BBDs: ADA Stereo Tapped Delay

Started by Noumenon, March 11, 2005, 10:24:45 AM

Previous topic - Next topic

Noumenon

So...

Finding a STD-1 is a project that takes years (or just luck). So I was thinking about building one instead, but the MN3011 isn't made anymore, so the question is if you could do a version of the chip using six single tapped BBDs.

Any ideas, comments?
Anyone interested in developing it with me?

Mark Hammer

If you can't find an MN3011 (they exist in pockets here and there, but they're not cheap), Matsushita also made the 5-tap MN3214 (never actually seen one sold anywhere...ever).  Not quite as long a delay but certainly optimized for flanging and chorus.

As many others cleverer and better-trained than I can tell you, there are some serious noise control and layout problems to be addressed when multiple clocks are running on the same board.  I don't profess to understand heterodyning, but I know it's not good.

Through the good graces of someone who comes here occasionally, I was able to come by an MN3011.  I am presently trying to complete a design for a scaled down version of the STD-1 that could fit in a 1590BB.  Depending on how successful my efforts are, this *might* turn into a commercial product or possibly a kit.  Of course, if I don't get down to work, it may just turn into smoke blown out of my pompous ass.

At the very least, I will certainly report back to the group here about what I see/hear/learn during the build.  The taps on the MN3011 are fairly far apart and there are a number of possibilities I've neverever heard in there somewhere.  Unfortunately, I have never actually HEARD an STD-1, so I don't know exactly what to expect.  I may veen be seriously disappointed.  If someone knew of a recording that used one that might help.

But it's a pretty damn tantalizing chip, I tell you.

Noumenon

Ok, regarding about a recording of the STD.-1, listen to any studio recording with Fredrik Thordendal (meshuggah) and you'll hear it. He always uses it to "spread" the sound. And live he has a foot controlled for the delay time to get that wacky pitch/whammy sound on the delays (which I LOVE). :)

Noumenon

Quote from: Mark HammerThrough the good graces of someone who comes here occasionally, I was able to come by an MN3011.  I am presently trying to complete a design for a scaled down version of the STD-1 that could fit in a 1590BB.  Depending on how successful my efforts are, this *might* turn into a commercial product or possibly a kit.  Of course, if I don't get down to work, it may just turn into smoke blown out of my pompous ass.

One question regarding this eventual kit, how would it be possible without the mn3011??

Mark Hammer

Well, first off, the "kitness" aspect is way off on the decision horizon.  Far enough off that I probably shouldn't have mentioned it....so please look at this light sir.  PAFF!!!

Running two separate BBDs with different clocking invites heterodyning and extra noise from the way the different HF clocks interact. Puretube/Ton and others assure that this CAN be overcome (and the Boss DimensionC shows this to be true), but not without a more complex design and higher parts count.  Running MORE than 2 or 3 clocks is likely to be quite problematic, unless you felt like making a big board for a rackmount kind of thing.

The "problem" is that BBDs tend to be multiples of each other in terms of storage: 128, 256, 512, 1024, 2048, 4096 stages, etc.  That means you can't drive them from a single clock without experiencing annoying resonances.  The MN3011 and MN3214 provide taps that are NOT multiples of each other, so they avoid problems with highly resonant feedback.

Well, let me qualify that.  If you were ONLY using feedforward and simply feeding the outputs of, say 256, 512, 768 (a 512 and 256 in series), and 1024 stages to alternate outputs, using the same clock to drive them all, perhaps you wouldn't have that big a problem with noise...assuming NO regeneration/feedback is involved.  Indeed, if you could get your hands on a bunch of 128-stage chips (MN3006), and create different delay times by placing a few in series, you could actually make something pretty lush in stereo.  Again, this assumes you aren't using and regeneration that might cause resonances to emerge.  I might point out that the MN3006 has a much higher dynamic range than most other BBDs, so maybe there is something to digging around for a stash of these.

Another route to consider is to check out the full-bandwidth all-pass filters that STM came up with in previous threads regarding through-zero flanging.  These provide fixed group delays and there is no clocking involved, hence no heterodyning to be concerned with.  So, a handful of MN3006's, some assignable group delay allpass filters and you may be able to accomplish an MN3011's task without an MN3011.

That sort of gives you an idea of why analog implementations of things like the STD-1 disappeared off the face of the earth once production of the MN3011 stopped, doesn't it?

