Forums

qpsk demodulation questions

Started by Unknown January 13, 2008
Hi,

I'm currently working on a QPSK demodulator, but I'm seeing some
things that don't look right, and I wanted to run them by some of the
people in this group. Is there a site where I can post up some
screenshots to show you what I'm seeing

1...The recovered signals on the I and Q branches are the inverse of
the actual transmitted data. A transmitted '1' is received as a zero.
Is this expected?

2...Both the recovered I and Q signals are very similar to the
summation of the I and Q channels in the modulator. I would have
expected the received I data  to look like the transmitted I data and
similarly for Q.

3...The signals on the I and Q branches are not clean, After mixing
and low pass filtering to remove the carrier, the signals still
contain spikes which are causing errors in the data. I read about
using a matched filter/averaging filter to clean them up. What about
an RRC at the output of the modulator and one on each of the
demodulator I and Q branches prior to the mixer....any suggestions?

4...Is any transmitted data normally scrambled with a PN code to
improve timing? The reason I'm asking is my TXed data contains strngs
of continuous ones and zeros. This is causing the PLL to lose lock. Is
it more usual to send a preamble sequence. The only problem I can see
with the preamble is that it would have to be continually transmitted.
Anybody been down this road before, I'd appreciate your advice.

Many Thanks
Bob

On Jan 13, 7:06 pm, sten...@yahoo.com wrote:
> Hi, > > I'm currently working on a QPSK demodulator, but I'm seeing some > things that don't look right, and I wanted to run them by some of the > people in this group. Is there a site where I can post up some > screenshots to show you what I'm seeing > > 1...The recovered signals on the I and Q branches are the inverse of > the actual transmitted data. A transmitted '1' is received as a zero. > Is this expected?
There is a 4 way ambiguity in the recovered phase, so this is possible, as are other combinations. You can use a preamble or differential encoding to resolve.
> > 2...Both the recovered I and Q signals are very similar to the > summation of the I and Q channels in the modulator. I would have > expected the received I data to look like the transmitted I data and > similarly for Q.
One would expect them to look like I and Q in the modulator, allowing for phase ambiguity.
> > 3...The signals on the I and Q branches are not clean, After mixing > and low pass filtering to remove the carrier, the signals still > contain spikes which are causing errors in the data. I read about > using a matched filter/averaging filter to clean them up. What about > an RRC at the output of the modulator and one on each of the > demodulator I and Q branches prior to the mixer....any suggestions? >
Match the receiver filter to the transmit pulse shape. If it is rectangular, use a rectangular filter in receiver.
> 4...Is any transmitted data normally scrambled with a PN code to > improve timing? The reason I'm asking is my TXed data contains strngs > of continuous ones and zeros. This is causing the PLL to lose lock. Is > it more usual to send a preamble sequence. The only problem I can see > with the preamble is that it would have to be continually transmitted. > Anybody been down this road before, I'd appreciate your advice. >
Scrambling is common in continuous modems like the old V.XX telephone standards. Preambles are used in burst modems. This should not affect the carrier PLL though. John
On 14 Jan, 10:55, John <sampson...@gmail.com> wrote:
> On Jan 13, 7:06 pm, sten...@yahoo.com wrote: > > > Hi, > > > I'm currently working on a QPSK demodulator, but I'm seeing some > > things that don't look right, and I wanted to run them by some of the > > people in this group. Is there a site where I can post up some > > screenshots to show you what I'm seeing > > > 1...The recovered signals on the I and Q branches are the inverse of > > the actual transmitted data. A transmitted '1' is received as a zero. > > Is this expected? > > There is a 4 way ambiguity in the recovered phase, so this is > possible, as are other combinations. You can use a preamble or > differential encoding to resolve. > > > > > 2...Both the recovered I and Q signals are very similar to the > > summation of the I and Q channels in the modulator. I would have > > expected the received I data &#2013266080;to look like the transmitted I data and > > similarly for Q. > > One would expect them to look like I and Q in the modulator, allowing > for phase ambiguity. > > > > > 3...The signals on the I and Q branches are not clean, After mixing > > and low pass filtering to remove the carrier, the signals still > > contain spikes which are causing errors in the data. I read about > > using a matched filter/averaging filter to clean them up. What about > > an RRC at the output of the modulator and one on each of the > > demodulator I and Q branches prior to the mixer....any suggestions? > > Match the receiver filter to the transmit pulse shape. If it is > rectangular, use a rectangular filter in receiver. > > > 4...Is any transmitted data normally scrambled with a PN code to > > improve timing? The reason I'm asking is my TXed data contains strngs > > of continuous ones and zeros. This is causing the PLL to lose lock. Is > > it more usual to send a preamble sequence. The only problem I can see > > with the preamble is that it would have to be continually transmitted. > > Anybody been down this road before, I'd appreciate your advice. > > Scrambling is common in continuous modems like the old V.XX telephone > standards. Preambles are used in burst modems. This should not affect > the carrier PLL though. > > John
Hi John, Thanks for the info. I'm going to apply some matched filters to try and clean up the waveforms. I'm also modifying the PLL to help it track better., so I'll see how these turn out. Many thanks for replying so quickly. Regards Bob
On Sun, 13 Jan 2008 16:06:11 -0800 (PST), stenasc@yahoo.com wrote:

> >4...Is any transmitted data normally scrambled with a PN code to >improve timing? The reason I'm asking is my TXed data contains strngs >of continuous ones and zeros. This is causing the PLL to lose lock. Is >it more usual to send a preamble sequence. The only problem I can see >with the preamble is that it would have to be continually transmitted. >Anybody been down this road before, I'd appreciate your advice. > >Many Thanks >Bob
Yes, generally a scrambler is used to ensure enough transitions so that the clock recovery loop doesn't lose lock due to a static pattern. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
On Jan 13, 6:06 pm, sten...@yahoo.com wrote:

> 3...The signals on the I and Q branches are not clean, After mixing > and low pass filtering to remove the carrier, the signals still > contain spikes which are causing errors in the data.
How are you determining the carrier input to your mixer? Is it being generated coherently?