ilovefuzz.com

Heya

So, there are an increasing number of digital pedals on the market that feature some means of adjusting what is variously called "sample rate," "sampling frequency," "clock," etc.

Usually this is an attractive thing to me because lofi

I'm quite familiar with sampling rate reduction (SRR) and bit crushing as relatively in-your-face ways of mangling digital signal and introducing aliasing, heterodyning (i think that is applicable with some implementations of SRR) and, maybe moreso with bit reduction/crushing, increased noise and distortion that is more akin to brickwalled squarewave than the gentler forms of distortion seen in mild gain-related clipping. This is all very cool to me, and since my first Geiger Counter yoinks ago i've used these effects fairly constantly.

More recently I've been noticing pedals, usually FV-1-based, usually delay and/or verb types (including glitch machines), offering controls like "clock" or "resolution" (tho i thought resolution had more to do with bit depth than sampling rate? ) that tend to impart an overall more subtle effect that i perceive as "graininess" or "crispiness" or maybe "crunchiness" that i also tend to associate with rack effects of the 80s or 90s (which usually weren't actively going for a lofi aesthetic but, rather, because markets and budgets are gonna market and budget were artifacts of whatever technical constraints were imposed by the available tech in the allowable cost category.

Some examples of pedals i'm thinking of here are the Chase Bliss Mood ("clock") (and I guess the CB/CFX Gen Loss as well (("gen") (tho this sounds pretty much like SRR to me)) and the Dr Scientist Atmosphere ("res"), but there are others like the Alexander Pedals Syntax Error ("sample") and Radical Delay DX ("clock"), and plenty of others I'm sure (and maybe it would be fun/useful to compile a list of these pedals here ).

But my main ask is - what appreciable difference beyond degree is there between applying an SRR effect to a signal and adjusting downwards the sampling frequency/rate/clock of a time-based effect? Are they just more or less extreme applications of the same thing? idk why, but my brain can't seem to work out the distinction, and so far the internet hasn't been helping (maybe because there isn't really a difference?)....

my poor befuddled brain thanks thee in advance for any assistance ye can offer so i can stop worrying and love the system slack

yeah, i figured it might be something like that - sort of the difference between implementing a feedback loop in a delay pedal and sticking an EHX Analogizer in a feedback looper pedal

thx

i'll dig up some youtube links later

ok here's one - check out the following time stamps:


02:42​ Plate (LFO: Resolution)
03:19​ Smear (LFO: Resolution)
10:13​ Aether (LFO: Resolution)

-Infinite Reverb-
11:25​ Alias (LFO: Sample Rate)
13:41​ Alias (LFO: Resolution, Sample Rate and Alias Mix)

[youtube]http://www.youtube.com/watch?v=1THR1iBVkBI[/youtube]

so, yeah, this is adding to my confusion (well, my brain problem situation is adding to my confusion, but this isn't helping)

as you can see, in this video the demoer has the Atmosphere's internal lfo varying Res (from the manual: "sets the resolution of the DSP engine that makes the reverb effects. Up all the way is full 32kHz sample rate, turn it down all the way to 16kHz sample rate for a lower, slower, longer, grainier reverb sound.")

and at one point the demoer has the lfo varying res *and* sample rate on a reverb algo called Alias which, according to the manual, allows the user to "[r]educe the sample rate of the reverb" with controls for sample rate and alias mix.

One obvious thing i do notice the res control (and other similar "clock") controls doing is changing the pitch of whatever is in the buffer, which isn't what happens when i change a sample rate knob on an SRR

also - add Eventide Rose (not FV-1) to the list of pedals that allow adjustment of global sample rate (as opposed to SRR)....

i'll throw up a demo (not mine) of the CBA Mood with its quantized clock adjustments and the CBA/CFX Gen Loss with what sounds to be SRR later.

I guess the Afterneath v.3 also seems to be doing quantized system clock adjustments of the drag control (which strikes me as being the same thing the Mood is doing with its clock control) in its extra modes (and non-quantized via the Drag knob itself in all versions)....

A lot of the earlier stuff (like the 80s-90s racks you mention) were pushing the technological envelope to get something passable in terms of printed spec. They were/are easier to push off the rails for sure.

I think an aspect of "that sound" with relation to sample rate has to do with working the converters beneath the Nyquist frequency, so the artifacts/hash coming off the D/A converter is now passing under the low pass filter - which is a fixed analog element tuned to roughly the halfway point of the sampling bandwidth.

So, as an example, if I have something like a Casio RZ-1 (8 bit machine running a 20kHz sampling rate) and modify the clock to run at 12kHz, the Nyquist knee drops to 6kHz so we have a healthy window of aliasing flying well under the 10kHz knee of the analog section. Nice and crunchy (I would ASSume, I have not performed such modifications on that machine - it has character enough). Furthermore, these "problem areas" were a lot closer to the audible spectrum, and have since been pushed out closer to the world of light speed.

