I / Q delay mismatch

In helping debug a QPSK receiver, I noticed that at the output of the
matched filter the I and Q channels are delayed by different amounts.
Is this what is called "I/Q imbalance"?  I have always thought that it
meant the I and Q mixers are not orthogonal to each other, or that the
amplitudes are not the same.

I would also love to hear about what may cause this phenomena.  I
thought that non-constant group delay may be the issue, but how does
it effect the I and Q channels so that one is delayed versus the
other?

Thanks in advance,
Julius

Hi Julius,

"julius" <juliusk@gmail.com> writes:

> In helping debug a QPSK receiver, I noticed that at the output of the
> matched filter the I and Q channels are delayed by different amounts.
> Is this what is called "I/Q imbalance"?  I have always thought that it
> meant the I and Q mixers are not orthogonal to each other, or that the
> amplitudes are not the same.

Yes, I think that's true. I.e., I would not call this "I/Q imbalance."
I/Q imbalance is memoryless. 

> I would also love to hear about what may cause this phenomena.  

Are you sure the problem is in the receiver? 
-- 
%  Randy Yates                  % "I met someone who looks alot like you,
%% Fuquay-Varina, NC            %             she does the things you do, 
%%% 919-577-9882                %                     but she is an IBM."
%%%% <yates@ieee.org>           %        'Yours Truly, 2095', *Time*, ELO   
http://home.earthlink.net/~yatescr

On 8 Mar, 19:01, "julius" <juli...@gmail.com> wrote:
> In helping debug a QPSK receiver, I noticed that at the output of the
> matched filter the I and Q channels are delayed by different amounts.
> Is this what is called "I/Q imbalance"?  I have always thought that it
> meant the I and Q mixers are not orthogonal to each other, or that the
> amplitudes are not the same.
>
> I would also love to hear about what may cause this phenomena.  I
> thought that non-constant group delay may be the issue, but how does
> it effect the I and Q channels so that one is delayed versus the
> other?

I don't have any experience with I/Q systems,
but off the top of my head, I would agree that
group delay might be the culprit.

As you know, group delay is defined as

g = -dphi/dw

meaning that two different filters must have
similar phase spectra in order to exhibit the
same group delays. The only filters where
that is guaranteed to happen, are symmetric
FIR filters of the same length.

Since you have two(?) matched filters, one
for the I channel and one for the Q channel,
these are almost certainly not symmetric
FIR filters, and hence have different
phase spectra, and so also different group
delays.

Provided this line of reasoning make some
sense -- and I am not at all sure it does --
I would suggest appending all-pass filters
on both channels with phase responses tailored
to equalize the total phase responses of the
two channels.

My NOK 0.02.

Rune

On Mar 8, 6:19 pm, "Rune Allnor" <all...@tele.ntnu.no> wrote:
> On 8 Mar, 19:01, "julius" <juli...@gmail.com> wrote:
>
> > In helping debug a QPSK receiver, I noticed that at the output of the
> > matched filter the I and Q channels are delayed by different amounts.
> > Is this what is called "I/Q imbalance"?  I have always thought that it
> > meant the I and Q mixers are not orthogonal to each other, or that the
> > amplitudes are not the same.
>
> > I would also love to hear about what may cause this phenomena.  I
> > thought that non-constant group delay may be the issue, but how does
> > it effect the I and Q channels so that one is delayed versus the
> > other?
>
> Since you have two(?) matched filters, one
> for the I channel and one for the Q channel,
> these are almost certainly not symmetric
> FIR filters, and hence have different
> phase spectra, and so also different group
> delays.

I would imagine that if this is being in done in DSP, then the filters
should be identical (and probably symmetric) FIRs; there's no reason
that they should have been designed differently from each other.


--
Oli

On Mar 8, 12:15 pm, Randy Yates <y...@ieee.org> wrote:

> Are you sure the problem is in the receiver?

Hi Randy, I was at first convinced that this can't be
a problem at the receiver.  But both TX and RX are direct-
synthesis, no IF, so now I am puzzled!

Thanks,
Julius

On Mar 8, 12:19 pm, "Rune Allnor" <all...@tele.ntnu.no> wrote:

> I don't have any experience with I/Q systems,
> but off the top of my head, I would agree that
> group delay might be the culprit.
>
> As you know, group delay is defined as
>
> g = -dphi/dw
>
> meaning that two different filters must have
> similar phase spectra in order to exhibit the
> same group delays. The only filters where
> that is guaranteed to happen, are symmetric
> FIR filters of the same length.
>

Hi Rune, thanks for the reply.  I thought about this, too.
Let's say that we have x(t) = x_I(t) + j*x_Q(t) to represent
the complex baseband.

