Subharmonic synthesis

[parmalee]

Contemplating putting together a subharmonic synthesizer along the lines of the old Dbx units (albeit simplified), or the Sub-o-matic described in the Stompboxology issue on frequency division (albeit modified).  The Sub-o-matic uses a three-stage bandpass filter from approx. 50 hz to 100 hz, then a Schmitt trigger, divider, synchronous detector--the usual--but it's tailored for line level, stereo system use.  On the other hand, the Dbx units with which I am familiar employ several filters (from 6 to as many as 9, I think?), along with far more circuitry overall, being as they were designed to produce a rather more clean sound, suitable for the average punter.  I am considering doing two separate bandpass filter stages--say, ~30 to 60 hz and ~60 to 120 hz--and using both dividers from the 4013.

I will be using it with a harmonium, heh, and perhaps a drum machine, so it will likely sound slightly... unusual no matter how I put it together.  But I am curious as to whether anyone here has tinkered around much with subharmonic synthesizers--either of your own design or one of the Dbx units, or even the DOD Meatbox.  And is the proposed ~30 to 60 hz filter perhaps a shade too low?  Obviously, results will vary considerably depending upon what one is plugged into, but that lower one might only yield some unpleasant rumbling or even psionic terrorism.

Of course, a line level, stereo one is also somewhat enticing, as such might afford one the opportunity to hear those Dewtron Mr. Bassman notes abundant on many an early Genesis record, but likely compromised by low frequency rolloff so that they could put out records with insanely long sides (28/29 minutes!).  I mean, why didn't they just make 'em double records with 10 or 12 minute sides?  That's what everyone does these days.

Thoughts?  Suggestions?

[Mark Hammer]

I have an old dbx "Boom Box" subharmonic synthesizer, like this one.

Never really used it much, if at all.  But for $25, it was a nice learning tool.  I took it apart, only to find that, like pretty much every commercial octave-down pedal, it was relying on a 4013 flip-flop to do the heavy lifting.  Most of the circuit, distributed over several boards, was filtering, and probably some gating, to assure the octave division a) sounded more or less natural and smooth, and b) was only applied where appropriate.  Worth noting that the unit was intended to provide selective subharmonic synthesis of an input signal that was wide bandwidth and consisted of multiple instruments and other sources (voice) in a mix.  As such it couldn't/shouldn't apply octave division to cymbals, singers, violins, etc.  This is different than our context, where we apply octave division to a single instrument with fairly limited bandwidth.

[parmalee]

Worth noting that the unit was intended to provide selective subharmonic synthesis of an input signal that was wide bandwidth and consisted of multiple instruments and other sources (voice) in a mix.  As such it couldn't/shouldn't apply octave division to cymbals, singers, violins, etc.  This is different than our context, where we apply octave division to a single instrument with fairly limited bandwidth.

That's kind of what I am aiming for here.  Having built countless octave divider units--most of the Boss OC2 variety, though I am partial to the slightly different method Merlin Blencowe employed with the U-Boat: switching at the zero-crossings, as opposed to the peaks.

Incidentally, that DOD Meatbox actually uses a specialized chip for the division, rather than the usual 4013: the Mitsubishi M51134P "Sub Hormonizer for Bass Emphasis"--datasheet complete with funny spelling error.  Alas, both the Meatboxes and the chips alone have become quite scarce and pricey, so I'll be sticking with the tried and true route.

Approximating the effect via software with recorded music yields interesting results:  while slowly mixing the effect in with a dry signal rather quickly goes from virtually no effect at all to a pleasant low-end emphasis to sounding like muddy, muddled crap.  Consequently, given the very narrow useful range, I'm looking for ways to introduce some variety.

For one, it seems that low pass filtering might suffice.  A bandpass with a bottom end high pass starting around 30 or 40 hz likely isn't filtering much of anything at all in most instances.  Also, I'm considering ways to sculpt the filtered and divided wet signal so that such that the effect is still there, but does not sound so much just a rumbling mess beyond a certain point when mixed with the dry.

In the Stompboxology article, the author notes that there is likely little need for employing multiple, very narrow bandwidth filters and dividers, as with most music, it's likely only one monophonic instrument (or vocalist) providing much in the way of "content" below that 100 hz point anyways.  Of course, he is neglecting those among us who like to play multiple bass pedal notes simultaneously, but a fair point nonetheless.