Please note, I may be absolutely incorrect in the following conclusion, and I am absolutely certain that there are probably (...) designers/builders who have circumvented this particular sandtrap...

Anyway, when we're looking at modern converters, we are looking at a *vastly* "superior" topology. Furthermore, a lot of the effects are done in software, which just bit feeds the hardware (which is still lock step with an xtal in circuit) with either reduced definition of steps (bit reduction) or reduced processor events (sample rate reduction) while the actual DAC circuitry is running in its design optimized state and so manipulation of digital format does not produce the same taste as yesteryear, because the lines that were crossed to get this 30 years ago have been moved.

missed this - thanks i'm going to take my time reading this and see if it penetrates my fog

Dandolin wrote:missed this - thanks i'm going to take my time reading this and see if it penetrates my fog

I was trying to avoid really breaking it down into the rudiments, as it (digital audio) is a pretty deep subject. I'm happy to belabor the nitty gritty more if you think that would help, I'm hardly an expert though so I tend to prefer just keeping my mouth shut (hahahaha).

always appreciate your input - and after one reading some things are starting to seem clearer in my head

I'm going to try to explain this without drawing lots of pictures, and maybe you know all this already, so hopefully I don't come across as a mr knowitall.

When you're sampling a wave, you can only reproduce the wave as accurately as the number of samples you took and the bit-depth of each sample. Let's say I had a bit depth of 5, a sample rate of 8kHz, and I was sampling a 1kHz sine wave. You would get eight thousand samples every second, each one representing between 0 and 4 (ignoring the fact that it would actually be in binary, but understanding that 0-4 is actually 5 numbers). It would look like this: 23432101 ...a thousand times a second. With that, you could recreate a stair-steppy sine wave at the correct frequency and amplitude (the x-axis is the middle of the resolution, or 2):



If you kept the sample rate the same, but decreased the bit-depth to 3 (0-2), it might look like 12221000. That sucks. Now you've got a sort-of square wave at half the frequency. Let's make the bit-depth bigger, say 10 (0-9). Now, you've got something like 46974202 - which is enough data to know that the sine wave is slightly asymmetrical. So, bit-depth (also called word-length) gives you important info about what the signal looks like *at the time of the sample*, and is an important part of resolution.

Taking the same example, (1kHz sine wave and a bit-depth of 10), but halving the sample rate to 4kHz, you'd get something like 4940 (a thousand times a second). This sucks too. It's a triangle wave at half the frequency again (sort of like when we halved the bit-depth). You can see how resolution is actually the combination of bit-depth and sample rate.

The other things to keep in mind is that there's a big difference between the *sampling* rate and the sample play-back rate. If you took my bit-depth of 10, sample-rate of 8kHz example from before (46974202), but played it back at half-speed, you'd get a sine wave of half the frequency, even though the "sample rate" of the data was 8kHz, the playback was only 4,000 of those samples in a second. This is how lot's of pitch-bending stuff works...playback speed. If you're playing stuff back slow enough, they'll often "interpolate" in-between samples to make it sound less jagged - so if the samples are usually 4 6 9 7 4 2 0 2, they'll add new samples to smooth it out, becoming 4567987643210123. That can give the impression of a higher sample rate despite not being samples of the actual signal.

The problem (like with most technical audio stuff) is with marketing and incorrect usage of terminology. A "tremolo" arm is for vibrato, and on many devices, "Sample Rate" is sample play-back rate. "Resolution" should be a combination of bit-depth and sample rate, but often is used to describe only bit-depth.

I don't know if any of that makes sense or helps. I wish we had a whiteboard in this virtual classroom.

Awesome explanation, thank you!

Whatever it is what it does, I do for sure love the Hotone KRUSH for those sonic experiments:

[youtube]http://www.youtube.com/watch?v=dcvAlg4xONc[/youtube]

[youtube]http://www.youtube.com/watch?v=FNa3uyWfL_Y[/youtube]

PeterBregman wrote:
The other things to keep in mind is that there's a big difference between the *sampling* rate and the sample play-back rate. If you took my bit-depth of 10, sample-rate of 8kHz example from before (46974202), but played it back at half-speed, you'd get a sine wave of half the frequency, even though the "sample rate" of the data was 8kHz, the playback was only 4,000 of those samples in a second. This is how lot's of pitch-bending stuff works...playback speed. If you're playing stuff back slow enough, they'll often "interpolate" in-between samples to make it sound less jagged - so if the samples are usually 4 6 9 7 4 2 0 2, they'll add new samples to smooth it out, becoming 4567987643210123. That can give the impression of a higher sample rate despite not being samples of the actual signal.

Thank you! The bit I've excerpted is especially helpful to me in making out the differences between what is actually being done and what the control labels/manuals say