Fluke 732A with popcorn noise SZA?

[Inverted18650]

Stream it, Ill watch. 

here now: https://xdevs.com/live/

[Echo88]

While a TC-adjusted and stable SZA might have a different drift direction compared to the LTZ and therefore offers value as a reference compared to LTZs which might drift in the same direction, im always for more interesting LTZ-boards.   

[Dr. Frank]

A new board makes no sense to me... It's clear, that the LTZ topology does not match the SZA one.

So you would only copy the existing SZA 263 schematic into a new layout, maybe a more compact one, but the two core problems will remain.

I would at first determine the correct value for R5 to have near zero T.C. for the SZA chip @ 45°C oven temperature.
The 2nd problem is the slightly incorrect divider ratio for 10.000 00 V, but I wouldn't replace R27 / R28, because the low T.C. of the output is the most interesting feature. Maybe the coarse trimming by R10 / R11 is sufficiently small, so not to affect the T.C. matching.

Probably the timely drift of the SZA from the 8842A is not bad at all.. you'll have to see before you scrap the whole thing, like you determine stinkers in your LTZ group.

PS: Maybe that initial drift of -2ppm/hr is also caused by hysteresis effects due to soldering heat, like on LTZ1000 chips.

Frank

[TiN]

How is it clear that LTZ no match to SZA chip? I must have missed that part as in my books both are in same performance level when *equally* biased.

Another available option is to scrap my second bad 5700A REFDAC A11 board to donor dual SZA reference hybrid off it and make board around it. Since we have already external oven, maybe inner hybrid oven can be turned off (omitting lot of parts to drive it) or at least ran at wee higher setpoint, like 55C (if my math right, this 732A have oven about 49-50c). But this path will need to redo output network if we'd want to use 13V output to get lower noise.

Does not look like it's slowing down  ...

[Dr. Frank]

How is it clear that LTZ no match to SZA chip? I must have missed that part as in my books both are in same performance level when *equally* biased.

The SZA / LTFLU chip and circuitry have a completely different topology than the LTZ1000, due to the fact that in the first case, the anode of the zener is connected to the emitter of the compensating transistor, and in the latter case it's connected to the base.
So there is no useful way to connect an SZA chip into a LTZ circuit.
Also, the SZA circuit outputs an amplified and buffered voltage (e.g. 10.000V), whereas the LTZ circuit outputs the buffered reference voltage (e.g. 7.2000V) only, and requires an additional amplifier to get e.g. 10.000V   

Another available option is to scrap my second bad 5700A REFDAC A11 board to donor dual SZA reference hybrid off it and make board around it. Since we have already external oven, maybe inner hybrid oven can be turned off (omitting lot of parts to drive it) or at least ran at wee higher setpoint, like 55C (if my math right, this 732A have oven about 49-50c). But this path will need to redo output network if we'd want to use 13V output to get lower noise.

The 732A has an oven temperature of (48 +/- 2) °C, ( see chapter 3-22), whereas the 732B is set to 45°C (see chapter 4-11).

Sorry, that's been my fault to confuse that.

Why mixing two different oven assemblies? Can't you use the 5700 reference separately.. it's also not intended to be as stable as a 732A/B, due to its higher oven temperature, i.e. typ. 2x higher drift rate.

I really would try to fix that 732A in its original shape.

Frank

[TiN]

So there is no useful way to connect an SZA chip into a LTZ circuit.

Sorry, maybe my posts were misleading. I did not mean to use LTZ in original 732A circuit.
Of course whole refamp circuit would be replaced with complete LTZ block to generate 7.xV and amplify that for 10V output, all on fully custom PCBA.
Essentially the only part that would be original in such a box is 732A oven box with heaters, external battery/charger management stuff and 10V->1.018V->1V divider circuit.
So whole path would be LTZ ref 7.2V -> 10V gain block -> 732A divider -> output terminals.

5700A ref is ovenized at about +62°C, but it's not hard to reduce oven temperature for lower setpoint. And stability of 5700A REF (6.7 ppm for 99% spec, annual) is very competitive to 732A (6.0 ppm annual). Of course actual instruments show better performance than spec.

Why mixing two different oven assemblies?

Smaller hybrid oven would allow tighter temperature control, while large original 732A oven would keep all gain circuitry around nice and stable. Essentially like double oven concept in OCXO's.
Onboard oven on hybrid already there and cannot be removed from the module, so we might as well use it for benefit.

I really would try to fix that 732A in its original shape.

Yea, I'd love that too, but those spikes from original SZA chip bother me too much.

[TiN]

Okay, it looks settled at about -18ppm finally.
Also running on battery power now, to see how much life does it have.

[martinr33]

These things can be microovolt stable over a couple of years. I have two of them facing off over a nanovoltmeter in my uncontrolled garage. Over 10 - 30C and six months, the delta is always less than a microvolt. Checked it just now and it is reading 24.9uV, was 25uV when it was checked against other units in the summer.  Local friends with 732as show similar results. I know of about 9 units in the Bay Area (anybody else around here got one)?



So I would definitely let it sit a little longer, and see how it starts behaving.

Over the same period, it looks like my 3458a walked about 0.7uV. It had been off a while, and it took about a week to get the readings to where they are now.

Anyhow, if this unit is stabilizing it is a keeper. Maybe your diligence in cleaning out the gunk was the trick.

Also, for lab units - the ones I found always have signs of serious external abuse. I like to pick units by the labels on them. Mine came from:

   - A Fluke cal lab (from when they were in multiple locations) (labeled as primary standard)
   - The Australian Air Force (labeled as "calibration not required")
   - Sandia National Labs (labeled as primary standard)

I don't feel so good about units with all the labels cleaned off.

[TiN]

Lucky for me, I know what these can be, having access data over a year with bank of 732s. Question is however will this particular 732A specimen will ever get that good.
Perhaps I should assemble little breadboard circuit with original SZA chip and run some tests standalone in controlled TEC oven box with teckit.