The lte Prach channel requires fft at a minimum of 2k resolution. To achieve that some decimation is done. This means the stream will go through some filters then followed by fft. The fft frame is well defined in terms of start/end (after cyclic prefix removal). However I am not sure if one should take into account group delay of filters to start fft frame. The standards does not necessarily describe such implementation.

Any thoughts?

Cheers

Kaz

How is the frame time synchronization performed?   If it does something like the typical cyclic prefix correlation with the end of the frame, and if the signal used for that has gone through the same filter, then what would be the source of any error?

Hi Slartibartfast,

The lte uplink channel goes from radio to cpri then to BBU(baseband unit).

I am working on BBU side.

From cpri the same stream (30.72 Msps for 20MHz lte) branches to (1) main channel of 15KHz carrier resolution and (2) separate PRACH channel with resolution of 1.25KHz. The Prach frame is 30720 samples (cp + sequence + guard time). This is too long and requires 24k fft unless decimated. We decimate by 12 and apply 2k fft giving 1.25KHz resolution. The start/end of cp is defined before decimation, so is start/end of fft sequence. I am just not sure about decimation filters in the chain with regard to sequence start. I can model the two cases (with or without decimation) but would like to hear opinions. 

What I was getting as was that if the signal going into the FFT was the same signal used for determining symbol synchronization, then group delay won't matter because it would be the same for both parts of the system.