ethernet data acquisition

[ASIM_BHATTI]

HI all

we need to digitize 32 analog channels simultaneously at atleast 20 Ksamp/s/ch.16-bit resolution is enough.Digitized data needed to be out on ethernet real time so that it can be sent over 600m distance.We need as much compact solution as possible since whole system will be placed in water sealed box.

Can NI help us to find out suitable solution?

Note: currently we are using four NI PCI 6123 cards installed on PC running window XP.The first problem is that these cards are quite large in size (and in price also) and second one is that first we have to read the data into PC and then this data is sent on ethernet.This approach often causes delays and what we need is rela time data acquisition and sending. 

THANKS

[Kyle_A.]

Hello,

One option I see for your system will be a cDAQ-9188 Ethernet chassis combined with 8 NI 9215 4 channel analog input modules.  You can synchronize the modules and they are simultaneously sampling, so you will have that covered.  32 channels can be sampled up to 25kS/s/ch with 16-bit resolution.  But the cDAQ-9188 does still need to be read into a PC first, so another option that may work better would be using a CompactRIO solution for stand alone real time processing, low cost and reliable.  You can select a real-time controller and cRIO chassis from here.

[ASIM_BHATTI]

Hello Kyle

Thanks for the reply. I am checking the options you mentioned. Is there a possibility to use NI PCI 6123 with some NI real time controller(compactRIO solution) so that data can be sent on ethernet real time?

THANKS

[Kyle_A.]

Hello,

To use the PCI-6123 card you would need to have the data go through the PC first before going through ethernet, which would not be real time.  The real time comes from the onboard processor on the cRIO which is separate from any other system process, like Windows OS, that would not be deterministic. 

[John_P1]

Just to clarify,

Sending data over ethernet is inherently NOT deterministic due to the buffering and packet overhead.  The RT Controller would allow you to run a deterministic OS on the target, but sending data back to the host would not be deterministic (what we usually refer to as "real-time" at NI).

I think by "real-time", what you really meant is whether or not you can send the data back to a host PC while the acquisition is running.  If this is the case, then the 9188 will work just fine.

A similar ambiguity comes up when you say the channels need to be sampled simultaneously.  Do you mean truly simultaneously, as in sampled off of the exact same clock edge?  Or do you simply mean that the acquisitions must be running at the same time, but the individual channels can be sampled sequentially.  Would a 4 µs delay between channels be acceptable?  Do you have settling time concerns such that a multiplexed board would not work for your needs?

If you do need truly simultaneous sampling at 20 kHz per channel, then the 9215 would work for this application, but it only has 4 channels per module so you would need to completely fill your 9188 chassis.

If multiplexed boards are tolerable, you could use 3x 9205s to get 32 channels at 20 kHz (11 channels on two boards, 10 channels on another).  These boards have an aggregate rate of 250 kHz, so the maximum rate you can sample on them is 250kHz / N Channels used.

If you are willing to compromise your sample rate a bit, you could use a single ENET-9205.  This would give you 32 channels at 7.8125 kHz. 

To summarize, ethernet is not "real-time", but you can send data back to the host while the acquisition is taking place.  Multiplexed boards are not "simultaneous", but you can multiplex through the channels fast enough to give the effect of simultaneous sampling for most applications.

Assuming you are not trying to implement feedback for a control system, I think the 9188 with either 8x 9215 or 3x 9205 would be your best bet.  The 9215s have the benefit of being truly simultaneous, and would not be susceptible to settling time concerns inherent to multiplexed boards.  If compromising your sample rate is an option, then the single-module ENET-9205 could save you quite a bit of cost and is much smaller than the full chassis.

Best Regards

[John_P1]

I just noticed in your first post that the delays caused by sending data over ethernet were unacceptable.  How much delay is acceptable?

[ASIM_BHATTI]

Hello

Thanks John for reply. By real time we mean no data loss should occur during the process of data acquisition and sending on ethernet.Actually what we want is a system (simple PC or some NI real time controller) which will continuosusly acquire data and send it over ethernet to a signal processing unit for processing.

We will prefer simultaneous sampling since we have to perform time delay beamforming etc but there is compromise b/w price and performance.Sampling rate can not be compromised however.

ENET-9205 will be more suitable choice if somehow we can synchronize two ENET cards. Can we do it?

Best Regards

[John_P1]

You can synchronize two ENET 9205s--the ENET-9163 (which is the part number for the sleeve that the module fits into) actually has two PFI lines on the back that can be used to share clocks and triggers.

However, if you put 16 channels on each 9205 that still only gives a max sample rate of 15.625kHz  (250kHz / 16).  If you can't compromise sampling rate than this is probably a problem since it is less than your desired 20 kHz.