Ultra Precision Reference LTZ1000

[doktor pyta]

For voltage measurement i use Klotz MY206 shielded  cable + 'Laboratory plug ø 4 mm CAT I, LS 4, Multi-Contact' from elfa distrelec.
For resistance I use PTFE insulated 4 wires in shield + LS 4 plugs (unfortunately there is no marking on the cable).

For the RFI testing I used 30cm copper (unshielded) twisted pair.

[plesa]

For voltage measurement i use Klotz MY206 shielded  cable + 'Laboratory plug ø 4 mm CAT I, LS 4, Multi-Contact' from elfa distrelec.
For resistance I use PTFE insulated 4 wires in shield + LS 4 plugs (unfortunately there is no marking on the cable).

For the RFI testing I used 30cm copper (unshielded) twisted pair.

The shielded cable on picture looks like Belden RG142 ( double shielded, silver plated), similar cable is also made by Habia.
Thanks for tip for Klotz cable, it can be good for its flexibility. Do you know its insulation resistance?

[doktor pyta]

What You see on the picture is 3mm 75ohm coax with inner and outer teflon insulation and it is silver plated copper. I have about 50m of that from some surplus auction and sometimes I use it. It is used as 1 wire (shield and wire shorted).

Let me use my good old HP4329A
Real measurements of insulation resistance of Klotz MY206 cable (test voltage: 100V, measured resistance is between shield and both inner conductors shorted, length: 1.5m, Temp=25.6'C ) R=1.5E+14 ohm. (that's nice result)

[plesa]

Let me use my good old HP4329A
Real measurements of insulation resistance of Klotz MY206 cable (test voltage: 100V, measured resistance is between shield and both inner conductors shorted, length: 1.5m, Temp=25.6'C ) R=1.5E+14 ohm. (that's nice result)

Thanks, that numbers forced me to look where I can buy it How satisfied you are with 4329A meter, can we expect teardown?

[doktor pyta]

User pa4tim did teardown on his website: http://www.pa4tim.nl/?p=2269
Very interesting description is in the HP Journal 1971 , Volume , Issue March-1971.

[VintageNut]

Hello Plesa

You can take this measurement with a KE2450. You have to use the rear triax cables. The current measure is the HI cable. There is an adapter with a part number 7078-TRX-BNC. This adapter connects the triax GUARD to BNC shield.

To take the measurement, you have to create an open circuit and press the REL. This measures the current offset and makes that current value a new zero.

After the REL, you force voltage and measure current.  For the above example of 1.5 x 10^14 ohms at 100V, the current is 0.6667 pA. That is well within the capabilities of the 2450 1nA range with 6 1/2 digits.

Be aware that the technique of making BNC shield at GUARD potential creates a shock hazard if you use high voltage to test insulation. For this reason I choose between 1v and 10V for the test voltage. The 2450 has no problem measuring 0.67pA at 1V or 6.7 pA at 10V.

I have performed this measurement for an ordinary solid core hookup wire and and the resistance is on the order of 10^13. I use an enclosure and make this enclosure at GUARD potential. This mitigates any current flowing from HI to LO in the air inside the enclosure.

[2N3055]

Keithley as app note Number 3262 about that...

[VintageNut]

Keithley as app note Number 3262 about that...

Unfortunately, that app note is missing the vital step to characterize and remove the current source offset and the current measure offset.

for the 10nA range, the 2450 has a source uncertainty of 100pA and a measure uncertainty of 50 pA. If you are measuring 0.67pA and there is a significant offset, the measurement is invalid.

To make an accurate measurement of this type, it is necessary to follow the steps of the app note for an open circuit condition and allow the measurement to stabilize. Note the current for later subtraction during the actual measurement or use the REL function.

Then perform the measurement as described in the app note with the added step of removing the offset that was previously measured.

I have successfully used a 2450 for this measurement an it is vital to remove the offset. The good news is that the offset is stable enough to be removed. The uncertainty of the measurement is much much lower than the worst case datasheet numbers.

[2N3055]

Keithley as app note Number 3262 about that...

Unfortunately, that app note is missing the vital step to characterize and remove the current source offset and the current measure offset.

for the 10nA range, the 2450 has a source uncertainty of 100pA and a measure uncertainty of 50 pA. If you are measuring 0.67pA and there is a significant offset, the measurement is invalid.

