[LogoShip]

Home

Daqarta for DOS Contents

Introduction

Downloads

Free Registration

Contact Us
Daqarta for DOS
Data AcQuisition And Real-Time Analysis
Shareware for Legacy Systems
(Use Daqarta for Windows with modern systems)

From the Daqarta for DOS Help system:
 

SPECTROGRAM:


SPECTROGRAM TOGGLE (Sgram):

Use ALT-S when FFT mode is active to enter Spectrogram mode. Here each vertical line on the trace represents one N-points FFT spectrum, with frequency running from low to high and the power at each frequency indicated by one of 15 colors, plus the screen background color for the lowest level. Time is recorded from left to right as a succession of these vertical color spectra.

When the line for the current time reaches the right end of the trace area, it wraps back to the left end and the process repeats indefinitely. The X-axis time labels are relative only to the 0 at the left end of the trace... they do not increase on subsequent passes. (If the labels kept increasing to show actual time since Spectrogram mode was begun, the labels would rapidly grow to an unmanageable size and collide with each other if the same resolution were to be maintained.)

The same triggering and window function options used for the normal spectrum mode are also used for the spectrogram, including trigger delay. Stimulus generation continues to operate normally.

In Sequential mode, the speed of the spectrogram sweep is determined by the sample rate and the N size, since these control how long it takes to acquire the samples needed for each line. Smaller N will result in a faster sweep, as well as finer time resolution. A faster sample rate will do this as well, but at high speeds the CPU speed will limit the processing time and hence the sweep speed.

If the trigger mode is set to Norm and the trigger events are infrequent, the Sequential X-axis will underestimate the actual spectrogram time.

In RTime mode, you can often increase the spectrogram sweep speed by toggling Trig to Free-run via the T-key. If your CPU is fast relative to the sample rate, overlap processing will take place to give finer time resolution and a faster display update.

Since the X-axis calibration in RTime mode is based upon actually measuring a few individual spectra, if the trigger mode is set to Norm and triggers are infrequent it may take a long time for the axes to appear when you activate Sgram mode.

You can change the colors representing each of the 16 power levels using CTRL-C, which invokes a special color control menu in Sgram mode. CTRL-S invokes the spectrogram control menu, which allows you to change the range of powers covered by these 16 levels.

There is no provision to save a spectrogram to disk as such. Instead, you should save the raw data using the DDisk option. This not only allows complete control over the spectrogram color and sensitivity after the recording, but also allows you to adjust overlap, window function, number of FFT points, triggering, etc.

For a good example of the spectrogram in operation, see the experiment described in the Frequency Modulation section of the Virtual Source.

A full-screen image of a Daqarta speech spectrogram is also available in our on-line image gallery.


SPECTROGRAM Y-AXIS:

When the spectrogram display is active, the Y-axis shows the frequency for any point in X-axis time. For N = 256 or 512, the Y-axis shows the whole Nyquist frequency range, which is half the sampling frequency. It will be identical to the X-axis in the normal spectrum mode.

When N = 1024, the Y-axis shows only half of the spectrum X-axis frequency range because it has only 256 screen pixels to work with, and an FFT from 1024 input samples gives 512 output frequencies. However, you can slide that half up and down in frequency with the X-axis control menu Xpand controls. (There is no "normal" Xpand in spectrogram mode.)

Or, more simply, use the PgUp/PgDn keys to this... they still control the Y-axis here, even though it's frequency instead of the usual trace magnification. You can even use SHIFT-HOME to force the display to show the lower half of the frequency range (or any preset position), and SHIFT-END to restore the previous range.


SPECTROGRAM CONTROL MENU:

While the spectrogram is active, CTRL-S invokes a menu to control the sensitivity of the color display. (Use CTRL-C to change the actual colors representing each level.)

Each of the 15 color levels is labeled with a threshold, in dB below full scale. (This makes them all positive numbers... in other words, a value of 20.0 dB below full scale is -20.0 dB relative to full scale, or 0.10 of full scale magnitude.) To compute each vertical line of the spectrogram from the corresponding spectrum at that time point, the spectrum output at each frequency is compared to these thresholds to find the highest one (lowest dB value) that it reaches or exceeds. The corresponding pixel is then shown in that color. Outputs below the lowest threshold are shown in the screen background color.

SPECTROGRAM SENSITIVITY:

The sensitivity range is controlled by setting the highest and lowest thresholds, and all others are automatically adjusted to form a linear sequence of constant-dB steps. The difference between the top and bottom is divided by 14 to get the step size between each of the 15 thresholds, rounded to the nearest 0.1 dB. This step size is then applied from the top down to determine the lower thresholds, including a new bottom value as a result of rounding and accumulation.

This adjustment method imposes certain path dependencies. For example, the default range is from 5.0 dB to 75.0 dB, in 5 dB steps. If you enter 0 for the top threshold, the range will become 0.0 to 75.6 dB in 5.4 dB steps. (75.0 / 14 = 5.357 dB, rounded to 5.4 dB per step. 5.4 dB times 14 steps = 75.6 dB.) If instead of entering the 0 dB directly you slowly scroll from 5.0 to 0, the range will be 0 to 70.0 dB in 5.0 dB steps. This is because after each small change, the computed step size was still so close to 5.0 dB that it rounded back to the same value.

In addition, unless FFT32 is active, there are limits due to the resolution of 16-bit data at the lower end, where an increase of one count in the raw data output may represent a large dB change. Since each dB threshold value must have a unique equivalent data value, Daqarta can't arbitrarily round to the nearest one without making sure that the next higher or lower threshold doesn't also round to the same data value. If the equivalent data values don't turn out to be unique for all thresholds throughout the dB range requested, Daqarta will keep the top threshold and readjust the bottom threshold upward (smaller dB value, hence finer resolution) until they are unique.

The implied resolution of 0.1 dB is thus an illusion for thresholds below about 70 dB. The actual thresholds used will be determined by the closest data values. The top threshold is limited to 68.9 dB to allow room for those below it.

However, by shifting to FFT32 mode, the greatly increased dynamic range of 32-bit FFT computations largely eliminates the above issues for all feasible applications. The top threshold can now go down to 90 dB, and the bottom down to 180 dB... WAY lower than any real input signal can achieve.

Because of this vast disparity in range capabilities, Daqarta maintains separate range settings for normal and FFT32 modes. When you shift between them, the current settings for the new mode immediately take effect.

You can use this to advantage even when you don't need the huge FFT32 dynamic range, but simply want to be able to flip quickly between two different range setups. Just set the desired limits, toggle FFT32 with SHIFT-F, and set the other.

With all range sensitivity settings, whether by directly adjusting a limit or by shifiting to a whole new setup with FFT32, any changes you make are implemented immediately for all incoming data... they have no effect on data previously acquired but still visible on the screen.

GO:

Questions? Comments? Contact us!

We respond to ALL inquiries, typically within 24 hrs.
INTERSTELLAR RESEARCH:
25 Years of Innovative Instrumentation
© Copyright 1999 - 2006 by Interstellar Research