Author Topic: Deciphering Fluke 732a  (Read 972 times)

0 Members and 1 Guest are viewing this topic.

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Deciphering Fluke 732a
« on: January 13, 2025, 07:48:54 am »
I was able to obtain a Fluke 732a without any history whatsoever (only sticker with thermistor resistance at oven setpoint  ;D) and so far I have not figured out yet whether it's a bad or good luck...

The good thing is that one of my 3458a's was reading its 10V output as 0.58ppm from exactly 10V which is a good start.

The unit had some problems though due to poor soldering/repairs that was done before me, so I decided to rebuild all the boards (incl. A5).
A3 and A4 are easy, but A5 (reference board inside the oven) is famous to be very messy to deal with.
I clearly had some poor contacts, so I decided to overhaul it as well (CC resistors and also caps).

Battery
Obviously, original batteries were completely dead, but after overhauling A3 and A4 boards I decided to give a try to LiFePO4 battery.
Luckily Amazon has a single 24V LiFePO4 battery, so there is no need to put any extra balancers, etc. for multi-battery setup.
Here is the battery I put into test Talentcell 24V 6Ah LF8011 https://www.amazon.com/dp/B0CNLKKL9C?ref=ppx_yo2ov_dt_b_fed_asin_title.
The battery was able to deliver about 5.7Ah in my discharge test (22V cutoff), which is almost as specified (they specify it as deep discharge down to 18V) and is quite a bit more than standard 4Ah Lead Acid.
Obviously you need to tweak R20 and R10 pots on A3 to make sure the charge and float voltages are good for LiFePO4 chemistry.
I put everything together, but have not yet tested it with a battery as I'm waiting while A5 will be ready too.
Here is how the battery looks inside the battery compartment - there is quite a bit free space there.
You don't even need to cut any new holes as the battery sits completely below the top plate and don't touch any metal with its terminals.
2481779-0
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #1 on: January 13, 2025, 07:50:44 am »
A5 (Regulator board)

One of the challenges with A5 board is that clearly something went wrong around CR5/VR2 area as PCB is charred and there were signs of rework in that area.
Also, U3 was replaced with LF353N in some weird DIY socket.
2481793-0

Here is the CR5/VR2/R9 area (already cleaned, it was worse).
2481801-1

There is clear PCB charring/discoloration, probably due to some failure/overload happened before and fixed.
The problem is that given the quality of other fixes, I'm kind of having a bit of distrust to this repair, so I decided to order new parts and replace CR5/VR2.



Has anybody looked into this part of A5?
It seems that this CR5 + R9 is a part of current compensation circuit...
Why R9 is a part of the ref amp set along with R5 and U2?
I understand the U2 and R5, but why the R9? It is to compensate and match the current set by R5?
« Last Edit: January 13, 2025, 08:22:35 am by dlebed »
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #2 on: January 13, 2025, 08:40:42 am »
So to replace the CR5/VR2 with trusted parts I ordered:
[VR2] 1N4574A - 6.4V temp. compensated Zener with 5ppm/C tempco - an impressive part with impressive price of $27
[CR2] A set of different CRDs from Semitec: E-101, E-301, E-501, E-102, see https://www.semitec-global.com/uploads/2022/01/P22-23-CRD.pdf

Interestingly, they all have different stabilization voltage and also a different tempco:
2481825-0

What I found is that the CR2 current regulating diode that I have on A5 board actually stabilizes above the rough 2.2V that it should be working at (18.6V - (10V + 6.4V) = 2.2V across CR5).
Here is a sweep of CR2 I have on the A5 board, you can see that it's not yet stable at 2.2V
2481829-1

Here is the sweep of CR2 + R9 in parallel, you can see that together they reach exactly 1mA at 2.2V
2481833-2

As for the Semitec's CRDs that I bought (10 each) I ran sweeps and figured out that I could put multiple of them in parallel to achieve lower tempco and very good regulation at 2.2V.
E-101
2481809-3

E-301
2481813-4

E-501
2481817-5

E-102
2481821-6

So the main question I'm currently battling is what is the actual role of the R9 and how to actually tweak the CR5 (I assume it will mostly affect the tempco of the U2 based on what I read about Fluke 731 tuning).
I'm thinking on putting 3 x E-301 in parallel, that will result in much better tempco and much better regulation at 2.2V.

