Analog delay sound without the delay

[Taylor]

I really like the sounds of analog delays, and PT2399 delays, but would sometimes like to get this lo-fi sound without the delay. There is an album by A Silver Mount Zion called Pretty Little Lightning Paw in which most of the instruments have this particular sound; in many cases it sounds like the entire mix has been put through an analog delay.

If I heard somebody else asking this question, I would recommend very soft clipping (like 5 pairs of diodes all in series, either in an opamp feedback loop or to ground) with a low-Q bandpass filter. Basically light saturation/compression, cut highs and lows.

That would be a cool effect, but I don't think it would go far enough towards the sound I want for this.

What is the analog delay chip with the lowest number of stages? This way we can clock it longer for lower fidelity without too audible a delay. I think I could handle 20ms delay and still find it usable.

Perhaps instead I could try to make a discrete BBD circuit, since I want it to be very short in time? Since I want this to be lo fi I could get away with the poor matching of transistors in the bucket brigade... Hmm.

[Hides-His-Eyes]

I love that album. You know how they got that sound?

After they mixed the entire thing, they played it through a crappy boombox into a pair of microphones. The whole mix.

#Myst'ry, and wonder
did light up the valley
to be beat back, by dark clouds
and a harsh shrieking wind

And the battle that raged there
for three awful months now
just to stop
for a minute
and then start up again

Your hands like birds in the trees
If the trees themselves were on fire
Your lips on mine make a choir
singing
[God is great]#

That's from memory, that song does weird things to me every time I hear it

[Hides-His-Eyes]

Just realised how unhelpful that post might have been

Could you use a compander into that soft overdrive? make every note equally distorted while preserving the dynamics?

[Taylor]

Well, I actually knew that they played it through a boombox, but didn't reveal that info because I thought people would just say to do that.

But in particular in the song you quote, it sounds to me like the voice, guitar, cello, etc. have all ben put through analog or tape delay. That track sounds much more lo fi than the first one for example.

I agree about that song, and whole album - there's something really special about it, at the risk of sounding silly.


Edit: hmm, I don't think I should be attempting this kind of heavy mental lifting at 4 in the morning. Discrete BBD sim try #1. Very wrong. 

http://tinyurl.com/26qs8v2

[markeebee]

If I heard somebody else asking this question, I would recommend very soft clipping (like 5 pairs of diodes all in series, either in an opamp feedback loop or to ground) with a low-Q bandpass filter. Basically light saturation/compression, cut highs and lows.

http://www.diystompboxes.com/smfforum/index.php?topic=81357.msg673733#msg673733

Where were you when I needed you?

[Penthode]

I think it is the sound of a Mutron Phaser http://en.wikipedia.org/wiki/Efrim_Menuck

[maarten]

What is the analog delay chip with the lowest number of stages? This way we can clock it longer for lower fidelity without too audible a delay. I think I could handle 20ms delay and still find it usable.

There were some very low ones, like TAD32, 32 steps, or TCA350Z, 185 steps, but probably these will be very hard to find. SAD 512 (yes, 512 steps) maybe is still to be had somewhere? Or TDA1022 (512 steps) - I recently bought some of those. SAD 1024 is 2 x 512 steps, usable either in series or just in half (or maybe in parallel?).  Also the MN3001 is 2x 512 steps, I think (never used those), MN 3004 is 1X 512.

Lowering the clock too much will make it become audible. So if you filter at say 7 or 9 kHz, your clockfrequency should not be lower than (preferably) three times 7 or 9 KHz.

Maarten

[Scruffie]

What is the analog delay chip with the lowest number of stages? This way we can clock it longer for lower fidelity without too audible a delay. I think I could handle 20ms delay and still find it usable.

There were some very low ones, like TAD32, 32 steps, or TCA350Z, 185 steps, but probably these will be very hard to find. SAD 512 (yes, 512 steps) maybe is still to be had somewhere? Or TDA1022 (512 steps) - I recently bought some of those. SAD 1024 is 2 x 512 steps, usable either in series or just in half (or maybe in parallel?).  Also the MN3001 is 2x 512 steps, I think (never used those), MN 3004 is 1X 512.

Lowering the clock too much will make it become audible. So if you filter at say 7 or 9 kHz, your clockfrequency should not be lower than (preferably) three times 7 or 9 KHz.

Maarten

I've got some of those TCA350s...

Shortest is the MN3006/3206 128 Stages, Found One Once... Think I still have it, not sure of there availibility in bulk.

Just Buffer the Clock and Push a 3207 Past its limits though for ease of chip source.