Noumenon

Yeah, thanks for the responses! Interesting reading :)

Maby we should just hope that someone will order a custom batch from panasonic, but that would probobly cost ALOT.

Mark Hammer

A-HA!!!!

I was reading through some back issues of Radio Electronics on the throne this morning (honestly, you'd think my intellectual model was Martin Luther...), and was reading through a 1994 article by Larry Klein on directionality in hearing and audio cues.  He notes that inter-ear differences in arrival time (i.e., how long it takes for the SAME signal to arrive at the left vs right ear) tend to play little or no role in sound localization above 1200hz, at which point other things start to matter.

One of the things designers have tried to create "stereo" signals over the years is the use of sum and difference outputs.  So, stick in a delay of some sort, take the difference between that and the dry signal for one output and the sum of them for the other.  This produces notches where peaks exist in the other complementary output.  The problem this produces, though, is that these two signals cancel when in mono, and can also reduce the stereo effect in a room-listening context when the listener stands in the "wrong" spot.

It occurred to me after/while reading Larry Klein's article that the problem the sum/diff method produces is that one is using IDENTICAL copies of the  original wet signal, such that all notches sit in the same spot where the peaks are.  The MN3011/MN3214 chips were blessed with the ability to produce notches and peaks in places that weren't related to each other, but one doesn't need multiple taps or delays to yield different wet signals on the left and right.

What you CAN do is impose a "group delay" via phase shift stages to one of the copies of the wet signal.  This will be functionally equivalent to a very small time shift that will increase as audio signal frequency goes up.  The number of stages doesn't need to be large since a small amount of phase shift will affect enough of the frequency content in the upper mids and higher to produce inter-ear differences in the distribution of notches and peaks in a stereo context.

To be fair, this is NOT the same thing as having delayed repeats showing up in one ear then the other then the other, as might happen with an MN3011 or digital unit.  But what it DOES do - potentially - is yield a stereo output from a single BBD that does not entirely cancel when mixed down to mono electronically, or when listened to through two speakers in a room from the "wrong" position.  If your goal is to have a richer stereo chorus, than that strikes me as a reasonable approach.

Since the group delay produced by the phase shift is fixed, not swept, FET matching and all of that crap that normally comes with building a phase-shifter is largely moot since *no* FETs are involved, just fixed resistors.  Make yourself two phase shift stages from op-amps, 10k resistors, 47k resistors to ground in place of FETs, and .01uf caps, and stick that subcircuit between the output of the lowpass filters after the BBD and the input of one of the mixing stages.  The signal available at the junction of the lowpass filters and allpass/phase-shift filters is routed to the other output mixing stage.  Voila, a more "mono-resistant" stereo chorus!

My guess is that Ton will pop up shortly with either a sober admission that he has done this already, or links to patent applications, or an enthusiastic grin and appreciative comment for the idea.  Then Steve will pop in with a comment about how Eventide did this already or something like that. :lol:

puretube


DiyFreaque

Hi,

I'm back from my Hellish biz trip, and caught this thread today - Mark has posted some (as usual) goooood stuff in this thread.

I mentioned this on the SDIY list a couple of weeks ago, and forgot to mention it here (where it just may have more impact).  In my endless search for the MN3011 (I have two of them, but want more....) I did stumble across 4 Star Electronics, a place that procures parts sitting in dusty warehouses from all over the world.  Apparently, they can get them, but there's a $300 minimum order, and I'm not sure if they would deal with small fry like me.

But, I fantasize about a group buy, or better yet, someone like, oh, Steve Daniels of Small Bear (cough, cough) snagging some and selling them for a profit (you'd be surprised how much I'd pay for one).  It's a thought.

If one could figure out what sort of demand there would be (especially from this list) and how much one would be willing to pay for the MN3011, it wouldn't such a risky venture for someone like Steve.  Of course, bear in mind that I'm speaking as a person who places the MN3011 on the highest BBD pedestal there is, and I am *NOT* a business man....

Cheers,
Scott

puretube

No time nowadays for me to dig up links, sorry...

look for Tom Scholz @ the patent office...
(R*ckman-likestuff with 3011 & 3014(?));

look for Logan, Jr. et al. (=Schroeder) (~`60s) for reverb mix/FB theory @ p.o. ("all-pass reverb.")