To make an accurate measurement of this type, it is necessary to follow the steps of the app note for an open circuit condition and allow the measurement to stabilize. Note the current for later subtraction during the actual measurement or use the REL function.

Then perform the measurement as described in the app note with the added step of removing the offset that was previously measured.

I have successfully used a 2450 for this measurement an it is vital to remove the offset. The good news is that the offset is stable enough to be removed. The uncertainty of the measurement is much much lower than the worst case datasheet numbers.

Sorry for not being clear... Your steps are absolutely gold.... I merely pointed out that even Keithley is marketing SMU for this kind of usage.. And as with any  manufacturers app note it's a starting point... they tend to be a bit optimistic...  At this level of measurements, measuring practice and discipline is all... Currents are miniscule, errors creep in...
And we are lucky that people like you are willing to share this kind of "ninja" knowledge with us...
Thanks!!

Take care!!

[VintageNut]

Keithley as app note Number 3262 about that...

Unfortunately, that app note is missing the vital step to characterize and remove the current source offset and the current measure offset.

for the 10nA range, the 2450 has a source uncertainty of 100pA and a measure uncertainty of 50 pA. If you are measuring 0.67pA and there is a significant offset, the measurement is invalid.

To make an accurate measurement of this type, it is necessary to follow the steps of the app note for an open circuit condition and allow the measurement to stabilize. Note the current for later subtraction during the actual measurement or use the REL function.

Then perform the measurement as described in the app note with the added step of removing the offset that was previously measured.

I have successfully used a 2450 for this measurement an it is vital to remove the offset. The good news is that the offset is stable enough to be removed. The uncertainty of the measurement is much much lower than the worst case datasheet numbers.

Sorry for not being clear... Your steps are absolutely gold.... I merely pointed out that even Keithley is marketing SMU for this kind of usage.. And as with any  manufacturers app note it's a starting point... they tend to be a bit optimistic...  At this level of measurements, measuring practice and discipline is all... Currents are miniscule, errors creep in...
And we are lucky that people like you are willing to share this kind of "ninja" knowledge with us...
Thanks!!

Take care!!

You are welcome. I am happy to share what I know. On balance, I am learning more here than I contribute.

The now obsolete Keithley 23X sourcemeters are quite capable of taking this measurement and cost much less on eBay than a 2450. I have performed this measurement with my KE236. The offset was very nearly zero for my KE236 on the 1nA range. The last digit represents 10fA.

[TiN]

Time to start thread "Measuring nanoamps like a ninja"! Jokes aside,  Perhaps worth to try K2400 (1uA min range) and 3458 too? 2400 can drive guard too, but no triax.

[Andreas]

What You see on the picture is 3mm 75ohm coax with inner and outer teflon insulation and it is silver plated copper.

Hello,

Thanks, that´s what I wanted to know: the transparent interesting looking cables.
And the spade lugs.

with best regards

Andreas

[Andreas]

Hello Andreas - can you send a pointer to the post from Lars you are referring to below ?

And also the trimmable 7V output is a good idea (and a basis for a 10V output with a LT5400 with the idea from Lars).
Andreas

Thanks
Flinstone

Hello,

I don´t know where I have seen the circuit from Lars.
It was a LT5400 with 2*100K in parallel and 2*10K in series.
So actually a 7:5 divider.

This together with a TL431 based heater could give a good basis for a 7->10V transfer.
See bottom of page:
http://www.techlib.com/electronics/ovenckts.htm

with best regards

Andreas

[VintageNut]

Time to start thread "Measuring nanoamps like a ninja"! Jokes aside,  Perhaps worth to try K2400 (1uA min range) and 3458 too? 2400 can drive guard too, but no triax.

I will start that thread to stop pollution of this thread.

Your question about 2400 and 3458A is a good question. My opinion is that if the instrument is using banana jacks on the front/rear connections(s) you will never achieve the good results possible with the instruments that have triax jacks.

I used a 2410 recently to try to force 1nA and measure leakage of a device. The results were very slooooooowwww. At 10 nA the results were much better.

A KE237 taking the same leakage measurement at 1nA was much faster than the 2410 and better results. 

[TiN]

I also has Agi 4142B with 41421B SMU, which begs to be used. It's lowest range is 1nA with triax and remote sense all the way with 20fA resolution

[branadic]

Hello Andreas - can you send a pointer to the post from Lars you are referring to below ?