[Mark Hammer]

The PAiA Rocktave pedal makes clever use of a 570 compander chip to improve tracking and also gate out sputters one doesn't want.  Perhaps there are some useful strategies there.

https://www.diystompboxes.com/smfforum/index.php?topic=20729.0

[parmalee]

The PAiA Rocktave pedal makes clever use of a 570 compander chip to improve tracking and also gate out sputters one doesn't want.  Perhaps there are some useful strategies there.

https://www.diystompboxes.com/smfforum/index.php?topic=20729.0

Ahh, thank you Mark!  Almost forgot about that one.  Also didn't know that it was ever marketed under the PAiA name--I just knew it from Anderton's book, Electronic Projects for Musicians.

Poring over the schematic for the Dbx 100 "Boom Box" now.  It replicates the filter, divider... six times--one for 50 to 60 hz, 60 to 70 hz, and so forth--and sums them prior to injecting them back into the mix.  Does seem a bit overkill, considering that instrumentation in that range is likely minimal (and singular and monophonic).  Also seems a bit odd that they didn't go with something like 50 to 55 hz, 55 to 65 hz, 65 to 85 hz, etc.

Have you ever listened to the isolated "wet" signal from the "Boom Box?"  If so, just how "sputtery" was it?  Given the abundance of almost (but not quite) redundancy in the circuitry, I would expect "not very."

Again, I'm aiming more for enhancement at the lower ends of the spectrum, as opposed to straightforward octave division.  Likewise, I'm hoping to achieve something that works adequately with polyphony--but NOT in the same manner as something like the EHX POGs (which can only be attained with DSP anyways).  Rather I want something that affords us a glimpse of the missing fundamental--or, in many cases, an imaginary missing fundamental--but only just sooo much, i.e., somewhere betwixt a subtle and a subtle-ish effect.

Some years back, I assembled something very close to Merlin's U-Boat, but having it switch on every third zero-crossing, rather than every other one.  It yielded a fairly clean (not synthy/squarish) fifth just below the fundamental--IOW, C4 produced a G3--and it was quite useable.

I think, some years back, R.G. proposed a variation of the Rocktave that made full use of the 4024 chip by mixing several of the divided outputs in an additive synthesis manner.  Which, of course, brings us back to some of the wonderful ideas illustrated in Richard H. Dorf's Electronic Musical Instruments.  But I guess I'm getting a bit side-tracked here...

[Rob Strand]

Also seems a bit odd that they didn't go with something like 50 to 55 hz, 55 to 65 hz, 65 to 85 hz, etc.

Some practicalities come into the design of the filters when you consider discrete part values and part tolerances.  If you take a 1/3 octave equalizer it steps at a factor of 1.2, so 50Hz to 60Hz, which is something that can be achieved without going overboard with special parts.

[DIY Bass]

I have an Ashdown Bass Octaver that from memory uses the Mitsubishi chip.  As an octave pedal it seems to work and track very well. 

[amptramp]

It sounds like you could use a graphic equalizer as the basis for a subharmonic generator.  It already has the individual tone-peaking elements for narrow frequency ranges and you could run a frequency divider off that.  Even a relatively complex signal should be divisible into 1/3 octave bands and there should be little interference between bands.  Use the signal level at each tone stage to set the amplitude of the divided signal.

You could use the trick mentioned by parmalee the triggers the flip-flop on every third zero axis crossing to get the fifth above the octave dowm - that is a nice trick that I hadn't seen before.

[parmalee]

Also seems a bit odd that they didn't go with something like 50 to 55 hz, 55 to 65 hz, 65 to 85 hz, etc.

Some practicalities come into the design of the filters when you consider discrete part values and part tolerances.  If you take a 1/3 octave equalizer it steps at a factor of 1.2, so 50Hz to 60Hz, which is something that can be achieved without going overboard with special parts.

I have this annoying habit of conveniently overlooking economical and logistical considerations when considering such things.  It's easy enough to do these days when obscure parts or components with most unusual values are literally at our fingertips, but I've gone through enough combo organs and suchlike (products made on the cheap) to know better.

I purchased a DOD Meatbox from Ebay which I'll likely re-sell after getting a sense of it.  As I'm tailoring this for a harmonium with the lowest fundamental being around ~65hz, diy is the best route.  Well, it's usually the best route.

Incidentally, having reviewed a fair number of schematics for subharmonic synths, what I notice is this:  they invest a lot of real estate to the bandpass filters, gating, and such, but when it comes to division, they pretty much inject a squarish signal into a 4013!  There's very little of the tailoring that you see in the typical octave divider circuit--like the OC2--for optimizing fundamental extraction.  I suppose that makes sense when working with multiple, very narrow bands, as likelihood for "errors" is greatly minimized, but it stilll seems a bit odd.

[Mark Hammer]

Entirely separate from the accuracy or reliability of tracking, one of the quirks of using a flip-flop of some form (e.g., Shin-Ei used discrete flip-flops,similar to the Boss switching circuit) is that the octave down is always a cycle behind.  It's not quite as bad as some of the earlier harmonizers/pitch-shifters that would simply slow down a sample to provide lower pitch, but the octave-down doesn't start precisely when the fundamental it is derived from starts.  And of course, the lower the fundamental to start with, the bigger a temporal gap between the onset of the note and the production of the divided waveform.

When the original EHX Attack Decay pedal came out, it used a BBD to stagger the input signal a bit such that the pedal could do what it needed to do and still line up the effect with the "real time" signal.  In view of the inherent stagger imposed by use of a flip-flop, I wonder if a few msec of BBD "stagger" could help a divided-down signal line up with the full-bandwidth source material a little better.  Doesn't actually have to be a BBD, either.  Could just as well be several "lag" stages of phase-shift (I always forget which is lead and which is lag, but the form where more phase shift is created the lower the frequency of the signal).

[parmalee]

Entirely separate from the accuracy or reliability of tracking, one of the quirks of using a flip-flop of some form (e.g., Shin-Ei used discrete flip-flops,similar to the Boss switching circuit) is that the octave down is always a cycle behind.  It's not quite as bad as some of the earlier harmonizers/pitch-shifters that would simply slow down a sample to provide lower pitch, but the octave-down doesn't start precisely when the fundamental it is derived from starts.  And of course, the lower the fundamental to start with, the bigger a temporal gap between the onset of the note and the production of the divided waveform.

When the original EHX Attack Decay pedal came out, it used a BBD to stagger the input signal a bit such that the pedal could do what it needed to do and still line up the effect with the "real time" signal.  In view of the inherent stagger imposed by use of a flip-flop, I wonder if a few msec of BBD "stagger" could help a divided-down signal line up with the full-bandwidth source material a little better.  Doesn't actually have to be a BBD, either.  Could just as well be several "lag" stages of phase-shift (I always forget which is lead and which is lag, but the form where more phase shift is created the lower the frequency of the signal).

Exploring just how far one can remove a thing from it's source--in both space and time--and still have it "sounding about right" has always been fascinating to me.  At one end, it is so far removed such that semblance becomes somewhat irrelevant; at the other end, you're getting into "uncanny valley" territory.

I mentioned above that switching-and-inverting at every third zero-crossing trick, in order to obtain a fifth just below the fundamental.  I don't feel like uploading an image, but just picture in your head a sine wave that is inverted at every third crossing--kinda weird looking, yeah?  Yet, somehow, it "sounds about right," even with minimal filtering.  (To do that with an octave divider, you have to use something like Merlin Blencowe's U-Boat design, as most others switch at the peaks.)

I can't say that I normally notice the lag, but...  as an experiment, I did once take a basic recording of some stuff played through a divider, and then I "shifted" it with software to remove the lag and I compared the two.  I did notice it when I did that, but the one that, had it been done in "real time," have been violating the laws of physics sounded kinda "funny."  Probably more to do with familiarity than anything else.

This little tangent brings to mind some thoughts I had about the EHX Freeze pedal--or rather, a question:  How is the EHX Freeze any different than taking a delay circuit with the PT2395 chip with separate DRAM chip, blocking the PT2395 from the WE (Write Enable) on the DRAM, grabbing what's in the DRAM and running it through some simple envelope circuitry?  As far as I can tell, fidelity-wise, they are about the same, but I am fairly certain that the Freeze employs DSP (don't actually know, having never seen the innards).

[parmalee]

I have an Ashdown Bass Octaver that from memory uses the Mitsubishi chip.  As an octave pedal it seems to work and track very well.

Going by the size of that thing (the Sub Octave Plus, right?), I'm guessing that that works with polyphony and is likely employing DSP.  I really like the idea of VU meters on pedals, even when they don't serve much of a practical purpose--makes it feel like you're doing something really serious.

Unfortunately, information on that Mitsubishi chip is scant.  The datasheet is 6 pages and most of it is simply detailing the physical dimensions of the chip itself.  But it's good to know that they're not departing from standard chip dimensions!

[parmalee]

As an aside, anyone remember that chapter on subharmonics in Henry Cowell's New Musical Resources?. Apart from being scientifically rubbish, unfortunately, it's still an inspiring read.  He talks of some obscure Russian composer/inventor who created an acoustic instrument which was capable of producing subharmonics--of it's own!  In all likelihood, said subharmonics were likely a product of the room and the environment (thanks Alvin Lucier!), but the idea was enticing all the same.

