Re: [Discuss-gnuradio] Question regarding correlating output and IQ samples in a reciever

[Marcus Müller]

Hi Abhinav,

On 01/27/2016 03:17 AM, abhinav narain wrote:

Hi Marcus,

I want the exact end of the preambles IQ samples, so that I can chop

it off from my trace and also find the exact locations where my bits

(transmitted sparsely) are. I want to calculate the SNR for the

decoder. So, if I know the samples exactly corresponding to data bits,

That "exact sample position" cannot exist, because if it did, you wouldn't need timing recovery.
So for example, your timing recovery might implicitely figure out (it won't give you that number, usually; you'd have to integrate over the timing error estimate to get an actual timing offset estimate) that your symbols have duration of 2.236472 samples and start with a sample offset of 114.060072; I don't see how knowing that would map the end of a preamble to any "exact" sample.

I can extract them and apply the S+N/N = E[noise +signal]/ E[noise]

equation.

I am transmitting preamble at double the amplitude than the actual

bits due to my experiment requirements (more fidelity at the

receiver), so I can't calculate the SNR over the preamble, hence I

will do it over samples corresponding to bit 1 pulses in the trace

corresponding to data transmission.

So what's your modulation scheme?

Generally, is only really a useful measure if you either have

1. a constant power modulation (e.g. PSK), or
   
2. whiten your over-the-air bits sufficiently (using coding), so that for (stochastically speaking almost any) data sequence, the signal power is the same.

Usually, even when you have 1., you do 2..

If you measure SNR based on a subset of bits that actually have energy, you're biasing your measurement, and it won't have any meaning for your real transmission, unless you've done the math and know that these bits represent a certain percentage of the power of an actual transmission. But as soon as you do the math, and figure out that some bits have higher Energy per bit , you'd try to minimize the average , so that you can just increase the signal amplitude without breaking your specifications, and get a higher average SNR; this implies spreading the energy over different bits more evenly, and that just leads to the whitening mentioned above.

Best regards,
Marcus

Thank you for your replies,

Abhinav

On Tue, Jan 26, 2016 at 3:12 AM, Marcus Müller <address@hidden> wrote:

Hi Abhinav,

point of timing recovery like you do it is that there's no IQ samples that *exactly and reliably* correspond to the symbol timing; that's why you need timing recovery. The question here is: what do you actually need? The IQ samples before timing recovery aren't useful to the decoder.

Just to assure you: AGC2 doesn't modify the number of samples, it just multiplies with a factor that changes so that average power stays the same.

Best regards,
Marcus

On 01/25/2016 06:26 PM, abhinav narain wrote:

Hi All,

My problem is that above calculation sort of works, and isn't it exact.

I have the following preable of 32 bits [1,0,1,0,1,1,0,0,1,1,0,1,1,1,0,1,1,0,1,0,0,1,1,0,0,1,1,1,0,1,0,1].

This can be seen in badpreamble.png, where for some reason I have trailing 0s or noise and the

start of preamble is missing.

If I shift the position to 8 bits lag in the bit file and then plot(8shiftedpreamble.png), the corresponding

IQ samples using above logic, it is almost right (alhough sitll not right.

Any suggestions how to get the exact IQ samples ?

Thanks,

Abhinav

On Sun, Jan 24, 2016 at 11:03 AM, abhinav narain <address@hidden> wrote:

Hi all,

This might be too simple, but I wanted to confirm as I am not sure how the

blocks manipulate the IQ samples and how AGC works, does it add more

samples for compensating the noise average or not.

I have a flowgraph(image attached) which dumps the IQ samples and the bits after using Polyphase Clock Sync. I want to find exact IQ samples corresponding to

a specific preamble in the bitsequence I have obtained in the output

file.

Let say the starting position of sample is= P, the sample rate=S ksps, samples \

per symbol=Y.

Is it correct to assume that the index of iq samples corresponding to

preamble in the IQ file will be at index P*Y ?

If preamble size is 32 bits then there will be 32*Y of them ?

Thanks,

Abhinav

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio