Thanks for the replies. I'm not certain industry would have followed my concept even if it works. What they like is something that's dirt cheap to produce and sells for a bundle. However I think I may still be wasting my time as I'm now thinking the difficult part is testing a new reference over years to qualify its performance. For a start you need to compare it to a very reliable reference and I don't have one (I'm aware of the Josephson standard)... Even if I build several refernces and note over time that they stay in close agreement I can't tell if they're very low drift or just all drifting together. Really each reference needs to be as independant as possible, right down to their design. Then hopefully the chance of each reference drifting just like all the others will be very low. It's all statistics in the end. Oh well... I've spotted these on eBay, but I'm suspicious about their quality and anyway you have to add the other ultra precise components : http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=260474645765 Alan