Warning: count(): Parameter must be an array or an object that implements Countable in /home/twobros2/bluesbros/kbj/wp-includes/post-template.php on line 284

Ken B Jammen

The Essential Ken B Jammen

The long and drawn out rambling about the weakness of digital processing.

December 8th, 2014

Digital signal processing has been around for quite some time.  In the early 1980’s you could get a digital delay on your floor board pretty easily.  Since then signal processing has taken a new step, from not only processing the signal, but making your guitar signal sound like you have an amplifier even when you do not.

Historically, the Boss DD-2 was the first digital pedal that Boss made and the world’s first stompbox Digital Delay. It is based around the chip made for the Roland SDE-3000 rack mounted digital delay. By luck, the chip just fits across the width of the compact pedal and for months the design time worked to squeeze the rest of the components on to the printed circuit board. The design is utilizing a 12 bit AD converter producing a flat frequency response between 40Hz and 7kHz.

So how does the sound get into the stompbox?   That is done with what is called an A/D/A convertor.  That is a computer processor that takes an analog signal and converts it into a digital signal and back.  Let’s discuss what a digital signal can be to a computer.   A computer can only process one of two things.  That is either a 0 message or a 1 message.  Think of it as either off or on.  Morse Code is a good example of binary communication.  Either a short burst or a long burst could be translated into letters and a complete message.

Obviously one on/off at a time does not provide a wide dynamic range.   That’s why we need more bit depth.  As in the DD-2, the 12-bit depth gives 4096 different levels per sample.  That may seem like a lot, but in reality it is not.  I worked for a short time in development of a new amplifier modeling system.  We totally rethought what the human ear could do.  What we found out is that every human ear is different.  What we discovered is that there is a portion of the population that cannot tell the difference between 12 bit digitizing at 12khz (very poor, less than FM signal) and the analog signal, and a portion (albeit small) of the population that can tell the difference between a 320khz at 64bit digitized signal and the analog signal.

So, throughout the history of digital effects there has been striving for realism.  Sampling theory would say that once 24khz sampling is reached that would contain every frequency audible to human hearing there should be no further quality improvement necessary.  Although the logic behind that theory seems sound, it is flawed.   A flute playing at 440hz, sounds different than a violin at 440hz, sounds different than a square wave synthesizer at 440hz.   What the ear hears is the difference in the shape of the waveform, not the frequency.  That’s why all of the instruments in the orchestra sound different.  So a sample rate of 24khz may get all of the frequencies, but not all of the characteristics of the waveform.  A 24kz sample rate can only get that the frequency of the 24khz tone is 24khz, and no waveform data from the 24khz waveform.   In our brief modeling study we determined that if

This logic helps us understand why effects in the higher note register sound less appealing than those is the lower pitch register.  A lower pitched note at 50hz can be sampled at 48 distinct points with a 24khz sampler.   Let’s take that same processor sampling a note at 4000 (4k)hz.  That processor will be able to get 6 points.  The general waveform will be sampled, with its note value, but tonal characteristics are lost.  What happens at the higher frequency is the waveform becomes interpolated or “guessed.”   That’s why although at the higher frequency digital sound processing sounds so processed, but lower frequencies sound great.

So we should see more digital bass modeling and less guitar modeling because the guitar modeling sounds worse that bass modeling?  Well, we don’t and that’s because the industry doesn’t want it.  It would take a computer as powerful as the world’s fastest computer to be able to sustain processing the signal at a transparent rate.  So we are not there yet, but we will get there someday.  At one point in time, small computers were the size of a closet and had the computing power of a small calculator.  It might not be today, but at some point we will have digital modeling that is transparent.

So what’s with all this rambling.  It goes to the summary point.  Digital modeling is adequate enough for you if… it sound adequate to you.   Where digitial processing is going is exciting, and it will be equal or better to analog at some point in the future.