posted 29 April 2002 02:07            

It consists of a stereo processing chain:

input -> PEQ -> Reverb -> Dyn -> Stereo -> Exciter -> output

These stages are:

Input: a gain control +/- 20 dB on the input stereo signal.

PEQ: an 8-band stereo "paragraphic" EQ with shelving filters at each end of the spectral range, and 6 user placeable peaking/notching filters. All filters have adjustable frequency, gain/attenuation, and Q. Each filter is given +/-20 dB range, Q can range from 0.1 (10 octaves!) to 10 (1/10 octave). The PEQ can be bypassed. Phase shift through the PEQ is modest (depending on how drastic your gains are) and very musical sounding. Left and right channels receive the same treatment from one set of VCS sliders, but audio is processed in separate signal chains.

Reverb: a sugaring reverb to be used lightly in mastering, made up of a Kyma EuverbLeft and EuverbRight. Control of left/right decay, reverb amount, and LPF cutoff is on the VCS, along with diffusion. The reverb can be soloed or muted.

Dyn: a 6-band stereo dynamics processor consisting of a noise gate followed by a compressor followed by a limiter. The 6-bands are as developed by the Kyma graphics EQ Sound, with the 8 KHz and 16 KHz bands combined into one. (These can be split for a 7th band if you wish -- all it needs is DSP power!). The noise gate has a user settable threshold, and the compressor has user settable gain, threshold, and ratio. The dynamics chains can be bypassed. Left and right channels are treated alike from one set of VCS sliders, but separate signal chains are maintained.

[I should probably add attack and release parameters on the VCS, but I'm out of room for now... This VCS takes a 1280 x 1024 screen to display, and all the faders are short things...]

Stereo: following each compression chain is a stereo spreader and phase panner for each of the 6 dynamics bands. The stereo processing can be bypassed. Spreading is accomplished by widening or narrowing each subband by means of subtraction or addition of the mono mix from left and right channels. Then the left and right are sent into two phase panners before recombining them as stereo.

Exciter: Bands 1K, 2K, and 4K are combined in stereo and then sent to an exciter which treats the left and right channels independently. The exciter can be adjusted for injection level and wet ratio. It also incorporates its own noise gates on the left and right inputs to prevent excessive noise buildup. Gate threshold is settable from the VCS. The exciter can be bypassed.

Output: an output gain stage with +/-20 dB range.

I have not provided a loudness boost or dithering at the end of this processing chain. Personally, I like the wide dynamics of older music. Compression can be applied in any one or all of the 6 subbands if you like, but there is no final output loudness booster.

The Capy probably already provides some dithering in its output DAC's. Not sure about its S/PDIF outputs. But this should be easy enough to check. If worse comes to worse, we can always record 24 bit AIF files directly from Kyma.

I don't have enough DSP horsepower here to run at 48 KHz, but at 44.1 KHz it sounds absolutely stunning!! I'm having a real blast going back over my old recordings and some ancient CD's and adding a little mastering spice to them. Just about anything prior to 1990 probably could use a touch up or two...

- DM

posted 29 April 2002 17:21            

mastering.kym

I'm still a little unhappy with my harmonic generation and my stereo subband panning... I realized this morning that my harmonics were being generated with a time delay (which didn't sound all that bad, in fact). But when I tried to straighten everything out in time then they started sounding a bit groady to me. I also had to insert some delay compensation in the higher bands to make up for a longer necessary compression delay in the lowest frequency band. After having done all this I am once again out of DSP gas.

For now, just to give a gander at the situation, I have a lite version (only 19 Capy/320 DSP's needed) that runs in realtime at 44.1 KHz. I had to remove the limiters, and the noise gate in the harmonics generation. These are relatively benign removals anyway, since noise in the harmonics generator with a gentle distortion curve should give only small amounts of audio -- and harmonics are added in trace amounts anyway.

The attached Lite version still needs my microsounds to make the parametric EQ bank work for you. That will be submitted a little later. But you can open the sound up (I hope!) and take a look at what's inside -- not rocket science here. The subband stereo panners are phase and amplitude panners so that at a pan level of 1 you get the original stereo placement with -6 dB of cross mix (great for headphone listening!). At pan = 0.5 everything is mono, and at pan = 0 everything is reversed left/right. (As I said, I'm not quite pleased with this behavior just yet...) The spread parameter subtracts (when positive) a slight amount of the mono mix from both left and right channels ahead of the panning. Negative spread adds mono mix making the subband more centered.

I put a 10 ms attack time on the 250 Hz subband, 1 ms attacks on all the others. Delay compensation has been added to the higher bands to keep time alignment.

