[sdiy] Frequency shifted from BBD?

[brianw]

On Oct 21, 2024, at 12:57 AM, Kylee Kennedy wrote:
> You get a temporary pitch shift when you change the delay speed on a BBD manually. So I bet if you set up a scanner that goes through several BBDs that are shifting the speed as it scans you could replicate this effect. Each BBD acts as a playback head on that device. Probably phlanging would be included if there isn't a hard switch between BBDs via the scanner steps.

You can't do quite the same things with a BBD that can be done with a multi-head tape or DDL.

Even with multiple BBDs, you can't "share the same tape" between them, because if the clock is different for each BBD then what's inside the chip on that BBD's "tape" is very different. The tape-based shifter is literally using the same tape for all four playback heads. Multiple BBDs cannot contain the same recording unless they're clocked precisely the same at all times, and if they're clocked precisely the same at all times then switching between them will not make any difference since they'll all have identical audio.

i.e. Similar things are occurring with each frequency-shifting device, but there's a distinct difference between moving the playback head at a different speed than the record head, versus varying the tape speed across two fixed heads (record and playback).


With the multihead tape, you have one head recording onto the tape at whatever speed the tape is running, and another set of rotating heads that can "play" the tape faster or slower than it was recorded, with automatic crossfades as each head approaches the tape and then pulls away. There are still waveform glitches every time the playback heads transition from one to the next.

With a DDL, you can set up the recording to write into memory at a fixed sample rate, and loop back when the end of memory is reached. Generally, the output from memory is also at the same sample rate, but the offset (delay time) between write and read can be changed. There are glitches because both input and output loop around the finite memory, and they have to overlap at some point, depending upon the amount of memory and the sample rate. Not only do they overlap, but changing the delay time to get the frequency shift also causes glitches (assuming a constant output sample rate).

With a BBD, there is no separation between the rate input and output, because there's just one clock (pair) connected to all (odd/even) buckets. There's also no way to have the output move closer or further from the input - the audio must pass through all buckets. I'm not aware of any multi-tap BBD (unless you put several BBD in series). A BBD is like a length of tape with one fixed record head and one fixed playback head, where the distance between the two heads can never be changed. The only variable is the clock speed, which is equivalent to the tape speed. With BBD, you get a frequency shift because by the time the audio comes out the other end, the clock rate happens to be different than it was when the audio went in. The same would be true on a tape loop with enough distance between record and play, if the tape speed changed between - all without moving the heads.

With a BBD frequency shifter, the clock rate has to be continuously varying. The "tape head" never moves or leaves "the tape" so there's no glitch in that sense. However, whether you're constantly increasing or constantly decreasing the BBD clock, you eventually run into either the upper or lower limit of the clock, and you have to jump to the other extreme. This causes both the input and output to glitch. The output glitches right away, and later the glitch at the input appears on the output after it's gone through every bucket. You end up needing multiple BBDs, not because they can play shared audio at different speeds, but because you need a crossfade when the glitch is occurring on one BBD, and you need to stagger the jumps in clock rate so they don't happen at the same time.

Brian

> On Sun, Oct 20, 2024 at 11:12 PM Didier Leplae wrote:
>> Yes! That’s surprisingly simple solution for pitch shifting on tape. I’m trying to think if such a brilliantly simple solution is possible for BBD though.
>> 
>> On Oct 20, 2024, at 7:45 PM, Ben Stuyts wrote:
>>> For all you frequency shifter and tape delay aficionados:
>>> https://youtu.be/U-4AMxkpyh0
>>> That’s brilliant! :-)
>>> 
>>> Ben
>>> 
>>> On 9 Oct 2024, at 07:07, Mike Bryant wrote:
>>>> 
>>>> 	• Besides the point that Tom raises, there's another consideration: Using an MCU to program a complex digital signal is often more wasteful (think: battery-draining) than judicious use of logic gates.
>>>> 
>>>> Sorry but this usually isn't the case.  Unless you are switching at a very low rate, discrete logic gates have to drive the capacitances on the PCB between each other whereas MCUs do all the work internally and just output the final waveform.   10 cent OTP MCUs cost less than many logic gates due to higher volumes and usually use less power and PCB area so system cost is lower.
>