The LTFLU (aka SZA263) reference zener diode circuit

[branadic]

Hello,

branadic sent me some pictures with a higher resolution. Every picture has around 10 MByte,
probably here are some people which can do an analysis of these pictures.

Here is the link:
http://www.mounty.de/LTFLU/index.html

And thanks again to branadic and his efforts.

Regards,

Andreas

I wonder if you already started setting up your LTFLUs?

[TiN]

Hm-hm..

[VintageNut]

Who is this Cal person and why couldn't he?

Looks like your dual LTFLU has arrived and your screwdriver is on the job.

[TiN]

I don't know about screwdriver (only one screw in whole thing), but camera is definately eating battery like crazy...

[zlymex]

Hm-hm..

That is the A11 module from a Fluke 5700A/5720A

[Dr. Frank]

I just took Flukes oldest RefAmp, see here: https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg393627/#msg393627 and modified it for 10V reference voltage.



That is, to replace the PWW 9k / 7k24 divider by 5k/10k for 10V,



and changing the other TF resistors which determine collector voltage and currents, so that these parameters stay unchanged.
This gives a similar circuit like in the 731B and 732A/B, and should behave identical, even if an SZA263 or LTFLU is used there.



Afterwards, I made slight disturbances on these 5 resistors by consecutively paralleling a 1M resistor, and then measuring the corresponding influence on the RefAmp, or the reference output voltage.

The relative decreases for each resistor was on the order of 0.15 .. 1%, and the relative changes of both voltages were attenuated, between several ppm and 3200ppm.



Dividing the relative resistor change by its corresponding relative voltage change, gives the attenuation or suppression factor for each resistor.
That means, any drift (i.e. time, temperature) of this specific resistor is attenuated by this factor.

As an example, a factor of 500 would reduce the effective T.C. of 50ppm/K for a TF resistor to 0.1ppm/K of voltage output.




This calculation was already done for the LTZ1000 circuit, which also gave attenuation factors between 75 and 700 for the different resistors.
Therefore, the 6.8V => 10V amplification resistors have to be ultra stable, as their drifts are attenuated by a factor of 3 only, whereas the other ones should also be PWW types at least, as their factors are all on the order of 170..500.
So the FLUKE type RefAmp circuit behaves - no wonder - very similar to the LTZ1000, in this aspect.

Its advantage is the direct 10V output, whereas the LTZ1000 has the advantage to deliver the raw 7.2V reference voltage as a low impedance output.

Frank

[lukier]

Nicely done Dr. Frank. The study of suppression factors is something very useful, might save someone a lot of money (by not paying $50 for every resistor in the circuit )

If you now parallel that with a 10V LTZ1000 based reference you should get something very stable. I remember in one of Fluke papers that SZA263 drifts down and LTZ1000 drifts up (or the other way around). Averaging the two could provide ultra stable 10V reference, limited only by the output scaling resistors.

[Kleinstein]

The output scaling resistors are like the more difficult part than the reference chip itself. For a more stable 10 V reference it might be worth of using a more stable 7 to 10 V scaling.

[Echo88]

@ zlymex: You dont plan on opening the A11-module and showing us its glorious guts? Pretty please.  After all, its the best realised PWM-DAC as far as i know. Meanwhile my little brain still tries to comprehend the Datron 4911 -PWM-circuit...

[Dr. Frank]

@ zlymex: You dont plan on opening the A11-module and showing us its glorious guts? Pretty please.  After all, its the best realised PWM-DAC as far as i know. Meanwhile my little brain still tries to comprehend the Datron 4911 -PWM-circuit...

I strongly suppose, that TiN on xDevs.com will do that (or done??)   

[elecdonia]

For many years I've owned several Fluke 8800A 5.5 digit multimeters.  I use them daily and I really like them a lot.   Over time I picked up several more 8800A parts units too.  The reference devices in several of my 8800A multimeters are labeled "SZA263" with the Motorola icon. But a couple of them look different (If anyone would like photos of the oddball reference devices I will post photos).

Recently I started a project to convert one of my "parts" 8800A multimeters into a reference voltage source, calibrated voltage divider, and buffer amplifier.

The 2 sections of the 8800A that I will be re-purposing are these:

The first part I will use is the internal +/- 1V references.  This is the SZA263 section.  The +1V and -1V each have individual trimpots for calibration.  The +/- 1V outputs come from voltage dividers with 1K5 source resistance.  So they can't be loaded.  On the other hand, the +/- 7V outputs don't have individual trimpots, but they do have a low output impedance.  I believe the trimpots can be used to bring either the +/-7V or the +/-1V to the exact target output voltage.

