[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Discuss-gnuradio] gr-ieee-802-11 Transceiver (was: gr-ieee802_11 Tr
Re: [Discuss-gnuradio] gr-ieee-802-11 Transceiver (was: gr-ieee802_11 Transceiver)
Sat, 11 Feb 2017 15:24:06 +0100
Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.2.0
what's wrong with simply putting a file sink in parallel with the USRP
sink in Wifi_tx.grc?
For correctly received packets, couldn't you just use the same
gr-ieee802-11 blocks to recreate the original packet in the receiver,
and then compare?
Just a bit of food for thought, not criticism as is: Also, EVM is a
relatively involved, as I find it very hard to define what "error" is in
an OFDM system. Is a phase offset on a single carrier part of the error?
I think not – phase is a function of path length, even in a no-noise
scenario, so, the receiver needs to account for that, anyway, and hence,
phase is a fact, but not an error.
In single-carrier systems, that might be easy to account for – simply
define the Error Vector to be taken after phase recovery instead of
before. Now, in OFDM systems, things get a whole lot more complicated –
do you take the error vectors of each subcarrier individually, and sum
them? Average them? RMS them? Pre- or post equalizer? Post-equalizer
makes more sense from a comm systems point of view (see the single
carrier example), but is totally unfair, because the "error reduction"
is not only an effect of how well the equalizer for each subcarrier can
correct a phase offset, but also of the quality of the phase estimate of
the overall OFDM bandwidth. Mathematically defending a specific type of
EVM as a relevant metric for the behaviour of the system isn't a trivial
task – in fact, systems where link quality is a commercial factor,
namely LTE, introduce a whole set of measurements that implement
different means of describing the error in a transceiver.
In the end, you could probably sum up the EVMs of the individual symbol
value in the equalizer (in lib/equalizer/*.cc) just before
"mod->map_to_points". I'd argue that a high value here is probably a
sign of a "difficult channel" for exactly the type of equalizer you use
(Comb, LS, LMS, STA), but that is not necessarily a problem – that's the
good ol' estimator problem: Cramér-Rao bound limits the
variance-reducing performance with an estimator which coverges on the
correct symbol (i.e. is unbiased), but for biased estimators, you might
get better; yet, you'd have a non-zero EVM even under perfect channels.
On 02/11/2017 01:09 PM, serefay wrote:
> Dear all,
> I am using Bastian Bloessl's gr-ieee802_11 OFDM Transceiver for my
> Thesis. I have used the Rx to captured the WiFi frames and saved them in
> a file and also Pcap file of them. Now I want to compare the frames that
> I received with the Original ones in order to check the change in
> transmitted signal. I wanted to use the frames I have captured and use
> TX to receive the original frames possibly. My aim is to compare the
> received one with the original signal in order to calculate EVM. Can
> anyone please help me? how should I use the wifi_tx.grc. How can I
> connect the set up flow graph to transmit Data Bytes of a frame to the
> transmitter in order to get the original data frame.Or is there any
> better way to do that.
> Any help would be appreciated.
> Kind Regards
> View this message in context:
> Sent from the GnuRadio mailing list archive at Nabble.com.
> Discuss-gnuradio mailing list