SAD1024 Woes - Any Suggestions?

[DiyFreaque]

Out of the blue, Thomas Henry sent me an old MAP Voltage Controlled Delay.  He was cleaning out his shop, and instead of tossing it, thought I might get a kick out of working on it.  He was, of course, quite right.

When I got the VCD, the only thing it would do was fart if I tried to blend in any delay.  A quick check of the inputs and outputs of the SAD4096 proved that it was pretty well shot.  Knowing the availability of the SAD4096 (I.E. there is no availability), I decided to see how the circuit would perform with a SAD1024.

The pinouts of the SAD4096 and SAD1024 are obviously quite different.  So, I removed the SAD4096 from its socket and jumpered the Vdd, Vbb, ground,  two clock signals and two signal pins via short wires to a breadboard, and arranged these signals so that I could insert a SAD1024 and see how it flew.

I've got six SAD1024s - three pairs obtained from three different sources throughout the years.  Up until this point, none had ever been tested.

Two of the SAD1024s are totally dead (one gets really hot, the other puts out a distorted signal only at a certain input level and bias on section A, section B does not work at all).

To begin with, the VCD does work very well with 512 stages - I can get a nice range of delay effects ranging from chorus to flanging (the clock has a wide sweep range of audio range to just under 1 MHz).

There's the rub - on the four remaining SAD1024s, section A works perfectly, but, of the four, I have never gotten section B to work.  I find this quite perplexing - I can't believe four SAD1024s would display the same failure mode.

My Vdd is 15V.  Vbb is 14V.  The clock signals are 0 to 15V, nice crisp square waves.  Ground is ground (I've checked continuity) and I've checked all signals on the actual pins of the SAD1024s, and on the VCD board to make sure the signals are getting back to it.  When using a single section, I am tying the unused inputs of the unused section to ground.  I am tying the unused signal outputs to Vdd.  The NC pins are all tied to ground.  After testing section A, and switching over to section B, I'm moving all inputs and outputs to the other section.  What was connected to signal input A is now connected signal input B, what was connected to the clock signal inputs for A is now connected to the clock inputs for B, and the connections for output signals of section A are moved to the output pins for section B. The inputs and outputs of section A are tied respectively low and high as mentioned above.  I leave Vdd, Vbb, and ground alone, and keep the NC pins grounded.   

I'm carefully observing which pins are clock input, signal input and signal output (IE, I know the pinout is not symmetrical in that regard).  I am following ESD procedures  (making sure I am grounded when handiling the IC, which I do as briefly as possible, and am avoiding touching the pins.  The breadboard and VCD are grounded at the same point I am).  Yet, section A works fine on all four of the SAD1024s and section B is bupkus.

Looking at the outputs of section B, some SAD1024s display a constant pedestal, the amplitude of which does not vary with clock frequency, while a few others display a pedestal that shrinks in amplitude as the clock frequency is increased.  In all instances, no signal is riding on the pedestals.  Probing the input pin of section B proves there is a signal on the input, the same proves clock signals are present, but the output is as described above.

I know the sections of the SAD1024 are supposed to be able to operate independently, but I'll be danged if I can figure out what is going on, unless fate has just struck me a highly coincidental blow.  As for the VCD, it doesn't bother me - it sounds absolutely great with 512 stages.  But, my other SAD's, which I was hoping to use on other projects, don't seem to want to work with section B.

Does anybody experienced in the ways of the SAD1024 have any advice?

Thanks,
Scott

[StephenGiles]

Scott, this may seem obvious but at this stage I would test all the SAD1024s in another unit, confirmed to be working. Could there be a fault on your breadboard perhaps? I had this problem a few years ago and threw away good ICs until I found that the continuity had been broken on a track under a set of 5 holes!

[DiyFreaque]

Thanks Stephen!

