Friday, November 16, 2007

Data Collection

Dr. Bussard in his lab notes says that the ionization and other quick gas processes happen in 1 to 2 microseconds. Nyquist says that to be sure of capturing the highest frequency of interest you have to sample at 2X that frequency. Even so, if you sample at the zero crossings you will miss it. So much for Nyquist.

What we want is not just to measure process completions, but also look at what is happening in those processes. We might gain some insights. To do that you have to sample at least 3X the frequency of interest and much better 20X.

That says we need to sample at 20 MSPS. Front your sampler with a nice high speed fiber optic transimpedance amplifier in parallel with a DC (relatively) amplifier for low frequency fidelity and you have a very nice front end for a lot of the instrumentation we will need.

Stuff that will be sampled less frequently can of course use a much less expensive design.

With memory as cheap as it is these days I think that you decide what your maximum data collection time is at high speed and keep that amount of memory local. Then you have some kind of stupid simple protocol (say RS-485 as the physical layer) to collect the data. 5 MBPS shouldn't be too tough. With a 60% efficient protocol you should be able to collect the data from 1 high speed channel in 100 seconds. About 2 minutes. Twenty such channels in under 35 minutes. If you used 100 Base T Ethernet in a command-response type network you should get transmission time to come it at around 3 1/2 minutes. The slower RS-485 might be better if you have to collect data in an operational situation. Command-response Ethernet in relatively quieter environments (do a shot - recharge everything - hit it again).

Of course to see what the actual timings are like you have to pass around a synchronizing signal. Getting everything to start within 10% of the actual start time means passing out nice pulses on cables that differ in length from each other by no more than 20 inches. Not too tough.

Update: 31 May 008 0337z

EMC2 has pulled the lab notes and they are no longer available on the www. You may be able to get a copy from EMC2.

No comments: