Fluke 730A DC Transfer Standard

[branadic]

Good evening,

I've noticed that there are almost no images of Fluke 730A on the web, except this



It just so happened that a unit arrived on my desk (/AH 13508). The batteries had leaked and were already removed, but other than that the unit was working.
I already replaced the HP connector by an IEC connector, etched away the residue that the NiCd had left on the aluminum of the battery compartement with go-devil, re-newed every single solder joint on the mainboard, gave the complete board a thoroughly clean, replaced the cables from the mainboard to the batteries and today removed the switch assembly to maintain it, as the springs were corroded and had left corrosion on the board. Most likely I have to replace all the springs of the switches, nothing impossible. Almost there to power it on myself.
What I found interesting though is that the "Mean Output" switch was factory locked, but I couldn't find a hint in the manual about it.

-branadic-

[Vgkid]

For the mean , there could be an averaging amp / or voltage divider.
edit: apparently it puts the in parallel 
http://www.ko4bb.com/manuals/47.134.114.28/Fluke_730A_Voltage_Reference_Operator_Manual.pdf

[branadic]

By locked I mean the switch was physically locked by a plastic washer instead of having a spring and that washer was designed for the switch, so nothing that looks like a hack by someone to prevent the operator from pushing the switch. Hence why I'm curious if there were options the F730A was delivered with.
During the process of maintaining the switches one of the rods broke. 
Super glue didn't fix it and epoxy based glue didn't work either. I guess the rods are made of resin, so I will try UV curing resin as glue next. In worst case I have to cast the rod, as 3D printing probably won't work due to the very small features on the rod.
More images about to come, once I've managed to complete my overhaulin'.

-branadic-

[alm]

The reference board looks quite a lot like the Fluke 731B, although clearly with a different layout. Except with the DH80417B which was also used in the 731A and some early 332/335 series calibrators instead of the SZA263. I have an early 731B which also has the DH80417B refamp (see attached pics).

[Rax]

More examples of early reference boards pre-SZA263 (this is from my Fluke 332D). I didn't take good pics of the reference itself, but one pic depicts 4ST1-2 (just as the manual states), which I assume to mean it's the DH80417B? 

[branadic]

The reference is A3, it is located inside the black component, which is an oven/thermostat made by Klixon.

-branadic-

[Rax]

The reference is A3, it is located inside the black component, which is an oven/thermostat made by Klixon.
-branadic-

Yes, it is - I just meant I didn't open my "4ST1-2" part to see if there's indeed a DH80417B inside it, or there were other variations using a different reference (which I don't think is likely).

Frank has that documented very neatly at https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg393627/#msg393627.

Interesting to observe about the F730A is that its reference is not ovenized.

[branadic]

UV resin didn't work to fix the shaft, it was fairly easy to remove it once cured. The resin the shaft is made of seems to be rather "special".

I then started making a mold using alginate. This worked, but since the mold releases water all the time neither UV curing resin nor epoxy felt happy with that and the casted shafts looked like a mess. What I learned in the process: Forget alginate as a solution for a mold, it simply doesn't work out.
Already ordered some RTV silicone (https://www.silikonfabrik.de/silikone/silixon/silixon40-silikonkautschuk-rtv2-giesssilikon.html?number=SILIXON40-500-0709), but wanted to finish the project this weekend. I remembered to have some RTV blue silicone and took that to make a mold. This works to some extend as it is not that liquid. However, it takes 24 h for it to cure, so this also would have been a "finish the project this weekend" blocker.

Eventually, I looked inside a Fluke 8000A I had sitting in the shelf and to my surprise found the very same switches. So for the time being the Fluke 8000A had to serve as a donor for the shaft and the springs until I have made and organized replacements. I also removed the lock on the Mean switch and replaced it with a spring.

The unit is mostly back together and I could turn it on without having to fear something goes up in smoke and yes, she is working  

There is still some work to do before I can call it finished, but I'm close to aquire some data like noise, short-term and mid-term stability.

-branadic-

[branadic]

I made some progress.

