Digital Vs. Analog Delay

[Bluesgeetar]

Well my input is,  I have an Ibanez PDM-1, BOSS RV-3, Korg SDD3300 (what a monster, 2 full rack spaces!), and a 1967 Maestro Echoplex.  I love all of them!  The SDD3300 has a preamp section to die for and it does things that only God could make happen before it.  The PDM-1 must be pretty good, it is Carlos Santanas only delay unit he uses and has been using since the late 80s it has 5   JRC4558 IC chips in it, yes that is 5 of them babies!  It sounds heavenly.  The RV-3 just sounds damn cool at the right setting.  Can sound just like a slow gear at the right setting.  The Echoplex is well, pure diamonds and gold.   Hell I love all of them!  They all have their own cool sounds and personality.

Has anyone figured out just what is it in Maestro effects that pass on that strange heavenly tone that all Maestro effects pass on to any signal going through them?  I seen this question brought up on many forums in the past.  Many folks report the same experience with Maestro effects.  A strange tone sweetening effect not found on any of their other old vintage effects.

[Ed Rembold]

Well, the simplest way I can think to educate your ears is to do the following.

You'll need a sinewave signal generator,  headphone amp, and a willingness to disconnect the dry mix resistor from the final mix stage, so that you'll be listening to Only the delay signal hrough your headphones.

Now start with 400hz, set delay time at max, and take a listen- you'll find wet only, to sound pretty close to dry.
bump down to 80hz, first sign of a problem,  low freq takes on a buzzyness
bump up to 4khz, now regardless of the LPF cutoff freq. there will still be still be some signal, and it will have "artifacts" not present in the dry signal,  and if you sweep your signal generator from say 1khz to 6khz
your ears will be amazed at the "un-natural" junk riding with the signal.
A poorly designed BBD based delay will also have "junk", but different sounding.

I hope this is clear,  but I'm no writer or debater,  I would expect makers of digital gear to take offence at my comments,  but that it Not my intention. digital has it's purpose as a looper, no argument......

analog BBD done right, is best for echo.

Ed R.

[Maneco]

Thanks very much for the information,i've learnt a lot from this topic

[Ed Rembold]

Mark,
From your earlier questions-
1) What it is I prefer in analog over digital, is analog BBD technology
suffers from only one of the two forms of "un-natural distortion" that digital suffers from.

Both suffer from "aliasing or foldover" distortion, which is created any time the input frequency is 'higher" than 1/2 the sampling frequency.
This creates the "tinkling, ring-mod" undertone most noted riding on the decay.  

In both technologies the cure is- never let the sample rate enter the "audio range" times 2. (or 3)  
this is easy now with modern digital, hard with BBD's, but possible.

Digital suffers alone with "quantizing" distortion,  which caused by the fact that each sample of audio is assigned a "number value". for example if the number values were 1,2,3,4 etc. but the audio value is actually 2.1 or 3.7, the digital process will round up or down, in this case to 2.0 and 4.0,
this causes 'quantizing distortion" on the playback.
This creates the "gritty" undertone,  audible at all frequencies,  not easily filtered.  And can be heard in any digital process under 8 bit.

2, 3 and 4)
Digital delay Can be done right- best example would be the Korg SDD series. I'd love to get my hands on one- to see for myself.
I've said enough,  I meant to offend no-one. my apologies.

Thanks, Ed R.

[Nasse]

I've said enough, I meant to offend no-one. my apologies.

Well I think I poured gasoline to the flames claiming there is some useless debate when people were discussin bout fine art of delay and echo effect

But anyhow hope discussion continues under this or other topic...

And original question, chips like those cheap and cheerful PT2955s, can they sound good...

[Ed G.]

I've said enough,  I meant to offend no-one. my apologies.

Thanks, Ed R.

I don't think anyone took offense, and I don't see where an apology is needed. Everyone has their opinions. You went a step further and gave some good information behind those opinions. This was a totally enlightening and interesting discussion, frankly, I'd like to see more like these.

