IQ AM DEMOD WITHOUT CARRIER SYNC?

Hello All

I have an internal SDR board which I'm hoping to re-use for demodulating
an AM data stream which is manchester bi-phase encoded data at 1Mbps.
However, despite understanding the basics, I'm not sure whether it is an
appropriate board. 

I would greatly appeaciate any experience you all may have in this. I dont
want to spend ages finding out Ive done something fundamentally difficult
because of my hardware, when it would be easier in the long run to
re-design the hardware!

I have a tuner/front end which outputs IQ which is then digitised.  The
setup can give me IQ ZIF at up to 8MHz bandwidth, or a signal centred on
2MHz and single ended. The output goes over USB to be processed by the PC.

Now I'm thinking theres 3 ways I could go about this:

1) I could set ZIF mode and the LO so that the signal appears at say 7M
within the 8M zif BW, then try to do an AM detect on the 7M carrier to get
the 1M envelope (which seems near the limit to me, would much rather have a
higher carrier to do detection on)

2) I could set ZIF mode and set the LO to match the carrier.  Then do a
magnitude (SQRT(I*I+Q*Q)).  But it seems to me that lack of carrier phase
lock would make this a poor choice? My hardware is fixed and if I were to
do a proper product detector with two paths and 90Deg phase shift etc it
would probably be more work than diode detecting a higher IF.

3) Centre the signal around 2M single ended and then process it using a
product detector with the full (phase error cancelling) setup on the PC. 
Don't really want to try this with a data rate of 1M as Ive previously
used the module I have for FM radio reception single ended with IQ creation
on the PC (yes I know I should have used both I&Q from the hardware but it
was a test), and I found that sample rate of 2Msamples/sec would pretty
much max out the ability of my 2.5G dual core machine doing broadcast FM
demod this way. So I'm not comfortable that the average machine is going to
be happy with demod of the 1M data stream this way.

So the question is, would I just be using the wrong hardware if I use the
existing board, should I dump it and start again, or is there a sensible
route?  

Many Thanks
Andy

On Mon, 06 Jul 2009 09:59:10 -0500, andyt wrote:

> Hello All
> 
> I have an internal SDR board which I'm hoping to re-use for demodulating
> an AM data stream which is manchester bi-phase encoded data at 1Mbps.
> However, despite understanding the basics, I'm not sure whether it is an
> appropriate board.
> 
> I would greatly appeaciate any experience you all may have in this. I
> dont want to spend ages finding out Ive done something fundamentally
> difficult because of my hardware, when it would be easier in the long
> run to re-design the hardware!
> 
> I have a tuner/front end which outputs IQ which is then digitised.  The
> setup can give me IQ ZIF at up to 8MHz bandwidth, or a signal centred on
> 2MHz and single ended. The output goes over USB to be processed by the
> PC.
> 
> Now I'm thinking theres 3 ways I could go about this:
> 
> 1) I could set ZIF mode and the LO so that the signal appears at say 7M
> within the 8M zif BW, then try to do an AM detect on the 7M carrier to
> get the 1M envelope (which seems near the limit to me, would much rather
> have a higher carrier to do detection on)

4MHz would probably be better, to chop the sidebands equally.  Manchester 
encoding doesn't need much bandwidth above the bit rate if you're really 
careful.

> 2) I could set ZIF mode and set the LO to match the carrier.  Then do a
> magnitude (SQRT(I*I+Q*Q)).  But it seems to me that lack of carrier
> phase lock would make this a poor choice? My hardware is fixed and if I
> were to do a proper product detector with two paths and 90Deg phase
> shift etc it would probably be more work than diode detecting a higher
> IF.

If you don't have much DC offset in your ADCs compared to the signal 
strength this should work well, particularly if you use a fast magnitude 
approximation.

> 3) Centre the signal around 2M single ended and then process it using a
> product detector with the full (phase error cancelling) setup on the PC.
> Don't really want to try this with a data rate of 1M as Ive previously
> used the module I have for FM radio reception single ended with IQ
> creation on the PC (yes I know I should have used both I&Q from the
> hardware but it was a test), and I found that sample rate of
> 2Msamples/sec would pretty much max out the ability of my 2.5G dual core
> machine doing broadcast FM demod this way. So I'm not comfortable that
> the average machine is going to be happy with demod of the 1M data
> stream this way.

If bandwidth is a problem, why are you contemplating the 4MHz solution?

> So the question is, would I just be using the wrong hardware if I use
> the existing board, should I dump it and start again, or is there a
> sensible route?

I think that using a PC for this is an oddball choice.  If the SNR is 
high then it should be very easy to build an AM receiver followed by a 
detector.  If the SNR is low, it's vital that you get the data, and 
you're not doing much with it after capture then a PC may do the job -- 
but it won't be easy.

