What's the shortest delay possible with MN3207?

[ElectricDruid]

Subject says it all really: What's the shortest delay possible with the MN3207 1024-stage BBD?

I know what the datasheet says - 200KHz clock max, giving 2.56ms delay. This is apparently a pretty conservative specification. I've got one on my breadboard that's running at 540KHz giving slightly below 1ms delay. This isn't a one-off, since I've tried several and they all do it.

However, I suspect this 1ms limit is caused by the clock driver chip (MN3102) not the BBD. The MN3102 chip has a constant deadband delay between the two anti-phase clock signals. When the frequency gets very high, this deadband delay means that the clock pulse is held low for almost all of it's period, until finally the pulses get so narrow they disappear. I *suspect* (but don't know) that if the deadband delay were reduced, it would be possible to drive the MN3207 at even higher speeds/even shorter delays.

Has anyone got any experience to offer? How high can you try?

Thanks,
Tom

[Mark Hammer]

You are correct.

The limits to clocking of the 3207, and its various cousins, is apparently the capacitance on the clock input pins (2 & 6) which the datasheet indicates as 700pf.  The MN3102 is a great little functional chip that accomplishes much in a small footprint, but it does not provide the current output to overcome the capacitance of those clock input pins.  This is why the max clock frequency is spec'd so low.

it is also why some of the more legendary flangers have used a Reticon SAD-124.  That chip is listedas having 110pf input capacitance, and able to be clock as high as 1.5mhz.

I know that with suitable buffering and current drive, the MN3007 can be ratcheted up to well over 1mhz clocking.  I would imagine the same is true of the MN3207 and its equivalents.

You can find some ideas of how to achieve the higher clock rates by looking at the John Hollis "Ultraflange", as well as the A/DA flanger.

[Fender3D]

Just a little side note:
You can drive the BBD well over 1MHz, and if you don't need go near 1,7~2 MHz you there's even no need buffering clock signals.
BUT bandwidth gain will suffer: @ ~800KHz MN3007 will lose ~4dB, the higher the clock the higher the gain loss.

[Mark Hammer]

Just a little side note:
You can drive the BBD well over 1MHz, and if you don't need go near 1,7~2 MHz you there's even no need buffering clock signals.
BUT bandwidth gain will suffer: @ ~800KHz MN3007 will lose ~4dB, the higher the clock the higher the gain loss.

I would imagine from the same underlying mechanism.  As the clock freq goes up, the input capacitance '"rounds off" the normally-square clock pulse such that the clean near-seamless handoff of each sample from one cell/bucket to another gets a little "gap" inserted.  The suggested buffering keeps the square wave square at higher frequencies.

[Thomeeque]

I did measure GAIN/CLOCK characteristic of my unit:

V_CC = 10VDC
R34 = 1k
IC4 = Tesla MAB311
f_CLOCK = 40kHz ~ 3MHz

Input = 200Hz / 525Hz / 1.5kHz sinus adjusted for 1V_PP at pin 3 (input) of BBD

GAIN is ratio between peek-to-peek voltages at R15 and at pin 3 (input) of BBD, measured by Velleman PPS10 oscilloscope.

FILTER-MATRIX mode, COLOR at 0.


(click pic for hires)

Results were quite surprising (for me) - characteristic depends on input signal frequency and for some (low) input frequencies gain drop stops around 1400kHz and gain slightly grows again (!!) with higher clock. For higher frequencies drop is continual and signal gets distorted above 2.4MHz (there is maybe space for bias voltage tweaking yet).

T.

EDIT: few notes:

- as you can see you can run MN3207 at 3MHz and still get signal passing thru and it seems you could still go higher.
- problem is that according to my measurements above cca 1MHz BBD line in EM3207 starts to work as a low-pass filter for the passing audio signal (as you can see until 1MHz all curves are more-or-less alike, above 1MHz signal with bigger frequency has bigger loss)
- note that EM3207 does not use what you call "deadband delay" ("Wave Forming" block inside MN3101/2) - maybe adding this block tuned for high frequencies could help
- there was a lot of "brainstorming" going on around this post in the original thread with very valuable input from the other guys, worth checking IMO.
- don't believe to my measurements (neither conclusions) totally, I'm not any expert and I did use equipment far from the lab level quality, it may be pretty wrong.

Btw. why do you ask actually, do you have some specific use on your mind?

