Subharmonic synthesis

[parmalee]

Hopefully yes... so there is still hope for a analog tap tempo!!

I think that a sort of crude and imprecise tap tempo can be built up from the fundamental detection blocks of Penfold's "guitar tracker" circuits, which use a 555 as a sort of tachometer.  Of course, "crude and imprecise" is hardly optimal for a tap tempo, but it's a start!

Taking the tap tempo idea several steps further, I've long had this fantasy about a sort of "intelligent" drum machine that can determine suitable places for kick and snare beats from a signal.  The drum sounds, of course, would be analog--and perhaps cymbals/hats of a very low sample rate--while the playing style would be closer to, say, Klaus Dinger than to someone like Jaki Liebezeit or Charles Hayward.

[Rob Strand]

Ahhh, FFT.  Yeah, I guess that would be a nightmare in the analogue domain.  Am I entirely wrong in thinking that what can be achieved in DSP is theoretically possible with analog, but might take an entire room--even if done with the smallest SMD--to functionally implement?

Think of a real-time analyser or a Vocoder.    Many band-pass filters.   No doubt someone has done with something using modulation/demodulation ideas.

[parmalee]

An update:

Just received a Boss OC-3 (the DSP one).  I split the signal from my harmonium preamp thusly:

Output 1 >>> 3rd order Sallen-Key high pass filter (Cf ~ 180hz) >>> sundry effects (runoffgroove's Tri-Vibe, something along the lines of MXR Dist+, delay...) >>> mixer

Output 2 >>> 2 3rd order Sallen-Key low pass filters  >>> Boss OC-3 >>> op amp gyrator for subtle accentuation >>> mixer (routed about 60% bass amp/40% acoustic guitar amp)

I'm using the OC-3's poly range mode and it's surprisingly effective at weeding out the higher notes--or maybe it's my filters.  So long as one keeps a few semitones between the lowest notes played and upper notes, it tracks remarkably well.  I can even play dense chords or tone clusters with my right hand without it muddying the bass end.

Frankly, I was rather surprised.  Two 3rd order filters is a lot--I can probably even cut back some--but it's hardly a "brick wall."  Apart from EHX's Bass9, I know of no other effects unit that offers this "splitting" function.  Yet I would think there would be some demand for such, no?

[parmalee]

Think of a real-time analyser or a Vocoder.    Many band-pass filters.   No doubt someone has done with something using modulation/demodulation ideas.

I've never tackled a vocoder--the abundance of high order band pass filters and the demand for very precise values is somewhat off-putting.  Now, in DSP, perhaps it's another matter entirely... ?

In my "research" I've encountered a number of audiophile-types discussing DSP crossovers.  Perhaps I am wrong here, but that would seem entirely unnecessary--and kinda futile, given the complications of another nature entirely which arise with super-duper filters with ultra-steep rolloffs.

Still, I'm curious.  A lot of people have complained about the Boss OC-3's inability to differentiate a complex, polyphonic signal, but I am finding it quite satisfactory.  Caveat:  I've only been playing around with this thing for a couple of hours, so I may encounter future grievances.  There is some lag, but it is barely discernable--I suspect that with very fast playing, it may prove frustrating.  Of course, harmoniums, with their in-built latency (much more apparent than with pipe organs), are not well suited for "very fast playing" anyways.

[Rob Strand]

I've never tackled a vocoder--the abundance of high order band pass filters and the demand for very precise values is somewhat off-putting.  Now, in DSP, perhaps it's another matter entirely... ?

My previous answer was really only saying that a vocoder is a case were you have a bank of band-pass filters in the analog domain.   An FFT is essentially a bank of band-pass filters so that's where you have an overlap between the analog and digital.  One difference is analog version only uses magnitudes but the FFT has phase information, which you can choose to ignore.

There's no tolerances to deal with in DSP so you can just create a whole stack of band-pass filters.    Some DSP graphic equalizers do just that.   The cost of a lot of filters is more load on the processor.  So people come-up with schemes trying to do the same thing with less processing but they are not all successful.   For example there's some very dodgy FFT based equalizers out there.  FFT's are used because they are fast.