Subharmonics have always been shrouded in mystery and mysticism, both of the woo-woo variety as well as in the spirit of sceptical inquiry--from Hugh Banton's HB1 (which can only be heard on the last VDGG Peel Session), with a speaker system that could produce an earth shattering 7 hz, to the 16 hz bottom C of the Bosendorfer Imperial, expertly deployed by the likes of Keith Tippett, Cecil Taylor and Charlemagne Palestine, when played in a room of just the right dimensions.

That's what I'm after.  For these purposes, the usual octave divider is simply too... obvious.  I want my little 3 1/2 octave Indian harmonium, with it's 5/6 size piano keys, to produce a sound that defies all physical constraints.

[ElectricDruid]

I want my little 3 1/2 octave Indian harmonium to produce a sound that defies all physical constraints.

This is a noble cause and we will help in any way we can.

[Mark Hammer]

As an aside, anyone remember that chapter on subharmonics in Henry Cowell's New Musical Resources?. Apart from being scientifically rubbish, unfortunately, it's still an inspiring read.  He talks of some obscure Russian composer/inventor who created an acoustic instrument which was capable of producing subharmonics--of it's own!  In all likelihood, said subharmonics were likely a product of the room and the environment (thanks Alvin Lucier!), but the idea was enticing all the same.

That book is sitting on my shelf.  Been a long time since I looked at it.  Busy making a raised planter box at the moment, but when my back gives out and I need to sit for a while and wipe the sweat off, I may reach over and flip through it.

[parmalee]

I want my little 3 1/2 octave Indian harmonium to produce a sound that defies all physical constraints.

This is a noble cause and we will help in any way we can.

Incidentally, it took me years to design a satisfactory dedicated miking and pre-amplification system for this harmonium.  It's a standard Indian folding harmonium, but I place it upon a small table and operate the bellows via foot pedal and bowden (bicycle brake) cable--essentially, like Nico's harmonium, though hers was the proper foot pedal variety reed organ/harmonium.  (There are plenty of turn-of-previous-century reed organs out there, like the ubiquitous Esteys, but they typically have only a single reed set and sound pretty thin, so customization was essential.)  I wanted something that would work well within quiet studio environs, as well as in a live environment, around other instruments, amps, pa systems, etc.  Also, I needed it to be sufficiently isolated so that I could run it through a variety of pedals, without external interference and without substantively altering the proper nature of it's sound.

I reviewed a number of systems that are used with accordions and found that the ones that actually sound good cost upwards of 800 to 1000 dollars!  Not really an option for me.

Anyways, I finally discovered that the source of my dissatisfaction lies with the electret mic elements (I use two):  It turned out that when miking so close to the reed sets, it was putting out like 130 to 140 SPL!  That just blew me away.  I just couldn't believe that such a little thing could be so freakin' loud.  But I replaced the electret with much pricier, and higher rated ones, and all my problems disappeared.

[parmalee]

Henry Cowell's New Musical Resources?

That book is sitting on my shelf.  Been a long time since I looked at it.  Busy making a raised planter box at the moment, but when my back gives out and I need to sit for a while and wipe the sweat off, I may reach over and flip through it.

It's an excellent read, along with pretty much anything by Partch or Cage.  Or Schoenberg or Adorno, for that matter (though the latter is awfully annoying and borderline racist at times).  Come to think of it, they're alll good.  Writing on music never gets old.

[Mark Hammer]

Hah! Found it.  Admittedly, my favorite is Charles Ives' "Essays Before a Sonata"

[amptramp]

There have been instruments that used multiplication of signals to produce sum and difference frequencies.  I have a Minshall Model E organ that has single-triode dividers for generating five octaves of keyboard sounds and uses the combination of a note and its musical fifth going into a stage driven hard into intermodulation territory to generate lower octaves.  For example G1 and C1 produce C0.  This avoids having extra tubes and gigantic (for the era, 1950 to 1951) capacitors to produce the lower octaves and gives the pedalboard a reasonable sound.

[skyled]

You might look into using a ring mod based pitch shifter. It won't be the same as dividers, but it will be polyphonic and there won't be tracking issues.

Just split your output signal, run one through a low pass filter to keep the bass end, then put that into a ring mod but only keep the lower sideband output. The EHX Ring Thing lets you do this, if you don't feel like building the Bode Frequency Shifter...