First of all, thanks to user beanflying the instrument is now standing on its own feets:
https://www.eevblog.com/forum/testgear/replacement-knobs-feet-and-fittings-for-test-equipment/msg2502039/#msg2502039

Next, I have checked the output voltages, which were quite different on all four channels. I was able to adjust at least 3 of them to what I assume to be 10 V. Reference board 2 however was way out of adjustment range. So I looked on the board and found that the reference boards are Rev. C, which has nothing to do with the schematic in the Fluke 730A manual. After some reverse-enginerring and comparison I found that the schematic matches the one given in the Fluke 731A manual, but uses Flukes hermetically sealed resistors as in Fluke 731B. After I found that I've added a 4R7 resistor in series to R7, which brought back the adjustment capabilities.
I was now able to also bring channel 2 to 10 V, which allows to parallel (Arithmetic Mean) all four reference boards without huge changes. From here we can now take some measurements, intercompare the channels and look at stabilities. But before that, I'm planning to add a temperature sensor (thermistor GA10K3A1) to the unit and to connect it to some unused pins of the Centronic connector. This way no extra holes have to be drilled.

-branadic-

[branadic]

Attached is a first result of a 1h stability measurement of all channels one after the other after an initial adjustment of the ouput voltages. There is room to improve and to do better, I know.

-branadic-

[branadic]

I today worked a little bit more on the adjustment and repeated the measurements, this time only half an hour on each channel. The outputs are now way closer. I think I will leave them just there and observe their drift from here.

Some of these 2x10 pin card edge connectors were ordered, so I can put each reference individually into the thermal chamber and we can check how they perform over temperature. The connectors should arrive soon.

Meanwhile, I can focus on measuring low frequency noise.

I wonder if someone else with such a unit has results to share, too. 

-branadic-

[Bill158]

I have one of these 730A.  The S/N is #331 and it has the round precision resistors.  Attached is the ".xls" file I have generated from a much larger set of data on all of my various voltage references.  The value axis is in uVolts or "0.1 ppm" per division.  This is because I collect all my data using the NBS 430 technical note as my guide.  This takes the errors out of trying to get a reading at 10 volts with the DVM (i.e. H3458A).
This data has been collected over 12 years or more against my most stable f732A  S/N 3435xxx.  This 732A has show it's self to be extremely stable for about 20 years now, using the NBS430 note, against 4 other 732As.
As you can see the maximum deviation has been about 1.5 ppm max and those outliers are due to mainly "MEAN" switch contact problems.  This unit needs some work for sure.  Exercising the "MEAN" switch takes care of this issue.  Some day Real Soon Now I will work on this.
Bill

[branadic]

I today performed low frequency noise (0.1 - 10 Hz) measurements of the 10V outputs and repeated it on each channel. Here are the results:

Mean:
rms = 179.20 nV
std = 178.87 nV
pp = 1.2344 µV

rms = 167.78 nV
std = 167.16 nV
pp = 1.1632 µV

Channel1:
rms = 335.30 nV
std = 335.28 nV
pp = 2.4805 µV

rms = 338.79 nV
std = 338.74 nV
pp = 2.1987 µV

Channel2:
rms = 297.15 nV
std = 296.80 nV
pp = 1.9943 µV

rms = 294.98 nV
std = 294.49 nV
pp = 2.3202 µV

Channel3:
rms = 286.60 nV
std = 283.83 nV
pp = 1.9611 µV

rms = 275.13 nV
std = 272.28 nV
pp = 1.9004 µV

Channel4:
rms = 297.94 nV
std = 297.96 nV
pp = 1.9335 µV

rms = 332.34 nV
std = 332.31 nV
pp = 2.3700 µV

So far, all measurements were performed with the NiCd batteries removed, replaced by some electrolytic caps and the unit running from mains. I'm still thinking about a proper battery repacement, be it NiMH or Supercaps. I was told that Conrad replaced the NiCd in one of his references with batteries plus zener diode.

-branadic-

[alm]

How did you measure the noise? AC-coupled LNA?

Much better noise figures than I've seen reported before on the DH80417B (though that was a 2 hour measurement). A bit less than sqrt(4) improvement in noise of the mean output compared to the individual channels. Which makes sense, given that not all noise will be uncorrelated.

