- From: Calin A. Culianu <calin_at_ajvar.org>
- Date: Wed, 2 Oct 2002 13:03:17 -0400 (EDT)
On Wed, 2 Oct 2002, Tim Holy wrote: > Hi Calin, > > I use a NI PCI-6071E board to do 64 channel acquisition, 10kHz per channel, to > record spike waveforms from neurons. (It's only 12-bit, so it may not be the > board for you.) While the board specs say it can do 1.25MS/s, implying that > we could do almost 20kHz/channel, the errors due to finite settling-time > increase when we go that fast. NI is very straightforward about this point in > their specs and (with the tiny bit of checking I've done) seem to be accurate > in their measurement of the errors as a function of scan rate. I certainly > wouldn't shy away from paying the extra $500 if it turns out to be the better > board (I don't know if it is, since I haven't tried any MC boards). Compare > it to the lost labor [& pigs :-)] for corrupt data. Well, if you've actually achieved such high rates with 64 channels simultaneously, I will assume that is the board to get (the 6071E). I think that if we are careful with our signal and use an optimal gain setting, a 12-bit signal should be fine. By the way where do you get specs on MUX settling times from NI? I got their 'spec sheet' for some of their boards, and it just seems like a glorified advertisement more than a spec sheet.... > > Your signals are well-buffered, right? I've occasionally seen weird things > with the multiplexer putting junk back on the channel when using e.g. just a > transistor to buffer signals. What do you mean by well-buffered? I am a programmer more than a hardware guy, so to me a buffer is simply a big chunk of memory you use for data. What do you mean by that word in a hardware/signalling context? Thanks so much for your very helpful advice, -Calin
Received on 2002-10-02Z16:03:17