LTZ1000CH or LTZ100ACH

[meggerman]

I am currently looking at a project to use either the LTZ1000CH or LTZ1000ACH, but I need to do some more testing to see how much difference the "special" heater in the LTZ1000ACH makes.
From what I understand is the LTZ1000ACH uses a method that is more efficient in heating the die - but surely any power input into a heater will result in the same amount of heat output.
Can anyone comment on what differences they can see between the 2 variations?
If the only difference is that the ACH will heat up faster than the CH then it would seem the CH is going to always be a better choice because of cost.

Pricing:
While there is stock available for the LTZ1000CH and LTZ1000ACH from the big 3: Arrow, Mouser and Digikey, the prices have risen sharply in the last week or so - if you are buying just a few and this I think this might be down to a larger gap in the price breaks.
It seems the prices directly from Analog Devices (qty=20 or more) is unchanged or only a small change.

Rob

[Dr. Frank]

The A version has a stronger isolation of its chip to its case, but the chip itself, and especially the heating element (250 Ohm resistor) are identical.

So the case gets less warm for the A version, which might be important when using excessively high oven temperatures like in the 3458A, which is about 95°C, but at the price of poor longterm stability. (typ. > 16ppm/yr.)

The standard part is the non-A version, which is used for the normal oven temperatures of 45 .. 60°C, which will give typical drift rates of < 1ppm/yr.

The A version on first thought consumes less power than the non-A version.

That may be true if you chose the same oven temperatures, but if you design for same maximum environmental temperature, the A version has to be set to about 10°C higher oven temperature, which neutralizes the supposed advantage of the A part, because a higher oven temperature requires higher power dissipation in any case.

It really depends on your design requirements, i.e. highest possible stability, vs. highest possible environmental temperature, which device you want to chose.

I'm THE great advocate for the non-A version, like in the REAL voltage references DATRON/Wavetek 4910 and Wavetek/Fluke 7000.
All of my 7 LTZ1000 references are meanwhile, or even instantaneously stable between 0.3 .. 1ppm/year at about 50°C oven (12k/1k).
Please also read our long, long 'Ultra Precison Reference LTZ1000' thread...

The non A version is also less expensive, around 40 vs. 50$.
In former times, Linear Tech. allowed sample ordering of 2 parts, or a minimum of about 100$, afair.. conditions may have been changed when Analog acquired LT.
Maybe you need to order also LT 1013, and chopper amps..for exceeding a minimum order limit.

Frank

[e61_phil]

That may be true if you chose the same oven temperatures, but if you design for same maximum environmental temperature, the A version has to be set to about 10°C higher oven temperature, which neutralizes the supposed advantage of the A part, because a higher oven temperature requires higher power dissipation in any case.

Is it really the case? The A has 5 times better thermal insulation (isolation? never get it right). Which means at five times more temperature gradient the power will be equal. But if you run the LTZ1000 at 55°C instead of 45°C at 25°C room temperature, the LTZ1000A should still need less power.

[pelule]

see ltz datasheet "The LT Z1000A should be set about 10°C higher than the LT Z1000. This is  because  normal  operating  power  dissipation  in  the LT Z1000A  causes  a  temperature  rise  of  about  10°C.  Of course  both  types  of  devices  should  be  insulated  from ambient. Several minutes of warm-up is usual"
link: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=2ahUKEwi_7LWDlMbhAhVSzRoKHQ1gAL8QFjAAegQIABAC&url=https%3A%2F%2Fwww.analog.com%2Fmedia%2Fen%2Ftechnical-documentation%2Fdata-sheets%2FLTZ1000.pdf&usg=AOvVaw33UF-yzTHHPfPyLPpG0yv0
/PeLuLe

[Andreas]

But if you run the LTZ1000 at 55°C instead of 45°C at 25°C room temperature, the LTZ1000A should still need less power.