The VCD works perfectly using the A side of each SAD1024.  I've checked the wiring on the breadboard literally dozens of times.  Could be the breadboard itself may be duff - I'll try it on a different one.  Even if changing the breadboard doesn't fix it, I see your point - a valid test would be to remove the VCD altogether and just rig a test circuit (like from the SAD1024 datasheet) and see how that goes.  I don't have anything else that uses SAD1024s.

Cheers,
Scott

[snap]

cheap breadboards love to play games with me
good breadboards don`t.
those contacts aren`t always contacts.
sometimes they are choppers.
sometimes they are isolators.

this fenomenon increases with repeated use of the breadboards.

[DiyFreaque]

Stephen,

I owe you a watery Kansas 3.2 lager next time you swing on through.  'Twas the breadboard.  Three of them work on section B, one still will not (that's the one I'll use for the ACD).  God Bless Ye.

Snap - I've got around 14 breadboards - the good ones are from Jameco, and they're all full of other projects.  Yep, this one was a cheap one, one I detest and just use for very small experimental things like this.  In this case, it wasn't even good for that.  First time I've been full-on stabbed in the ass by even one of my cheapos.  First time I ever threw one away, too.

Cheers,
Scott

[snap]

got about the same amount of breadboards:
good ones (used again and again and again for different complex circuits),
cheap ones (used for a single circuit, noticing while tweaking
that they decrease in performance).

those smaller cheapos now have "chalkmarks" on them
for each time they are being re-used,
because even simple 2 transistor side-projects don`t want to work
properly without heavy tossing.

[StephenGiles]

Glad to have been of help Scott. As a Kansas man, please tell me this - why is Arkansas not spoken like Kansas? Is it an old Indian name perhaps?

[Mark Hammer]

People have mentioned the E-bay re-seller who peddles old MXR boards, among them a board for the old green Analog Delay.  As many are quite aware, finding BBDs for these things is getting to be a bit like finding 54 Tele's under grandparents' beds with the tags still on them.  What this particular re-seller has done is to work around the BBD issue by means of an adaptor board using a PT2399.  (I'm not sure how they address the tracking filter, but that wouldn't be an issue with one of Thomas' boards).  Perhaps you might want to consider a similar fix, and save those SAD1024's (now that you know more of them actually work) for somethng where a 1024 is truly called for.

BTW, your breadboard issue is yet one more scenario where this Stabilant stuff I keep harping about can be a lifesaver.

[DiyFreaque]

why is Arkansas not spoken like Kansas? Is it an old Indian name perhaps?

That's a good question, for which I do not know the answer.  I do know that 'Kansas' is derived from the Kansa tribe (Kansa meaning People of the South Wind, like the song).  Perhaps Arkansas has a different root?  In any event, 'Arkansas' is pronounced 'Ar-kansas' here in Kansas when in reference to the Arkansas River.  Only in Kansas is that prononunciation held true - it's even pronounced that way on the local news.  In any other state, it's 'Ar-kansaw', just like the state.  One can tell a true newcomer to Kansas with just one reference to the river as the "Arkansaw" river.  Often, on national news, if the Arkansas river is mentioned, it takes a while to dawn on me they're speaking of the same river that passes two miles north of my home.

As many are quite aware, finding BBDs for these things is getting to be a bit like finding 54 Tele's under grandparents' beds with the tags still on them.  What this particular re-seller has done is to work around the BBD issue by means of an adaptor board using a PT2399.  (I'm not sure how they address the tracking filter, but that wouldn't be an issue with one of Thomas' boards).  Perhaps you might want to consider a similar fix, and save those SAD1024's (now that you know more of them actually work) for somethng where a 1024 is truly called for.

(A) Because, as I mentioned in the first message, it sounds absolutely dynamite - if you heard it, you would understand.  A PT2399 will not flange like this.

(B) To put a PT2399 into this circuit would not be a trivial matter.  A means to control the clock frequency of the PT2399 would have to be added.  Audio level conditioning would have to be altered.  There is precious little room for these things without adding a bracket and another circuit board.  It would not have the delay range of either the SAD1024 in use now nor the original SAD4096 (longer delay but not as short compared to 4096).  Putting in the SAD1024 involves only creating a small board holding the IC with the appropriate wiring changes in the form of jumpers, one that can easily be plugged into the original 4096 socket.  I might open up the filters a tad bit, but that again is trivial.