[branadic]

The measurement was done using pipelies 0.1 - 10 Hz, 80 dB LNA in front of an R9211E, so yes, AC coupled.

I'm not sure why Frank reported worse results, it could be partially due to less quality resistors in Fluke 332? 

-branadic-

[branadic]

Today a second unit (SN 302), that was donated to me, arrived. It has even older turn counting dials, but they could have been replaced at some point in the past by a previous owner, as the top cover has some cutouts to make them fit, that don't look very original to me.

This unit obviously is revision A that is shown in the figures given in the manual. There are some bodge resistors on the back, for which I need to figure out if they are an original afterthought or something that was added by a former owner.
Attached some very first impressions with the Mica resistor goddies on the reference boards.

The device was missing the feets, so some 14 h later proper replacement was available. Furthermore, it was missing one of the Guard covers, which I've recreated from 1.5 mm aluminum.

-branadic-

[branadic]

I was asked to provide a comparison of the noise measurements above to my LTZ1000 references running on batteries. The noise floor of the LNA is known to be 100 nVpp with shorted input.

LT1000-1 (10.008658 V):
rms = 257.86 nV
std = 257.84 nV
pp = 1.5744 µV

LTZ1000-2 (10.008774 V):
rms = 283.23 nV
std = 283.22 nV
pp = 1.9169 µV

LTZ1000-3 (9.989901 V):
rms = 268.25 nV
std = 267.62 nV
pp = 1.6186 µV

I also attach results for the 10 V output of my ADR1000 voltage reference running on batteries for the measurement.
rms = 152.91 nV
std = 133.30 nV
pp = 919.21 nV

-branadic-

[doktor pyta]

The reference board seems familiar: https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/msg5208300/#msg5208300

[W6EL]

Greetings,

I have a Fluke 633A which seems to have two of the same reference boards. Mine are similar to the ones in the manual's illustrations, like your second set. I would guess that the dividers might be set differently on the output side though.

I am currently doing a test over the course of the day -- just one reading every hour or so, manually -- but I am curious as to how stable these things are without ovenization.

I'd like to hear your thoughts on the additional stability gained by parallel units and/or ovenizing. I too might order some extra sockets and build a little box inside this enormous 633A to house a pair in a controlled thermal environment.

I'll post clear photos of the reference boards once this set of tests is complete.

--E

[W6EL]

Alright, here are the reference cards from the Fluke 633A. They do look very similar to your set.

In the Fluke 633A, one card is run from about -10V up to ground, and the other card is form ground up to about 14 volts. I'm still trying to understand the 633A circuits, but I think ultimately I will ovenize the pair and "average" them in parallel. It seems like a good project for this rather spacious box. I could then create these odd-ball reference voltages that the 633A provides using the 10V reference output (but with more PPMs of course).

Oh hey, I also noticed that these boards say "730A" on them. So, there you go. The Fluke 633A is probably a great source for replacement reference boards, although they do not seem to show up on ebay much. Also 622A, and 644A likely.

[W6EL]

Also, I made some readings using my (uncalibrated) Fluke 8505A. All the equipment is in my garage, without any temperature regulation. Basically somewhere between outdoor (SoCal) temperature and indoor temps (two shared walls with the house). Pleasant but not exactly a controlled environment.

Note that the scales are not the same in these plots, so judge carefully. The Fluke 8505A uses the same reference as these reference boards, although the circuit is not exactly the same.

Standard deviation on both + and - outputs is about 17 microvolts. Span on the + was 54 microvolts, span on the negative 10V output was 58 microvolts. For these stats I only used the last half or so of the datasets, so as to not include the first day of "warm up" and all that.

[W6EL]

Today a second unit (SN 302), that was donated to me, arrived. It has even older turn counting dials, but they could have been replaced at some point in the past by a previous owner, as the top cover has some cutouts to make them fit, that don't look very original to me.

This unit obviously is revision A that is shown in the figures given in the manual. There are some bodge resistors on the back, for which I need to figure out if they are an original afterthought or something that was added by a former owner.
Attached some very first impressions with the Mica resistor goddies on the reference boards.

The device was missing the feets, so some 14 h later proper replacement was available. Furthermore, it was missing one of the Guard covers, which I've recreated from 1.5 mm aluminum.

