EVM in OFDM

Measuring the RMS EVM of the underlying modulation in an OFDM signal is
straight forward.  However the accuracy of the EVM appears to be affected
by the number of null tones present in the OFDM FFT.  In other words if
there are no null tones, then the EVM measurement is as expected and as the
number of null tones increases the EVM measurement diverges from expected.

Does anybody know of any literature they could point me to that may help
explain what I'm seeing?  

Thanks...

On Jan 5, 10:45&#4294967295;am, "rbb" <Rory.Buchanan@n_o_s_p_a_m.onsemi.com>
wrote:
> ...
>
> Does anybody know of any literature they could point me to that may help
> explain what I'm seeing? &#4294967295;
>
> Thanks...

Would you care to state what you are seeing?

Dale B. Dalrymple

>
>Would you care to state what you are seeing?
>
>Dale B. Dalrymple
>

The EVM measurement is off depending upon the number of null tones.  There
must be some mathematical way of explaining this offset and as such some
way of compensating for it. 

On Jan 5, 12:21&#4294967295;pm, "rbb" <Rory.Buchanan@n_o_s_p_a_m.onsemi.com>
wrote:
> >Would you care to state what you are seeing?
>
> >Dale B. Dalrymple
>
> The EVM measurement is off depending upon the number of null tones. &#4294967295;There
> must be some mathematical way of explaining this offset and as such some
> way of compensating for it.

"off" is your interpretation of what you are seeing.

Would you care to state what you are seeing? And doing? And what why
you think that is "off".

Dale B. Dalrymple

>
>"off" is your interpretation of what you are seeing.
>
>Would you care to state what you are seeing? And doing? And what why
>you think that is "off".
>
>Dale B. Dalrymple
>

I have an ideal OFDM modulator that modulates a message.  The message is
then passed through an AWGN channel for a given SNR value.  An ideal OFDM
demodulator demodulates the message.  I then measure the EVM of the
post-modulator/pre-ifft transmit signal vs. the post-fft/pre-demodulator
received signal.  

I can then use the above system to sweep an SNR range & measure EVM values.
 If there are 0 null tones in the FFT then the EVM I measured matches what
I would expect given the SNR value.  However as I introduce the null tones
the EVM values do _not_ match as when there are no null tones.  The EVM
measurement is still linear with SNR, but there is an offset from the
_ideal_.  

My suspicion is the IFFT/FFT with null tones is causing the problem, but as
I typed this out I'm thinking maybe my SNR calculations are off in the
channel.

I'm not sure if I explained that very well or not...

On Wed, 05 Jan 2011 15:18:12 -0600, "rbb"
<Rory.Buchanan@n_o_s_p_a_m.onsemi.com> wrote:

>>
>>"off" is your interpretation of what you are seeing.
>>
>>Would you care to state what you are seeing? And doing? And what why
>>you think that is "off".
>>
>>Dale B. Dalrymple
>>
>
>I have an ideal OFDM modulator that modulates a message.  The message is
>then passed through an AWGN channel for a given SNR value.  An ideal OFDM
>demodulator demodulates the message.  I then measure the EVM of the
>post-modulator/pre-ifft transmit signal vs. the post-fft/pre-demodulator
>received signal.  
>
>I can then use the above system to sweep an SNR range & measure EVM values.
> If there are 0 null tones in the FFT then the EVM I measured matches what
>I would expect given the SNR value.  However as I introduce the null tones
>the EVM values do _not_ match as when there are no null tones.  The EVM
>measurement is still linear with SNR, but there is an offset from the
>_ideal_.  
>
>My suspicion is the IFFT/FFT with null tones is causing the problem, but as
>I typed this out I'm thinking maybe my SNR calculations are off in the
>channel.
>
>I'm not sure if I explained that very well or not...
>

I expect it's just that your calculations are off wrt the channel.
Don't count the null subcarriers.   There are several ways to get
tripped up on this.

I'd bet if you measure the error vector of each received constellation
in the populated subcarriers and compute EVM only from that you'll be
back where you need to be.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com