T.

[oldschoolanalog]

Using a CMOS clock & buffer combo will help. But that will take up more real estate. The up side is the chips are readily available and cheap.

[ElectricDruid]

Btw. why do you ask actually, do you have some specific use on your mind?

Yes, I do. I was thinking of building a flanger. For that application the datasheet range of the MN3207 isn't really suitable, but I suspected that in fact it'd be fine, and I thought I'd ask here.

I'll post something if I get anywhere with it.

Thanks all,
Tom

[Mark Hammer]

Boss used the 3207 for the BF-2 for many years, and managed to sell a bunch of them to happy musicians.  All you're really talking about is nudging it down to a shorter delay range.  They did that too, but the HF-2 shortened the delays by moving to a smaller capacity (512-stage) chip: the MN3204.

[ElectricDruid]

Boss used the 3207 for the BF-2 for many years, and managed to sell a bunch of them to happy musicians.  All you're really talking about is nudging it down to a shorter delay range.

That's exactly what I had in mind, Mark, hence the question. If I can clock it faster than the datasheet says, it'll make a good flanger.

 They did that too, but the HF-2 shortened the delays by moving to a smaller capacity (512-stage) chip: the MN3204.

I've also got a MN3209 256-stage chip, so if the MN3207 doesn't come off, I'm going to use that instead. I ought to get better bandwidth with the longer delay line though, so it's to be preferred if it works.

Thanks,
Tom

[Fernando]

Hi, I know it's an old thread, but it's about the only message I found around the subject!
Most are interested on maximum delay time : ) and most other delay ICs, including the PT2399, have a too long min delay time for my purpose.

So I ask, any advance made driving the 3207 with higher clock freqs to get shorter delays?
Coolaudio version is available and cheap, so I'm interested.
Going down to 1ms would be very good, although 200us or even less would be great.

I'm looking to get very short delay times and this is the IC with the shortest delay times among the easily available ICs
Very happy to read you achieved 1ms delay and that it could be possible to go lower.

So, any update about this matter?
Any circuit designed for this?

Another option would be the SAD512, but not easy to get or cheap, not so unobtanium as the SAD1024 but scarce.
The SAD512 can go down to 200 microseconds

What I'd like is to build a voltage controlled delay for my modular synth and use it for pseudo-physical-modelling sound/music making. I can't design circuits, just build, but could try to find help to design it from a working guitar effect version.

Best,

Fernando

[Mark Hammer]

Jack Orman has a circuit for a "Super Buffer" (http://www.muzique.com/lab/superbuff.htm).  The use of paralleled op-amps provides a higher current low-impedance signal which is able to successfully travel long cable distances and still deliver full bandwidth.

In similar fashion, those pedals that extract very short delay out of BBDs use a buffered clock signal with several paralleled invertors to provide higher current clock pulses.

Now, the clock pulse doesn't have to travel quite the same distance as a guitar going to the mixer at the back of the house.  But the bandwidth of the guitar is fairly narrow, and certainly nothing like  1mhz.  If the clock frequency is that high, even modest capacitance on the BBD input pins, and a "travel distance" of probably no more than an inch, can compromise the squareness of the clock pulses.  The same way you don't need to buffer content under 1khz for it to reach an amp 20ft away, you also don't need to buffer a 50khz clock signal travelling an inch into a 700pf clock pin.  But the buffering matters at much higher frequencies, whether we're talking bright single-coils feeding an amp 20ft away, or a 500khz clock feeding a BBD an inch away.

FWIW, the BF-2 flanger uses an unbuffered MN3102 and MN3207, and Boss's service notes state that it has a minimum delay time of 1msec.  I imagine that's rounded off, but the MN3102 craps out around 100khz.

[Rob Strand]

I've never run these things over 1(+ a bit) MHz but the I'd expect the bandwidth to drop.

The suggested buffering keeps the square wave square at higher frequencies

This may help bandwidth.  If the sample duty cycle is effectively narrower the average level of the signal will drop.

Another thing that might help is to put a cap to ground at the analog input, like an RC filter with a low R (1k to 2k ohm).  Opamp output impedances are fairly poor at high frequencies and I suspect the output dips at the sample points. (If you have ever looked at the voltage dip at the input of some ADC's you will know what I mean - it can be quite bad!).

