Timing Submenu:

• Spec Mode
• Slow Mode
• Step Mode

• Slow x / Step Factor
• Units
• Smooth TC
• Quant
• Order
• Rand / Copy
• Shift

Timing Submenu:

Spec Timing Mode:

Slow Timing Mode:

Contrast this with simple low-pass filtering, which would result in a proportionally smaller level as the bandwidth was reduced. However, although the average output level would be smaller, the maximum possible output level would not be affected... it would simply be less probable. This is because the random sources have occasional runs of the same value, as any true random source should. So, if you filtered the noise to a low average level and then boosted the output, there would be occasional high-level runs that exceeded the output range. When that happened the output would be limited to the full-scale value, and the suppposedly low-frequency output would have high-frequency content due to the "clipping".

But in Slow mode the output level is not affected as you make the signal slower. However, due to the linear interpolation, the output is NOT equivalent to a large low-pass filtered version of the raw waveform. Each of the original data points is present, but spread out in time by the Slow factor and simply connected by straight lines.

If this is not suitable for your purposes, use the Smooth function below. Set the Smooth TC to about the same value (in samples, not time) as the Slow factor and you will have a fairly decent approximation to a "big, slow" version of the source. The level will be reduced slightly, but WAY less than the comparable reduction from filtering. The Smooth function is indeed a simple low-pass filter, but here you only need a relatively mild filtering action to remove the straight-line distortions, not to provide the entire low-pass action.

For comparison, you may want to try setting the Slow factor to 1x and attempting to do all the low-pass filtering with the Smooth control. Or you may want to try using the Noise Band source with the frequency range set very low.

Step Timing Mode:

For an example of how this can be used, consider the COMPOSER.S3A setup. This setup plays "Oriental"-sounding music, where each note frequency is determined by a randomly selected value from a region of the EXPNOTE.DAT Arb file in Round mode. The Step factor controls the duration of the notes by setting how long each random value should be held. The random values are then used to control the EXPNOTE.DAT selection by means of Phase modulation.

The Step approach can be used to define how long each stage of a test sequence should last. An Arb file in Step mode could provide the sequence of values if a predefined sequence must be repeated exactly. Or one of the Noise sources could provide random values, either to be used directly as the test frequency (or amplitude, etc), or to be used as above with Phase modulation to select a value at random from an Arb sequence.

Slow / Step Timing Factor:

In Step mode, this factor will be either in samples or time units depending upon the Units control below. In Smpls mode it is a simple multiplier giving the total number of output samples for each input sample value. In Sec mode it is the duration of each held step.

In Dynamic RTime mode the Sync interval is controlled by the Step size set here, instead of by the main Freq control.

Units (Smpls / Sec):

Smooth TC:

As with a real RC filter, the output of the smoothing function exponentially approaches the average value of the input. It is mainly intended for smoothing the Slow or Step outputs, but it can in fact be used as a low-pass filter in Spec mode as well. When the TC value is much higher than the Slow or Step factor (or much higher than 1 for Spec mode), there will be a large reduction in the output amplitude. This is normal for a low-pass filter since the average value tends toward a constant, usually zero with normal bipolar signals.

For Slow mode, try setting the TC to the same value in samples as the Slow factor. In Step mode, you will probably want a much smaller TC (if you want one at all) just to round off the sharp corners of step transitions.

Quantization:

The Quant control specifies the number of steps between negative and positive full-scale outputs. When it is set to 0 (default) it gives the maximum number of steps, regardless of the DAC bits. When it is set to 1, the output is binary: It goes from one extreme to the other in a single step. Note that if you want some number L levels in the output, set Quant to L-1.

Quant has a qualitatively different effect depending upon whether you are using Step or Slow mode. This is best seen with White Noise and Quant = 1 for a binary output. With the Step factor = 100 you will see random binary transitions, but each interval will be some multiple of 100 samples. With Slow factor = 100 you also see random binary transitions, but there is no particular width multiple. That's because the Slow wave may sometimes, for example, just barely peek over the binary threshold for a few samples, then return. You'll have to decide which approach is most appropriate for your work.

Order:

Rand / Copy:

By default, this starts at Ind, which means the generator for the current component page is Independent of the rest. As you scroll up, the Ind changes to A0, which means the A component page of DAC 0, then to B0, C0, D0, A1, B1, C1, D1.