[RickL]

This thread came along at a good time for me. I just picked up a beat DD-2 at a pawn shop for $40 so I figured I could safely muck with it without drastic consequences if I did something horribly wrong.

On the DD-2 the outside lugs of both the echo volume and repeats/feedback are connected together. Lug 1 (CCW) is ground. Putting a 0.01uF cap across these lugs gives a slight rolloff of the treble on the repeats but it doesn't accumulate (each repeat doesn't have less treble). I tried increasing the cap size, as high as 1 uF, which increased the treble rolloff but also decreased the number of repeats with the feedback control at maximum. Using a 1 uF cap I got maybe three repeats vs dozens with no cap.

Putting the cap between the wiper of the echo volume and ground had no effect and between the wiper of the feedback and ground was the same as between the two outside lugs.

If I could get the repeats to loose increasing amounts of treble without the loss in numbers of repeats I think I would mod it permanently but as it is I don't think it's worth the bother.

Incidently there is just enough room in the DD-2 to mount a mini toggle just above the echo volume control, beside the "effect on" LED. If the psuedo analog mod doesn't work out does anyone (Mark?) have any other suggestions for a worthwhile mod to this pedal?

[Ansil]

in a nutshell i just set two of them side by side and fed  the tape through both heads..  used onboard preamps

[Nasse]

:? Many "better-than-average" digital units seems to leave something desirable features not included, like high enough sampling rate, eq/filtering settings for various echo/delay effects and feedback/multitap facilities...

So an ideal delay machine would be one with high delayed signal quality, and steep and versatile filters with huge adjustable range of corner freq...

Anyone remembering Graig Anderton "Mudguard" project (steep highpass filter to remove "mud" from your reverb/delay, and the euro/brit "instro/surf" and rockabilly players swear by lopass filtering...

[smoguzbenjamin]

Rick, My DD-2 gives me almost infinite repeats with maximum feedback You could try playing with the trimmers inside

[RickL]

I get almost infinite repeats without the cap mod too. I would like to know why adding the cap so drastically cuts down on the number of repeats. My guess is that by cutting out some of the signal with the cap there is less to amplify on each repeat. I think if I tweak the trim pot for more repeats with the cap mod it will get uncontrollable when the cap is switched out, or at least severly reduce the usable amount of pot rotation.

[smoguzbenjamin]

I believe there's a clock trimmer there that will allow you to set the delay time a bit shorter/longer. I tweaked it a bit, but make sure you mark where the trimmer was at first! When I was done tweaking it I got a little longer delays, not much though. I'm too lazy to do any really extensive modding :mrgreen:

[troubledtom]

ed r.
 you post a very tight case. i have 2 roland sde 3000 delays that i wouldn't
trade for anthing. some very hard hitters still use them . i'd say they're about 20 yr's old. i'd also say you know more tech stuff on this subject but check it out. i've had pro's offer me what i paid for them, of course they are in mint condition. $1000. bucks each.
http://www.theorangepeel.net/inside/SDE3000.html
       peace,
         tom,,,,,,,tom......tom,,,,,tom...to......t............................. :wink:

[smoguzbenjamin]

24W for a delay! WOW! That's more than my practice amp :lol:

[Mark Hammer]

Rick,

Yup, cutting bandwidth also cuts overall amplitude.  Some designs, of course, have recirculation trimmers where you can offset amplitude losses with a trimpot.  Ideally, what one wants is a full schematic could indicate where to insert a proper active unity-gain lowpass filter, maybe even a 2-pole job.  Admittedly, the cap-to-ground solution is primitive.  I've had good luck with it.  Not sure why you seem to get an all-or-nothing outcome.

Ed,

Don't be too strongly persuaded by the textbook sampling formulae.  The reason one wants sampling of at least double the highest frequency is that it is difficult to accurately reconstruct a waveform without having at least one sample to reflect the positive half-cycle, and another to reflect/sample the negative half-cycle.  

