I am evaluating the use of WM3000 for an application where more than 100 nodes will share the time (through a TDMA slotted scheme), each node transmitting 500 bytes every 100ms.
In theory, this leads to an overal thoughtput of 4Mbs (100 nodes * 500 bytes * 8 bits * 10 times/second), which should fit into 6,8Mbps per second throughput indicated in DW3000 datasheet. But it is not clear to me if this 6,8Mbps throughput is available at application level, or if It will be reduced by Read Solomom Forward Error Correction (seems to be 48 bits every 330bits), preamble, sync …
A UWB packet includes a lot more than just the data, the structure is given in section 3 of the user manual and is a preamble of selectable length, a SFD marker, a phy header, the data payload and a checksum. There is also an optional STS marker.
You wish to send 1000 packets per second which gives you 1 ms per packet. With the shortest possible preamble a 500 byte packet has a duration of around 685 us.
This gives you about 300us of space between packets. Assuming your can get your TDMA synchronisation between nodes accurate enough this seems like a reasonable target.
You may however hit range issues depending on what distance you need. Increasing the preamble length will give you more range but the increase is very roughly 1 us per symbol so 256 would be the upper limit before you run out of time.
Also the regulatory power limit is calculated as the average over 1 ms for your worst case transmissions. With very short packets this averaging means you can turn the power up and still remain within the limits. When sending longer packets like you will be using the transmit power has to be lower to maintain the same average, this in turn drops the range and limits the benefits of the longer preamble.
The spread sheet linked in the forum post below can calculate the packet duration for various configuration settings. It was intended for the DW1000 rather than the DW3000 so ignore any power numbers. However unless you enable the STS header the packet structure is the same and so the packet duration parts of the calculation are still good.