-branadic-

Looks like R24 is broken on one unit, just in case you hadn't spotted that yet.

[branadic]

Looks like R24 is broken on one unit, just in case you hadn't spotted that yet.

No, it's not broken, that is one of the bodges I mentioned earlier. There is a parallel resistor on the bottom of the board, while the top one was lifted on one side.

Meanwhile, all electrolytic caps and carbon composite resistors have been replaced. The unit was already equiped with an IEC connector, which I replaced by an IEC connector with mains filter (1A) on both units. I yesterday performed very first measurements on the arithmetic mean, refboard 1 to 4 and arithmetic mean again over a course of 30 min each. As we can see at least one refboard seems to be suspices showing rather large noise.

-branadic-

[W6EL]

branadic, what are you using to make these measurements?

[branadic]

The 30min measurements were performed using Advantest R6581 8.5-digit multimeter.

-branadic-

[W6EL]

This is quite impressive.

I wonder if those little pops you saw on one channel could be from a fresh capacitor. Maybe it will need a day of burn-in to stop?

I like to reform my caps before use. Sometimes you will see some odd stuff if you monitor the microamps during a reform. I imagine such a thing could show up in your data, especially at these levels.

Or, something else. Who knows.

[dietert1]

Looking at the diagram one could say a single reference is about 4 uVpp and the mean about 2 uVpp - as expected from statistics. To me it appears as if the Channel 3 reference may be ok and its visible noise was a measurement problem, as the mean behaves normal. The red curve with its low noise indicates there is room for improvement. Still a nice sup-ppm reference.

Regards, Dieter

[branadic]

To me it appears as if the Channel 3 reference may be ok and its visible noise was a measurement problem, as the mean behaves normal.

What you mean by that? While I agree in theory on the observation, the setup was left untouched between the measurements, but only the buttons on the front panel were pushed from arithmetic mean to channel 1 ... 4 and to arithmetic mean again. The cable was connected to the output binding post prior the very first measurement with some waiting time and left as is.

-branadic-

[dietert1]

For example there could be a bad contact in the switch. Or the averaging output of each reference module is separate from the output used for the switch. I seem to remember this from the schematics of the Valhalla unit you showd before.

Regards, Dieter

[branadic]

I think there is a simple test to find out if the switches are dodgy. Instead of selecting each reference by the switch, I could set the unit to arithmetic mean and set all other references to open via the voltage output switch, except the one I want to measure. This way I can see the contribution of the selected reference to the mean.

-branadic-

[alm]

I think there is a simple test to find out if the switches are dodgy. Instead of selecting each reference by the switch, I could set the unit to arithmetic mean and set all other references to open via the voltage output switch, except the one I want to measure. This way I can see the contribution of the selected reference to the mean.

Or how about using the Centronics-type connector at the back which has each individual reference on a separate pin?

[branadic]

Easy answer, there is no Centronics connector at the back of this unit
I wonder if the connector at the second unit is a factory modification. 

-branadic-

[branadic]

I've repeated the measurement with the front push switch in the arithmetic mean position and by setting the reference to be measured to the 10V position using the rotary switch, while the other three references were set to open. As we can see, the result looks identical to the former one. This way we can rule out the reference selection push switch to be the reason for the large noise on channel 3. We can also see signs of popcorn noise on this reference board.

-branadic-

[branadic]

To further rule out switches or voltage regulators, I rearranged references in the unit, that is reference 4 went to position 2, 2 to 3 and 3 to 4. I then repeated the measurement (after some short adjustment that was necessary as output voltage changed on all of them, most likely due to the supply voltage).
So I'm now sure it's the reference board and not a switch contact or similar.

-branadic-

[branadic]

While working on a replacement reference board I found that the schematic for Rev. A given in the manual has some mistakes at the card edge connector, so attached a corrected version of the schematic.

-branadic-

[branadic]

I received some card edge connectors, which helps to pull the reference boards and to perform some t.c. measurements on them. Attached a very first result of reference board 4, which is the one with the large noise. I used a limited temperature range of 10 ... 23 ... 36 °C. I will go on with one after the other, so we get an idea of how they perform.