Of course, there is no divine law which says that the peak of the positive half-cycle will occur at the precise moment the clock tells your gadget to grab a sample NOW! (and ibid for the negative half-cycle).  For that reason, of course, one wants as many samples per wave-form as possible with as much resolution as possible.  The resolution in terms of word-size (8-bit, 10-bit, 16-bit, etc.) reduces the likelihood of quantization error as you correctly point out, and the resolution in time (which is essentially what sampling *rate* is) assures that the actual changes in time to the signal are faithfully depicted.  All waveform changes should be accurately reflected as far as their "height" is concerned (no serious rounding up or down), and no changes should occur in between samples.  The more samples you take, the less the likelihood of any changes occurring between samples - i.e., keep watching and you don't miss anything.

The thing one needs to remember, though, is that the clock rate is fixed, but the frequencies the device is sampling vary all over the place.  Whether analog or digital, a 20khz sampling rate does a piss-poor job of "describing" a 10khz waveform sitting atop other content.  On the other hand, a 20khz clock rate (and whatever delay interval that corresponds to with the chipset in use) does a pretty decent job of representing a 500hz waveform, with little error (or rather, MUCH much less).  Indeed, it is a truism that for any sampling rate waveform fidelity will increase as audio input signal frequency decreases - the slower the waveform changes, the easier it is to keep up with.

This is all the long way of saying that the quantization error and other sources of distortion stemming from the act of sampling at a fixed rate with a fixed word-size, create problems in proportion to the bandwidth one is aiming for.  Aim low enough (e.g., 1khz bandwidth with brickwall filters) and the problems all go away, assuming the word-size is acceptable (e.g., I wouldn't expect any such "filter miracles" with 4-bit sampling).  This is why imposing steep lowpass filtering to "warm up" a digital delay renders it acceptable and indistinguishable from analog to many users.  What that "magical rolloff" might be with vary depending on a bunch of parameters, but I have no doubt that there will always be a filtering configuration that achieves the goal of blurring the lines between A and D.

Analog samples WILL leak if contained too long (i.e., clock rate too slow), so they are subject to their own type of quasi-quantization error.  And if you recirculate the signal, that error is multiplied.  But of course, you can repair/mask that with filters.  Mike Irwin designed an ultra-long analog delay for Modcan, and he tells me that if you set the bandwidth ridiculously low, but use very good filtering, you can have the thing recirculate for a half an hour and "still recognize" the original signal in there.  Of course, a human brain recognizing it as somewhat similar to what you started out with is a very different thing from having it look similar on the scope.

Admittedly, there is a tendency to aim budget delays a little higher than they ought to be aimed (i.e., striving for more bandwidth than the filters, sample rate, word-size, and application might call for) and that WILL introduce objectionable qualities to the sound.  I guess the suitable analogy is this: take a picture with a webcam and blow it up to front page headline size and it will look absolutely terrible in comparison to a blow-up of even the cheesiest disposable 110 camera.  Make them both B&W 1.5" x 1.5" images, though (like the little head and shoulders shot in a column byline), and whatever image quality differences do exist between these technologies will completely disappear.

And Ed, you could only offend me by losing your enthusiasm.  

[Ed Rembold]

explaination of "apology"....

Sometimes I forget that when I post something here, that I'm not talking to a few friends, but rather to the world.

We have vetran dsp programers,  expert digital designers,  owners of excellent rack gear.  etc.

And my focus at times becomes too narrow,  I tend to just focus on the little world of analog pedals.

Great comments/points were made by all.

I have heard/tested all the common available digital delay pedals.
but not all the best rack digital delays.
I do still own a Marshall JFX1 (digital rack delay) I used to think it was great, until I was able to compare it to an analog delay of equal delay time- then and only then, could I tell what I was "missing",  and then and only then, did it's "digitalness" become So apparent.  
Ed R.