Ultra Precision Reference LTZ1000

[GregDunn]

Hmmm... I hope switching on/off a DMM connected to the reference output isn't a potential issue because I've done that a couple of times.  Shouldn't be...

I've left it open circuit for now, and checked the output a few times in the last couple of days, but haven't seen it happen again.  The output seems in spec at the moment.

[ArthurDent]

I'm using a different opamp but I also connected a transistor to buffer the opamp output and protect it in case I did something stupid. Where the transistor is inside the feedback loop it has no effect on the output accuracy. Only needed to add 3 components.

[Andreas]

It usually takes several ma to induce SCR latchup.

But this can happen if you use some types of switchmode power supplies with capacitors between input mains line and output.
(not fully galvanic isolated).

You should mention latchup as usually the power supply current increases and the OP-Amp will get hot in this case.

with best regards

Andreas

[Andreas]

Hello,

I built 2 LTZ1000 (non A) devices. (12K heater setpoint)
So now we can do a direct comparison of startup-times

Here a cold-start of my LTZ#5

blue: 14V power supply after regulator (switching on is done via "on"-Pin of the LT1763)
yellow: zener voltage (unbuffered) but the buffered output follows within ms (2nd picture)
green: heater voltage (stable after ~400 ms)
red: zener current (voltage on 120 ohms resistor).

When I zoom in for the zener voltage the output is stable (within some mV) after 200 ms.

First picture Here on LTZ1000 #7 with LT1013A (CERDIP) as OP-Amp

blue: 14V power supply after regulator (switching on is done via "on"-Pin of the LT1763)
red: zener current (voltage on 120 ohms resistor).
green: Voltage on Gate of FET BF245C (output current regulator)
yellow: heater voltage (stable after 90 seconds!)


2nd picture:

blue: 14V power supply after regulator (switching on is done via "on"-Pin of the LT1763)
red: zener current (voltage on 120 ohms resistor).
green: Voltage on Gate of FET BF245C (output current regulator)
yellow: unbuffered zener voltage (not much change after 30 ms)

with best regards

Andreas

[Dr. Frank]

Hello,

I built 2 LTZ1000 (non A) devices. (12K heater setpoint)
So now we can do a direct comparison of startup-times




First picture Here on LTZ1000 #7 with LT1013A (CERDIP) as OP-Amp

..
red: zener current (voltage on 120 ohms resistor).
...
with best regards

Andreas

Hello Andreas,

could you please determine more precisely (in mV) the voltage difference across the 120 Ohm resistor, between start @ 10.22msec, and later, when oven has stabilized, @ 30..90sec? What was your room temperature?

From these numbers it's possible to determine the real oven temperature quite precisely.

Thanks!
Frank

[Andreas]

Hello Frank,

good Idea.

but since this measurement was only made to have a look at all regulator cirquits (for oscillations) I did neither measure the PCB temperature (which might be different from room temperature) nor pay attention to use shielded probes for all channels.
But I will repeat the UR1 measurement with a better setup the next days.

Today I am measuring T.C. (with long legs) of the LTZ#7.

with best regards

Andreas

[Andreas]

Hello,

today I repeated the power-on measurement of UR1 on the new LTZ1000 devices:

LTZ7:
start with 24.7 deg C PCB-temperature (NTC populated on PCB) and about 24 deg C room temperature.
I increased the resolution by adding a 400 mV offset so I could use the +/-500mV range and 16 Bits ADC-resolution.

result
start with 523 mV end with 468mV = 55.2 mV difference or 27.6 deg C above 24.7 deg c = 52.3 deg C operating temperature if I calculated it right.

T.C. measurement shows linear T.C. -57 ppb/K up to 40 deg C environment temperature.

LTZ8:
start with 24.6 deg C PCB-temperature

result
start with 533 mV end with 478 mV = 54.5 mV difference or 27.2 deg C above 24.6 deg C = 51.8 deg C operating temperature.

But when I look at the T.C. measurement of LTZ8 it seems that the heater regulation gets instable above 33 deg C environment temperature.
Linear T.C. up to 32 deg C is around -26 ppb/K.

with best regards

Andreas

[Andreas]

Hello,

just some measurement values in comparison:

LTZ#7 is populated with a LT1013A in CERDIP for the basis cirquit
LTZ#8 is populated with 2*LTC2057
Both LTZ1000 (non A) with datecode 1443
temperature setpoint = 12K
the output buffer is LTC2057.

on #8 I will try to increase the temperature setpoint to see if the non-linearity can be shifted above 40 deg C
33 deg C is too marginal for my "lab" as this temperature is reached in summer.

with best regards

Andreas

[Dr. Frank]

Hello,

today I repeated the power-on measurement of UR1 on the new LTZ1000 devices:

LTZ7:
start with 24.7 deg C PCB-temperature (NTC populated on PCB) and about 24 deg C room temperature.
I increased the resolution by adding a 400 mV offset so I could use the +/-500mV range and 16 Bits ADC-resolution.

result
start with 523 mV end with 468mV = 55.2 mV difference or 27.6 deg C above 24.7 deg c = 52.3 deg C operating temperature if I calculated it right.