-branadic-

[branadic]

Meanwhile, I've measured t.c. on reference board 3 and board 2 is currently running. What I found interesting is, that board 3 with lower noise also has a larger second order component in its t.c. curve. Is that just a coincidence or some correlation that someone else has observed in general and can confirm?
I think we will know more once I've measured all four boards of F730A Rev. A showing different noise levels.

I've also ordered a set of Sub-C NiMH cells to build a battery pack for F730A Rev. B first.

-branadic-

[Dr. Frank]

While working on a replacement reference board I found that the schematic for Rev. A given in the manual has some mistakes at the card edge connector, so attached a corrected version of the schematic.

-branadic-

This schematic still has some faults inside, especially concerning the trim pot R10, its value and its location. This pot should be located between  R11 and R12 and should have on the order of 50 Ohm, or should be drawn as two separate resistors, as you have correctly drawn it in your schematic
Frank
You have

[branadic]

Frank, R10 isn't a trim pot, but a wirewound resistor network on a single mica card.

-branadic-

[branadic]

What I found interesting is, that board 3 with lower noise also has a larger alpha in its parabolic t.c. curve. Is that just a coincidence or some correlation that someone else has observed in general and can confirm?

So while the evidence gets stronger during the current temperature sweep, I've searched the web and found at least one source that seems to confirm my current observation

ZENER-DIODE NOISE GENERATORS

Zener diodes operated in the breakdown region have been investigated as transfer noise standards in the r.f. region. A close correlation has been found between the variation in excess noise ratio (e.n.r., dB) and the temperature coefficient of Zener diodes.

while this is an interesting read too
Noise as a Diagnostic Tool for Quality and Reliability of Electronic Devices

-branadic-

[Kleinstein]

There is correlation between the noise, TC and zener voltage.  Lower voltage zeners (in the relevant 5-7 V range) tend to have less noise and a more negative TC. The noise however also depends on the construction. In combination there is also a correlation noise and linear TC. On the other side the high voltage zeners can have a lower differential resistance and are thus less critical with the current stability.

With some of the zener reference there are reports that Zeners (and AFAIR also resistors) with more popcorn type noise also showed more long term drift than the less noisy ones. This somewhat makes sense as the popcorn type noise can be seen as current paths that switch between active and inactive. Beside the random changes causing noise there could as well be more permananet activation / deactivation of the same paths.
So there is some incentive to look for the low noise ones, not just for the noise, but also to hopefully avoid the rather drifty ones.

The zener references here show a good trim of the linear TC (selecting the right resistors in the ref. amplifier) and only a difference in the 2nd oder TC and thus the curvature. The curvature may still correlate with the noise, but that is likely more indirectly, e.g. via different voltages as a common cause for both. With just one sample of a slightly more noisy ref. this are not really much data. It could be as simple as a lower ref. current because of a different voltage to start with.
With a relatively high linear TC (at least for parts) there is a chance that temperature fluctuations cause low frequency noise.

[dietert1]

How are these TC measurements done? What is the temperature parameter?
As far as i understood, each reference board is used with a cable extender, i.e. outside of the 730A instrument and inside some temperature chamber. So the nice temperature adjustment for 20 .. 26 °C might be subject to additional heating when running inside the instrument, with all covers closed.

Regards, Dieter

[branadic]

The reference boards are pulled from the F730A and connected to the meter as well as a linear lab power supply set to 13.2 V (that is what the regulators in the F730A are set to) using a card edge connector. The board is sitting inside my thermal chamber (incubator controlled by an Arroyo 5305). The temperature you are looking at is the temperature of the chamber, with the temperature sensor of the controller being close to the DUT. By the way, inside the instrument there is plenty of space between the cards.

Attached is an update of the diagram with the result of reference card 2 included.

-branadic-

[branadic]

Attached is the t.c. plot with all four reference being measured. To me it still looks like the noise really correlates with the opening of the parabolic t.c. curve as reference board 1 showed the lowest noise of 'em all followed by ref 2, 3 and eventually 4.