I will also use the 8800A input buffer amplifier.  It has a very high input impedance, near 0 bias current, and when operating at a gain of 0.1 it can handle inputs of +/- 20V.  On the other hand, the maximum output voltage available from the buffer amplifier is +/- 2V.  But that's high enough for the things I plan to do with it.  The buffer amplifier can be set to have a gain of 0.1, 1 (unity gain), or 10 by logic signals.  There is also a separate 100:1 input voltage divider that can be switched in front of the buffer amplifier.  It has a 10 meg input resistance, however.

I will be putting control switches onto the front of the 8800A panel in place of the original  LED display and control levers:
One switch will have 3 positions, to select input from the front binding posts, or from the internal +1V or -1V reference voltage.
Another switch will configure the buffer amplifier gain to 1/10, 1, or 10.
And finally I will have a switch to enable the 100:1 input divider

The result will provide SZA263-based references voltages of +/- 1V and +/- 100mV
If I use the 100:1 input divider then I can also get +/-10mV and +/-1mV outputs
And finally it can operate with an external voltage reference or even a standard cell because the buffer amp has such a high input resistance

When I get things working I'll start a new thread for this project.

[TiN]

Sounds great project. Keep us posted!
To my shame, I didn't do much yet with my Fluke boards..

[Kleinstein]

Just for curiosity, I had a look at the schematics of the Fluke8800. The reference circuit be itself is a little odd. While the SZA263 can be a very good reference, the circuit in the 8800 is strange, in that there are quite a few resistors involved to set the 1 V and also the 7 V voltage. The 7 V is not directly the internal 7 V of the reference itself. Also the OP (LM301) used is not very good by today's standard. The second strange point is, the the reference current is set by the 18 V supply and not as usual bootstrapped from the reference itself - so this is not a super stable reference circuit. As the SZA263 relies on the suitable individual resistors to get the best performance, it could be hard to change that.

The input amplifier is good for having the large range and stable divider. However the time 0.1 output is more like the divider just behind the amplifier. So it is high impedance, just like the 1 V output of the reference. So t would need an extra buffer. The x1 and times x 10 modes should be able to give more than 2 V too and could be low impedance.

[elecdonia]

Just for curiosity, I had a look at the schematics of the Fluke8800. The reference circuit be itself is a little odd. While the SZA263 can be a very good reference, the circuit in the 8800 is strange, in that there are quite a few resistors involved to set the 1 V and also the 7 V voltage. The 7 V is not directly the internal 7 V of the reference itself. Also the OP (LM301) used is not very good by today's standard. The second strange point is, the reference current is set by the 18 V supply and not as usual bootstrapped from the reference itself - so this is not a super stable reference circuit. As the SZA263 relies on the suitable individual resistors to get the best performance, it could be hard to change that.

Most of the available schematics are for the oldest version, circa 1975.  By 1978 Fluke changed U7 to LM308H (in T05 can).  I was surprised to see Fluke using the LM301A myself!

Actually the 18V supply regulator itself is referenced to the 7V.   Therefore the 18V should be rather stable.  It is a somewhat byzantine circuit design overall!

The input amplifier is good for having the large range and stable divider. However the time 0.1 output is more like the divider just behind the amplifier. So it is high impedance, just like the 1 V output of the reference. So t would need an extra buffer. The x1 and times x 10 modes should be able to give more than 2 V too and could be low impedance.

Yes, I was thinking about adding a unity gain buffer for the output.  Actually there is already an available opamp in the 8800A that could be re-purposed for this.  It is U3, and it is already configured as a unity gain buffer.  In the original design it is part of the dual-slope integrator A-D converter.  But it could be rewired to perform as a unit gain output buffer with high input resistance.

[elecdonia]

Comparison of voltage reference section of 2 different vintages of Fluke 8800A 5.5 digit multimeter:

Older unit (circa 1975) uses LM301A in reference circuit

Newer unit (circa 1979) uses higher performance LM308H in reference circuit

[Dr. Frank]

Comparison of voltage reference section of 2 different vintages of Fluke 8800A 5.5 digit multimeter:

Older unit (circa 1975) uses LM301A in reference circuit

The reference amplifier, U9, in the older unit probably is the old T.I. reference, which was initially used in the 332/335 instruments also.

https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg393627/#msg393627

Frank

[elecdonia]

Comparison of voltage reference section of 2 different vintages of Fluke 8800A 5.5 digit multimeter:
Older unit (circa 1975) uses LM301A in reference circuit

The reference amplifier, U9, in the older unit probably is the old T.I. reference, which was initially used in the 332/335 instruments also.
https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg393627/#msg393627
Frank

Yes that is the reference device in my oldest Fluke 8800A.  Same markings:  DH80417.
I have 4 other Fluke 8800A multimeters:  They all have a Motorola branded SZA263.  According to date codes my most recent 8800A was manufactured ini 1986.