When you have advanced the selection to the current page, the control header will say Rand. If you hit ENTER then, the current page random generator will be randomized. That process uses the current time, date, and system status to start the generator at a random place.

If you advance to a page other than the current one, the header will change to Copy. If you hit ENTER then, the current page random generator will become an exact copy of the one selected. Unless or until you specifically change that, the two will stay locked together. This allows you to provide the exact same random sequence to different component pages, even in different DACs. It is also the first step to creating time-shifted random sources, controlled by the Shift option below.

If you continue to advance the Rand/Copy control, you will get to Copy All. If you hit ENTER there, it copies all generators to match A0.

The final possibility is Rand All, which randomizes all the generators to break up any tracking due to prior Copy operations.

Rand/Copy is unlike other Daqarta controls in that it is not "live": No action is taken until the final ENTER. This is because these operations are not generally reversible. (There is no UNrandomize!)

Also, note that there must be some change activity in the Rand/Copy item or no action is taken. For example, let's say you are in page B0 and have previously set Copy A0. Perhaps in the interim you have changed something and want to renew the A0 copy. If you just hit ENTER to select the existing Copy A0 and then hit ENTER again to accept it, nothing happnens because no change was perceived by the menu handler. Instead, you should scroll from A0 to any other setting and then back to A0 before hitting ENTER to accept it.

Shift:

Why might you want to do this? One reason might be to obtain comb-filtered noise. Comb-filtering results when you add a delayed version of a signal to itself. At all frequencies for which the delay represents an integer number of full cycles, the delayed and original signals will be in phase and those frequency components will be doubled in level. At all frequencies where the delay represents an odd number of half-cycles, the delayed and original will be 180 degrees out of phase and those components will cancel. The resulting spectrum thus has a series of alternating peaks and dips, with the separation between adjacent peaks (or dips) equal to the reciprocal of the delay time.

For example, if you set the delay to 10 samples and the sample rate is 20 kHz, that constitutes a 500 microsecond delay which represents 2 kHz. The spectrum will thus have peaks at 0, 2, 4, 6... kHz and notches at 1, 3, 5, 7... kHz. If you listen to this noise, it's pretty hard to tell it from normal white noise. But when you scroll the Shift value you hear an effect that sounds like a jet plane taking off. That's because when a jet takes off, the noise from its engines reaches your ears via different routes; some directly, some after reflecting off the runway and nearby buildings. So you are really hearing comb-filtered noise, and when the jet moves the delay changes as the noise source gets farther away.

You will need to set each source's Level to 50%, so that their sum will never exceed the full-scale output limits. Instead of adding, you can get the same comb-filtered phenomenon if you set one of the Levels to -50%. Now when the Shift delay is 0, you will be subtracting one identical random noise from the other, and the output will be 0. Even a single sample shift either way from this point and the output jumps back up to cover the full-scale range.

Shift works in both directions; you can set negative delays just as easily as positive. And you can change Shift on either or both noise sources... it's only the relative delay that makes a difference.

Comb filters with variable delays have been used in Rock music, and were the basis of the "phaser" effects pedals used with electric guitars. (Many of those used only a crude approximation to a true comb filter, however, in that they only provided a couple of dips and peaks.)

Of course, another obvious use for shifted noise is to simply give a delayed version of a random stimulus, perhaps to simulate an echo. For example, if the two random sources are used as AM modulators, the two will track each other with the selected delay.

The Shift value is always given in samples, not time units, regardless of the Smpls / Sec setting. This is the number of samples provided by the random source, not the number of output samples. When Step or Slow modes are active, the Slow/Step factor effectively multiplies the delay time, since there are no new random samples created between the original data points... just interpolated or repeated data.

Also, some Noise Waves use more random values than others. The White source only takes one value per data sample, whereas Gauss and Pink take two. Band, on the other hand, only uses one random value per eight data samples. This means that you can have Gauss and Pink sources that track, or two different Bands, but you can't have tracking White and Gauss, for example, or Pink and Band.

Shift only applies to the source, not the Band filter, so two different bands of noise can still track perfectly using Shift.

• Back to STIM3A Timing Menu.
• Back to STIM3A Main Contents.
• Continue to STIM3A Burst Menu.
• Jump to the STIM3A Index
• Daqarta for DOS Help Contents.
• Daqarta for DOS Help Index.
• Daqarta home page.
• Daqarta download options.
• Free Daqarta for DOS Registration

Questions? Comments? Contact us!