And also the trimmable 7V output is a good idea (and a basis for a 10V output with a LT5400 with the idea from Lars).
Andreas

Thanks
Flinstone

Hello,

I don´t know where I have seen the circuit from Lars.
It was a LT5400 with 2*100K in parallel and 2*10K in series.
So actually a 7:5 divider.

This together with a TL431 based heater could give a good basis for a 7->10V transfer.
See bottom of page:
http://www.techlib.com/electronics/ovenckts.htm

with best regards

Andreas

Thanks, that's a pretty good hint. I guess you will have this solution populated on your next pcb revision?

[Andreas]

Thanks, that's a pretty good hint. I guess you will have this solution populated on your next pcb revision?

Hello,

I am not certain: from the thermal design (isolation),
it might be better to have either a sub-board or a external cirquit
(at the expansion D-Sub).

with best regards

Andreas

[lars]

About my LT5400 based amplifier that Andreas mentioned:
Well I have probably described it somewhere but don´t remember when.

My LTZ1000A based board is just a simple prototype board with the usual LTZ design from the data sheet using 13k:1k, 120 and 70k. To make it even worse I use  a socket for the LTZ, the resistors are metal film from Farnell and the junk box. My plan was to make two more boards with resistors from Ultrohm, but in that case I was one of the two unlucky receivers of humidity sensitive WW.

The first LTZ measured about 7.16V and the second 7.13V (the third I haven´t tested yet). As 7.16 and 7.13 was close to 5/7 of 10V I saw a chance to test an idea I had to use the LT5400. So I added first a divider from about 7.16 to 7.14 buffered with an op and added another op with the LT5400 with 10k x2 + 100k x2. See attached schematic. As op-amp I used an OPA2188 as I had some in my junkbox. As I use a prototype board I have connected the LT5400 with just thin wires (AWG30).

The prototype board is mounted in a Hammond alu-box and that box is mounted in a Prema alu-box from Reichelt (about 160x100x56mm if I remember correct). In the bottom of the Prema box I mounted a 40x40mm cheap peltier and on the peltier the Hammond box. So far I have only used the peltier to test the TC of the 10V out. From 20C to 40C it changed about -2ppm so about -0.1ppm/C.

The box have three Pomona 3770 binding posts 0V, 7.14286 and 10V. And yes the 7.14286 actually was that after I trimmed the 10V to 10.0000V. I only have 6 digit DMMs so that is a problem.

From July last year to September this year I have powered up the LTZ about 1 hour every 2-3 week. So it haven´t been on very much. I attach a graph showing the difference to one of my SVR-boards that are powered continuous. As the temperature during the tests have varied from about 18 to 27C I have made a slight first and second order temperature compensation of the data.
From september I have had the LTZ powered continuous. As can be seen the output have started to drift downwards. The first year the drift was about +0.7ppm but of course this can be the SVR drifting down. Now with the downward drift the question is if it is the LTZ or some other part drifting. I have an 10k NTC in the box and the interior is about 3C above room. As I have only a 6 digit DMM I just can say that the outputs are 9.99997V and 7.14284V so about the same ratio (from the LT5400). One reason the LT5400 is on the last op-amp is that it was easier to have the 10V and 7.14V out. Otherwise it isn´t good from a noise stand point. 

Another mistake I have done is to use a trimpotentiometer in series with the 250ohm input to the first opamps. The 250ohm and 90kohm are 8E16 WW from the junkbox.

Other notes. The OPA2188 can´t handle capacitive loads so a redesign of the output is on the list. The box is also susceptible to EMI. A simple test with a loop from a signal generator moved the output. If I some time will take me time I should make a program to run my signal generator together with a DMM all the way from 0.15-2000MHz. Just a quick check with 10dBm out and 80%AM and some frequencies from 10 to 1800MHz showed that both the LTZ, SVR-boards and the famous 2ppm from Ebay showed terrible sensitivities to the signals. 1800MHz (mobile) was far from worst. One source of error for me are when I do tests of OCXO´s at 10MHz especially if they have square wave signal out.

The LTZ-box is powered with 12V from a lab-supply and takes about 20mA at 23C room temperature                                       

Lars

[julian1]

Does anyone know how temperature stable the two 70k ohm collector pull-ups in the ltz1000 reference designs need to be?

The reference circuit without the heater stabilisation uses a single 30k, and the 3458A board chose two 75k from memory - so I wonder how critical the temp-stability of the current is across these transistors needs to be?