Hello,

not much according to my measurements:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg2303835/#msg2303835

Difference is about 10-30% mostly depending on datecode of the LTZ1000(A)
(be carefully to compare only devices with either LT1013A or LTC2057 against each other as the LTC2057 needs more power).

with best regards

Andreas

[e61_phil]

I was trying to say: Higher temperature doesn't mean higher power in any case

Let's assume 25°C ambient temperature:
LTZ1000A (55°C-25°C)/400K/W = 75mW
LTZ1000 (45°C-25°C)/80K/W = 250mW

[Andreas]

Hello,

its not so easy

Most of the heater power is dissipated in the external transistor. (from 14 V power supply I have only around 5V at the LTZ1000 heater).

the 25 deg C environment temperature (for you)
is not the environment (PCB) temperature of the LTZ which is around 8 deg C higher. (with no forced cooling)

with best regards

Andreas

[e61_phil]

Hello,

its not so easy

Most of the heater power is dissipated in the external transistor. (from 14 V power supply I have only around 5V at the LTZ1000 heater).

the 25 deg C environment temperature (for you)
is not the environment (PCB) temperature of the LTZ which is around 8 deg C higher.

with best regards

Andreas

Yes, you're absolutely right and if you cover the LTZ1000 the gradient will be much smaller. But that doesn't matter. The LTZ1000A will consume less power than the LTZ1000 if it is only running 10K higher. My point was just, that the 400k/W will not be eaten up by just 10K higher temperature.

If you run your LTZs with only 2.5K over the thermal resistance the power will be equal. But I don't think anybody will do that.

[meggerman]

Hi Dr. Frank,

The standard part is the non-A version, which is used for the normal oven temperatures of 45 .. 60°C, which will give typical drift rates of < 1ppm/yr.

The A version on first thought consumes less power than the non-A version.

That may be true if you chose the same oven temperatures, but if you design for same maximum environmental temperature, the A version has to be set to about 10°C higher oven temperature, which neutralizes the supposed advantage of the A part, because a higher oven temperature requires higher power dissipation in any case.

It really depends on your design requirements, i.e. highest possible stability, vs. highest possible environmental temperature, which device you want to chose.

Yes, this is exactly what I was hoping for: better long term stability and a lower cost part.
I am looking to insulate the whole reference board and run everything at the same temperature as the LTZ1000 die.
It may have applications in very cold environments where outside temperatures are as low as -50C and high stability with low power consumption, is important.

Rob

[splin]

I was trying to say: Higher temperature doesn't mean higher power in any case

Let's assume 25°C ambient temperature:
LTZ1000A (55°C-25°C)/400K/W = 75mW
LTZ1000 (45°C-25°C)/80K/W = 250mW

The datasheet states 400K/W and 80K/W (typical) in one place but later includes this graph:



Anybody made any measurements to see which numbers are right? Would those numbers include heat lost via the leads, using a typical PCB layout?

Like Dr Frank, I can't see much, if any advantage to the LTZ1000A for meterological applications - there's nothing to stop you adding more insulation to the LTZ1000 to match the A version but not vice versa. IE. you have no choice but to use a higher minimum operating temperature above ambient, 10C, than the 2.3C (or 5C?) of the LTZ1000. The LTZ1000 is quite a bit cheaper to boot.

There has been some discussion that the A version suffers from less hysteresis but I can't recall if was just conjecture or real.

Ironically, to minimize power consumption you need to set the operating temperature as high as possible to allow the maximum amount of insulation to be used, and still maintain regulation at the highest ambient temperature. The heater power required at lower ambient temperatures is thus reduced because of the improved insulation levels.

Eg. assuming Tambient max = 38C, min = 10C and zener dissipation is 29mW (7.2 x 4mA)

Operating temperature = 55C,  max thermal resistance allowed (Rth) = (55 - 38)/.029 = 586K/W. At 10C ambient, heater power = (55 - 10)/586 =  77mW.

Operating temp = 100C, max Rth = (100 - 38)/.029 = 2138K/W. At 10C ambient heater power = 42mW. Achieving 2138K/W may be a bit of a challenge with the non A version however.

Operating temp = 43C (LTZ1000 only) max Rth = 172K/W. Power @ 10C = 192mW.

The advantage of the A version here is perhaps that Rth is specified and presumably doesn't vary much (but not specified of course!), whereas characterising your own design insulation, and more importantly manufacturing it consistently enough may be quite tricky. So for applications where long term drift is less important than power consumption, and the volumes are too high to characterise individual units but too low for significant thermal design and characterisation expenditure, the A version may be better.

[SilverSolder]

Urgh, what's going on with the pricing of the LTZ1000?  Has Martin Shkreli gone into the electronics business?!?   

[maginnovision]

Urgh, what's going on with the pricing of the LTZ1000?  Has Martin Shkreli gone into the electronics business?!?   

Digikey shows about 10$ more than few months ago but that's not a huge issue considering.

[splin]

All of my 7 LTZ1000 references are meanwhile, or even instantaneously stable between 0.3 .. 1ppm/year at about 50°C oven (12k/1k).

Are they continuously powered? I seem to remember you saying something to the effect that electricity in Germany is too expensive to keep things powered on unnecessarily - and for environmental reasons as well I assume.

[Dr. Frank]

All of my 7 LTZ1000 references are meanwhile, or even instantaneously stable between 0.3 .. 1ppm/year at about 50°C oven (12k/1k).

Are they continuously powered? I seem to remember you saying something to the effect that electricity in Germany is too expensive to keep things powered on unnecessarily - and for environmental reasons as well I assume.

Hi Splin,
You remember that correctly. 

Yes, the first two are nearly continuously running since 2009, or even longer, and the 4 recent ones since October 2017.

At 25mA / 12V supply or about 600mW primary per reference, that is 5kWh / year, that may be about 10€/year for all of them.

As our extremely wise government (Dr. Merkel) continuously increases cost for electrical energy by this stupid and uneffective 'regenerative energy' change, I meanwhile gave up saving electrical energy at all costs.   

But I don't let my 3458A run continuously, that's correct.
At first due to  saving energy, but also to not consume lifetime of several 'hot' and unobtanium components inside, like the ELxxxx comparators.

Frank

[TiN]

I have my 3458's running 24/7/365, but they run at reduced oven temps. Also one unit is LTZ1000CH.
But in my design I prefer to use ACH's.
They also have linear and easily correctable tempco, while CH chips require much dicking around with trims and external isolation just to find that "sweet spot". (My rule of thumb for good LTZ tempco = less than 0.05ppm/K).

[meggerman]

Hi TiN

I have my 3458's running 24/7/365, but they run at reduced oven temps. Also one unit is LTZ1000CH.
But in my design I prefer to use ACH's.
They also have linear and easily correctable tempco, while CH chips require much dicking around with trims and external isolation just to find that "sweet spot". (My rule of thumb for good LTZ tempco = less than 0.05ppm/K).

Is the sweet spot a specific temperature to run the device at or a combination of things, like zener current, to get the lowest level of noise and stability over a period of time?
Certainly I will be testing with both variations to see how it the devices compare against each other.
So I should be aiming for 0.05ppm/K for 7V = 350nV per degree Kelvin.
Has anyone done close temperature monitoring of both parts to see what fluctuations there, are once the device is up to temperature?
The supply rails for the 03458-66509 circuit are +18V and -15V, the example circuit mentions +15v and -ve rail below 0V vref, is there a reason the rails are so high/low compared to the output of the reference?

Prices for today (2019/04/11):
Digikey Qty=1 LTZ1000ACH#BPF = $78.14 LTZ1000CH#BPF = $61.40
Mouser Qty=1 LTZ1000ACH#BPF = $72.37 LTZ1000CH#BPF = $61.65
Arrow Qty=1   LTZ1000ACH#BPF = $68.45 LTZ1000CH#BPF = $58.32
ADI Qty=20    LTZ1000ACH#BPF = $54.50 LTZ1000CH#BPF = $42.85

Past prices (using google cache)
Arrow Qty=1   (2019/03/9)   LTZ1000ACH#BPF = $57.31
Arrow Qty=1   (2019/03/24) LTZ1000CH#BPF = $45.05

ADI prices for 2019/03/29 were the same as for today, so its just the distributors that have increased their prices for low quantity orders.
Although Arrow are nearly always lower cost for this device in a single part, they do not offer price breaks on this and no back ordering. But they said they may be able to offer a better price through their online chat.

[Andreas]

Anybody made any measurements to see which numbers are right?
Would those numbers include heat lost via the leads, using a typical PCB layout?

Hello,

from my measurements with power consumption I have a large stray between different date codes (of the A-version)
So I guess that the typical will depend on production lot.
Somewhere I have read (guess it was the data sheet of the thermal converter) that the thermal resistance can be up to 600 K/W.

And of course the data sheet does not contain any measurement conditions like long/short leads, PCB horizontal/vertical, still air/forced cooling.

I am looking to insulate the whole reference board and run everything at the same temperature as the LTZ1000 die.
It may have applications in very cold environments where outside temperatures are as low as -50C and high stability with low power consumption, is important.

running the board at the die temperature will not work: the T.C. of the LTZ without on chip heater is at around 50ppm/K.
In very cold environments the A-Version might give less power consumption.

with best regards

Andreas

[Kleinstein]

While the LTZ1000 is a precision part in most aspects the specs for the heater resistance are very loose (200-400 Ohms). So some units may need more current than others to get the same heating effect.