[elecdonia]

One thing that continues to surprise me about the SZ263 7V reference as implemented in the Fluke 8800A multimeter series is this:
These multimeters don't use an oven.  The SZ263 is simply sitting right there on the main PC board alongside all the other components.  From photos posted in this thread it appears that the newer 8840 and 8842 are similar:  No oven.
Yet the TC seems to be excellent.  I've heard that Fluke factory-selected the appropriate resistors to control the TC.  They must have had a great system for doing that at the factory.

[Dr. Frank]

One thing that continues to surprise me about the SZ263 7V reference as implemented in the Fluke 8800A multimeter series is this:
These multimeters don't use an oven.  The SZ263 is simply sitting right there on the main PC board alongside all the other components.  From photos posted in this thread it appears that the newer 8840 and 8842 are similar:  No oven.
Yet the TC seems to be excellent.  I've heard that Fluke factory-selected the appropriate resistors to control the TC.  They must have had a great system for doing that at the factory.

The SZA263 can very well be trimmed to near zero T.C. by the zener and transistor currents (in contrast to the LTZ1000).

That's the basic principle of the 731A.


Fluke was great at T.C. compensating, either resistor pairs and reference amplifiers, also.
 
Frank

[Kleinstein]

Trimming to zero TC was quite common in the early days.

If you really want, one could also adjust the LTZ1000 to near zero TC without the heater. It just would need a rather low current and thus might not be very practical.

Edit: I was wron on the LTZ, it it would need a very high current and thus might not work to get zero TC with just a different current.

[elecdonia]

My experimental modified Fluke 8800A is gradually coming together.

I am getting excellent results using the internal buffer amplifier in gain = x1 mode to buffer the +1.000000V and -1.000000V from the SZA263 reference circuit.  I've got a very stable output and since it has a low output resistance I can use the +1V and -1V with devices that have 10 meg input resistance.
 
Problems:
The old-fashioned 2-layer PC board layout may have issues with DC offset voltages up to +/- 10 uV at different VSS (ground) points around the PC board.  I can't zero out the DC offset from the buffer amplifier for all 3 levels of gain (10x, 1x, and 0.1x).  I can get any one of them < +/- 1 uV but I can't get all 3 gains to zero without individual tweaking of each gain level.  Worst case DC offset is about 100 uV at gain 10x.

I think there could be another issue too:  The non-inverting  input of the buffer amplifier has bias current compensation, but the inverting input does not.  This might be an issue when using 10x gain.  The buffer amplifier input stage uses bipolar transistors.  I suppose Fluke went with the bipolar input stage because an FET input would have too much DC offset voltage drift?

Oh well I'm having a lot of fun chasing microvolts.

I'm also thinking about getting some modern opamps!   That could be a really simple way to upgrade ancient gear!

This has gone far enough that I will start a new thread for my next post on the project.  Not sure if I should put it in Metrology or in Projects category.  Any recommendations from experienced forum members?

[Kleinstein]

A dual layer board is not a problem per se - it is a question on how good the layout is in using star ground and avoiding thermal EMF at resistors.
Not having bias compensation on the inverting input could be the reason to have different offsets.
A good OP like LT1012 could likely outperform the discrete input stage.

[TiN]

Interesting to note, there is non-A LTFLU-1CH inside of today's 8842A teardown vid from Dave. I thought there are only LTFLU-1ACH version in existence. Perhaps same difference, as LTZ, different die attach material?

Now I need to buy 8842A too! 

[Kleinstein]

As the LTFLU does not have an internal heater, the higher thermal resistance mounting does not make sense. I would more expect a lower grade (e.g. not so well tested) type of reference. This would make kind of sense as the 8842A is supposed to be only 5.5 digits - so it might not need the very best reference, still a lower grade LTFLU might still outperform an LM399.

[VintageNut]

I have been living with 2 x 731B and a DMM7510 for over a year. Both of the 731Bs have a nonovenized SZA263 and the DMM7510 has an ovenized LTFLU.

The DMM7510 measuring the best 731B over large temperature swings in my house/office/lab never wanders more than 1ppm p-p. The std dev is usually less than 1/4 ppm.

Each of the 731B cost me about USD $250 and the DMM7510 is USD $4,000.

I am extremely happy with this 10V setup. No desire to spend any more $$$ trying to get a better 10V. I would go so far as to say I have the best 10V in my neighborhood.