Edit - Another thought - will not the pn junction in the compensating bjt and the temp-sense bjt, be around 2mv/C - and mostly independent of current? On that basis the changes in current due to temperature affects on the pull-up resistance shouldn't really matter.

I was looking for 70k resistors from the Vishay S-series, but neither element14 nor mouser stock them. Then I wondered if these really needed to be tempco stable.

[2N3055]

Datasheet for ltz1000 has info.. For a100 ppm change in those two resistors, 0,2 and 0.3 ppm change in ref voltage... Influence is resistor tempco divided with 300-500.. So I guess not so critical...

[Dr. Frank]

Does anyone know how temperature stable the two 70k ohm collector pull-ups in the ltz1000 reference designs need to be?

The reference circuit without the heater stabilisation uses a single 30k, and the 3458A board chose two 75k from memory - so I wonder how critical the temp-stability of the current is across these transistors needs to be?

Edit - Another thought - will not the pn junction in the compensating bjt and the temp-sense bjt, be around 2mv/C - and mostly independent of current? On that basis the changes in current due to temperature affects on the pull-up resistance shouldn't really matter.

I was looking for 70k resistors from the Vishay S-series, but neither element14 nor mouser stock them. Then I wondered if these really needed to be tempco stable.

Don't waste your money on expensive Vishay MBF resistors, especially not for the 70k ones.
Several people measured their influence, consistently.

R2 influences by 1/250, R3 by < 1/1000, only. That's the reason, why HP in the 3458A used Thin Film.

R4 and R5 (12..15k / 1k) affect the most (+/- 1/75), but the overall T.C. can be trimmed to near zero by changing the compensation resistor.

Therefore, PWW resistors @ <= 5ppm/K are fully sufficient.

Also, do not waste any more time on re-thinking again the whole stuff, what we have already discussed many, many times.
This circuit really bears no miracle any more, i think.

Frank

[Echo88]

On another note: Has anyone on this forum actually build a working PWM-circuit which convert the 7.2V from a LTZ to 10V? Ive seen Diligent Minds simulating one suitable circuit on Page 24.
I dont want to talk about pros/cons regarding PWM vs. Resistor dividers, just results.
Also: Does the active filter design, which is used by Datron 4910/Fluke 5700A/Diligent Minds have an actual name?

[mimmus78]

R4 and R5 (12..15k / 1k) affect the most (+/- 1/75), but the overall T.C. can be trimmed to near zero by changing the compensation resistor.

Yes but if I remember correctly (this thread is too just big) Andreas did some experimenting on this and the result is that the compensation resistor has similar effect on Vout as the divider with the disadvantage that you cannot use anymore TC tracking resistor for nulling temperature effects on Vout.

The most cost/effective way is still to have two good stable resistors for the divider with similar TC and don't use any temperature compensation resistor for LTZ1000.

[Kleinstein]

How much the temperature set point influences the output voltage depends on the TC of the circuit without temperature stabilization (but still high temperature). This can be quite different between samples and also depends on other details, like the value for the "70K" resistor (R2) and the zener current. In principle there could be the option to tune the unregulated TC (via currents or possible series resistor) to get a small TC to start with, so that temperature regulation is not that critical any more. In this case the two temperature setting resistors could be much lower sensitivity  - up to the point that thin film ones could work. However the down side is, that one would need extra measurements and an individually matched resistor and thus would have more lead time and not every LTZ1000 might have a suitable operation point.

The 400 K compensation resistor in the DS is a different thing: it is like more compensating an possible effect of heater power. AFAIR the data from Andreas showed that this compensation is not working that good, as the residual TC is more linear and the "400 K" resistor gives nonlinear (e.g. square root like) adjustments.

[Andreas]

The 400 K compensation resistor in the DS is a different thing: it is like more compensating an possible effect of heater power. AFAIR the data from Andreas showed that this compensation is not working that good, as the residual TC is more linear and the "400 K" resistor gives nonlinear (e.g. square root like) adjustments.

Hello,

I also would only trim small deviations caused by LTZ1000 itself with R9.
The outer resistors are not on the same temperature as the heater and the temperature sensing transistor.
So a too large tempco will cause issues if you do not have a constant room temperature.
But as Frank already said PWW resistors with 3-5ppm/K are fine.

with best regards

Andreas