(C) If I want a PT2399 delay circuit, it makes more sense to build one from ground up - I've breadboarded plenty of them, have a dual one on perfboard.  I don't need another.

(D) I have a SAD1024 that only has one section working.  Why not put it in here, stick a fork in it and call it done?

Cheers,
Scott

[Joe Kramer]

Perhaps you might want to consider a similar fix, and save those SAD1024's (now that you know more of them actually work) for somethng where a 1024 is truly called for.

(A) Because, as I mentioned in the first message, it sounds absolutely dynamite - if you heard it, you would understand.  A PT2399 will not flange like this.

The Electric Mistress, Boss BF-1, MXR, Ross, and A/DA all have one thing in common, and they pretty much comprise the entire Flanger Hall Of Fame, which might as well just be called the SAD1024 Hall Of Fame.  I'd say if the idea is flanging, the 1024 would definitely be "called  for."

[Mark Hammer]

The Electric Mistress, Boss BF-1, MXR, Ross, and A/DA all have one thing in common, and they pretty much comprise the entire Flanger Hall Of Fame, which might as well just be called the SAD1024 Hall Of Fame.  I'd say if the idea is flanging, the 1024 would definitely be "called  for."

I used to think that myself too.  Until Mike Irwin persuaded me with his A/DA experiments using an MN3007.

The Matsushita - Reticon difference has to do with how easily the chips are clocked at very high speeds without 'fancy" circuits, and that in turn has to do with the input capacitance of their respective clock pins.  Datasheets show the clock pins of the Reticon chips as having a capacitance of 110pf and those of the comparable Matsushita units as 700pf.  The Reticon chips tended to be designed to be "video-ready".

How does that make a difference in flanging quality?  "Jet plane" flanging requires sweeping downward from a very very short delay - as close to 0msec as possible.  Very short delays requires very fast clock speeds (not to be confused with LFO speeds).  Just like the way cable capacitance acts like a treble-cut filter when the cable is long and fed with an unsuitably buffered source, the input capacitance of the clock pins acts like a treble-cut filter on the clock signal.  For at least some portion of the clock's entire operating range, a square wave/pulse generated by the clock is seen as a square wave by the BBD chip.  After a certain frequency, though, that capacitance adds some lag and the clock pulse starts to behave like (or rather be seen as) a trapezoid wave and eventually a triangle wave.  Keeping in mind that the sound quality will depend upon an absolutely instantaneous "handoff" from the one FET/cap path within the BBD to the other complementary one (driven by the complementary clock pulse), once that clock pulse starts to go from pure squareness to something less, functioning will start to suffer as those brief delays between handoffs start to occur.

The much lower input capacitance of the SAD-1024 allows it to, on its own with very little help, clock up to 1.5mhz and beyond without any trouble.  The Matsushita chips tend to peter out around 100khz (as the datasheet says) when they don't have any extra help.  The MN3101 and 3102 chips do a nice job generating complementary clock pulses, but do NOT provide the requisite buffering and extra current oomph needed to get past that 100khz hurdle set by the 700pf input capacitance on the BBD.  The Matsushita datasheets do not discuss this.  They merely assume you will use the Matsushita clock driver because it is the most convenient solution.

As Mike's experiments confirmed, however, sticking a suitable buffer between the clock and the MN3007 (usually a trio of paralleled invertor sections such as those in a 4049) allowed it to be comfortably clocked up to and beyond 2mhz, yielding some VERY short delays.  Just about every "legendary" flanger uses a clock generating circuit different from that shown in Matsushita datasheets.  It will often involve several CMOS chips.