The question is whether this R9 is in any way responsible for CR5 tempco compensation (which I doubt) and should CR5 be tuned to be exactly 1mA?
Will having a better regulation and tempco for CR5 in isolation actually improve things or make them worse (if R9 supposedly compensating for CR5 in some way)?
« Last Edit: January 13, 2025, 08:45:20 am by dlebed »
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #3 on: January 14, 2025, 03:20:13 am »
I was able to acquire the original Fluke CRDs (Fluke 334839) and run sweeps, interestingly they're quite different from the Semitec's CRDs.
BTW, if someone will ever need that rare Fluke 334839, shoot me a message, I might have some leftovers. I don't think you'll be able to easily find them on the market.

Semitec 1mA CRD has negative tempco, while Fluke 334839 has positive.
Also, for 1mA CRD, Fluke's has better regulation at ~2V than Semitec's one.

Fluke 334839 CRD sweep
2482415-0

Here is zoomed in version around 2-2.5V for Fluke 334839 vs Semitec E-102.
Each sweep was bidirectional, so you see two lines with the same color.
The bigger the gap between the lines the bigger the tempco - the gap is due to self-heating and tempco effect.
You can see a better regulation of Fluke's CRD and lower tempco.
I feel the same result could be achieved with 3xE-301 in parallel.

2482419-12482423-2
 

Offline bdunham7

  • Super Contributor
  • ***
  • Posts: 8363
  • Country: us
Re: Deciphering Fluke 732a
« Reply #4 on: January 14, 2025, 06:07:09 am »
I also have a 732A I'm going to be going back into at some point.  I'll probably be interested in 1.0mA CRDs at some point.  Just a few thoughts and a question for now.

IDK why R9 is part of the ref amp set because it, along with CR5, sets the zener current through VR2--and CR5/VR2 are not part of the ref amp set.  IDK why Fluke chose to set the voltage here in this way instead of a buffer and divider deriving it from the output voltage.  I suppose a 5ppm/C zener stacked on the 10V reference is pretty good and might actually be hard to match the other way.   It's all in the oven anyway. 

I find that this particular wiring diagram is harder than expected to understand and I often refer to the first interconnect diagram which is simplified, but actually shows how things are set up.  For example, it is very hard to understand what is going on with the output circuit and U1 unless you attach the snippet from the interconnect diagram that I've attached. 

The question is where are the discussions of 731B tuning?  I've seen them referred to before and what people say about them seems to imply some very in-depth discussion of zeroing the tempco of the reference by finding just the right current.  However, I've asked but have never seen those detailed discussions.

I'm not sure what to say about the rework, you can still buy an LM358H AFAIK.  The browned boards seem to be a common issue and is probably just because those parts get a bit warm. 
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #5 on: January 14, 2025, 07:19:36 am »
Yes, it's quite weird how R9 is a part of ref amp set and CR5/VR2 somehow aren't.

The only explanation for R9 being a part of ref amp set I could come with so far is that it works together with R5 to tweak the U2 collector current.
Probably CR5 provides the VR2 nominal current (1mA) and then the R9 provides extra current to feed into U2 collector, so R5+R6+60 set the U2 collector current based on 16.4V nominal voltage and then the R9 provides this extra current in parallel with CR5.
In my case R9 is 45.3k, which will deliver little current - it will be about 50uA (assuming the voltage over R9 is 2.2V).

It looks like yes, you're stacking 6.4V 5ppm/C zener (6.3V@1mA is mostly what I see in reality for my parts) on top of even more stable 10V, so the zener tempco contribution is only ~1/3.

Clearly, the CR5 that I have on my board is far from what it should look like according to I-V curves (regulation around 2.2V is 10x worse than Fluke's original part), so I'll replace at least it (maybe even with Fluke's original one).
Though with that R9 connected to a regulated but not so stable 18.6V is puzzling me... Once I'll put things back I'll try to measure the supply rejection for 10V output vs 18.6V supply.

As for the Fluke 731B tempco tuning, looks like it is tuned by adjusting the collector current of SZA263 using R5/R6 resistors:


See the SZA263 thread around here https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg2808660/#msg2808660
Also here is the summary PDF with voltage vs temperature plots for different collector resistors for SZA263/LTFLU
https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/?action=dlattach;attach=882150

P.S. my LM358H (true metal can) is on the way and I'm going to replace that socket nonsense by soldering the original part.
Not sure about browned board around zener / CRD, these things should see about 1mA current at < 10V, the amount of power which should not char the board.
Clearly, something happened with CR5 probably...
« Last Edit: January 14, 2025, 07:41:54 am by dlebed »
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #6 on: January 14, 2025, 07:28:35 am »
BTW, here is a quick note on the foam insulation.

I was able to get access to the A5 board relatively easily (though my unit was made around ~1985), but the foam is crumbling and leaving particles all around.