[Echo88]

Where did you get your mentioned >16ppm/°C typical spec Frank?

[Dr. Frank]

Where did you get your mentioned >16ppm/°C typical spec Frank?

The drift rate at that temperature of 95°C is nowhere specified, but it is an Arrhenius calculation, according to e.g. P J Spreadbury: 'The Ultra-Zener.. is it a portable replacement for the Weston cell?' , Meas. Si. Technol. 1 (1990).

They demonstrated, that the drift of the LTZ1000 doubles with each 10°C increase of the oven temperature. At 55°C they measure typically -2ppm/year. 95°C would yield a 16 times higher drift.

From my experience, it might be a bit better, like 1ppm/year @ 50°C (my 7 LTZs all run between 50..53°C @ 12k/1k) and maybe the 15k/1k gives a bit lower oven temperature, like 90°C.

So my >16ppm/year typ. @ 95°C is really a minimum estimation.

Btw.: HP has to pre-age their abused LTZ1000A references for the 3458A and 34470A, and then sort out (*) the better ones having 8ppm/yr., or even 2ppm/yr.

They additionally have problems with hysteresis (see AN-18), which gives an apparently higher drift rate, under certain circumstances.

Frank

(*) yield is probably < 50%

[Kleinstein]

The drift rate at that temperature of 95°C is nowhere specified, but it is an Arrhenius calculation, according to e.g. P J Spreadbury: 'The Ultra-Zener.. is it a portable replacement for the Weston cell?' , Meas. Si. Technol. 1 (1990).

They demonstrated, that the drift of the LTZ1000 doubles with each 10°C increase of the oven temperature. At 55°C they measure typically -2ppm/year. 95°C would yield a 16 times higher drift.
....

So this mean one can expect about a 2 times higher drift rate for the LTZ1000 A due to the about 10 K higher set point needed. This could be a bit less if good insulation is used around the non A version.

Due to the square law for the heater, half the power needed means about 70% of the current needed. So there is some power saving for the A version possible, but not that much if external insulation is added.  A first point to save power and keep the drift low is choosing an oven temperature that is not too high for the planed use.

Just as a crazy idea: The heat lost from the transistor to drive the heater could in theory also be used to heat the reference - so maybe have the transistor on top of the LTZ instead of somewhere far away. This may need a modified compensation however, to really work at the low power end.

[splin]

The drift rate at that temperature of 95°C is nowhere specified, but it is an Arrhenius calculation, according to e.g. P J Spreadbury: 'The Ultra-Zener.. is it a portable replacement for the Weston cell?' , Meas. Si. Technol. 1 (1990).

They demonstrated, that the drift of the LTZ1000 doubles with each 10°C increase of the oven temperature. At 55°C they measure typically -2ppm/year. 95°C would yield a 16 times higher drift.

From my experience, it might be a bit better, like 1ppm/year @ 50°C (my 7 LTZs all run between 50..53°C @ 12k/1k) and maybe the 15k/1k gives a bit lower oven temperature, like 90°C.

So my >16ppm/year typ. @ 95°C is really a minimum estimation.

What is the drift experience in ppm/year for those of you who have equipment that is a) powered up (mostly) continuously and b) have it calibrated periodically? It would be interesting to see if Dr Frank's hypothesis is born out in pratice and to compare LTZ1000 to LM399 based instruments; I'd include zener reference meters such as Datron and Solartron but I expect that very few, if any many meet a) or b) let alone both. Personnaly I'd expect unmodified 3458As to drift considerably less than 16ppm/year as the vast majority will be well aged.

The 3458A's LTZ1000A runs at 95C which is almost the same as the LM399 so it would be interesting to know how 3458As drift compared to say 34401As as well as newer instruments including 34470A, DMM7510 and 34465A.

[Andreas]

Hello,

my picture of the ageing drift is a bit different.
I think that there are several contributions to drift.

a) the ageing of the zener itself (most probably following Arrhenius law)
  mostly depending on diffusion effects and impurity of the silicon.

b) Interaction of the cement with the die  (epoxy which is used to fix the chip into the housing).
  Here we have very large tolerances from batch to batch and over time.
  (thickness of epoxy, viscosity after hardening, changes of epoxy composition due to RoHS and other legislation)
  I also think that effects like hysteresis are mostly dependant on the epoxy which is used to fix the chip.
  If the sealing of the housing is not 100.0000% hermetically, also humidity can change the epoxy (swelling).

c) Interaction of the epoxy PCB with the chip through the leads.
  This one should be small for metal can housings especially with longer leads
  but can lead to seasonal changes due to humidity changes/swelling of the PCB.