At another level, though, I thnk you are correct.  Many of the "great" flangers use a Reticon chip.  The gist of my post here is that the "greatness" lies not in the sonic properties of the Reticon chip itself, but rather in the tendency of designers to use clock circuits that got in the way of great flanging when working with Matsushita chips by turning to Matsushita clock chips a little too easily.

I guess the lesson is to not be TOO misled by the appnotes, or rather, to think beyond them.  Perfect case in point is the current thread re: the SSM2166.  The datasheet assumes the chip will be embedded within a more complex circuit and in any event be fed a low impedance mic signal that will require lots of gain.  We collectively neglected the fact that its input impedance is only 175k and that the "default" wiring of the input stage delivers too much gain for a guitar player's needs.  Datasheets and appnotes CAN be very helpful, but sometimes they can lead you to a cognitive trap.

[StephenGiles]

I wasn't aware that Mike has done the deed with the MN3007 as well. I hope he's OK these days.

[DiyFreaque]

Besides the Flanger Hoax, the only other currently produced flanger I can think of that is nearing legendary status, if it's not already there, is the Modcan flanger.  Designed by Mr. Irwin, natch.  Having never seen the schematics, I think I can still safely  assume he used clock buffers (though not necessarily the MN3007 - I've heard it's a lower number staged device - don't know).

Interestingly, the MN3010 version of the A/DA does *not* use the buffers.  At least on the schematics I've got. This is interesting, because they did use buffers on the STD-1, which isn't driven to near the extreme clock frequencies the flanger is.  I can only imagine, because  the extra stages of the MN3011 represent a greater load capacitance,  the effect was much more exaggerated, prompting those A/DA gurus to put in the buffers.  Mike did put the buffers in the SAD1024 remake of the flanger, as shown on Stephen's drawing of it.  The ACD has the buffers as well.  It was intended for the SAD4096, which again would represent a greater capacitive load than the SAD1024.

I wasn't aware that Mike has done the deed with the MN3007 as well. I hope he's OK these days.

Me, too.

Cheers,
Scott

[Mark Hammer]

I certainly don't know enough about any of the chips involved, but some of the more pleasing flanger designs use a CD4047 to drive the BBD.  That's what the A/DA used and it is, incidentally, what the old Morley/Tel-Ray flangers used too, as well as (gasp) the Anderton Pedal-Flanger (http://hammer.ampage.org/files/Anderflange1.PDF).

[DiyFreaque]

VCD uses the CD4041.

(Edit): As the buffer (as does STD-1).  4046 is the clock driver.

Cheerio,
Scott

[Joe Kramer]

Many of the "great" flangers use a Reticon chip.  The gist of my post here is that the "greatness" lies not in the sonic properties of the Reticon chip itself, but rather in the tendency of designers to use clock circuits that got in the way of great flanging when working with Matsushita chips by turning to Matsushita clock chips a little too easily.

The pity is that there's practically no commercial non-1024 flangers worth beans IMHO (have not heard the Hoax or the Modcan), so any comparison would be a task.  But it would be interesting to hear a side-by-side shoot-out between SAD- and MN-based implementations, all circuit variations being made equal.  It may be that, given the same stout clock signals and the same milli/microsecond ranges, they would sound the same.  Or not.  I wouldn't be surprised at either conclusion actually.  They may sound different but both good.  I'd have to hear it for myself. 

BTW, I seem to have gleaned from somewhere (sorry, can't recall the source) that a nice hard square wave is good for BBDs up to a point: TOO hard and the antiphase edges verge on overlap, creating a problem as bad or worse than too soft an edge.

[Mark Hammer]

BTW, I seem to have gleaned from somewhere (sorry, can't recall the source) that a nice hard square wave is good for BBDs up to a point: TOO hard and the antiphase edges verge on overlap, creating a problem as bad or worse than too soft an edge.

Never heard that one before, but it's an interesting notion.

We've got all this math revolving around sampling and sampling theorems.  Somebody somewhere must have contemplated what the tolerable limits are for gap/overlap between successive samples as a function of sample rate.