The IP340/341 is a simultaneous sampling input board. It is capable of sampling blocks of 8 channels every 8µS and storing the result in a FIFO buffer. The problem with many systems is that they cannot read the data as fast as it is being sampled. The result is that the CPU can not clear the FIFO full bit. Another result is that the data can become unsynchronized which results in erroneous readings.
You must do some calculations to determine the maximum rate at which you can read the data. First, the IP module FIFO can be read with two wait states (500nS per sample). If the IP module is on a non-intelligent carrier board, there is another access delay caused by the carrier. (For example: a CompactPCI carrier such as the AcPC8625 has a maximum access delay of 650nS) The total access delay in this example is (500nS + 650nS) 1150nS. This means that the maximum throughput to read 8 samples is 9.2µS.
Following are the maximum access delays for several Acromag Carriers.
In addition to the access delay, there are other delays which can slow the system considerably and prevent achieving maximum throughput. First, in a multiprocessing system like VME, there could be additional delay time required to acquire the bus. In other systems, the CPU is overloaded doing other things and does not read the data at machine speeds. Another factor is the Operating system. Windows based systems are very slow and can be unpredictable. Real Time Operating Systems are much more deterministic.
It is important to test and measure your delay so that you can set the conversion rate at a speed the system can handle.
In a VxWorks based CompactPCI system with a 700 MHz Pentium processor, the maximum sample rate measured was 48uS for 8 channels without errors. In the same system, with a PowerPC MVME2700, the maximum conversion rate was slightly worse. You can see that the carrier and IP340 only contribute 9.2µS of the 48µS sample rate. The rest of the delay is contributed by the processor, operating system, and the priority of the tasks being performed.
There are two ways to achieve your desired throughput. First, go with an intelligent processor like a Motorola 162 which eliminates the need to go out on the VMEbus. The data is acquired locally and processed locally. There are several DSP carriers available as well that do the same thing. Another alternative is to use the PMC340 module, 32 bit reads can be done is burst mode. Access times are much faster with the PMC340, and it can be plugged directly into a CPU. Nearly every CPU has a PMC socket.