What's new
What's new

OT: Bandwidth for PC data acquisition and control

WizardOfBoz

Diamond
Joined
Sep 30, 2006
Location
SE PA, Philly
Wild shot in the dark here: Anybody running PC-based A/D and D/A converters? What are the normal sampling frequencies? What type of bandwidth limitations might I run into using the PC for control?

Thanks!
 
I believe it depends on CPU useage. I have seen a process PC's CPU get loaded up to 100% where it won't reliably sample an event which happens in 1/4 second. Standard guideline IIRC is to be able to sample 10x as fast as the "event" you are trying to "catch"
 
Wild shot in the dark here: Anybody running PC-based A/D and D/A converters? What are the normal sampling frequencies? What type of bandwidth limitations might I run into using the PC for control?

Thanks!

There is not enough information given answer your question properly.

There are two parameters. Bandwidth, and latency. For the feedback loop in a machine control a 20µsec latency is nominal.

Bandwidth itself is simply sample-width-in-bits *2 (for Nyquist) * sample frequency. Bandwidth alone is not sufficient to specify the requirement for a feedback coupled control. For that latency must be taken into account.

Then you also need a RTOS. And not just the OS, also the BIOS. Modern (standard) BIOSes for Intel chipsets implement a SMI (system management interrupt) that is not compatible with a true RTOS (real-time operating system).

To better answer your question, one would need to know 'what is your tightest feedback loop'?

Cheers
<jbc>
.
 
2-10 Khz is fairly easy, 4 Mhz to 2 Ghz is doable on the input side with the right board and hardware DMA under Msoft operating systems.
Standard 640x480 colorvideo will eat up close to 20 Mbytes/second and even win 95 will do this with the right hardware and drivers.
Check the info at CyberResearch .com, mostly it's just a matter of money and what kind of slot you plug into.

Grabbing the data and servicing it are two different animals.
As JBC noted the x86 family was not designed for real time work and laptop chipsets will make your life worse.
Without knowing what you want to do or how you are going to program it, this is also a wild shot in the dark answer. :)
Bob
 
Trying to control plasma membrane voltage in a nerve cell. I think I need about 1Mhz for A/D conversion and probably less than 0.01 millisecond latency. A paper talks about using standard PC software for the control algorithm but I'm thinking I need a dedicated RT processor.
Thanks!
 
Trying to control plasma membrane voltage in a nerve cell. I think I need about 1Mhz for A/D conversion and probably less than 0.01 millisecond latency. A paper talks about using standard PC software for the control algorithm but I'm thinking I need a dedicated RT processor.
Thanks!

This could be done with a pre-Core architecture Intel desktop. Anything up to the and including the P4 (List of Intel Pentium 4 microprocessors) chip. Computers using Core chips implement SMI for cooling management. The SMI interrupt is not maskable and so breaks Real-Time constraints. On some recent hardware you can disable the SMI by flashing your own BIOS, for example with CoreBoot. Tricky!

Look into RTAI-Lab. It is designed for your application, has Scilab/Matlab bridges and Comedi interface drivers.

Are you doing this by yourself, or are you a member of a team? My suggestions would be easier for someone with specialist computer knowledge. It is still a quadruple-buttload of work to get it going properly.
 
Are you doing this by yourself, or are you a member of a team? My suggestions would be easier for someone with specialist computer knowledge. It is still a quadruple-buttload of work to get it going properly.

The goal is to understand available systems. Current dedicated amplifiers for measuring picoamp-level currents have an A/D sampling frequency of 250KHz. There are papers where people claim to do real-time control at a sampling frequency of 125KHz, using a Simulink-type modeling approach with function blocks. I was trying to figure out in a gross way how they did that.

Although I'm fine putting together a system like this, I agree that a specialist would be worthwhile. My contribution would be to 1) create the appropriate block diagram for measurement and control, and 2) writing a simulation model for the cell under study, and analysis of the data obtained, using the model I created.
 








 
Back
Top