Ultra Precision Reference LTZ1000

[TiN]

LTZ1000 circuit itself does not require external oven to reach its typical performance ~0.05 ppm/K. Battery is preferred because it avoids risk of ground loops during measurement, but its not needed for burn-in or aging. Decent linear power supply like those old HP ones would be sufficient.

[alex-sh]

LTZ1000 circuit itself does not require external oven to reach its typical performance ~0.05 ppm/K.

That's helpful - thank you

Battery is preferred because it avoids risk of ground loops during measurement, but its not needed for burn-in or aging. Decent linear power supply like those old HP ones would be sufficient.

OK, I need 15V input (LT1763 can take it up to 20V and then dropping it down to 12V). What battery are you using please? I have 1800mAh 18.5V 5S Graphene battery, but it would not be enough I suppose. How is the battery charged? Swapping for another one?

[Hermann W]

what battery are you using please?

For aging, I take a simple plug-in power supply. For measurements I use a linear power supply and for transfer a tool battery. Because the battery can have up to 21V, I have built a small adapter with 78L15. This battery (1.5Ah) lasts for 2 days but there are bigger ones.

[BU508A]

For my LTFLU-1 prototype I have made something similiar with Li-Ion batteries from Makita:



The advantage is: you can get a cheap chinese version of these sockets on ebay,
rip out the electronics and put in some nice LDO.

Example:
https://www.ebay.com/itm/283470670352

BU508A

[alex-sh]

what battery are you using please?

For aging, I take a simple plug-in power supply. For measurements I use a linear power supply and for transfer a tool battery. Because the battery can have up to 21V, I have built a small adapter with 78L15. This battery (1.5Ah) lasts for 2 days but there are bigger ones.

I have a 15V 4A PSU for SkyRC MC3000 charging device, but I think p-p is more like millivolts.
The battery you have on the photo is from a power tool? 

[Hermann W]

The battery you have on the photo is from a power tool?

Yes, a cheap one from the hardware store. They have nominal 18V. That means 5 LiIon with a maximum of 21V after charging. The voltage regulator with the matching capacitors filters a lot.

[3roomlab]

I know 6 1/2 DMM should not be used to track LTZ1000, but given that I do not have any better one yet, what are the best settings used for monitoring LTZ1000 (or LM399)? Currently, I have it set-up as follows:

1) Filter on (Moving average, count - 100, window 2%)
2) NPLC: 12 (maximum) 

What do you think?


Thanks
Alex

I am curious
when should 1 use moving average and when to use repeating average in this case? I am biased to use repeating because the result/data can be used for other math in spreadsheet.
If I collect the data in moving average, the plot is already displayed/calculated as a trend and I cannot use it for Stddev calculation or alan-variance, am I right?

and since the time scale of LTZ1000 shifting ppm is in hours days weeks and months, and the limit (is that the limit?) of the machine is 12NPLC, shouldnt the data be collected as repeating in order to process a longer trend period in spreadsheet? eg : 1000 nplc math?

(sanity check : I have not been sampling much and may have forgotten some things)

[martinr33]

If you turn off the filter and use the fewest NPLC that gives you 6.5 digits, you can do all the processing in your spreadsheet.

Conversely, at these settings the display is pretty useless.

[alex-sh]

Another question - did anyone use LT3045-A Ultralow-noise (<0.8µVrms) LDO to feed 12V to LTZ1000?
Is there any disadvantages?

[niner_007]

Another question - did anyone use LT3045-A Ultralow-noise (<0.8µVrms) LDO to feed 12V to LTZ1000?
Is there any disadvantages?

I believe TiN did in his FX reference, he used the lower current version LT3042. LT3042 and LT3045 require very careful layout design, not doing so would basically degrade or invalidate its performance.

[pansku]

Just finished building 2 units with the Dr. Frank PCB 

I'm only on page 20 on this subject, but I got the impression that these should be powered on for about a month continously to burn in? As I'm limited to just a Keithley 2015 before we get some 3458A's to the lab I work in, is the correct way to do this by measuring the difference between 2 units on mV range and when they don't drift against each other that much, then I'll start trimming to desired output? After that time we should have the HPAK meters, put the boxes in a weather cabinet, run from lets say from 15 to 35 degrees °C and trim the 400k resistor? And lastly finish the project by having them calibrated against 732B(?) at the company cal lab.

Does that sound like a solid plan? I'm currently using 13k/1k and non-A LTZ1000. These references are going to end up in homelab with high'ish oven set point as the tempereture can reach around 30°C on the hottest summerdays inside.

Obligatory picture 
Don't mind the missing power jack for now. I'll grab it later with the Hammond boxes when I'll order some more components to other projects.

[MiDi]

As I'm limited to just a Keithley 2015 before we get some 3458A's to the lab I work in, is the correct way to do this by measuring the difference between 2 units on mV range and when they don't drift against each other that much, then I'll start trimming to desired output? After that time we should have the HPAK meters, put the boxes in a weather cabinet, run from lets say from 15 to 35 degrees °C and trim the 400k resistor? And lastly finish the project by having them calibrated against 732B(?) at the company cal lab.