-branadic-

[dietert1]

Yes, if we assume that inside the instrument the temperature will be 5 or 10 °C higher, then references 3 and 4 will be operating in a region with much higher slope than reference 1. One could try to adjust their zero TC temperatures a little higher.

Regards, Dieter

[Dr. Frank]

Hello,
it would be of interest, whether branadic measured the temperature inside or outside the package (I mostly measure the inner temperature on such experiments.)
For the LTZ1000 reference, the temperature difference between inside and outside, with a thermal isolation around the circuit, is at about 6..7°C, if I remember correctly, and all my ovens run at around 50°C.
The A version consumes about 18mA @ 12V supply, whereas the non A version consumes 22mA, i.e. these are 220 and 260mW.

As branadic wrote to me, one module / reference of the 730A draws <6mA @ 13.2V, i.e. 80mW or 1/3 of the LTZ case.
So I expect a rise of the inner temperature of only 1..2 °C.

I guess, that Fluke has taken that into consideration, and set the zero T.C. point to the intended nominal room temperature.
This is evidently 25°C, as the 730As T.C. is specified (symmetrically) 0.5ppm/°C from 20 .. 30°C.

Frank

[branadic]

Hi,

it would be of interest, whether branadic measured the temperature inside or outside the package (I mostly measure the inner temperature on such experiments.)

My thermal chamber is way too small to fit the whole 19" rack into it, so

The reference boards are pulled from the F730A and connected to the meter as well as a linear lab power supply set to 13.2 V (that is what the regulators in the F730A are set to) using a card edge connector. The board is sitting inside my thermal chamber (incubator controlled by an Arroyo 5305). The temperature you are looking at is the temperature of the chamber, with the temperature sensor of the controller being close to the DUT. By the way, inside the instrument there is plenty of space between the cards.

As the Rev A unit also suffers from not having a connector on the rear panel I have no temperature sensor inside the rack. It would require me drilling a hole into the unit, which I haven't done just yet.

-branadic-

[branadic]

In the meantime I received some battery cells and am in the process of welding/building battery packs for both units. I also received some switch assemblies and was able to repair the power button on Rev. A unit.
I also added a connector at the back and an internal 10k thermistor (GA10K3A), to monitor internal temperature in the future.
Furthermore, the kind previous owner of the Rev. A unit sent me more reference boards so I can select the ones with lowest noise to complete the unit into a fully working and usable device. More noise and t.c. measurements to be performed soon.

-branadic-

[branadic]

From the reference boards I received I selected 4 out of 8 that showed lowest noise for F730A Rev. A. I had to add a small 3Ω resistor to one of the reference boards to get back into adjustment range for the 10 V.
After some adjustments of the 10 V on all channels I captured 1 hour stability data for the averaged output and it looks rather promissing, given the unit in this configuration is just running a few hours and needs some time to stabilize. Note: This measurement was made with the unit powered from mains, as some battery cells are backordered, so I couldn't complete the fourth battery pack for both units.

Next step is to check and if necessary adjust the 1 V, 1.018 V and 1.019 V as preparation for a volt comparison with Frank.

-branadic-

[branadic]

More cells arrived and I could finish the battery packs for F730A Rev.A. I used NiCd batteries for this one.

-branadic-

[branadic]

Just wondering if anyone else has more insides on that topic?
There is this table:



F730A: DH80417B
F731A: DH80417B
F731B: DH80417B/SZA263

all (three) with the same performance in terms of t.c. and noise, which indicates that DH80417B, no matter who manufactured them (Halske&Siemens, Dickson, TI), were similar in performance.



With F732A, still using SZA263, Fluke ovenized the reference and associated components for the first time and achieved 20x better t.c. and 2x lower noise. Did ovenizing the reference prevent air drafts/thermal fluctuations, that showed up as noise in the former models?
With F732B Fluke switched to LTFLU, wich improved noise by about sqrt(4) due to 4 zeners in parallel.



In LTFLU we can spot the two resistors on the die, that could act as heating elements, but I can't spot a temperature sensor. So question is, why would you design them in?

Are my aforementioned conclusions reasonable or is there more to it?