T.C. measurement shows linear T.C. -57 ppb/K up to 40 deg C environment temperature.

LTZ8:
start with 24.6 deg C PCB-temperature

result
start with 533 mV end with 478 mV = 54.5 mV difference or 27.2 deg C above 24.6 deg C = 51.8 deg C operating temperature.

But when I look at the T.C. measurement of LTZ8 it seems that the heater regulation gets instable above 33 deg C environment temperature.
Linear T.C. up to 32 deg C is around -26 ppb/K.

with best regards

Andreas

Hello Andreas,
thank you for these measurements.
I also found 50..52°C on all of my 7 LTZ1000 / 12k/1k references.
I also found similar T.C. figures like you.
My interpretation concerning the 'instability' above 33°C R.T. is, that the visible increase in your last diagram is not necessarily a sign that the oven regulation gets unstable, that is to my opinion an increasing T.C. of higher order, which depends on the degree of compensation. for the -52ppb/K, that anomaly is not visible.

52°C is high enough to have about 38°C R.T. before the margin for the oven control is too small.. All of my newer LTZs can be operated at  up to 40°C in that sense, but their individual T.C. compensations, and the resulting reference voltage deviations at higher temperature, limit the usable temperature between 35°C to 40°C.
(At the May meeting, we should have a look on the variety of graphs of my T.C. measurements, difficult to describe here)

That also depends on the thermal packaging. My electronics package heats up 7.5°C above R.T., and I have no additional isolation around the LTZ1000. That means, 40°C R.T.  gives 47.5°C inner temperature, and that is barely 5°C regulation margin, so that's the tightest upper limit.

In the end, 12k/1k ~ 52°C oven, is good and safe for each sample, intended for the nominal cal lab of 18..28°C.
(Aren't the 732s specified in a similar manner?)*
Above that, the behavior of each compensated circuit has to be determined.

Frank

*Yes, the 732B is specified for T.C. <= 0.04ppm/K in 15..35°C. Oven temperature is about 45°C

[Andreas]

Hello,

I have additional thermal isolation around the LTZ (to reduce air current induced noise) and need more than the usual 18 to 28 deg C range.

So I tried increasing the setpoint by adding a PTF56 resistor with 25K2 in parallel to the 1K setpoint resistor.

the ratio changed from 13:1 to 13.476:1 and starting from 7.10881 V output voltage base voltage decreased by 19.1 mV or +9.5 deg C setpoint change.
Output voltage increased by +2.99 mV so + 420 ppm or 44.1 ppm/K  (unheated) raw T.C. of the LTZ.

The non-linearity above 33 deg C disappeared (was shifted above 42 deg C).
Average T.C. -30 ppb/K

q.e.d.

Andreas

[dkozel]

It's taken longer than I'd hoped to get to this point, but just about everything is soldered up. Trying to figure out some unit tests before adding the LTZ, but hoping to add those later today.

[branadic]

Thanks, but there must be much more references out there as far as I remember.

Found this, looks like they put together what is published in this thread (plastic caps, design ideas, ...)

https://islandlabs.eu/ltz1000.html

-branadic-

[jlmoon]

Aren't these the A Version? 

[maginnovision]

Thanks, but there must be much more references out there as far as I remember.

Found this, looks like they put together what is published in this thread (plastic caps, design ideas, ...)

https://islandlabs.eu/ltz1000.html

-branadic-

I have two of those. One A one non A. The non A aged > 10000 hours the A now about 3-4k. 213 is A, 136 is not. Temp was  23C +-10C over the time.

[Andreas]

Hello,

bad surprise (at least for me).

while shortening the legs on 2 of my LTZ1000A samples increased the T.C. this time with LTZ1000 (non-A) the T.C. decreased.

On LTZ#7 I had with long legs -57 ppb/K. I shortened the legs by ~7mm and wanted to get into the +/-20ppb/K range.
Instead I got -138 ppb/K making T.C. even worse as before.

Similar picture on LTZ#8: long legs +25K2 in parallel gave -30ppb/K (so near target). So I shorted only by ~5 mm.
Instead of zero I got -45 ppb/K.

Anyone with similar experiences for LTZ1000/LTZ1000A?

with best regards

Andreas

[maginnovision]

I just finished my boxes. The refs are isolated in RF boxes with some capacitors on Vin then those boxes are soldered shut(two spots to make sure shielding is continous. Those boxes are put into some Hammond boxes with a wire soldered to the RF box for earth connection and closed up. My reading offsets dropped by 50%. Was -40uV from nominal, down to -20uV.

