x-axis for waterfall plot

Hi

I am not an experienced Q(w)t user, but learning. I did write a software defined radio for a simple card, under Linux (Slackware) with Qt > 4, Qwt > 5.
Until now I used a spectrumdisplay for displaying the spectrum of a range of 48k around the selected frequency (depending on the chosen samplerate).

Using spectrogram stuff I managed to create a waterfall display, showing the spectrum evolving in time. However, I need some help for setting up the appropriate X axis frequencies for this display.
a. on a 512 element display (power of two handy for the FFT) it shows a block of 512 wide, but the screen is 600 wide
b. the x axis labels are 0 .. 512 (or -256 .. 256). I need to be able to map these labels
onto something like 14010 .. 14070, depending on the samplerate chosen.
anyone can help me?

Quote:

---

a. on a 512 element display (power of two handy for the FFT) it shows a block of 512 wide, but the screen is 600 wide

---

Or any other size !

In QwtRasterData::value() you have to return a value for any requested position (not only those in your value matrix) calculated from your matrix (resampling). There a different ways how to do it maybe the most simplest one is using a next neighbour strategy.

Quote:

---

b. the x axis labels are 0 .. 512 (or -256 .. 256). I need to be able to map these labels
onto something like 14010 .. 14070, depending on the samplerate chosen.
anyone can help me?

---

Implement your own type of QwtScaleDraw and reimplement QwtScaleDraw::label(). Look into the cpuplot example.

Uwe

ad a:
That is understood, I need to know the number of pixels and map my array data
on the pixel raster.
However - probably the most stupid question of the last year - I cannot find the
number of pixels. rasterHint gives systematically (-1, -1) and is therefore not helpful.
googling on "qwt rasterHint" gives 8 hits, none of them helpfulhttp://www.qtcentre.org/forum/images/smilies/confused.png

so, I am stuck

any help??

all the best and thanks in advance
janK

Quote:

---

Originally Posted by janK

That is understood, I need to know the number of pixels and map my array data on the pixel raster.

---

No, you don't need to know the number of pixels.

Your rasterdata object needs to calculate a value for any coordinate it is requested ( + return a valid bounding rectangle to avoid requests for values, where you can't return something useful ). When the resolution of the plot is higher than your value matrix, next neighbor will return each value several times, if it is lower you will have undisplayed values. ( more advanced resampling strategies can avoid this effect, but slow down the rendering process).

Sometimes it is helpful to know the pixel raster to resample in advance. ( F.e. when you have a huge amount of values on disk you can't have in memory at once ). That's what QwtRasterData::initRaster()/discardRaster() are good for.

QwtRasterData::rasterHint() returns the raster of your matrix. It is used to find the best resolution for the image of the spectrogram.

Uwe

Hi

I did some experimenting and could not help noticing that the size of the requested
raster extends the requested x-axis. It seems to extend the x-axis to the next multiple
of 100. a request for a raster for a matrix with n rows gives a raster of
(n % 100 == 0)? n : (n / 100 + 1) * 100. Based on this I made a value function
projecting the n rows on the displayed raster,

At least it works, so my next job is creating the appropriate X_axis for the waterfall display

So far, thanks

jan

I'm not sure what you mean by "requested raster", but from your posting I guess your scales are simply the result of the autoscaler.

You can adjust the behaviour of your autoscaler by changing the attributes of the scale engines. ( or disable autoscaling and assign the scales you like).

Uwe

sorry to bother again, one step further, still needing help

first: by adding an x axis of the right size, the raster size is correct. I.e. in my case
I generate FFT spectra consisting of 512 values, by setting the x axis of the plot
to 512 I do get the whole plot.

a linear shift of the values of the x axis is possible, i.e., when I have some spectral
info from e.g. 15000 KHz, I can set the middle of the x axis of the plot to 15000.
However, the 512 values in my plot may represent 96 Khz in m1y spectrum. Apparently
I have to scale the value in the x axis, most likely with qwtscalediv.
The specification of the mapping (i.e. what is being projected upon what) is not
clear to me. Either the lowerbound .. upperbound map on the qwtvalue list or vv, furthermore the structure of the qwtvaluelist is unclear.

all the best
jan

I am unable to solve the x-axis issue.
Briefly stated, my problem is as follwos:
a I create a spectrogram (basically according to the example), with width 512 elements
b. app 10 times a second I generate an update where the topline in the matrix representing
the spectrogramdata is replaced, with as result a waterfall plot.

c. I am unable to create an appropriate X axis. My requirements are
1. the 512 elements represent 96 kHz, so I need to scale. One thought is to
instruct the spectrogram that it has a matrix of 96000 x 50 by implementing
a suitable value function
2. the middle element of my 512 elements has a known (though in time varying)
frequency attached to it. I have been experimenting with drawscale etc using
material from the examples and ( as far as I could understand it) the references
however, in those cases that I had more or less an x-axis to my taste, the spectrogram
would display all blanks. This makes sense because I need to be able to create a
mapping from my frequencyrange (e.g. 14000 +/- 48) to the interval 0 .. 512 (the
implied x-axis from my raster)
it is a pity that I could not find an introductory text, but I really need help here

all the best
jan
:confused:

stupid me,

Of course a (may be "the") solution was very simple. I seem to have misunderstood
the notions of raster and the data for the raster. The bounding block is now labeled
with the intended data (i.e. frequencies of something), and the "value" function
now maps the array data onto the pixels (domain -> domain). No complex doing with
X axis, just straight formward.

Next stop will be a plotmarker

anyway it was a useful exercise and my waterfall plot looks very nice (and is useful as well).
all the best
jan