posted 24 September 2001 05:07            

What are these Sounds? They are stereo multiband compressors with a special VCS input control.

As some of you may have read, I started a company named The Euterpe Group to manufacture high-quality DSP based hearing correction systems for people with hearing loss. But I can't possibly hope to help everyone needing help. So if you have a Kyma system then you can render recordings and live playing through your Kyma/Capybara and one of these RED sounds to get at least partial restoration.

These RED sounds are freebies to Kyma owners. They aren't Crescendo algorithms by a longshot, but they do sound pretty good anyway. (Crescendo is our code-name for the psychoacoustic anti-recruitment compression algorithm that operates over the critical bands of human hearing).

The RED Sounds make use of linear amplitude compression scaled so that they approximate the required psychoacoustic compression for louder sounds. Psychoacoustic compression is actually quite nonlinear and at faint sound levels they become increasingly compressive. In contrast, these linear compressors fail to reproduce the faint levels properly and they tend to sound just a bit too bright overall when compared to Crescendo compensation.

Furthermore these RED sounds utilize octave bandpass filters instead of 1/3 octave or critical band filters. Also, as presented here, the same compression ratios are used for both left and right channels, though you could edit them to provide for hearing asymmetries.

The RED-5 and RED-7 Sounds utilize the Kyma GraphicEQ (FIR) filters. They have throughput latencies on the order of 12 ms or so. The RED-5 clumps together the 250 Hz and 500 Hz bands, as well as the 8 KHz and 16 KHz bands. Doing this allows this stereo hearing compensation to run on a simple 4-DSP Capybara. The RED-7 utilizes separate compression on all 7 of the GraphicEQ filters.

The RED-7.5 version demonstrates a 7.5 octave stereo filterbank using the 1-pole filters discussed in the other topic. Its throughput latency is only about 1 ms. It tends to sound just a bit brighter than the RED sounds based on the Kyma FIR GraphicEQ Sound. But it still sounds pretty good anyway.

To use these RED sounds you need to have your hearing tested and get a copy of your audiology report. That report shows your threshold hearing elevation in dB at standard frequencies of 500, 750, 1K, 1.5K, 2K, 3K, 4K, 6K, and 8K. It might also show the value for 250 Hz. Take those values and, using the average threshold elevation for your left and right ears, at frequencies of 250 Hz, 500 Hz, 1 KHz, 2 KHz, 4 KHz, and 8 KHz, enter those threshold elevations in dB in the VCS control.

Once you do this, the leading script of the RED sounds has a subroutine to compute the required compression ratios for your hearing.

Unlike pro-audio multiband compressors where for effect one frequently uses compression ratios of 3 or greater, all the compression ratios for hearing restoration are relatively subtle. They always lie between 1 and 2. If you ever reach a compression ratio of 2 it means you have profound loss of more than 90 dB threshold elevation at those frequencies --- in other words, you have profound deafness at those frequencies, and no amount of compression will ever restore your hearing in those bands.

The equation used for developing the compression ratios is:

R = 90 / (90 - (0.5*(TdB-5)))

In other words, 5 dB are removed to account for sensitivity improvements of binaural hearing, and the audiology values are scaled by 0.5.

Some audiologists, who don't understand mathematics, may tell you about the 50% rule (my 0.5 factor here) saying that it is used because sensory deprivation makes one too sensitive to withstand a full correction. There may well be a psychological component, but I can tell you more correctly that the 0.5 scaling simply makes the linear compression at the loud end lie more closely to the required psychoacoustic compression.

Finally, these compressors work in amplitude or SPL (Sound Pressure Level) space, not in terms of Loudness. We hear Loudness not SPL. The famous Fletcher-Munson curves illustrate that difference for the case of isolated pure sine-tones. Music is more complicated than pure sine-tones and the Fletcher-Munson curves are not so useful there. But in any event, amplitude compression is not quite what is needed. Rather one needs Loudness compression. But the linear amplitude compression in RED is better than nothing.

Why am I giving this away? Well for one, it only runs on a Kyma as presented here. Secondly it is a far cry from a real Crescendo algorithm. A real Crescendo performs Psychoacoustic Loudness Compression in each of your Critical Bands. The sound is superb!

But RED is better than nothing. If you were to buy a Kyma system just so that you could run RED, then you would be wasting money. For the same money or less, you could have had a real dedicated Crescendo and much better sound restoration.

So I see no purpose to selling RED on the Kyma.... I may as well give it away for persons already owning a Kyma, and maybe I can reach a few more people needing help.

BTW -- We have a version of RED-6 that runs on the little Nord Micro (hence the name). This is a nice little unit that can be used for dedicated listening. It has line-in and line-out as well as a built-in headphone amp with volume control. It also has a pretty good audio stage and uses 96 KHz sampling with 24 bits. (But Capybara is nicer!!)

Cheers,

- DM

[PS: Most people have hearing loss if they are older than 40 years... But most often it is not severe and it affects mainly the high end above 4 KHz. Your compression ratios will tend to be between 1 and 1.2. Try RED and see if it helps you hear things you didn't even realize were missing!

Also, see our Web site at www.euterpegroup.com for an explanation of recruitment and why simple EQ is all wrong for hearing correction.]