modulation

[lavalava]

Do you have an example of how to downcovert an output audio signal from a modem and then demodulate it in real time?  I should plug the audio output from the modem into the audio input port of the computer sound card, right?  And how do I tell Labview to read in from the input audio port?

Also, I'm not sure how to enter some of the parameters in the digital demodulation toolkit. I also need to apply a Nyquist filter.  Could you please provide me with an example? 

Here is an example of the modem spec:
1200 bauds
center frequency at 1400hz, the pulse width is 1200hz

How would I input these information into the modulation toolkit?  I'm using version 3.0 with labview 8.6 developer edition.

thanks,

[Steve_B]

Hi lavalava,

We have an example that you can you to continuously read in a sound wave from your sound card which can be located in the Example Finder (Help>>Find Examples) and then categorized under Hardware Input and Output>>Sound as Continuous Sound Input.vi. This example is set up to continuously read in data from the sound card and you can specify how this is done. If you only have 1 sound card I would recommend you try using Device ID 0 to see if this is how your system needs to be configured. What type of demodulation do you need to perform? I would assume that the signal is an FSK modulated signal. The modulation toolkit is looking for a complex waveform or I and Q data to perform the demodulation, but the waveform that is returned from the sound card is only a real waveform. There may be some additional demodulation options or perhaps we can convert the data to work with the Modulation Toolkit VIs. There are some filtering VIs that can be found on the Functions Palette under Signal Processing>>Filters and you can choose your filter type and its settings.

[lavalava]

thanks Steve,

I'm looking at that examples right now and also all the block diagrams  in signal processing, however, I couln't spot anyone of them that does "root raised cosine" which is a form of Nyquist alorigthm.  Would you know where can I find this block?  And also, how do I manually downcovert the audio signal coming out of the modem to DC?  The modem is putting out at about 1800-3600Hz, so that's in the audible range for the sound card to handle.

I've also used the downcoverter block from modulation toolkit but that was quite useless, I couldn't goup the constellation together quite well.  I need the error vector magnitude lesser than 10% rms value but the way it's showing is around 25% which is unacceptable.  This is why I would like to manually downconvert it myself to see if there is a way to work around it.

Thanks 

[Steve_B]

Hi lavalava,

We do have the MT Generate Filter Coefficients Vi that can be uwsed to generate the coefficients necessary for a raised cosine filter. You can use these coefficients as an input to the MT Demodulate ... VIs. This is the method we use in our examples to implement a raised cosine filter. It would be helpful if you were able to post some recorded data to the forum along with the modulation scheme and the expected data. The demodulation VI that you were using would allow you to look at the constellation diagram and you can double click on the VI and look at its block diagram to see the math behind the function. Is it possible that the errors you are seeing are quantization errors due to the sampling through the sound card?

[lavalava]

Hello, Steve:

Here are the two sample files:

sample22kHz was recorded using a SigmaTel Audio sound card built into the Dell System

sample44kHz was recorded using a Ediro UA-20 Audio processor (USB)

[lavalava]

Audio Sample 2 file is the sample using what you suggested by using MT to create root raised cosine coefficents and capturing in continuous mode.  This file was based off one of the example that came with the toolkit, I modified to take in the audio input -> downconvert -> feed it into MT adaptive equalizer using LMS method, I couldn't get any good results with this one.  

Audio Sample 3 file usings MT down convert then plot the constellation.  When I pick filter, it automatically guess and apply the best filter.  The vector error magnitude(EVM) is appearing to be around 25%.  When I used DFD Raised Cosine Design filter from DFD toolkit to filter the audio input first then feed the results into the MT downconvert, it improves very little but the end result EVM still appears to be around 25% rms.

Not sure what I did was wrong, normally we use an old 1990 Agilent 89410A vector signal analyzer.   Set it to Digital Demodulation -> any PSK/QAM -> Baud/Symbol Rate: 2400 -> Root Raised Cosine with a roll off factor of 0.25 -> Carrier/Center Frequency: 1800Hz -> Bandwidth: 3000Hz.  And it plots the constellation with closed to perfection and the audio sample capture in that Vector Signal Analyzer was no where as good as most up to date sound card we get them for the computer nowaday.  So I'm not sure if it's I need some additional filtering.  One would think MT does it all, but then again I could be wrong.  Unfortunately, the Agilent machine stops at 32 QAM and we need go up to 64 and be able to do all these in real time, so it would be nice to use Labview and the sound processor.

[lavalava]

Also, the modulation scheme used for the 2 sound files recorded above was 8PSK.

The quantization error you mentioned, might be a factor but I'm not sure, both sound cards that I"m using as I was told by my colleagues should be alright for these sort of things.  The SigmaTel is capable of sampling up to 44kHz at 16bits while the Edirol is capable of sampling up to 192kHz at 24bits.

Thanks 

[lavalava]

bump ...help, I'm stuck ...any suggestions would be greatly appreciated as I've tried just about every sound cards from creative to via to asus to sigmatel to intel and still no lucks with getting clean constellation.  Can someone please suggest me a better alternative methods?  Is it because of quantizatione error?  If so how do I get around it? 

[Paul_C.]

Hi,

Steve is out of the office today and I will do my best to help.  We typically use the MT Demodulate PSK.vi to apply a complex filter to the complex waveform before plotting it to the constellation plot.  In addition, can you please include the appropriate settings if you would like me to try the examples included if they aren't already set.  These settings include the number of symbols, symbol rate, whether its differential PSK, etc.  My recommendation would be to remove the filtering at the beginning of the vi in the audio sample3.vi and try using the MT Demodulate PSK.vi to filter the waveform.  You can generate the filter parameters with the MT Generate Filter Coefficients.vi.  These filter parameters are used as the inputs to the MT Demodulate PSK.vi.  You can also use the MT Generate System Parameters.vi to generate the appropriate parameters for the demodulation of the 8PSK signal.  I've tried the 22KHz waveform file with guesses on the settings but so far I've had no luck.  Please try wiring the complex baseband waveform (after the MT Downconvert Passband) into the demodulate PSK with the appropriate settings prior to plotting it with the format constellation.  It looks like the results might just be noise on the signal or quantization noise introduced by the sound cards. 

I hope this helps,

Paul C.

[lavalava]

Hi Paul,

Here are the settings that I normally set to the Agilent 89410A Vector Signal Analyzer for those 2 files:

8PSK (normal, not differential,  and no phase offset)

Baud or Symbol Rate: 2400 kHz

Center/Carrier Frequency: 1.8kHz

Bandwidth: 3kHz (starts at 300hz: ends at 3.3kHz)

Alpha/BT: 0.25

Reference Filter: Raised Cosine

Measurement Filter: Root Raised Cosine

Windows: Flat Top

Average: Off 

This is a narrow band baseband modem that I'm testing.