Elektor (D) 2/`79 & 5/`79 for theory, 2/`90 for praxis.

look at the 3011 datasheet delaytimes-table:
19.8 to 166.4 ms (@ 10kHz clck), with no multiple relation amongst each other...
in the 3011, this is achieved by tapping the delayline at appropiate stages,
while the whole line gets the same clock.

roll yer own #1:
series-circuit 6x 3007, clock individually such that 1st chip delays 19,8ms, the second according to a.m. table`s delay time difference between 1st and 2nd tap, and so on...

voila: 3011 cloned functionally.

like Mark has mentioned before: due to heterodyning, you can (rather: will) get bad artefacts/whine-ing.

to get rid of that, there are hints/tips/tricks in the forum`s vault...

roll yer own #2:
parallell them, clocking individually to get the delay times as stated in the table...

oops - getting off topic... more later...

puretube


Noumenon

well, if theres was about to be a group buy, count me in! :)

Don't know what a high vs low price on the mn3011 would be though.
If it isn't to high I'd probobly order like 10 or more

Noumenon

Quote from: puretubelook at the 3011 datasheet delaytimes-table:
19.8 to 166.4 ms (@ 10kHz clck), with no multiple relation amongst each other...in the 3011, this is achieved by tapping the delayline at appropiate stages,
while the whole line gets the same clock.

roll yer own #1:
series-circuit 6x 3007, clock individually such that 1st chip delays 19,8ms, the second according to a.m. table`s delay time difference between 1st and 2nd tap, and so on...

voila: 3011 cloned functionally.

That was what I was thinking about.

Quote
like Mark has mentioned before: due to heterodyning, you can (rather: will) get bad artefacts/whine-ing.

to get rid of that, there are hints/tips/tricks in the forum`s vault...

how does one to that, and, why does it occur?

puretube

Quote from: Noumenon
how does one to that, and, why does it occur?

can someone jump in here, plz???

(sorry, Noumenon, don`t have time at the moment to help with that complex problem... there have been some topics on it...
some other  guy will surely soon pop up and lead you to some links
here in the forum ... -  or have a little patience  :) ).

Noumenon

hehe, I'm trying the old sarch technique. haven't had any luck yet though :)

Mark Hammer

Probably the most produced pedal at risk for heterodyning is the Boss Dimension C chorus.  This uses two BBD-delayed signals that share a single LFO.  The LFO is inverted for one of them so that you have two clocks going in opposite directions, hence rarely and only momentarily at the same speed.

Boss uses companding and lots and lots of low pass filtering to keep these two delay signals "playing nice" with each other.

Certainly another aspect is to simply keep the clock sections away from each other and the clock lines VERY short.  This is normally achieved by sticking the clock chip right up against the BBD.

scratch

How about using 6 x PT2399's instead of 6 x MN3007 (certainly cheaper),
could set all of the 6 'stages' with different delays just using a trimpot, could even modulate all of them using a single LFO and LDR's, and you wouldn't have clock problems ... Just a thought.
Denis,
Nothing witty yet ...

ExpAnonColin

Wouldn't it be possible to get heterodyning or other bad effects from the PT2399's clock frequency, or subharmonics, even thuogh it is much higher?  The PT2399 has a clock too...

Mark, your post count is 3011.  How fitting.

-Colin

StephenGiles

I think perhaps a redraw of the circuit is in order(not by me), on A4 it is eye screwingly small and even on A3 (2x) it's not much kinder on the eyes. I would estimate it would need to be at least 4 times bigger and clear to make a project feasable - to minimise the possibility of errors. Now using 6 delay chips would of course increase the amount of circuitry - I say gloomily and enhance the chances of a complete %^&* up. Not that I'm trying to put anyone off - unless...............it's built on veroboard!!!!!
Stephen :lol:  :lol:  :lol:  :lol:  :lol:  :lol:  :lol:
"I want my meat burned, like St Joan. Bring me pickles and vicious mustards to pierce the tongue like Cardigan's Lancers.".

puretube

Noumenon: guess you`re counted in...

Scratch: sure doable, but with the drawbacks Mark (?) mentioned earlier,
about the "resolution" of low-volume digitized signals, and the repeatedly being A/D`d at lowering levels...

Colin: yes, but to a lesser degree...

Stephen: "the VEROVERB"  :lol: ...

Mark: cc. Dim C: yes;
       more clocks make it worse, or rather more complex: keep delay A apart from delay B, both apart from C;
plus: possible hetero-products in turn can interfere with a 3rd clock.
Not only keep the "units" small and physically apart, but also severely "star-ground",
and "star-supply" the (rail-) RF-decoupled "units"...

more later, Ton.