search

OpenBCI / OpenBCI_Cyton_Library

Repository for OpenBCI Cyton Arduino Libraries
MIT License
88 stars 89 forks source link

Board ADS running too fast, producing more than 250 samples per second. #25

Closed andrewjaykeller closed 8 years ago

andrewjaykeller commented 8 years ago

@jfrey-xx reported with his boards that they seem to be running too fast.

jfrey-xx commented 8 years ago

...not a brand new issue, here are my initial concerns: http://openbci.com/forum/index.php?p=/discussion/313/ads1299-sample-rate-drift-or-jitter :p

yj-xxxiii commented 8 years ago

Even 1% faster !!! Why does this matter ?

Anyway it is not a surprise since the internal clock accuracy of the ads1299 is +/-0.5% at 35°C and +/- 2.5% between -40&85°C.

y.j.

wjcroft commented 8 years ago

Mentioning Joel @biomurph and @conorrussomanno here.

(1) the real issue is the lack of the crystal on the current mainboard; understandable as a cost saving measure in the early days of the first Kickstarter. But much less justifiable at the current pricing.

(2) if the 8 channel board is ever revised, suggestion: put in the traces / pads so an external crystal could be EASILY added with no SMT soldering skills. That way those users that needed the extra accuracy could upgrade on their own.

(3) Ensure that the accuracy of the Ganglion sample rate (256 sps?) is adequate. Optionally have crystal solder-in available as a DIY addon.

Regards, William

andrewjaykeller commented 8 years ago

Yann Renard allowed me the open vibe driver that has a drift correction scheme. I think the new time synched packets will allow people to easily offline determine the rate of their ADS clock at the time of.

wjcroft commented 8 years ago

AJ, thanks. I'm aware of the OpenViBE drift correction, if you look through the threads on the forum. Nothing can beat the utility of an accurate sample rate. There is a whole paper there to read by David Hairston, have you seen it?

Accounting for Timing Drift and Variability in Contemporary Electroencepholography Systems http://www.dtic.mil/dtic/tr/fulltext/u2/a561715.pdf

Our drift is worse than any of the other systems listed there. So may be a concern from a marketing point of view.

yj-xxxiii commented 8 years ago

Could it help to read a counter of the PIC32 which is crystal driven, and to write this high resolution time stamp in the auxiliary data field? y.j. P..S. : I never tried to study evoked potentials without recording a hard trigger signal on a dedicated channel. This way a 1% drift is not a problem...

andrewjaykeller commented 8 years ago

Added timestamps to the data packets which will allow you to take record, start time stamping, take the (last time stamp - first time stamp) / samples to find the number of samples captured in comparison with the Driver. Works well if you time sync up with a Driver using new protocol and continually resync every minute or so.