I think that effect b) has the largest influence at least when the chip is new and the epoxy is not fully hardened.

Practically I have not a large difference in ageing between my 24/7 running LTZ#1 LTZ#2 and LM399#2 (LM399#3 is a bit off with the drift) references.
Although the LM399s are operated at a much higher temperature (90/50 deg C) and should age a factor 16 more than the LTZ1000A.

Hints to the diagram: the jumps on the LTZ1000A references are due to accidently shorts on the unbuffered output.
All measurements are made with a LTC2400 based ADC with temperature compensated AD586LQ reference
(so this reference at room temperature has nearly the same ageing drift than LM399#3) . X-Axis is in days.

with best regards

Andreas

[splin]

Hello,

my picture of the ageing drift is a bit different.
I think that there are several contributions to drift.

a) the ageing of the zener itself (most probably following Arrhenius law)
  mostly depending on diffusion effects and impurity of the silicon.

b) Interaction of the cement with the die  (epoxy which is used to fix the chip into the housing).
  Here we have very large tolerances from batch to batch and over time.
  (thickness of epoxy, viscosity after hardening, changes of epoxy composition due to RoHS and other legislation)
  I also think that effects like hysteresis are mostly dependant on the epoxy which is used to fix the chip.
  If the sealing of the housing is not 100.0000% hermetically, also humidity can change the epoxy (swelling).

c) Interaction of the epoxy PCB with the chip through the leads.
  This one should be small for metal can housings especially with longer leads
  but can lead to seasonal changes due to humidity changes/swelling of the PCB.

d)  Drift in the oven temperature sensor and associated components - Q2 Vbe in the case of the LTZ1000. Probably also Vbe for the LM399 - Q4 and the 11.2K/1K shown in Fig. 27 of the D/S?

In the case of the LTZ1000 the drift could (in theory) be directly observed by periodically measuring Vbe whilst operating the device at a closely controlled temperature with the heater off. The collector current must be the same at each test so a calibrated collector resistor should be used.

In the case of the LM399 I guess the ambient temperature could be slowly increased until the heater current (or more accurately the temperature stabilizer circuit current) drops to a minimum to determine the oven temperature setpoint. I don't know how stable the circuit is when it nears dropout however.

I don't suppose anyone has done any such measurements but it would give more insight into the sources of drift for these devices. Assuming the reference zeners have a TC of 50ppm/K then 1ppm of drift implies 20mK drift maximum (if it were the sole drift mechanism) so such measurments would be extremely difficult on a well-aged, low drift part but may be feasible during the first 1000 hours or so.

[meggerman]

Hi,

Practically I have not a large difference in ageing between my 24/7 running LTZ#1 LTZ#2 and LM399#2 (LM399#3 is a bit off with the drift) references.
Although the LM399s are operated at a much higher temperature (90/50 deg C) and should age a factor 16 more than the LTZ1000A.

Hints to the diagram: the jumps on the LTZ1000A references are due to accidently shorts on the unbuffered output.
All measurements are made with a LTC2400 based ADC with temperature compensated AD586LQ reference
(so this reference at room temperature has nearly the same ageing drift than LM399#3) . X-Axis is in days.

It's interesting to see how you measure the drift - so you are using an LTC2400 ADC.
This is always a bit of a puzzle for me, how do you measure something that is so accurate to start with - apart from asking someone at NPL to check it
Having access to a HP 3458a is always handy if you one or more - which I do not - at the moment.

I was looking to compare the difference in output of a whole series of pairs of LTZ1000 devices and perhaps a set of LM399's and measure this difference using a couple of 6.5 digit multimeters (HP 34401 + 3457a) via the GPIB interfaces. So by measuring the difference between 2 devices means you can use the millivolt scale on the meter.
Is it possible there is a slight temperature hysteresis of LTZ1000 (rather than the LTZ1000A) as a result of the device heating other nearby components (including solder junctions, connectors etc.) and affecting their values.

Certainly looking at output from Q2 Vbe (as Splin has suggested) using an external heater over a range of temperatures might provide some details about how linear it's output is.
So the end goal is low noise output, ultra low drift, low power consumption and long term reliability - the end product should be made from parts that last for decades - so no electrolytics.

Rob