-branadic-

[Kleinstein]

For the LTFLU one could still use the BE voltage to "measure" the temperature - no ideal, but it can be done. Depending on how good the symmetry is, the heater may not be uniform enough and possibly cause more trouble than good. An external heater can work as well.
The chip internal heater could still be used temporarily for finding the right current / number of transistors used and fuses to blow. So chances are it is used during production. Here a heater close to the reference can speed up the test and is thus very well come.

[branadic]

The chip internal heater could still be used temporarily for finding the right current / number of transistors used and fuses to blow. So chances are it is used during production.

Yes, it is somehow used during the production, we can see the scratch marks at the associated pads.

-branadic-

[dietert1]

I haven't seen any LTFLU with bonded heaters. Anyway, one can use Ube of the transistor to measure LTFLU chip temperature. In 2019 i implemented that for our LTFLU reference builds. The circuit includes a 100x gain amplifier inside the oven, so there is a temperature measurement with about 210 mV/K available. The signal is used for oven tuning and for TC adjustment.

See: https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg2578746/#msg2578746
Later i changed the feedback network from 10K/1M to 1K/100K (metal film resistors).

Regards, Dieter

[branadic]

Frank also provided an answer here

These RefAmps all are relatively tightly specified concerning their initial reference voltage span, but especially for the span of the transistors collector current (20..200µA) where zero T.C. can be achieved, over defined operating temperature span, with zener current being 3mA.

Therefore it's clear, that they have to heat the device in situ, in turn to trim the transistor and zener to the correct above mentioned parameters.

I was more interested in how the voltage standards based on DH80417B/SZA263/LTFLU improved in spec and what was changed design-wise to get there. Anyone with more inside on that topic?

I haven't seen any LTFLU with bonded heaters. Anyway, one can use Ube of the transistor to measure LTFLU chip temperature. In 2019 i implemented that for our LTFLU reference builds. The circuit includes a 100x gain amplifier inside the oven, so there is a temperature measurement with about 210 mV/K available. The signal is used for oven tuning and for TC adjustment.

See: https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/msg2578746/#msg2578746
Later i changed the feedback network from 10K/1M to 1K/100K (metal film resistors).

I've seen your posts, but couldn't find a comprehensive report what was "finally" achieved including all the details. Unfortunately, some aspects are spread over multiple pages and threads, but others seem to be missing. Would have liked to see it all together including images of the build to enjoy the achievements made.

-branadic-

[dietert1]

No, i posted schematics, images and some pdf-papers and hundreds of people used them. In comparison to those Fluke references my reference builds are superior as i used low temperature TEC ovens. I'd guess it's worth looking.
And yes, there isn't a final result. The last result i showed was from October 2022, where i logged the difference between two of those LTFLU builds. There were seven consecutive days where the daily averages stayed constant to 0.01 ppm. I'm pretty much convinced the ultimate LTFLU-1 performance will be about that.

Regards, Dieter

[branadic]

I had a Fluke 730A transfer standard Rev.B of a fellow voltnut with me for a few weeks, on which I've measured t.c. and that I'd like to share. Attached are the results.

Today I finish the t.c. measurement on my Fluke 730A transfer standard Rev.A, before I switch to Fluke 730A transfer standard Rev.B as well.

-branadic-

[alm]

Interesting how ref 2 and 4 are different. Ref 2 with the hysteresis and ref 4 with the much higher TC. Ref 1 and 3 look better than I expected.

[branadic]

I have the impression that the Mica card wirewound resistors with the coating applied don't necessarily like rising temperature slopes from cold to warm and experience some sort of stress or wire-movement that leads to the observed hysteresis. I can see a similar effect on the Fluke 730A Rev.A reference boards (see attachement). Surprisingly, one unit shows rather low noise (green trace) compared to the others and one shows a rather low t.c. (cyan trace), besides a large hysteresis effect at the rising temperature slope from 10 °C to 23 °C.

-branadic-

[branadic]

I today received a very first sample of a copycat of the xformer that is used in F730A. Why I had it made? Because it allows for 4 floating references like in F730A, while primary and each secondary is shielded individually. The primary is 2x 115 V and the secondaries are 24.8 Vac with no load. I'm planning a group order so in case you are interested, please send me a message via mail or PN. More details are about to come.

-branadic-