Hello,
Many thanks for all replies! I would have guessed that it is only to be expected that the TC of the LT1021 is device dependent,
because if there were a systematic TC, LT would have undoubtedly compensated that to make their product less temperature dependent.
So, perhaps (at best) there might be a correlation between devices of the same batch,
but in general, the TC measured over devices should be noise with average zero. Right?
That means indeed that the TC needs to be determined per device.
So, before (slowly) becoming a volt-nut, I first need to become a temp-nut! (Resistance is futile....
)
[Side note: I have designed and constructed a 0.1 ppb GPSDO, so I am also on my way to becoming a time-nut!
]
What follows shows my current thinking (which is undoubtedly flawed), and I would be very grateful to receive your comments.
It so happens that I have a small metal box that used to be the enclosure of an OCXO, with two ground pins (connected to the box itself), and four signal pins.
I also have an LM35A, would that have sufficient accuracy to serve as the sensor in the measurements needed to assess the TC of the LT1021?
The initial idea is to put some isolation in the box, and in the isolated space put the LT1021, the temp sensor, and a heating element of some kind.
Then hook the sensor up to an ADS1115 (also available in my spare parts box), connect the ADC and the heating element to a controller (Arduino Uno or such),
and that should result in an oven that I can preset to an arbitrary (within reason) temperature. I need to take care to make a sufficiently stable supply for the sensor and ADC, of course.
(How about using an LT3042 for that purpose?) The build should be such that I can measure all four of my LT1021 (indeed in a metal can, not the plastic version).
Then there is the question of how to measure small changes in voltage as produced by the LT1021.
First of all, a stable reference is needed. This is almost a chicken and egg issue, but the most important thing is that the new reference is stable.
I do have an LM399H. With an opamp (LT1012 is ok here?) I could make a stable (but not very accurate) 10 Volt reference.
In fact, I could trim the LM399-based reference to the untrimmed LT1021.
Then I would need a difference amplifier to track the differences between the LM399 and the LT1021 references (another LT1012?). Gain set to 100 or so?
Come to think of it, I have two LM399... Hmm... If I would build two such references, I could also measure how the difference develops between these two.
Just to get a better view on the drift of the LM399 references themselves. Would that make sense?
Anyway, after measuring all four LT1021, I could make temperature-compensation circuits for all four, and measure them again.
Then the best could be selected to be built into the oven permanently. So all trimming, compensation, etc. should be done at the target temperature. Say, 40 C?
Then, finally, the reference-in-the-oven can be trimmed. For trimming of my egg, I will need access to some professional reference chicken.
Or are there any other fun things to do with 6(!) references? (Bob Pease had a setup with 4x4 LM399, measuring drift within each group of 4, and across the four groups.)
The main purpose of all this is to learn, and also to see how far stability and accuracy can be pushed with moderate means.
Any comments and ideas are highly appreciated!
Home
Help
Search
Login
Register