This is a real monster of a patch in terms of DSP usage, but it sounds incredible on older recordings. Even if you are light on DSP's you should still be able to process audio, just not in real-time, by recording the Sound, or by using disk caches inside the overall Sound.

I'll submit the details on the 8-band parametric EQ shortly...

- DM

[I think the really neato thing about doing mastering in Kyma is the ability to make time varying parameter adjustments in the timeline. All the other mastering tools I have seen are either entirely manual (knobs and buttons) or else preprogrammed for static application of a fixed set of parameters over the entire duration of a piece of audio. Kyma frees us from this binding and makes possible some really incredible and adaptive mastering! Yeaay Kyma!]

posted 29 April 2002 18:27            

userprog.zip mastering.kym

[Note: I had to Zip this file because the Forum manager wouldn't allow me to upload files named xxx.ASY. Just unzip it -- there is only one file inside and it is very small... I also added individual channel mutes to the 6 band compressor/stereo thingy]

To use the parametric EQ filters you need to go to your file menu and select on "Choose Microsounds...". Apparently it can only load microsounds from a file named USERPROG.ASY. So if you already have microsounds in use, be sure to put this file in a separate directory so as not to overwrite your existing microsounds file. When (or if...) it asks you whether or not to Edit or Reload the microsounds from this file, choose Reload. Capy will then reinitialize and "Assemble" the microsounds in this file.

I haven't bothered to encrypt it, so if you can read DSP assembly language feel free to have a look at what it does. The microsound named DirectIIR implements a direct form IIR filter using externally supplied coefficients. The documentation in that microsound file tells you what is expected. Filters with up to 8 2-pole sections can be accommodated, but we only need 1 2-pole section for each of the parametric EQ filters.

The parametric EQ is created by providing a low-shelf filter at the low frequency end, a high-shelf filter for the high frequency end, and 6 peaking EQ filters for anywhere in between. There are no particular limits on the placement in frequency of any of these filters. The high-shelf filter can become unstable at positive shelf gains, but its primary purpose is to roll-off excessive high frequency energy, and so the gain supplied will generally be negative (= attenuation).

All of these filters allow hotvalues for the filter frequency, Q, and dB gain or attenuation.

Enjoy!

- DM

[Here's a quickie conversion table between Q values and expected bandwidth (approximate) at the -3 dB amplitude levels...

Q ....... Bandwidth (octaves)
-- ...... -------------------
0.1 ..... 3.3 (= 3 octaves)
0.3 ..... 1.9 (= 2)
0.5 ..... 1.3 (close enough to 1.5)
0.707 ... 0.95 (= 1)
1 ....... 0.69 (= 2/3)
1.414 ... 0.5 (= 1/2)
2 ....... 0.36 (= 1/3)
3 ....... 0.24 (= 1/4)
5 ....... 0.14 (= 1/7)
10 ...... 0.07 (= 1/14)

]]

posted 29 April 2002 21:56            

mastering.kym

These both use cascaded IIR filters from Kyma's filter offerings (Butterworth filters) to define the 4 passbands for dynamics and stereo field adjustment. One of them uses 4-pole filters and the other uses the Behringer trick with 1-pole filters. You be the judge of which sounds better. The 1-pole filters allow more adjacent channel leakage, but the 4-pole filters have greater band-edge phase cancellation.

In both of these you can freely adjust the band frequencies F1, F2, and F3. The band extremes are held at 20 Hz and 20 KHz. The FIR version with 6 bands does not allow this kind of tweaking. So there are advantages to this approach too.

The high-frequency exciter now takes the total post dynamics mix and sends the left and right channels through an adjustable high-pass filter (nominal range 500 Hz - 6000 Hz).

I think these all sound terrific. Differences between them are subtle. Overall, I think the great sound comes really from the oustanding quality of the Capybara!

Enjoy!

- DM

posted 30 April 2002 15:18            

I really don't know what I'm doing most of the time. I use Kyma as a learning tool. Oh, I understand a great deal about math and physics and how filters work and so on... But what I don't know diddly about is subjective "sound". That's why I so value the feedback of you audio pro's!

This sound collection is my first serious attempt to understand the "mastering process". I have heeded the frequent admonitions to use various effects sparingly and not overdo them.

When I pay careful attention to these suggestions I find that using the mastering tools developed here really add substance and finish to prior recordings. It reminds me of the first time I donned pilot's sunglasses. (Some sport hunters use them too.) They are amber colored and all of a sudden the whole world became more 3-D through them. Same here too with the sounds resulting from this mastering.

Cheers,