[p_wats]

I use a modded Clari(not) to give a light tape-like warble and thickness. All you have to do is add a switch to lift the fuzz, then go wet signal only with some light modulation. Sounds great---kind of like you've been recording over the same tape for years.

[Processaurus]

Remember analog delays resemble digital audio, in that they have a sampling rate and come out of the BBD in a stairstep.  One of the aliaser effects like the nyquist aliaser or cmos aliaser followed by a sharp lopass filter would do an identical mangling of the sound, minus the delay.  One could even introduce noise on the input to cause errors in the S/H, to simulate the noisy process of long BBD delays.

Something like that could be grafted into an existing analog delay, just by replacing the BBD chip with a simple S/H circuit, using the existing clock to run the S/H.

I like using the Lo-fi delay model on the DL4 minus delay for the reason you're getting at, just as a gross, distant pre-amp to garble the sound.

[Processaurus]

Totally different approach, but if you used a real BBD and introduced errors into the clock, like on some semi-irregular basis one of the clock cycles is longer than the others, it would put unique little burbles into the audio.

[Processaurus]

If I heard somebody else asking this question, I would recommend very soft clipping (like 5 pairs of diodes all in series, either in an opamp feedback loop or to ground) with a low-Q bandpass filter. Basically light saturation/compression, cut highs and lows.

You can get very soft clipping with zener diodes in the feedback loop, like this:
--------|<|------|>|-------

[Taylor]

These are all really great ideas, thanks. I will probably explore them all.

I do still want to mess around with a discrete BBD, just for the sake of an unusual circuit design.

[Taylor]

http://tinyurl.com/26qs8v2 discrete BBD sim. Very Wrong.

It really has nothing to do with a discrete BBD, because I put that together while most of my brain was already in bed, but look at that waveform...! Funky.

[PRR]

> Discrete BBD sim try #1. Very wrong.

You need a 2-phase drive for switches. Holding one phase at +9V won't do. Wobbling the other phase +5V to -5V is not workable with the +9V switches. For the simple BBD you want switch cap switch cap switch cap .... with caps grounded. (The integrated BBD may do it different.) Your switch rate should be over twice the signal rate; you appear to have both set to 40Hz? And your input is labeled Out, which may not fool the sim but confuses people. You have two different value caps; I suspect 10uFd is way-high and of course even 10nFd is more than you get on the chip. I think the buckets are a few pFd; to get anything out you need a source-follower at the end.

try this Falstad sim.

[Taylor]

Yeah, the only reason I even posted that link was for myself, so I could click it when I was awake and make it right. Thanks for that - at 400hz clock and 40hz audio, input looks pretty much like output. Lowering the clock gives us a nicely drunk-looking waveform.

Any thoughts on what kind of MOSFET would be similar to what's inside a BBD? Or does it not really matter? All n-channel, right?


These patents show BJTs, strangely...

http://www.freepatentsonline.com/3402355.html
http://www.freepatentsonline.com/3546490.html

[PRR]

> input looks pretty much like output

BUT 4/(400Hz/40Hz) late (4 stages at 10 clocks per signal cycle).

You ARE correct, the commercial BBD has caps crossed on clock lines and series switches. It also has a grounded input cap, and the added switches are biased to 14/15th of Vdd(?). If I ever knew how this worked, I've forgot, and the old papers are dust.

_I_ think the "sound" is the 512 cascaded followers. If each one is 0.01% THD, it comes out 5% THD, with an IM richness (or harshness) you can't get with any sane number of stages. A boardful of unity gain hex inverters might be similar.

In delays there is an additional high-frequency distortion due to low sample rate and limited bucket-cap charging time.

A 1024 stage delay clocked at 500KHz is only 2 second delay. Like the listener were standing 2 feet further away. Delays approaching 40mS will mess with live speech/music, but a pure 2mS delay is fairly invisible, so the im-pure BBD 2mS delay should not have annoying latency. And while BBD chips are no longer cheep, they are available, and probably cheaper parts and labor than boardfuls of CMOS.

[Taylor]

probably cheaper parts and labor than boardfuls of CMOS.

Yeah, true.

[PRR]

> A 1024 stage delay clocked at 500KHz is only 2 second delay. Like the listener were standing 2 feet further away

Darn fingers!! 2 MILLI seconds.

The boardful of CMOS inverters sounds like an interesting project, but labor has to be awful cheep.

[Taylor]

Sweet sassy mo-lassy...

http://www.effectsdatabase.com/model/zvex/instantlofijunky

Zach, since you're reading this thread, can I get a credit on the bottom of that thing like Chuck Zwicky on the Mammoth? 



Just being silly of course.