Since replacing the foam with a new one is even messier process than desoldering the A5 board from all the wires, I decided to try to improve the insulation condition.

I was able to solve (for now) the crumbling problem by applying diluted (1:1) wood glue (PVA-based) to the foam surfaces.
The foam is anyway quite rigid already, but the PVA improved the surface condition and there is no more crumbling.

On top of that I'm playing with some extra surface coating.
First try will be aluminium tape.
It should seal the surface completely and might provide some extra insulation - we'll see.
2482607-0
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #7 on: January 16, 2025, 06:29:01 am »
I did some measurements on typical CR5/VR2 parts, here are the numbers:

Original VR2 (6.4V Zener) that I have in my 732 (Motorola part): Zener impedance at 1mA is 55Ohm, cold-warm voltage diff @1mA: 770uV (1mA => 10mA for 5 seconds (warmup) => 1mA)
1N4574A from Microchip (5ppm/C, 6.4V): Zener impedance at 1mA is 44Ohm, cold-warm voltage diff @1mA: -160uV

Fluke 334839 CRD (CR5): 2V to 3V current change: 5.4uA (impedance in 2-3V range is 184kOhm), this will result in 5.4uA * 44 Ohm = 237uV Zener voltage change

R9 which is 43kOhm for me, over 2.2V nominal will deliver 50uA of current into VR2, which results in 50 * 44 = 2.2mV Zener voltage change

It looks like that the CR5 CRD is there to deliver 1mA current to match the spec of compensated Zener VR2 to ensure that the VR2 has minimal tempco.
The R9 is added in parallel to match the current that will be fed into U2's collector.
« Last Edit: January 20, 2025, 08:31:57 am by dlebed »
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #8 on: January 20, 2025, 08:53:23 am »
I finished overhauling my 732a and I'm very pleased with my first test results.

I put the original Fluke's 334839 CRD that I was able to source into CR5.
Also, VR2 was replaced as original was had some charring on it.
Also, all CC resistors and tantalum and film capacitors (with PP ones) were replaced.
Thermistor accessible from front terminals was replaced with PT1000 RTD.

Initial Results:
10V output is within 0.5ppm from its value before overhaul (I even did a full IPA cleaning of A5 board).
10V load regulation improved from ~2-3mOhm to -0.04mOhm (-0.02ppm change at 5mA load).
Oven temperature is very good at 48.27C (according to RTD).

Unheated SZA263-based reference tempco in 732a
Tempco of 10V output during 732a heatup is about 0.28ppm/C in my case, this is measured during 732a warm-up.
What is interesting is that it's quite linear during the entire warm-up from 25C all the way to 48C.
Also, temperature rise is also linear with minor overshoot.

Heating over time (blue - temperature, orange - 10V output change in PPM).
2486043-0

10V ouput (in ppm) vs oven temperature scatter.
Curve Fit: 2.778E-01 * x + -8.116E+00
2486047-1

Oven control has little overshoot and looks very stable.
2486051-2

Oven temperature regulation is very good at about 0.01C / C (100x attenuation) (oven vs ambient), it might be even better as my 3458a measuring PT1000's resistance might be a limiting factor.
This will result in 0.28ppm/C / 100 = 0.0028ppm/C ovenized 732a tempco which is much better than 0.05ppm/C spec (18x better to be precise).
2486055-3
« Last Edit: January 21, 2025, 03:49:09 am by dlebed »
 

Offline dlebedTopic starter

  • Regular Contributor
  • *
  • Posts: 50
  • Country: us
Re: Deciphering Fluke 732a
« Reply #9 on: February 02, 2025, 07:29:43 am »
After running reworked 732A connected to 3 x 3458a and 1 x DMM7510, it seems that Fluke 732A is in good shape (at least after watching 10V output for 4 days in a row) and is not drifty.
Data is collected from 3458a's (A, C, D) with ACAL done before taking each sample and DMM7510. Samples are taken with 4 hour interval.
PPM change of average value over time (+ linear fit of -0.0015ppm/day drift) is plotted below.
At this point, even with 3 x 3458a's I'm still measuring 3458a's drift and not 732a's, but its good to see they're in agreement.
3458a's tempco is 0.15ppm/C with ACAL, while 732a is specified at 0.05ppm/C max.


P.S. I was rushing too much to put things together, so I did not take any photos of its internals after rework.
I probably will open it again after initial verification is completed to take photos and post before vs after comparison.
« Last Edit: February 02, 2025, 07:32:22 am by dlebed »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf