LTZ1000/LTZ1000A thermal resistance and temperature setting

[splin]

The thermal specs are 80K/W and 400K/W respectively; however I can't square that with the chart showing 'Die Temperature Rise vs Heater Power'. From that the LTZ1000A appears to be around 375K/W which is close enough but the LTZ1000 works out at 175K/W. Why the disparity?

As to temperature setting, the datasheet says:

'Production variations in emitter-base voltage will typically cause about ±10°C variation. Since the emitter-base voltage changes about 2mV/°C and is very predictable, other temperatures are easily set.'

But does this include the effect of the large spread between minimum and maximum current gain of the control transistor (80 min, 200 typical and 400 max)? Would it be worthwhile measuring the gain and selecting the collector resistor R3 accordingly? The datasheet seems to be rather vague in this area.

Has anyone measured the actual operating temperature of their LTZ1000(A)? I guess that might not be easy, especially with the LTZ1000A - would it be better to measure the collector current against temperature when heating the device externally?

I'm asking because for long term stability it is best to run the device at the lowest temperature that can be maintained at the maximum ambient so 10C of variation is quite significant. The LTZ1000 part appears to have the advantage here as well as being cheaper. Apart from the higher power consumption are there other disadvantages - such as bigger thermal problems due to higher lead temperatures?

Thanks, Splin

[MK]

tucked away in some of the other ltz1000 threads there are comments from bob deakin that the 1000 has worse temperature hysteresis than the 1000a.

[Andreas]

Hello,

in my opinion the transistor gain only affects overall heater control loop amplification. (Including Op-Amp).
And so the static deviation from setpoint together with the variations of open loop gain of the Op-Amp.
I would invest into the "A-Type" of the LT1013. (or another high gain amplifier).

Most probably the transistor is optimized (maximum gain) for a collector current of around 100uA.
Although I have measured that decreasing the resistor (to 50K) increases the unheated tempco of the zener section slightly.

Its more easy to measure the heater voltage than directly at the LTZ1000 where any measurement probe will change the operating conditions. At maximum environment temperature there should be at least some heater power.

In any case you should have a good thermal shielding of the LTZ1000 from both sides of the pcb to equalize pin temperatures and thermal EMFs.

With best regards

Andreas

[janaf]

The external transistor for the heater current is inside the control loop, so it should not influence the static value of heater current at all. I guess that's what Andreas was saying too.

Depending on what board you have, practically, you can, for testing, add a low value multi-turn trimmer in divider mode between R4/R5. If you measure the total current to the circuit and the output voltage, it's quite easy to see when the temperature regulation collapses; output voltage and current start wobbling. Then trim up to stability with some reasonable margin and measure the R4/R5 resistance ratio.

Depending on op-amps, resistor values, supply voltage etc., the whole circuit might use something a ballpark 25-35mA with the heater working normally and down to 15-20mA when the temperature setting is really too low.

But there is another mega-thread for the LTZ1000....

[JoeN]

But there is another mega-thread for the LTZ1000....

What is the source of the fetish/worship associated with this device?  It seems no one even carries it (http://www.findchips.com/search/LTZ1000), except for ebay sellers, most of them selling really beat up looking used ones (www.ebay.com/itm/1x-LTZ1000ACH-Ultra-Precision-Reference-LTZ1000A-/111325506311).  One American selling new devices, doesn't say where he got them, it's a little suspicious to me:  www.ebay.com/itm/Linear-Technology-LTZ1000ACH-Ultra-Precision-Reference-/251804783798.  What applications is the LTZ1000 found in?  And why are there no distributors?  Or is this something that Linear mostly sells directly, and why would they go that route on this one particular device?  Thanks.

It looks like this can be had directly from Linear at a lower price than the Chinese joker selling used garbage, a little more than the American guy considering he is offering free shipping.  Still, this seems the safest option.  http://www.linear.com/purchase/LTZ1000.  Should one purchase the leaded or unleaded version?  Why do you think the substantial difference in price exists?
 

[nctnico]

You can order the LTZ1000A from Linear directly.

[JoeN]

You can order the LTZ1000A from Linear directly.

I noticed.  And it's cheaper than the Chinese junk sellers and you will get it in 3 days rather than 3 weeks.

[janaf]

Digi-key used to stock but no more. Yes, safer to buy from LT direct. There are fake devises sold. I got one 

The fetish/worship? I think it's interesting of several reasons. First, it is simply the best voltage reference IC without competition. I's used in essentially all top of the line voltage transfers, voltage standards, 7.5 and 8.5 digit DMMs etc.

It's also interesting because when you look for sub-ppm behavior, you kind of leave the world of simply predictable behavior, physics, chemistry, thermodynamics join in. And things take years to verify....

Back to reality: You can get a decent estimate of heater current by measuring voltage over the low side 4148 diode, then a datasheet voltage/current diagram should give you a good idea of the current. You can also measure the resistance of the heater and estimate current that way. The LTZ1000 I have measured had around 250 ohm heater resistance at room temp, increasing by about 10% up to 20mA, while letting the device self-heat. 

[splin]

Hello,

in my opinion the transistor gain only affects overall heater control loop amplification. (Including Op-Amp).
And so the static deviation from setpoint together with the variations of open loop gain of the Op-Amp.

Ah - thank you - I can see that now.

I would invest into the "A-Type" of the LT1013. (or another high gain amplifier).

I have some OP777As which are similar - slightly better offset voltage and drift, slightly lower input bias current and noise, but the typical large signal voltage gain is lower at 2.5V/uV compared to 8V/uV.  I'm not sure if that's a problem for the amp that drives the zener, but using a high gain transistor such as a ZTX696B with a minimum hfe of 800 @ 100mA should compensate in the heater control loop.

Just looking at the LTZ1000 datasheet, it states that both amps contribute less than 2uV drift (.28ppm) over a 50C temperature range. The typical offset drift for the LT1013A is .3uV/C so 50C = 15uV. Assuming that this error is equally split between the amps (although I can't see any particular reason to believe it will), 15uV offset error equates to .14ppm reference output error or .0093ppm/uV.

The op-amp driving the zener's output is approx 7.8V, therefore input error due to limited gain is 7.8/8 = .975uV. Using the OP777 instead increases this to 7.8/2.5 = 3.12uV or an extra (3.12-.975)*.0093 =  .02ppm, so probably not worth worrying about. Also the minimum gain for both devices is 1V/uV. Is this analysis reasonable?

Can you see any problems with using OP777s?

Most probably the transistor is optimized (maximum gain) for a collector current of around 100uA.
Although I have measured that decreasing the resistor (to 50K) increases the unheated tempco of the zener section slightly.

Its more easy to measure the heater voltage than directly at the LTZ1000 where any measurement probe will change the operating conditions. At maximum environment temperature there should be at least some heater power.

In any case you should have a good thermal shielding of the LTZ1000 from both sides of the pcb to equalize pin temperatures and thermal EMFs.

I was considering using a short piece of insulated copper tube with a diameter such that it just fits inside the leads and using some insulated copper tape around the outside of the leads to help equalize lead temperatures. The insulation would need to be thin for this to be effective. It's clearly not necessary, like PCB cut-outs, but since its easy to do why not experiment - unless anyone knows why it would be a bad idea?

Actually, rather than copper I would probably use some insulated pyrolitic graphite sheet which I happen to have: http://industrial.panasonic.com/www-data/pdf/AYA0000/AYA0000CE2.pdf

This would have the advantage of a much lower heat capacity and thus a much faster thermal reponse.

Splin

[splin]

Depending on what board you have, practically, you can, for testing, add a low value multi-turn trimmer in divider mode between R4/R5. If you measure the total current to the circuit and the output voltage, it's quite easy to see when the temperature regulation collapses; output voltage and current start wobbling. Then trim up to stability with some reasonable margin and measure the R4/R5 resistance ratio.

Depending on op-amps, resistor values, supply voltage etc., the whole circuit might use something a ballpark 25-35mA with the heater working normally and down to 15-20mA when the temperature setting is really too low.

That looks like a good way to do it - thanks.

I wonder how much margin is needed - does regulation degrade (i.e. the drift slope, ppm/C, increases) as ambient approaches the operating temperature or does it simply fail when it gets too close?

Clearly it will take longer to respond when the temperature difference gets small as the device takes longer to cool so presumably the minimum cooling rate must exceed the maximum rate of change of ambient, but no doubt that's not the end of the story?

[Dr. Frank]

'Production variations in emitter-base voltage will typically cause about ±10°C variation. Since the emitter-base voltage changes about 2mV/°C and is very predictable, other temperatures are easily set.'

...

Has anyone measured the actual operating temperature of their LTZ1000(A)? I guess that might not be easy, especially with the LTZ1000A - would it be better to measure the collector current against temperature when heating the device externally?

I'm asking because for long term stability it is best to run the device at the lowest temperature that can be maintained at the maximum ambient so 10C of variation is quite significant. The LTZ1000 part appears to have the advantage here as well as being cheaper. Apart from the higher power consumption are there other disadvantages - such as bigger thermal problems due to higher lead temperatures?

Thanks, Splin

Well, the dU/dT of UBE is very predictable, i.e. -2mV/°C.
The variation of the absolute value of UBE is what causes the 10°C variation.

