image
image
image
image
 

image
A Technical Note on 'Convo Boy' Audio Quality

 

This technical note explores a direct comparison between the audio output quality of Elevayta 'Convo Boy' (V2.02) and another commercially available premium level convolution plug-in. Using identical input (a stereo Dirac delta function) and an identical impulse response (IR) (96 kHz, 32 bit, 4 second duration), a rigorous 'apples to apples' comparison shows the audio output of 'Convo Boy' to be essentially identical in quality to that of the premium level alternative (the difference being inaudible).

Furthermore, in these comparisons, all plug-ins tested (including 'Convo Boy') perform re-sampling of the IR from 96 kHz at 32 bits resolution down to the host sample rate and bit depth. Comparisons have been made when re-sampling the IR down to 48 kHz, 16 bits and also 44.1 kHz, 16 bits. A comparison of the audio output quality, between all plug-ins and IR sample rate conversions, show that 'Convo Boy' yields a practically identical output quality to that of the premium level alternative (differences lie at/below the 16 bit dynamic range of 96 dB).

The convolution processing and IR re-sampling technology used in 'Convo Boy' is therefore very competitive (even very slightly superior in these tests) compared to the others available on the market.

Note: When implemented correctly, digital convolution is an exact mathematical process. There are many different ways to write software code that realizes the same mathematical result. The high level of agreement between the results shown here is therefore not unexpected.

 

 

Figure 1 - Sample of (100% wet) stereo audio output response to stereo Dirac delta input. Top traces 'Convo Boy'. Bottom traces from a demo version of another premium quality convolution plug-in. Identical IR, gain and latency settings were used in both plug-ins. The IR was internally down-sampled from 96 kHz, 32 bit to 44.1 kHz, 16 bit. (i.e. 'Match SR' enabled in 'Convo Boy').

 

 

Figure 2 - Spectral analysis of the difference between rendered audio outputs from 'Convo Boy' and another premium quality convolution plug-in. Dirac delta function input and identical IR, gain and latency settings were used in both plug-ins. The IR was internally down-sampled from 96 kHz, 32 bit to 48 kHz, 16 bit. (i.e. 'Match SR' enabled in 'Convo Boy'). The difference is not audible since the difference signal energy is at -112 dB across most of the audio spectrum.

 

 

Figure 3 - Spectral analysis of the difference between rendered audio outputs from 'Convo Boy' and another premium quality convolution plug-in. Dirac delta function input and identical IR, gain and latency settings were used in both plug-ins. The IR was internally down-sampled from 96 kHz, 32 bit to 44.1 kHz, 16 bit. (i.e. 'Match SR' enabled in 'Convo Boy'). The difference is not audible since the difference signal energy is less than -96 dB across most of the audio spectrum.

 

Comparing each plug-in output with the original IR

 

Using a Dirac delta function as input to a convolution processor should yield an exact copy of the IR on the output.

Comparing the output of both 'Convo Boy', and the other premium quality convolution plug-in, with the original IR (after down-sampling the IR to 44.1 kHz using an independent reputable wave editing utility) makes it possible to determine an absolute measure of the effect of each plug-in with respect to audio fidelity.

 

 

Figure 4 - (top) displays the first ms of the original stereo IR function after re-sampling from 96 kHz down to 44.1 kHz using a high quality 3rd party re-sampling utility. (middle) shows the first ms of the output response of 'Convo Boy' using the same IR function (with internal re-sampling) for a Dirac delta input function. (bottom) shows the first ms of the output response of another premium quality convolution plug-in using the same IR function (with internal re-sampling) for a Dirac delta input function.

 

 

Figure 5 - Spectral analysis of the difference between rendered audio outputs from 'Convo Boy' and the original impulse response (IR) function (in response to a Dirac delta input).

 

 

Figure 6 - Spectral analysis of the difference between rendered audio outputs from another premium quality convolution plug-in and the original impulse response (IR) function (in response to a Dirac delta input).

 

Discussion

Using a Dirac delta function as input to a convolution processor should yield an exact copy of the IR on the output.

Under such conditions, Figure 4 shows the output response of 'Convo Boy' to be essentially identical to the externally re-sampled impulse response function. The output response for the other premium quality plug-in, under identical conditions is also essentially equivalent (apart from some small differences at the start).

Figures 5 and 6 show a spectral analysis of the difference between the rendered audio output of each plug-in and the externally re-sampled IR. Both are essentially identical, with 'Convo Boy' doing very slightly better overall (DC and spike at 20 kHz are both lower).

 

What to Conclude?

Comparing the output of 'Convo Boy' with the original IR (after down-sampling the IR to 44.1 kHz using an independent reputable wave editing utility) shows it to be essentially the same, but not absolutely identical. Likewise with the other premium quality convolution plug-in, its output is very much the same as the original IR but not identical.

So, neither of the plug-ins tested can claim to be the absolute 'gold standard' in audio precision (although they come close). However, it is apparent that the audio quality of 'Convo Boy' is certainly as good as any other premium quality convolution plug-in currently available.

 

I am putting together a full description of the method used for this study. You can request a copy by mailing the link below:

 

 

Note: The report will not contain information on the names of tested products or manufacturers, other than 'Convo Boy' and Elevayta.



[Go Back]

image
image
image