Forums

Sampling Time Difference in OFDM Demodulator

Started by insecuritate October 24, 2006
Dear all,

in my diploma thesis i am realizing an ofdm modem. i am working with two
adsp bf533 ez-kit-lites. one as transmitter and one as receiver.
for d/a and a/d conversion i use the implemented audio codec. this works
with nominally 48khz.
and here is my problem. these 48khz are only a nominally value. the "real"
values differ slightly.
after receiving my ofdm signal and detecting the start of a symbol i have
to do the fft to demodulate my subcarriers. but due to the sampling time
offset my fft vektor is distorted (because of the slightly faster / slower
sampling i don't have the original frame in my N-FFT).
I can't change the sampling rate of the codec.

Hope you can help and offer me a way to adjust this sampling time offset
per software... possible solutions?!

regards,

thomas pilgrim


insecuritate wrote:
> Dear all, > > [...] > > Hope you can help and offer me a way to adjust this sampling time offset > per software... possible solutions?! > > regards, > > thomas pilgrim
Hi! This paper is a good introduction to the topic, and also contains references for further study. Best regard, Trond Danielsen
It is not enough to do symbol synchronization. You must also perform
sample synchronization, meaning, you want to match the receiver's
sample clock to the transmitters sample clock. Otherwise you will see
your symbol frames slowly fall out of sync.

You can do this digitally by calculating a sample timing phase offset
in the data and using it to resample or interpolate the original
samples.

-V

Trond Danielsen wrote:
> > Hi! > > This paper is a good introduction to the topic, and also contains > references for further study. > > Best regard, > Trond Danielsen
Hi Trond, Is that paper you mentioned downloadable somewhere? Cheers, Guenter
>It is not enough to do symbol synchronization. You must also perform >sample synchronization, meaning, you want to match the receiver's >sample clock to the transmitters sample clock. Otherwise you will see >your symbol frames slowly fall out of sync. > >You can do this digitally by calculating a sample timing phase offset >in the data and using it to resample or interpolate the original
A few weeks later i have implemented synchronization algorithms to detect start-of-frame by performing maximum-liklihood-estimation and phase correction using pilot symbols. this works very well. Now i have the problem mentioned above: "My symbol frames slowly fall out of sync" and i don't have any idea to cope with this problem because the sampling frequency of my DAC/ADC is fixed, i can't modify it. are there common approaches for this problem?! hope you can help. regards again, thomas
> > Now i have the problem mentioned above: "My symbol frames slowly fall out > of sync" and i don't have any idea to cope with this problem because the > sampling frequency of my DAC/ADC is fixed, i can't modify it. > are there common approaches for this problem?! > > hope you can help. > > regards again, > > thomas
Synchronous sampling is when you can adjust the rate of the sample clock to match that of the transmitter. Asynchronous sampling is when the sample clock is at a fixed rate--the condition you have described. In this case you need to re-sample or interpolate the original samples to match the sample rate of the transmitter. Before you can do that you need to determine the sample-period offset error (the difference between the receivers local concept of the sample rate and the transmitters remote concept of the sample rate, neither of which are exactly the sample rate). Measuring this error is the job of the Timing Error Detector (TED) which feeds back to the Timing Error Corrector (TEC). The TEC is some form of interpolator. The TED is one of a number of techniques. Sometimes there is a pilot tone to facilitate TED; other techniques use non-linear functions or pilotless techniques. I recommend a book: "Synchronization Techniques for Digital Receivers" by Mengali and D'Andrea. There are also a number of papers on timing recovery for MCM and OFDM. Good luck. -V