Due to the 15MHz maximum sampling rate, streaming applications or large finite acquisitions should consider the required throughput needed for their tasks. The PXI(e) Chassis selection is especially important if you are considering multiple cards streaming back to the host or have other modules on the same PCI(e) bus section of the chassis.
To continuously transfer large amounts of data, the entire PXI Express system must be designed with sufficient data bandwidth. The bandwidth between the PXI Express chassis and host computer can limit data throughput. When selecting a PXI Express embedded controller or MXI, confirm the controller has sufficient bandwidth to handle the needs of the entire system. This means checking the per-slot bandwidth (for example, at least 500 MB/s slot bandwidth) and a sufficient system bandwidth to operate all devices installed in the system. If a PXI Express remote controller is selected, it should use a x4 or higher connection. A x1 PXI Express remote controller does not have sufficient bandwidth to allow all channels to acquire/generate at maximum rates.
Example – Data Bandwidth for a Single PXIe-6386/96
The PXIe-6386 and PXIe-6396 use a Gen 1 x1 PXI Express connection to the PXI Express chassis backplane. These modules will operate with a x1 connection, but the amount of data that can be sent to/from the host computer will be limited. Tasks using multiple channels at higher sample rates may experience buffer errors.
Maximum streaming rates:
- PXIe-6396: 15MS/s x 8 ch x 32-bits = 480MB/s
- PXIe-6386: 15MS/s x 8 ch x 16-bits = 240MB/s
In this example, with a Gen 1 PXI Express x1 connection and the theoretical bandwidth specification of 250MB/s, you are likely going to hit buffer overflow errors at maximum sampling rates, when streaming with either module. However, in general, the application areas that would require 15MS/s do not require continuous streaming.
Example – Data Bandwidth for Multiple Cards on a PXI segment
For example, a NI PXIe-1078 chassis, the backplane architecture of which is pictured below, only has Gen-1 x1 PCI Express links from each slot. Slots 5-9 have Gen-1 x1 links back through a PCIe switch which has a Gen-1 x4 link to the controller. In either case, the actual link bandwidth of a slot is limited to 250MB/s, with possible further limitations on slots 5-9 depending on other modules sharing that PCIe switch. So, if you had 5 PXIe-6386 modules in slots 5-9, all sampling at 12.5 MS/s, that would add up to be 1GB/s of data (200MB/s/module) all needing to be streamed back to the host, through the Gen 1 x4 link to the controller slot from the 5-9 slots. This would max out the theoretical limit of that link back to the controller, even if not streaming at maximum rates on all modules. The same limitation is true if you were combining several different modules on that PXI segment, all either streaming data to or from the controller slot.
Figure 1: PXIe-1078 Backplane Architecture
In contrast, the NI PXIe-1085 (24GB/s variant) has Gen-3 x8 PCI Express links from each slot to the Gen-3 PCIe Switches, which have Gen-3 x8 and Gen-3 x16 connections back to the controller slot. Choosing a chassis with this high of throughput, will allow for plenty of headroom for other modules or multiple PXIe-6386's or PXIe 6396's running through that PCIe switch from the same bus segment.
Figure 2: PXIe-1085 Backplane Architecture
Please note: The PXIe-6386 and PXIe-6396 will still be limited by their own Gen 1 x1 PXIe link to the backplane even on a chassis with more than enough bandwidth, so streaming at maximum sample rates on these modules will still likely cause buffer overflows.