Drift should be measured against trusted reference.
Two drifting references could have nearly same drifting rate...
If the TC is unknown/untrimmed this could appear as drift, unless you measure them at constant temperature against each other.
A month of burn in should decrease the drift to acceptable rate.
It depends how accurate you want to trim the output to 10V...
Finally you have to determine value with precision reference - 732B should be good enough 

[Andreas]

Hello,

additional to the points of MiDi:
Trimming the T.C. will affect output voltage. -> you will have to do this first.

Fortunately you can do T.C. measuring of the both references against each other (in 100mV Range of your DMM) as soon as the first large drift has gone.

not all references are stable after 1 kHr.
my LTZ#8 still drifts with  ~0.5 ppm/kHr against my stable LTZ#2 after 2 kHrs.

with best regards

Andreas

[pansku]

Drift should be measured against trusted reference.
Two drifting references could have nearly same drifting rate...
If the TC is unknown/untrimmed this could appear as drift, unless you measure them at constant temperature against each other.
A month of burn in should decrease the drift to acceptable rate.
It depends how accurate you want to trim the output to 10V...
Finally you have to determine value with precision reference - 732B should be good enough 

Hello,

additional to the points of MiDi:
Trimming the T.C. will affect output voltage. -> you will have to do this first.

Fortunately you can do T.C. measuring of the both references against each other (in 100mV Range of your DMM) as soon as the first large drift has gone.

not all references are stable after 1 kHr.
my LTZ#8 still drifts with  ~0.5 ppm/kHr against my stable LTZ#2 after 2 kHrs.

with best regards

Andreas

Thanks for the info. Too bad I only have access to LM399 based instruments atm as the cal lab is too busy to lend me a 8508A and a transfer standard  But hey, can't complain if they can provide me with a traceable calibration for free.

Seems like I have a lot more research to do while waiting for them to stabilize. This project and the tester build I get to participate at work (     ) will surely provide a valuable lesson in precision electronics measurements.

[TiN]

pansku
You are on the right track, other than using LTZ1000CH. I found ACH easier to work with, and more linear in TC (before you trim them to <0.05ppm/K).
Avoid using connectors out of the LTZ boards to reduce risk of thermal EMF issues But what do I know...

[Dr. Frank]

Just finished building 2 units with the Dr. Frank PCB 

I'm only on page 20 on this subject, but I got the impression that these should be powered on for about a month continously to burn in? As I'm limited to just a Keithley 2015 before we get some 3458A's to the lab I work in, is the correct way to do this by measuring the difference between 2 units on mV range and when they don't drift against each other that much, then I'll start trimming to desired output? After that time we should have the HPAK meters, put the boxes in a weather cabinet, run from lets say from 15 to 35 degrees °C and trim the 400k resistor? And lastly finish the project by having them calibrated against 732B(?) at the company cal lab.

Does that sound like a solid plan? I'm currently using 13k/1k and non-A LTZ1000. These references are going to end up in homelab with high'ish oven set point as the tempereture can reach around 30°C on the hottest summerdays inside.

Obligatory picture 
Don't mind the missing power jack for now. I'll grab it later with the Hammond boxes when I'll order some more components to other projects.

Glad to see another new born pair of references.. welcome in the club.

These non -A references are quite fine, but you've set their oven temperature to probably 60..65°C, instead of using 12k/1k for 50..55°C.
Therefore you can expect a typical drift rate of about -2ppm/year, after the initial higher rate drift phase of 3..6 months.
Note: A 'burn-In' would mean that you'd run the LTZ1000 or LTZ1000A at elevated room temperatures, say over 120°C to let interior epoxy / silicon crystal anneal or create sped-up drift of exterior components.
(This process might be necessary on the A version.)

That's not necessary at all, because the LTZ1000 simply will show an initial drift of about -1..-3ppm/ 3 months, and then the drift will change to the typical rates mentioned above.

So you can directly build the completed tuner box , thermal isolation and Jacks and start trimming the overall T.C., as the references voltages will change during that process.

Your 6 1/2 digit meter is quite fine for tracking the initial drift, if you don't let it run continuously.. any reference which is not powered, will probably drift at nearly zero rate, especially when it's already aged.
So you will compare in the end 3 different references with 3 different drift rates.. in the beginning you LTZs will also differ, but might show the same rate later.

Later on, your new LTZ references will take over as fix points, hopefully you can transfer absolute readings into your lab by comparison.

The T.C. trimming process by bootstrapping is a bit more difficult using a LM399 based instrument, so you need to take care for constant environmental temperature.. Have you got basements in houses in Finland?

Frank

[pansku]

Glad to see another new born pair of references.. welcome in the club.

These non -A references are quite fine, but you've set their oven temperature to probably 60..65°C, instead of using 12k/1k for 50..55°C.
Therefore you can expect a typical drift rate of about -2ppm/year, after the initial higher rate drift phase of 3..6 months.
Note: A 'burn-In' would mean that you'd run the LTZ1000 or LTZ1000A at elevated room temperatures, say over 120°C to let interior epoxy / silicon crystal anneal or create sped-up drift of exterior components.
(This process might be necessary on the A version.)