-- 
www.wescottdesign.com

>On Mon, 06 Jul 2009 09:59:10 -0500, andyt wrote:
>
>> Hello All
>> 
>> I have an internal SDR board which I'm hoping to re-use for
demodulating
>> an AM data stream which is manchester bi-phase encoded data at 1Mbps.
>> However, despite understanding the basics, I'm not sure whether it is
an
>> appropriate board.
>> 
>> I would greatly appeaciate any experience you all may have in this. I
>> dont want to spend ages finding out Ive done something fundamentally
>> difficult because of my hardware, when it would be easier in the long
>> run to re-design the hardware!
>> 
>> I have a tuner/front end which outputs IQ which is then digitised. 
The
>> setup can give me IQ ZIF at up to 8MHz bandwidth, or a signal centred
on
>> 2MHz and single ended. The output goes over USB to be processed by the
>> PC.
>> 
>> Now I'm thinking theres 3 ways I could go about this:
>> 
>> 1) I could set ZIF mode and the LO so that the signal appears at say
7M
>> within the 8M zif BW, then try to do an AM detect on the 7M carrier to
>> get the 1M envelope (which seems near the limit to me, would much
rather
>> have a higher carrier to do detection on)
>
>4MHz would probably be better, to chop the sidebands equally.  Manchester

>encoding doesn't need much bandwidth above the bit rate if you're really

>careful.
>
>> 2) I could set ZIF mode and set the LO to match the carrier.  Then do
a
>> magnitude (SQRT(I*I+Q*Q)).  But it seems to me that lack of carrier
>> phase lock would make this a poor choice? My hardware is fixed and if
I
>> were to do a proper product detector with two paths and 90Deg phase
>> shift etc it would probably be more work than diode detecting a higher
>> IF.
>
>If you don't have much DC offset in your ADCs compared to the signal 
>strength this should work well, particularly if you use a fast magnitude

>approximation.
>
>> 3) Centre the signal around 2M single ended and then process it using
a
>> product detector with the full (phase error cancelling) setup on the
PC.
>> Don't really want to try this with a data rate of 1M as Ive previously
>> used the module I have for FM radio reception single ended with IQ
>> creation on the PC (yes I know I should have used both I&Q from the
>> hardware but it was a test), and I found that sample rate of
>> 2Msamples/sec would pretty much max out the ability of my 2.5G dual
core
>> machine doing broadcast FM demod this way. So I'm not comfortable that
>> the average machine is going to be happy with demod of the 1M data
>> stream this way.
>
>If bandwidth is a problem, why are you contemplating the 4MHz solution?
>
>> So the question is, would I just be using the wrong hardware if I use
>> the existing board, should I dump it and start again, or is there a
>> sensible route?
>
>I think that using a PC for this is an oddball choice.  If the SNR is 
>high then it should be very easy to build an AM receiver followed by a 
>detector.  If the SNR is low, it's vital that you get the data, and 
>you're not doing much with it after capture then a PC may do the job -- 
>but it won't be easy.
>
>-- 
>www.wescottdesign.com
>

Thanks yes I forgot to mention something important.
The ZIF method at baseband would use the PC to AM demod via mag method,
however the alternative plan was to use the highest carrier possible (7M)
and do an analog envelope detect (diode and cap), before the ADCs.

So to summarise there is
1) 7M carrier, AM demod the 1M manchester signal via diode and cap (bit on
the edge here)
2) carrier around zero, analogue demod IQ in software on the PC with
potential problem with drift of lo relative to carrier
3) Centre around 2M, as you point out this will probably be a bust on
sampling/processing for my app.  But I could dump this to disk at 4Msamples
no processing, which would allow me to batch test the performance of a more
complex product detect with 90deg phase shift to see if the lo drift in my
front end is significantly degrading performance in 2.

At the end of the day...I'm starting to think I need a front end that will
put out a higher IF and then just do a detect before i jump into heavy
processing on a PC platform

Andy

Tim Wescott <tim@seemywebsite.com> wrote:
(snip)
 
< 4MHz would probably be better, to chop the sidebands equally.  Manchester 
< encoding doesn't need much bandwidth above the bit rate if you're really 
< careful.

I believe that sounds right.  But that is still twice what other
coding (modulation) methods require.  That is, like AM-DSB it is
wasteful in its use of bandwidth.  Also, the spectrum has a large
peak at the bit rate that is important in considering EMI from UTP
cable.

The coding used by fast ethernet has much of its energy at 1/4 
the transition rate (of 125MHz), for about 31MHz.  

-- glen