The document contains a comparison between
Dirac and WinMLS. Both the currently released versions Dirac 2.7 and WinMLS 2000, and the beta versions of the Dirac 3.0 and WinMLS 4
are compared. If WinMLS is mentioned without specific mention of a version
number, all WinMLS versions a 23523p1524x re referred to.
Note that the comparison is based on the WinMLS help file, information
retrieved from the WinMLS website (www.winmls.com),
and the use of an evaluation version of WinMLS 2000 release 3.10c.
As both products aim to replace MLSSA, it is not surprising that both contain more or less the same functionality. Both products can be used for impulse response measurements. In the following, only the most significant differences are described.
The user interface is by definition the most visible defining part of the software. Even if the functionality of 2 software products is exactly identical, they can still exhibit large differences in their user interfaces.
Looking at WinMLS one could conclude that the intent of the designers is to give the user the largest possible amount of controls by which to influence the behaviour of the software. The WinMLS user interface allows the user to modify many aspects of the program. On the one hand, this means there is a very steep learning curve to master the software. On the other hand it gives the user a lot of flexibility to perform tasks in many different ways.
Dirac in contrast is designed to limit the number of choices a user has to make. The intent is to have the software decide what to do where possible. This allows mere mortals to perform successful measurements with a very limited amount of training.
As a result of having many options associated with a given function, very often a number of dialogs are involved in performing a function in WinMLS.
An example is viewing the band filtered impulse response: In WinMLS one has to select the Plot Type Settings dialog, from which the Filter Settings dialog can be reached where the user can select any desired centre frequency, bandwidth, filter order and filter type. In Dirac one simply presses the filter up/down buttons or selects the correct filter from the filter dialog. Depending on one's point of view, one could say that Dirac does not allow the user to choose the filter order, filter type etc. Or one could say that in Dirac there is no need to set these filter parameters.
Confounding the problem of the many options is the fact that menu items and toolbar buttons are not always disabled when their use is in fact impossible. The user is not guided to the relevant options, but confronted with all options including those that make no sense in a given situation (or that make no sense in any situation).
Dirac contains a comprehensive soundcard calibration that contains the following functions:
Determination of optimum mixer settings and D/A levels for maximum SNR and minimum distortion.
Determination of the mixer's volume control gain steps.
Measurement of the soundcard's frequency response and its subsequent correction through inverse filtering of the stimuli.
Measurement / correction of the sign (polarity) of the soundcard's gain.
Measurement / correction of the soundcard delay.
Once the soundcard is calibrated successfully, the user only has to ensure that no overload occurs during a measurement.
In WinMLS 2000, most of these functions have to be performed manually. WinMLS can measure the volume control gain steps, but it does not take the possibility of clipping within the soundcard into account. In a measurement as many as 4 volume controls and the supplied D/A level are of importance. WinMLS leaves it up to the user to set the volume controls and D/A values to the correct levels. For the correction of the soundcard's frequency response, a loop back measurement has to be performed for every possible stimulus length. This loop back measurement must then be deconvolved with every measurement to correct for the soundcard's response. There is no option in WinMLS to pre-compensate the stimuli.
WinMLS requires an extra connection between the soundcard's input and output, during a measurement, to determine the correct source-receiver distance.
All in all, the above shows that WinMLS does not really control the Mixer/Soundcard.
WinMLS contains an absolute input (and output) level calibration. One can also specify the microphone characteristics, allowing for SPL measurements. Dirac will only have the input level calibration capability in version 3.0.
WinMLS 2000 does not provide for (linear) sweep measurements, whereas Dirac does. Both WinMLS 4 and Dirac 3.0 contain exponential sweep measurements.
WinMLS 2000 does not support external MLS, sweep or noise measurements.
Basically, WinMLS and Dirac provide the same room acoustics parameters that are described in ISO 3382. WinMLS will only calculate the reverberation times (EDT, T30 and Tuser) in third-octave bands. All other parameters are only available in full-octave bands.
In WinMLS it is possible to define custom parameters such as T15, C20 and IACC60.
WinMLS 2000 calculates a single STI value based on the original Houtgast and Steeneken weights. Both Dirac and WinMLS 4 calculate the male and female STI values as specified by IEC-60268-16.
In WinMLS, the (RA)STI values can only be measured using an MLS excitation, which makes it very difficult to measure these parameters in situations where the speaker position is very far removed from the listener position as an external stimulus must be used in these cases.
WinMLS lacks male and female speech filters, which makes it necessary to filter the soundcard's output using an equalizer that can deliver a very high gain or attenuation.
WinMLS has a number of features intended for loudspeaker measurements, such as phase plots, and (in WinMLS 4) impedance measurements.
Although WinMLS can show the average of multiple curves (not measurements), it is rather difficult to set up this feature. This set-up needs to be repeated for every set of curves. Individual parameter values that cannot be calculated will be set to zero and thereby invalidate the results.
WinMLS does not provide an option to edit the measured impulse response. This option is often useful to extract short impulse responses from long measurements, to reposition the start of the impulse response to the start of the file, to remove an initial spike due to cross talk etc. The use of a proprietary file format (.wmb) makes this extra difficult because there are no .wmb editors available.
In WinMLS it is not possible to listen to the measured impulse response, which is a very important option the check the quality of a measurement.
Starting with version 2, WinMLS has used third-party tools for graphing results. In version 2 GNUplot was used. In WinMLS 2000 and WinMLS 4 a library called TeeChart is used. The advantage of using a third-party tool is that it saves a lot of development time. The disadvantage is that this (general purpose) tool seldom fits the exact requirements of the software. This results in the rather nonsensical possibility to display Gantt-charts of the reverberation time, or pie charts for the frequency response. Also, the charting tools in WinMLS have an enormous amount of options to set all kinds of fonts, colours, gradients, line widths etc. etc., that can only confuse a user or lead to unproductive tweaking.
The graphing code in Dirac was written to exactly fit the requirements. An example is the labelling of the frequency axis. Where Dirac always displays the octave-band centre frequencies with tick marks for the third-octave band centres, WinMLS per default displays a base 10 logarithmic frequency scale.
An advantage of WinMLS is that it can display multiple time- or frequency-domain curves in a single graph. Also, it provides additional time domain views such as the step response, double integrated (position from acceleration), differentiated (speed from position) and double differentiated (acceleration from position). These views are mainly of interest for vibration measurements, not for room acoustics.
WinMLS uses a proprietary file format (extension: .wmb), which makes it difficult to use the data in other programs. Although these files can be converted to .wav format, all level information is lost in the conversion.
The creators of WinMLS are watching Dirac closely, which can be seen from the fact that some paragraphs of the Dirac manual have been copied verbatim into WinMLS documentation. Also, the introduction (in WinMLS 4) of a grouping of the parameters in reverberation, levels, spaciousness etc. seems to have been inspired by Dirac.
WinMLS 2
WinMLS 2000
WinMLS 4 (Currently in beta, release date unknown)
Dirac 1.0
Dirac 1.5
Dirac 2.0
Dirac 2.5, 2.7
Dirac 3.0 (Currently in beta)
The most significant difference between Dirac and WinMLS lies in the user interface. The WinMLS user interface is very complex and therefore error prone, requiring a steep learning curve.
Many important aspects of WinMLS do not seem to be designed with a user in mind, such as the soundcard calibration and the speech intelligibility measurements.
The overwhelming array of options associated with every function actually hides the fact that these functions are not set up in a way to help the user perform measurements.
A hands-on evaluation of both programs will make these differences very apparent to a prospective user.
|