Therefore, I have assembled the LT circuit, and first disabled the heater part.
Then I attached a fast DMM, with MiniMax function, the 34401A, to measure the UBE of the voltage reference transistor, that is Q1.
I noted the ambient temperature, that's the same, the circuit is initially residing on.

Powering up, the DMM measures the UBE, let's say 0.560V, which will rapidly decrease by self heating of the reference circuit.

These 0.56V are stored as a maximum value, representing the measured room temperature.

You can repeat that measurement several times for higher reliability of the R.T. value.


If you now engage the heater (12k/1k for 45°C in my case), the UBE will further decrease with relatively precise -2mV/K, and you now can easily calculate the real stabilization temperature.

I do not remember the outcome of this measurement, but the calculated setpoint was quite precise.

It's also not that dramatic to have 5..10°C more.
Both of my LTZ1000s drift less than -1ppm/yr with blindly assembling 12k/1K, and only big variations as 65°C or these 95°C of the 3458A will noteworthy degrade the performance.

Therefore, I don't recommend to trim the oven temperature exactly to 45°C.. the exemplar variation also causes different stability figures.
Only make sure, that the temperature is not less than 45°C, because of regulation stability.

And more important is, to avoid big temperature excursions, due to soldering, or due to wrong oven temp, or due to oven regulation faults, because of the big hysteresis of the device.

Frank

[janaf]

I just posted some results from measurements on the heater of the LTZ1000A in the "big thread".
https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/

[MK]

'Production variations in emitter-base voltage will typically cause about ±10°C variation. Since the emitter-base voltage changes about 2mV/°C and is very predictable, other temperatures are easily set.'

...

Has anyone measured the actual operating temperature of their LTZ1000(A)? I guess that might not be easy, especially with the LTZ1000A - would it be better to measure the collector current against temperature when heating the device externally?

I'm asking because for long term stability it is best to run the device at the lowest temperature that can be maintained at the maximum ambient so 10C of variation is quite significant. The LTZ1000 part appears to have the advantage here as well as being cheaper. Apart from the higher power consumption are there other disadvantages - such as bigger thermal problems due to higher lead temperatures?

Thanks, Splin

Well, the dU/dT of UBE is very predictable, i.e. -2mV/°C.
The variation of the absolute value of UBE is what causes the 10°C variation.

Therefore, I have assembled the LT circuit, and first disabled the heater part.
Then I attached a fast DMM, with MiniMax function, the 34401A, to measure the UBE of the voltage reference transistor, that is Q1.
I noted the ambient temperature, that's the same, the circuit is initially residing on.

Powering up, the DMM measures the UBE, let's say 0.560V, which will rapidly decrease by self heating of the reference circuit.

These 0.56V are stored as a maximum value, representing the measured room temperature.

You can repeat that measurement several times for higher reliability of the R.T. value.


If you now engage the heater (12k/1k for 45°C in my case), the UBE will further decrease with relatively precise -2mV/K, and you now can easily calculate the real stabilization temperature.

I do not remember the outcome of this measurement, but the calculated setpoint was quite precise.

It's also not that dramatic to have 5..10°C more.
Both of my LTZ1000s drift less than -1ppm/yr with blindly assembling 12k/1K, and only big variations as 65°C or these 95°C of the 3458A will noteworthy degrade the performance.

Therefore, I don't recommend to trim the oven temperature exactly to 45°C.. the exemplar variation also causes different stability figures.
Only make sure, that the temperature is not less than 45°C, because of regulation stability.

And more important is, to avoid big temperature excursions, due to soldering, or due to wrong oven temp, or due to oven regulation faults, because of the big hysteresis of the device.

Frank

Use this link http://www.ti.com/ww/en/bobpease/assets/www-national-com_rap.pdf

then scroll down to the "whats all this vbe stuff" for how best to use your known single point "temp/Ic/Vbe"

[janaf]

Another practical answer: use 1:12.5 for room temperature, with some margin. Maybe 1:12.0 if you are in a well controlled temperature environment but before buying expensive high stability resistors, you can test with standard 1% metal film resistors to see if the your circuit is stable at your temperature.

I have just posted a diagram in the "big thread".

[macfly]

Hi volt-nuts,

coincidentally I have made measurements to find out, what would be the best individually  resistance divider for my 8 pieces of LTZ1000.
I used two circuits  (see document), one to find the UBE of the temperature sensing transistor and another one, to measure indirectly the self heating of the device via the transistor Q2.  A waste-product of this is the thermal resistance of the LTZ's.
The measurement showed me just as well, that probably some of my LTZ's are fake-parts.

There are possibly some questions about what I have done, but for this evening, I am out of business.

Regards,

macfly