Banks of band-pass filters come-up a lot so people have spent of time on them in DSP.   This is an example of where an introductory DSP book might have a small section of a chapter on Filter banks but for state of the art you need a whole book on Filter Banks and a few technical papers.

https://en.wikipedia.org/wiki/Filter_bank

In my "research" I've encountered a number of audiophile-types discussing DSP crossovers.  Perhaps I am wrong here, but that would seem entirely unnecessary--and kinda futile, given the complications of another nature entirely which arise with super-duper filters with ultra-steep rolloffs.

There's two types of filters in DSP IIR and FIR.  The IIR filters are more or less behave like the analog filters.   You do come-up with the same old problems, perhaps with the exception of part tolerances.  The FIR filters don't really have a (practical) parallel in the analog world.    The thing about FIR filters is they can be designed with linear-phase regardless of the filter slope or order.   Linear-phase means no phase distortion, it just looks like a time delay.   Most of the basic stuff for DSP crossovers was well established before 1990.    FIR filters with low frequency cut-offs require very long filters and more processing, to get around that people come-up with multi-rate and poly-phase filters (and there's whole books on those subjects).

[parmalee]

There's two types of filters in DSP IIR and FIR.  The IIR filters are more or less behave like the analog filters.   You do come-up with the same old problems, perhaps with the exception of part tolerances. 

The FIR filters don't really have a (practical) parallel in the analog world.    The thing about FIR filters is they can be designed with linear-phase regardless of the filter slope or order.   Linear-phase means no phase distortion, it just looks like a time delay.   

Most of the basic stuff for DSP crossovers was well established before 1990.    FIR filters with low frequency cut-offs require very long filters and more processing, to get around that people come-up with multi-rate and poly-phase filters (and there's whole books on those subjects).

Would this partly account for the reason that DSP stuff sometimes sounds kinda "off?"  Our ears are accustomed to some small level of phase distortion with analog signal processing, whereas the absence of such sounds "unnatural?"

(I'm referring to the middle portion of your quote--for some reason, the forum software does not allow me to boldface text within quotes.)

[PRR]

(I'm referring to the middle portion of your quote--for some reason, the forum software does not allow me to boldface text within quotes.)

I've always had trouble with bold on this forum. Underline works.

[Rob Strand]

Would this partly account for the reason that DSP stuff sometimes sounds kinda "off?"  Our ears are accustomed to some small level of phase distortion with analog signal processing, whereas the absence of such sounds "unnatural?"

There's too many factors to judge a cause.    My personal conclusions is the phase-distortion isn't easy to hear.  Most studies end-up showing the same conclusions.   If your filters have very high Q's then ringing can be audible.   Filter's that ring have a high phase-shift but it's the ringing not the phase shift that's the cause of the problem.

When you design crossovers the main goal is the signals from the drivers add to produces an even response.   That actually takes effort and is probably the problem with crossovers.     It it also desirable to have the even response over a region of space not just at one magic listening point.

A common design technique for IIR filters is to start with an analog filter as the target.   Typically the IIR filter produced doesn't quite match the analog filter in response or phase.   A crossover based on this "naive" design approach probably wouldn't produce a flat response.  I suspect a lot of band-splitting plugins have this problem.

Supposing the above problems are fixed.    One trick you can do with DSP is remove the phase distortion even for filters which have phase distortion.   The trick only works on pre-recorded signals, not real-time signals.    There is a way to play the signal source reversed in time then combine that with the signal played normally (it's called non-causal filtering).    Most experiments like this conclude you can't hear a difference.

A general conclusion about good sounding loudspeakers is they tend to have smoothly changing beam patterns.    A sharp slope crossover could prevent that from happening.  That would happen in a highly likely scenario when a narrow beam-width  low frequency driver crosses-over to a wide beam-width high frequency driver.    So while a sharp crossover might help the common problem of the two drivers producing a flat response it could mess with the beam pattern.    That would be the case for IIR or FIR or analog filters.

[skyled]

I'm glad that you were willing to try a DSP based solution. How does it compare to the sound you imagine in your head? Do you have any audio clips for us?
I might have to add a harmonium to go along with my Hammond...