[RandallMcRee]

I just finished my boxes. The refs are isolated in RF boxes with some capacitors on Vin then those boxes are soldered shut(two spots to make sure shielding is continous. Those boxes are put into some Hammond boxes with a wire soldered to the RF box for earth connection and closed up. My reading offsets dropped by 50%. Was -40uV from nominal, down to -20uV.

Sealed? No ventilation? I've read in this very thread (somewhere, can'f find it now--sorry!) that no ventilation is bad because the LTZ1000 will lose regulation at some point...

[maginnovision]

I just finished my boxes. The refs are isolated in RF boxes with some capacitors on Vin then those boxes are soldered shut(two spots to make sure shielding is continous. Those boxes are put into some Hammond boxes with a wire soldered to the RF box for earth connection and closed up. My reading offsets dropped by 50%. Was -40uV from nominal, down to -20uV.

Sealed? No ventilation? I've read in this very thread (somewhere, can'f find it now--sorry!) that no ventilation is bad because the LTZ1000 will lose regulation at some point...

No, they're not hermetically sealed, just shielded from noise.

[3roomlab]

Sealed? No ventilation? I've read in this very thread (somewhere, can'f find it now--sorry!) that no ventilation is bad because the LTZ1000 will lose regulation at some point...

you must be talking about misterdiodes post.

it was actually a discussion about LTZ power disspation as heat, and restricting it with thermal isolation. as heat can get into LTZ, the idea was to have high thermal isolation. to which misterdiodes commented, if the isolation is too high, the LTZ heater regulation will go nuts.

if you simulate on femm, it is so easy for thermal conduction to over ride all the insulation done. from FR4 or some plastic box to air, the insulation advantage could be around 10x (in W/mk terms). but from FR4 to copper, the conductive advantage is more than 100x 30x. so for most designs, the copper trace can easily out conduct heat around isolation slots. the isolation gain becomes more apparent when the "moat" is large enough. but all this is only important if the environment temperature fluctuates. so if everyone puts their LTZ in Franks basement, I think the results will be very nice and striaght. but for everyone else, they need to buy a wine chiller or something.
which is very nicely done by Mr Jorn, which I always always love to repost as example.
https://www.eevblog.com/forum/metrology/dmm-adc-noise-comparison-testing-project/450/

in the CERN document about ADC 7717 (2017 pdf), they spoke of regulating the temperature of the entire LTZ circuit including the 7717 ADC so that this fluctuation is stopped. but I think they did not go with it because they said it in the document, it consume too much power. they are not nuts enough, otherwise, what is a few watts. <-- I think I misread this one, wanted to re read the pdf, cant find it. strange it stuck in my mind.

[Magnificent Bastard]

"[CERN] ... are not nuts enough, otherwise, what is a few watts."

[MiDi]

I've been using my reference to check calibration of my DMMs (particularly newly repaired ones) and as such I have had to move the output cables around to connect to different units.

Something I've seen happen: when removing and re-connecting cables, sometimes the reference output will drop to about 2.9VDC and stick there.  Cycling power to the reference brings it back immediately.  I've been careful to ground myself before touching the reference's case and not touch the bare leads but it happened anyway.

Is this a characteristic of the LTC1052 going into protection mode?  The datasheet only talks about overload recovery and doesn't address any sort of latch-up considerations...

Today I had same issue on my #2, but it happend without changing anything.
There were 2 DMMs connected to 10V output for couple of hours when the output got stuck at a couple of volts.
After power cycling everything seems to be fine again.
 

[branadic]

I have my 9x LTZ1000 references running for almost 1000h and haven't observed something similar by now.

-branadic-

[GregDunn]

I've left mine running continuously since it arrived (though without a load) and it's only happened a couple of times.  I just went downstairs to check the output a few minutes ago and it's rock solid at 10.0000V.  The only thing I can think of is that hooking it to a live circuit instead of a DMM may have transferred a "floating" voltage to the output and upset the buffer - but that's mere speculation.

Maybe I should leave one of my 8800As powered up and monitoring it for a while, just for the sake of completeness.

[pelule]

description sounds like a soft latcht-up. Is there any pin which may high (even short time) a pin higher/lower voltage than a supply voltage.
/pelule

[GregDunn]

The supply on mine is 15V - the only way I see that a higher voltage could be applied to any part of the circuit would be a stray voltage applied to the output terminals.  I'm not saying it doesn't happen, but it would need to be an issue with an AC-powered piece of gear, either improperly grounded or having a static charge accumulating at the input...  I pretty much only connect it to my Fluke, HP and Keithley bench DMMs.