Intersymbol Distortion

[AndyCN]

Okay - so I've been running a few tests.

I suspect that the serialiser might not be sending the correction signals in the right way.
/I and /Q should occur slightly ahead of I and Q.

The result is that an IQ plot shows a spread of points and the spectrum appears to have excessive spurs which I can't cancel.

IQ Plot showing the Output after the Digilite is turned on ( presumably a PRBS)


IQ Plot Showing USB Mode


As far as I can understand from the F4DAY original article - in which he shows a system clock of 8.192Mhz , and a symbol rate of 1.024MBaud - I think the correction signal should occur one quarter of the symbol time ahead of the main signal ( if that makes sense ).

I also note that the serialiser appears 'bursty' - i.e. if I set it to output USB only - it occasionally sends an unmodulated carrier. ( approximately every 3 or 4 seconds ).

Anyone dispute / agree ?

[G4EWJ]

The serialiser does output a pseudo random sequence after a reset. It repeats every 4096 bits, so it doesn't give a very good spectrum plot, but it's fine for a constellation plot.

The original F4DAY article calls for the inverted, weighted value of the following I or Q to be added to the current I or Q.

"To correct the total response of the channel filter (5th order Butterworth filter in the transmitter, plus root raised cosine filter in the receiver), it is necessary to create and apply to the QPSK modulator the dual pre-distortion signals I* and Q* which minimize the inter-symbol interference. After in-depth studies of this phenomenon, we see that the correction signals I* and Q* must be precursory signals with levels reversed compared to their references I and Q with an advance of one symbol duration."

http://jf.fourcadier.pagesperso-orange. ... iter_e.htm

I've only recently noticed the carrier burst in test mode. I'm still looking into it. I think if you stop DigiLite Transmit running, the effect goes away, but I'll check that. The serialiser stops outputting data for a brief period when it re-starts or when its settings are changed. I'm working on a digital Nyquist filter for DigiLite and it uses this break in output to resynchronise itself to the I and Q streams. The serialiser must incorrectly think that its settings have been changed every few seconds.

You may also notice that the serialiser will not come out of test mode unless there is a active recording file for DigiLite Transmit to process.

Brian