That's not necessary at all, because the LTZ1000 simply will show an initial drift of about -1..-3ppm/ 3 months, and then the drift will change to the typical rates mentioned above.

So you can directly build the completed tuner box , thermal isolation and Jacks and start trimming the overall T.C., as the references voltages will change during that process.

Your 6 1/2 digit meter is quite fine for tracking the initial drift, if you don't let it run continuously.. any reference which is not powered, will probably drift at nearly zero rate, especially when it's already aged.
So you will compare in the end 3 different references with 3 different drift rates.. in the beginning you LTZs will also differ, but might show the same rate later.

Later on, your new LTZ references will take over as fix points, hopefully you can transfer absolute readings into your lab by comparison.

The T.C. trimming process by bootstrapping is a bit more difficult using a LM399 based instrument, so you need to take care for constant environmental temperature.. Have you got basements in houses in Finland?

Frank

Hi Frank,

Unfortunately the ambient conditions are really poor at the moment. I live in a student apartment with no basement available, even worse it is high up in southwest-side so one of the hottest places in the old building during summer. Typically outdoor tempereture +2...3°C unless I'm running a portable ac which in turn can create a worst case temperure delta of almost 10°C between day and night 

Fortunately I work part time in a power electronics RnD lab so I do have access to stuff like weather cabinets and, in a bit over a month, a pair of 3458A's  I will propably leave these boxes running somewhere in that building to age because all the linear power supplies I have at my tiny 25m² apartment make a lot of audible noise.

Technically speaking I could use the cabinets (-40° to +130°C with humidity adjustment) for burn in (not the aging I spoke of earlier with the wrong term). The lab at work is at a pretty constant tempereture. Those could maybe be used for TC trimming with the Keithley?

I still have one more PCB left and I'll try sneak in a order from VPG at work. That would allow me to build one more unit with 12k/1k and using possibly VHP101 resistors for the booster (and maybe even for rest of the critical parts) at almost no cost to me. I think I'll have to gather some experience with these first.

[e61_phil]

These non -A references are quite fine, but you've set their oven temperature to probably 60..65°C, instead of using 12k/1k for 50..55°C.

Isn't it the other way around? The A is better isolated and will heat up itself a little bit more. Therefore, you will need a higher heater temperature on the A type.


@pansku: I have also problems with my room temperature. I bought a small thermal chamber and a camping cooling box. At the moment I'm only using the thermal chamber to have stable 23°C and put the DUT and the reference into it. The meter isn't that crucial because it only compares the two references.
My plan is to control the camping box to fixed 23°C and use the thermal chamber for different temperatures to measure TCs.

[niner_007]

These non -A references are quite fine, but you've set their oven temperature to probably 60..65°C, instead of using 12k/1k for 50..55°C.

Isn't it the other way around? The A is better isolated and will heat up itself a little bit more. Therefore, you will need a higher heater temperature on the A type.

How can one calculate the ratio to temperature yield for the LTZ1000? Is there any documentation on that? I remember on the voltnut board someone said they posted a chart here on the eevblog forum but I could not find it

[niner_007]

Has anyone tried to place a big thick copper plate, on the bottom of an LTZ1000 PCB? With thermal non conductive insulation. The goal would be to equalize and eliminate the temperature gradients, only thing I don't know if the thermal insulator would cause any noise/leakage on its own.

The better option being a ceramic PCB of course, like in the modern Fluke DMMs. The best option would be a multilayer ceramic PCB.

[Dr. Frank]

If you have a single sided PCB, that'll be a good idea thermally-wise, but that may cause short circuit.
Another technique would be to have a thermal isolation all around the bottom layer, e.g. by styrofoam... see my mechanical / thermal design around page 67, or so.

Best solution is to have a broad circular copper ring around the LTZ1000 on the top layer, when all its electrical connections are on the bottom layer.
See John Pickerings design inside the Wavetek / Fluke 7000 reference.

Btw. alll these other slots are nonsense.

Frank

[MegaVolt]

Tell me a mistake in my reasoning
A typical circuit stabilizes the current through the zener diode.
If we, knowing this current, add a precise resistor in series with the Zener diode, then we can get 10 V out by applying only 1 exact resistor.
Where am I wrong?

[niner_007]

Best solution is to have a broad circular copper ring around the LTZ1000 on the top layer, when all its electrical connections are on the bottom layer.

Will do thanks!

Actually, what would be interesting is machining a small copper ring, with a thickness of 4 mm and diameter just around the LTZ, and soldering it, top and bottom, while routing the connections on the inner layers, might be overkill, PCB can be 4 - 6 layers

[exe]

If we, knowing this current, add a precise resistor in series with the Zener diode, then we can get 10 V out by applying only 1 exact resistor.
Where am I wrong?

Where/how to get such resistor? . Making a stable precision resistor of a non-standard value seems to be quite a challenge.

[MegaVolt]

Where/how to get such resistor? . Making a stable precision resistor of a non-standard value seems to be quite a challenge.

Creating a divider requires two exact resistors. In my opinion one exact resistor is cheaper than two.