.05% is sort of useless for a reference. Handheld meters have better accuracy. How long would you age them? Would you include your data from ageing? Would you measure and include noise specifications? For comparison I paid about €200 for an ltz1000 ref aged >10000 hours, have data from that time, noise specs, calibration data(including copy of Cal cert from device used for Cal) and 1 free calibration.
The initial accuracy of the reference is nearly irrelevant. The uncertainty of the calibration is what is important. In other words, the output could be 10.050012V, but this value is known to (say) +/-5ppm at a k-factor of 3 (which is ~99.97% uncertainty). 5ppm would be difficult to achieve with a non-heated reference that is not kept in an oven at a continuous temperature-- and even with that you are limited by the output 1/f noise <= 1Hz. You might be able to get an LM399 (heated) reference to hold still for 2ppm, but below that you would need to change to an LTZ1000. So, it kind of depends on what you want to do with this reference-- if it is just for standard jelly-bean handheld DMMs, then the specs can be relaxed-- if is for calibration of a 6+1/2 digit DMM, then the 2ppm device will be needed, and for better than 6+1/2 digits, then an LTZ1000 will be needed.
Remember, the artifact (voltage reference) must have at least 4X less uncertainty than the instrument you are calibrating with it.
I don't know what those numbers have to do with anything, he said .05%(500ppm) not 0.05ppm. Unless I'm wrong about that. If .05ppm I'd buy one.
The original question and answer:
I think a lot would also depend on the type of calibration that you can offer. Can you specify?
For example, Doug Malone specifies an in-cal 3458a and the burn-in time. He also allows for a re-calibration after some amount of time.
It depends, at the moment I could offer 0,05% initial accuracy, if I could sell one hundred devices I'd purchase a traceable fluke 732
and null against that.
To clarify, at this moment I can't offer a calibration with useful uncertainties, let a alone traceable to any national standard.
I could order 0,05% accurate parts so that would be something.
The idea I have is to make something sturdy and compact that can be mailed around in a cal-club type fashion,
so it will gather calibration points just like rolling stones gathers moss, or something like that.
And if it gets lost or damaged a replacement is readily available (for purchase..).
No can do on 0.05ppm accuracy, however I might know a competitor that is willing to sell you that.