit, and in what condition was it sold?
You'll need the Transfer Switch, or similar assembly, if making ppm precision transfers.
What do you plan to replace this missing piece?
Frank

[Muxr]

What a beautiful piece of kit.

[quarks]

Great score
Thanks a lot for sharing

[Echo88]

Beautiful device! Did i miss something or is this your Fluke 5700A for which you need some more parts?

[HighVoltage]

Your scores, teardowns and repairs are just amazing.
Thanks for sharing this with us.

[TiN]

Well, to be precise, this is not my score exactly, I just stealing the fame here.  There are some distant plans to use this box for future AC transfers/calibrations.
Same as F5700A on photos above.

My DIY 5700A which I prefer call X5700A need way more parts before it can look like one covered before in post #2000. 
However it's getting there, slowly. Got CPU and rear I/O board today, and few more boards ETA tomorrow.

We do have 792A transfer switch. Photos of which will be coming soon. Stay tuned

[e61_phil]

Very nice teardown! Thanks




Is there a marking within the 792A which notices the BeO content? Most of the APEX high voltage opamps also contain BeO, but there is a marking on the IC and a notice in the datasheet.

[TiN]

No marking was spotted on the unit, but BeO warning clearly present in instruction manual.

[Echo88]

Im interested in using the PowerPack-schematic as a basis for a LTZ1000-power-supply (similar to a 732B, which schematics i was unable to find). But i dont understand the necessity of the monostable relays K301/302, since my understanding is that they only act just like the latching relays K303/K304. Can anyone show me my error in reasoning?

Edit: Also they use 5V-relays for K303/K304 (K301/302 are okay) which arent suitable for the 12V, even if its just a short impulse?

[sietseage@ziggo.nl]

Nice fluke  wow

Verstuurd vanaf mijn SM-N910F met Tapatalk

[CalMachine]

Awww yissss!!     Thanks again for another great post.  I always love reading your work.

After she's complete, send er on over and we can make some comparisons off my 4920!

[carl_lab]

What an excellent teardown with superb photos of that very nice piece of test gear... 

[timgiles]

Hmmm my 100Mb/100Mb swedish internet suffered when I first loaded the page ;-)

Nice set of photos and explanations - thanks for the time/effort spent.

[chinapp]

why you give a input of 2.2VDC but the 3458a reading is 1.88V ?

[TiN]

That is how this box works. Actual voltage value is not important, it's the difference between AC voltage and DC voltage what matters. Because of this 792A is called transfer standard. It "transfers" accuracy of DC measurement to AC signal. Thanks to advancement in DC voltage standards (ppm-level quantum JJA effect standards, solid-state references like Fluke 732B, 732A, Datron 4910) we can measure DC signals very accurately. Even off the shelf hi-end DMM, like HP 3458A or Keithley 2002 can measure 10 VDC at 8ppm accuracy without sweat or usually better.

However same HP 3458A have accuracy "only" 70+20 ppm for AC voltage in frequency band 40Hz-1kHz and much worse 1%+0.01% for 1MHz! So the idea of using 792A is to supply known DC voltage at it's input, let's say same 10VDC. This voltage we measure by HP 3458A to 8ppm, directly. 792A then transfers it to some random value, on setup in photos above is 1.886276 V. Now we change input signal from known 10V DC to unknown 10V AC signal, which we want to measure. Result is 1.886286 V, which is -5.3ppm. Even with 792A transfer error, this is many times better than 3458A can measure directly.

[TiN]

As time went, so called AC volt-nut team got the whole kit complete now.

Range resistor



One might think it's just two resistors, 500 KOhm and 500 Ohm inside an fancy box, but truth is bit more complicated.



Application note covering details of whats inside this box posted by Fluke, THIN FILM DESIGNS FOR 1000V AC RANGE RESISTORS. 5790A using same approach, but with bit of compromise on design to save cost.
Well worthy read, with some photos inside

To avoid risk of upsetting sensitive resistor, we didn't disassemble the divider.

Transfer switch



This transfer switch allow connection of two inputs and quick swap of signals, with minimal switching errors.





Switch is sturdy and robust, built like a tank.



Inside we see same porcelain construct insulator with slider switch, as in range switch assembly of 792A.



Output 5-way binding posts have massive copper base to keep thermal variations small.

[lukier]

Nice switch! But, for me, not nice enough for $1200 . For that money one can not only get the copper binding posts but the whole case made out of precious metals Except the ceramic isolators is there something special in the rotary switch, like rare contact material or similar? Resistor on the other hand probably required serious R&D.

It looks like the LO binding posts in the switch are in contact with the case, is it true? What if one has ground potential higher than earth?

[TiN]

Contacts and slider look like silver plated to me. Very similar off-shelf 5-way binding posts are about 25$/piece. N-type connectors maybe another 50$. So that's 250$ for connectors. 100$ or so for CNC case. Switch maybe another 200. But pricing for such test equipment pieces rarely come directly from BOM, but based on R&D and manufacturing cost for low quantity amounts. It's not the product made in million pcs qty to spread high R&D cost over the units.

Other significant price impact would be from testing and validation cost for such precision instrumentation. Somebody has to pay for all that time to plot neat ppm-level charts we all see in application notes and 792A manual.

What if one has ground potential higher than earth?

Then don't touch it I'd expect major field of use for this switch is for AC/DC transfers using listed 792A procedures, in which case there should be no common ground potential.

[Vgkid]

That appears to be one of multitechs rf switches.
http://www.multi-tech-industries.com/rswitches.html
Which remind me of the "jvc" switches.

[TiN]

Vgkid, these indeed look very similar. Could be even same manufacturer, with Model 90 and Model 85 custom ordered by Fluke.
I wonder how much those would cost? Yea, can't afford if you have to ask.

[bitseeker]

I missed this earlier. Awesome stuff, TiN! Lots of good learning to be had in there.