How do manufacturer's calculate thier specifications for devices.

[mendip_discovery]

I am curious how they come up with the specs they state.

How would I go about testing a device to work out the specifications? Is it much like Uncertainty Budgets where you do a fair bit of statistics and RSS maths etc?

Part of this is I would like to move away from the manufacturer's specifications for some of my test gear but my assessor was keen to point out it is not just a case of taking a few ppm off the spec. The other part is because I don't know it and would like to know as its entertaining trying to fathom the specification stated by some cheap piece of tat a customer has bought from the internet.

[sahko123]

For something like test equipment often times the real life performance is actually better than stated.

For devices many specs are promised "by design" or sometimes by statistical binning to achieve various grades of products. Most times if they use statistical methods they would actually have the distribution for that parameter such as opamp bias current or offset voltage.

[mendip_discovery]

For something like test equipment often times the real life performance is actually better than stated.

Yes, that is quite common. I just would like to know how it is calculated so I could in theory create my own specification by doing my own testing. For example, plenty of meters have a spec for +/-%5C with a spec for every 1C outside the spec. If working to a lab temp of +/-2C could I reduce the spec by a multiple of the 1C spec?

I just like to be clear on how I went about generating a new spec. and not just be grabbing numbers until it fits my current curve.

[KT88]

In case you need an answer to EVERY question - UG-311: Reliability Handbook from Analog:
https://www.analog.com/media/en/technical-documentation/user-guides/UG-311.pdf

Cheers

Andreas

[David Hess]

Motorola has an application note which discusses the non-destructive and *destructive* testing which is used to determine specifications.

[mendip_discovery]

I was hoping for info on meters/test gear rather than at the component level.

[Kleinstein]

The principle way is to make a statistical model for the error sources and than estimate the contributions to the errors. The estimates can be from component spec, measurements of sample units and other special test. This can also include measurements on each units, to check some relatively easy to measure parameters (e.g. initial offsets or bias currents).
Experiance with earlier models can also help to estimate those parameters.
Not everything can be reliably measured and is strict science.  Especially long term drift estimates with newer parts are difficult and there is some room to get more or less stringent or confident in the performance.
The statistical model can than be used to calculate uncertainty limits to a certain failure properbility.
The final specs are than usually a simplified model, e.g. specifying the limits as part of the full scale and actual value to make them easier to use and more comparable.

Depending the company they may not do whole strict math (or do it similar with different naming or context - math sometimes has several ways to lead to the same result, giving different named to intemediate results). Worst can the specs are more from marketing and the dev. team only checks for obvious errors.

I just would like to know how it is calculated so I could in theory create my own specification by doing my own testing. For example, plenty of meters have a spec for +/-%5C with a spec for every 1C outside the spec. If working to a lab temp of +/-2C could I reduce the spec by a multiple of the 1C spec?

It does not work this way: The TC specs are upper limits for a quite large range.  It is quite common that the TC is highest at the extremes of the temperature range. They may very well have a more detailed model (e.g. TC depending on temperature range, include higher order terms) for the TC in the calculation.  So chances are they assume less effect from the TC in the specs for the +-5 C range. The official specs may not show all the details used in the calculation.  There may be an internal tighter TC spec for the close range, that is just not published.
So subtracting the TC part, even if done in the RSS way would be usually too optimistic.

[mzzj]

I was hoping for info on meters/test gear rather than at the component level.

Instead of manufacturer 1 year specification use your own history data. Calibration history from at least 3 to 5 last calibrations gives you pretty good idea on stability.

For temperature coefficient you have to measure your own equipment. Either use thermal chamber or track the changes over longer perioid of time and plot lab temperature vs. output value. (gets tricky with better equipment as it is more difficult to distinquish between the temperature coefficient of DUT and reference used in the test)
In some cases you can also apply correction factor based on temperature: for example your reference resistor is specified at 2ppm/Cel level but you have measured it at 18, 20, 23 and 26 degree oil bath and know that the temperature coefficient is  1E-6*t+3E-7*t^2
Now you can compensate the resistor value for temperature effect to better than 0.1Cel and  resulting uncertainty is tiny fraction of manufacturer specification. 

[mendip_discovery]

Thanks that is more like it. I am sure that even some of the big players have before mislabled the specs to give people the idea they are much better than reality. Then there are other units that year on year do much better than expected.

I was playing with a Fluke 714 yesterday and it was interesting to see they had to update the specs for the unit in the manual. Though the website still reports a different spec for the unit. This being a unit for K-Type thermometers and with my recent look into the rabbit hole I am sure many of these specs are a little holistic, or just omit some of the other sources of error.

Part of me just wants to know how Agilent came up with the specs for the 34401A and could I recreate some of that maths so that I can insert my data in to see if I can reduce the specifications myself. I am also interested in creating a spec for my 5V voltage source I have here. I guess to a certain extent I will just need to go down the route of characterisation of these things.

[alm]

The big difference is that the manufacturer had to come up with a spec that they could meet for all 34401As made for over more than a decade.

You would be characterizing just one unit. I don't think the math would be very similar. You are basically, in GUM terms, treating the drift etc of the instrument as type A uncertainty (based on measurements) instead of type B (based on prior knowledge). But for drift you would need to collect a fair amount of data over time for this. I think for the 732x voltage reference Fluke mentioned somewhere that they can estimate and forecast the drift based on 8 calibrations over time?

[mzzj]

Thanks that is more like it. I am sure that even some of the big players have before mislabled the specs to give people the idea they are much better than reality. Then there are other units that year on year do much better than expected.

I was playing with a Fluke 714 yesterday and it was interesting to see they had to update the specs for the unit in the manual. Though the website still reports a different spec for the unit. This being a unit for K-Type thermometers and with my recent look into the rabbit hole I am sure many of these specs are a little holistic, or just omit some of the other sources of error.

Part of me just wants to know how Agilent came up with the specs for the 34401A and could I recreate some of that maths so that I can insert my data in to see if I can reduce the specifications myself. I am also interested in creating a spec for my 5V voltage source I have here. I guess to a certain extent I will just need to go down the route of characterisation of these things.

Thermocouple inputs/outputs in many of the Fluke products are pretty sketchy. Their idea of cold junction compensation seem to work less than stellar in many cases.

DMM specs usually lack the information of coverage factor k in the stability data. Calibrators and high-end reference multimeters (Fluke 8588A )spec also specify the stability data with k=2 95% probability and k=2.6 99% probability.
AFAIK some lower end multimeters actually use even larger coverage factor than 3 but finding this information is not easy.