The shortest pulse duration which you can transmit is going to be
a) the sampling rate of the converters
b) the USB interface
c) the bandwidth of IF/RF components
I don’t know your exact setup. So, let me provide an example of what
I transmit and receive in an always-on fashion using a single USRP in 4
Byte/sample mode (2 for real, 2 for complex) Therefore, for each sample
that must be transmitted and received, 8 bytes will traverse the USB (4
for Tx, 4 for Rx). The USRP is limited to 32 MB/s across the USB.
Therefore, I can only handle signals 4 MHz wide. Because the USRP does
complex sampling, 4 MHz becomes the maximum sampling rate I can use to
generate my signal at baseband. (This signal will be interpolated to
128 MHz on the USRP.) Because the fastest I can generate samples is 4
MHz, the smallest interval between samples is 1/4e6 = 250 ns. That is
(theoretically) the shortest pulse width.
Now, anytime you signal using one sample you will suffer more system
distortion than if you used, say, two. This is because the converters
will act as a really wide low-pass filter. However, with that said, I
am able to do it. I believe the minimum interpolation factor is 16.
Therefore, in a transmit-only mode, I believe the minimum pulse width
would be 1/8MHz = 125 ns.
I haven’t had coffee yet to day. So, caveat emptor on these
calculations, but I hope they help.