[ElectricDruid]

So I ask, any advance made driving the 3207 with higher clock freqs to get shorter delays?
Coolaudio version is available and cheap, so I'm interested.
Going down to 1ms would be very good, although 200us or even less would be great.

I'm looking to get very short delay times and this is the IC with the shortest delay times among the easily available ICs
Very happy to read you achieved 1ms delay and that it could be possible to go lower.

So, any update about this matter?
Any circuit designed for this?

Yeah, the summary is that without much effort the MN3207 can run up to 1MHz (0.5ms).  It can go higher still if you buffer the clocks, but expect the BBD output to be lower.

I designed a flanger around it which uses a 500KHz max clock (limited by the PIC I used, but the BBD output starts to droop after 500KHz). This replaces both the LFO and the MN3102 clock chip with a single 8-pin PIC to make a combined LFO+clock for flangers:

http://electricdruid.net/flangelicious-a-super-dooper-flanger/

Physical modelling / Analog Karplus-Strong synthesis had also occurred to me as a use for this. My version of your idea was a MIDI controlled BBD clock chip, so you feed it MIDI notes and it produces a clock at the right frequency to make a resonant delay line at that note (or some octave of it, depending on whether you use 512-stage, 1024-stage, or 2048-stage delays). A MIDI-controlled tuned delay, basically. Turn the resonance up, play the radio or your iPod through it, and play a tune on whatever audio you like. I've done some of the preliminary work for this, but haven't built a prototype up yet.

HTH,
Tom

[bool]

Slightly OT (but somewhere along the lines I suppose) : so which would be a recommended (optimal?) non-MN3102 clock driver for the 3207 in your experience?

[ElectricDruid]

Slightly OT (but somewhere along the lines I suppose) : so which would be a recommended (optimal?) non-MN3102 clock driver for the 3207 in your experience?

Depends entirely on what you're trying to do. Are you building a delay, a flanger, a chorus, summat else?

I was trying to design a minimal-parts-but-fully-featured flanger, and my FLANGE PIC chip is the best way I could come up with to do that. For a standard chorus, the MN3102 is perfectly good, and many pedals use it. But it's also relatively straightforward to build a biphase clock using CMOS and several pedals do that too. I think the Morley Flanger might be an example done like that, IIRC.

Tom

[bool]

The target would be a flanger (in a tighter, tinny, close-to-phaser range, but not tru-zero ... let's call it a "hi-range" flanger, with high regen). I have a couple of 3207's, and have been vaguely interested in CD4013 based clocks. So if you happened to stumble on a particularly efficient circuit (low part count + good sound vibe), imho "we're all ears" ... I mean eyes.

[Rob Strand]

and have been vaguely interested in CD4013 based clocks

The VOX flanger used a CD4013 in a kind of funky way.

[ElectricDruid]

and have been vaguely interested in CD4013 based clocks

The VOX flanger used a CD4013 in a kind of funky way.

Having a look through my flanger schematics, the EH Electric Mistress, Boss BF-1, and MXR Flanger use 4013 clocks too.
Craig Anderton's Hyperflange uses other CMOS (404-something-i-can't-read and 4041), and the Morley Sapphire uses 4007 and 4047. Jon Hollis' Ultraflange uses 4049 (buffers) and 4046(clock).

Most of these circuits require two chips for the clock - one for the actual clock, and then one for either buffering or providing the biphase output. And then you need an LFO. So I dunno if that makes any of them a "particularly efficient circuit".

If you hadn't already got a MN3207, I'd say use a MN3209 instead. That shoves the sweep range up two octaves without any other changes.

Tom

[Rob Strand]

EH Electric Mistress, Boss BF-1, and MXR Flanger

Thanks for the mind jog.   
Haven't looked at the BF-1 schematic for ages.

[bool]

...
Most of these circuits require two chips for the clock - one for the actual clock, and then one for either buffering or providing the biphase output. And then you need an LFO. So I dunno if that makes any of them a "particularly efficient circuit".
...

This is the precise reason I wasn't overly enthusiastic but only "interested".

BUT I'm actually pretty convinced that using CMOS chips (and not the 3102) is the way to go. So if there really exists some "efficient" circuit, I'm still hoping for it

OTOH if there is something that could be done with a 7555 or any other generic mos 555 variant (one as a vco ant the other as inverter), that would be even better imho. These have much sturdier outputs than 4xxx chips.