Comparing LT1021 to LT1236 10V Refs - Which One?

[Majorassburn]

Both the LT1021-10 & LT1236-10 "C" versions are .05% basic accuracy 10V references.
If you had to choose either one for your reference, which would you choose and why?
I chose the LT1236 and find no faults with it but I have not measured long term drift due to lack of time.
What say?

[Alex Nikitin]

It is the same chip, the only difference I can see is that LT1021 is available in a metal can package (which can provide a much better stability in time and with a humidity change).

Cheers

Alex

[Andreas]

Hello,

LT1021 is either trimmed for tempco or for basic accuracy.
LT1236 is trimmed for both.

So either LT1021 as "B" grade (best tempco, accuracy can be adjusted)
or LT1236 as "A" grade  (both basic accuracy + tempco)

But always I would prefer a metal can package. (low PCB stress and no humidity influence so lowest long term drift)
next would be LS8 ceramic SMD package (dead bug mounted on PCB or mounted on ceramic substate module or FLEX PCB)
Then plastic DIP8 package.
The worst is SMD plastic package. (more influence from PCB stress and humidity than DIP8 package)

with best regards

Andreas

[iMo]

There is the "LT1236LS8" in "smd ceramic hermetic" package. But it is in 5V version only.
It has got a separate datasheet, you have to search for above string on the ADI pages.

[Majorassburn]

Hello,

LT1021 is either trimmed for tempco or for basic accuracy.
LT1236 is trimmed for both.

So either LT1021 as "B" grade (best tempco, accuracy can be adjusted)
or LT1236 as "A" grade  (both basic accuracy + tempco)

But always I would prefer a metal can package. (low PCB stress and no humidity influence so lowest long term drift)
next would be LS8 ceramic SMD package (dead bug mounted on PCB or mounted on ceramic substate module or FLEX PCB)
Then plastic DIP8 package.
The worst is SMD plastic package. (more influence from PCB stress and humidity than DIP8 package)

with best regards

Andreas

Thanks for that answer. I did notice the trim choice. What bugs me is when you look at the "Output Voltage Temperature Drift" graphs for both the 1236 and the 1021, the 1236 looks AWFULLY unstable compared to the nice, flat 1021 graph over an 80C temperature range.  Am I interpreting those graphs correctly? 

In practice, my LT1236A's & B's seem acceptably stable (+/- 100uV) from about 74F-82F but I have not tested them beyond because that range would be their typical operating environment. If they drift as badly as the data sheet graphs indicate, I might think about changing to the 1021.

[Andreas]

Am I interpreting those graphs correctly? 

No. Every reference behaves very individual. The 1021 datasheet has a "idealized" curve.
In practice: if you have thight requirements for a special temperature range you will have to measure each device and select them.
(or do some oven / temperature compensation).


For the 5V version:
My LT1236AILS8-5 have typical 2-4 ppm/K relative linear over 10-40 deg C. Some are even much better.
One LT1021CCN8-5 (DIP8) has around 7.5 ppm/K over 10-40 deg C.

with best regards

Andreas

[iMo]

My aprox 40 pieces of epoxy LT1021-5 and -10 had TC from -1 (1pc.)..0(1pc. I smoked  ).. +5ppm/C, w/ majority around +3. Measured with the simple quick finger method (put your finger on the package for say 1 minute, the voltage "diff/8" with say 23-31C), long time back, no graphs to see. The biggest issue with the LT1021 in epoxy I see is the hysteresis (off/on).

PS: The moral of the story - "Never ever start your experiments with your best voltage reference"..