Now, we know that X_I(f) = even(X_f) and X_Q(f) = odd(X_f)
are the appropriate Fourier transforms.

Suppose now that I have a "channel" that is
  H(f) = \exp(j \phi(f))
to represent the delay terms.

Then this will effect X_I(f) and X_Q(f) identically.  So
unfortunately this doesn't explain why the I and Q channels
are delayed by different amounts.  :-(

Thanks,
Julius

On Mar 8, 12:37 pm, "Oli Charlesworth" <c...@olifilth.co.uk> wrote:
>
> I would imagine that if this is being in done in DSP, then the filters
> should be identical (and probably symmetric) FIRs; there's no reason
> that they should have been designed differently from each other.
>
> --
> Oli

Hi Oli, you are correct that this is done digitally, in Matlab
actually.  I just read the ADC output and did the mixing
and match filtering in Matlab.  I thought about the possible
effect of jitter in the ADC, or when writing the data into
the buffer for Matlab, but since this is the *modulated*
version, then a delay will look rather obvious in the eye
diagram, I thought.

Thanks,
Julius

"julius" <juliusk@gmail.com> wrote in message 
news:1173379347.923060.209270@c51g2000cwc.googlegroups.com...
> On Mar 8, 12:19 pm, "Rune Allnor" <all...@tele.ntnu.no> wrote:
>
>> I don't have any experience with I/Q systems,
>> but off the top of my head, I would agree that
>> group delay might be the culprit.
>>
>> As you know, group delay is defined as
>>
>> g = -dphi/dw
>>
>> meaning that two different filters must have
>> similar phase spectra in order to exhibit the
>> same group delays. The only filters where
>> that is guaranteed to happen, are symmetric
>> FIR filters of the same length.
>>
>
> Hi Rune, thanks for the reply.  I thought about this, too.
> Let's say that we have x(t) = x_I(t) + j*x_Q(t) to represent
> the complex baseband.
>
> Now, we know that X_I(f) = even(X_f) and X_Q(f) = odd(X_f)
> are the appropriate Fourier transforms.
>
> Suppose now that I have a "channel" that is
>  H(f) = \exp(j \phi(f))
> to represent the delay terms.
>
> Then this will effect X_I(f) and X_Q(f) identically.  So
> unfortunately this doesn't explain why the I and Q channels
> are delayed by different amounts.  :-(
>
> Thanks,
> Julius

If the I and Q channels have different spectral content and if the channel 
isn't linear phase (constant delay) then the channels might be delayed 
differently.  Not that this is what's going on ......
That is, if the group delay isn't flat.

Fred 

julius wrote:

> In helping debug a QPSK receiver, I noticed that at the output of the
> matched filter the I and Q channels are delayed by different amounts.
> Is this what is called "I/Q imbalance"? 

It is called the screwed up PLL.

> I have always thought that it
> meant the I and Q mixers are not orthogonal to each other, or that the
> amplitudes are not the same.

There is no need to think. Extra knowledge is only harmfull.

> I would also love to hear about what may cause this phenomena.

As usual: lousy programming, bad or broken hardware or both.

> I
> thought that non-constant group delay may be the issue, but how does
> it effect the I and Q channels so that one is delayed versus the
> other?

Don't think about that. Fix the PLL first.



Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

On Mar 8, 1:01 pm, "julius" <juli...@gmail.com> wrote:
> In helping debug a QPSK receiver, I noticed that at the output of the
> matched filter the I and Q channels are delayed by different amounts.
> Is this what is called "I/Q imbalance"?  I have always thought that it
> meant the I and Q mixers are not orthogonal to each other, or that the
> amplitudes are not the same.
>
> I would also love to hear about what may cause this phenomena.  I
> thought that non-constant group delay may be the issue, but how does
> it effect the I and Q channels so that one is delayed versus the
> other?
>
> Thanks in advance,
> Julius

This sounds like an implementation error of some kind. Gain and phase
imbalance don't cause delay differences. How much delay is there?
Could
it be the transmitter is modulating I(t) and Q(t-d) or vice versa?

John