Why is pitch-to-CV still behind?

[OriginalCrash]

This applies to pitch-to-MIDI, too... Why is it that converters are still quite glitchy? Some require abnormally high action or don't work well on low notes. Why is it that a tuner can track notes this quickly, but pitch-to-___ converters can't? Anyone have any insight?

Brian

[OriginalCrash]

Home from work bump to see if anybody has anything...

Brian

[Bofee]

I suspect that your tuner isn't as fast as you think.

[bongomania]

The tuner comparison is informative. Most tuners take at least a second to grab onto the lower notes. The first and only tuner I've used that doesn't have any delay on the low notes is the Turbo Tuner. So in fact tuners generally do not do a great job of tracking low frequencies, and the only reason they seem good at it is because when tuning, we don't care if it takes a second for the pitch to register. When tracking with MIDI etc. even a tenth of a second is a big deal, sounds terrible.

Tack on to that the fact that all a tuner has to do is light up some LED's, whereas a MIDI brain has to convert the pitch and amplitude data into MIDI CC's and then transmit the data to a remote device which then interprets the MIDI data and sends a command to its tone generator etc. etc.

The best low-frequency tracking I've ever heard was from pitch-to-CV devices with short signal paths.

Maybe if the technology of the Turbo Tuner can be adapted to MIDI conversion then we'd be on to something. But otherwise bass strings are so "all over the map" in terms of the readable frequencies coming off them from moment to moment that it's hard for any device to read accurately in a short enough span of microseconds.

[Chunk Systems]

Bongomania is right. Tuners will generally take a couple of hundred milliseconds to lock, whereas for pitch->MIDI you need tracking that locks in tens of milliseconds. Bear in mind that the period of a low E is somewhere around 25ms, and you can't really do any better than one cycle. Of course you should be able to track higher-pitched notes faster.

The main issue, however, is that for pitch->MIDI conversion to need to know which octave the note is in, whereas for a tuner you don't care. This is actually the hardest part for bass guitar signals because the first harmonic is extremely strong and the harmonic content changes over time. "glitchiness" is usually where the system is uncertain whether its tracking the fundamental or tracking the first harmonic. You end up building all kinds of locking mechanisms in to avoid this uncertainty, but you want to be sensitive enough so that you catch it if the player plays something in one octave and then jumps up an octave since this is a common figure in many genres of bassplaying. This limits your ability to be able to apply hard locking constraints.

There are other issues surrounding when notes start and stop as well. With an instrument like bass there are various string and fret noises which make detecting the start or end of a note less than straightforward.

Finally, no two basses or players are exactly alike. So you can build something that tracks perfectly on a particular bass when played with a particular technique, but that might not transfer well to another bass with a totally different harmonic character played by somebody with a different technique.

[rasthedoctor]

Wow - thanks both Bongo and Chunk Systems. I wish all answers (generally) were so well thought out and articulated so a layman, like myself, could understand.

[OriginalCrash]

Yes, thank you... very informative. I just have a dream of really harnessing my FreqBox by splitting my signal, then running one line into the the audio in (with hard sync off) to drive the internal envelope follower and the other into a pitch-to-CV converter and then into the frequency control input on the FreqBox. That should get rid of the inherent distortion caused by hard syncing and give you a pure waveform at whatever note you play on the bass... and still maintain the dynamics of what you're playing.

Brian

[bongomania]

As I noted above, pitch to CV is much easier and more reliable. All you need to do is convert the incoming analog frequency into a DC voltage; it will convert almost instantly.

[JtheJazzMan]

dont forget when you first pluck a string, the pitch is all over the shop in that inital attack, so it always will take some time to lock onto the fundamental.

perhaps if a device could be "tuned" to a particular bass tone, it could intelligently cut out that initial interference, but pie in the sky i guess

[OriginalCrash]

So you think that pitch-to-MIDI is the main difficulty but pitch-to-CV, while far from perfect, is doable? I just ask because that's really all I'd need. And as long as it tracked as quickly as the Turbo Tuner, it'd be good for me... The FreqBox probably doesn't track perfectly in real time when in hard sync mode, but I have no complaints about it (I'd just prefer a purer waveform).

Brian

[Chunk Systems]

Originally Posted by bongomania

As I noted above, pitch to CV is much easier and more reliable. All you need to do is convert the incoming analog frequency into a DC voltage; it will convert almost instantly.

Everything that I said about pitch->MIDI also applies to pitch->CV. The extra time required to encode a MIDI message and transmit it is negligible compared to the time it takes to work out what the incoming frequency is.

In other words turning analog frequency into a DC voltage is far from an instantaneous process. Imagine for a moment that you had a pure sine wave coming in without all the time-varying harmonics and other junk you get in bass signals, conversion to a DC voltage or to a number in a DSP would still take time.

There are broadly speaking three methods to convert a frequency to a DC voltage: 1) period analysis (measure how long it takes for one cycle of the wave to complete) 2) frequency analysis (count how many waves in a fixed amount of time) 3) feedback methods (generate a guess at what the frequency is, compare against the original and then adjust your guess - the most common way to do this is with a Phase Locked Loop (PLL)).

All of these methods take time to produce an answer: 1) you have to wait for at least one period to know how long a period is (unless you know the exact shape of the wave ahead of time, which you generally don't) 2) you have to wait for at least one timeslot to know how many cycles occurred within it and 3) it takes time for "lock-in" to occur.

When you get into DSP, there are more advanced methods you can apply as well, such as FFT and autocorrelation analysis. These methods also take time to collect data and time to compute the answer, especially when we're talking about the low frequencies that a bass guitar generates.

In practice, you'll find that any device which does a halfway decent job of tracking bass needs to use a combination of several of the above techniques and also needs to have a bunch of advanced stuff to solve the issues to do with harmonics and note start/end. It's almost always the uncertainty introduced by the harmonics which causes the glitchiness and the latency.

[bongomania]

Hmmm... I understand that this is something you know a lot about--my statement was based on end-user experience with modular synths, in which it never seemed like there was any delay converting external audio sources into CV. But I suppose the general abstraction of the music I was making may have colored my observations.

[OriginalCrash]

Everything that I said about pitch->MIDI also applies to pitch->CV.

Hmm... I do have to wonder perhaps how technical you're being... how high your standards are, I suppose. I'm not criticizing... But I just have difficulty rectifying what you're saying with the tracking performance of certain units. I believe you're saying that the time consuming part is not the conversion to MIDI or CV but the pitch recognition. That seems to imply to me that anything involving pitch recognition would have slow tracking. And I understand your point that a tuner does not have to know the correct octave of the note. But what about my FreqBox in hard sync mode? It tracks very well (near perfectly from a practical, playing in the real world point of view IMO... I don't know how that converts to tens or hundreds of milliseconds) with good technique. Same with what I've read of the Boss OC-2. So how could units like these track effectively but a pitch-to-CV converter present difficulty? Again, I'm not trying to discredit or criticize you (far from it... as I see it you're a valuable source of info I know little about ), I'm just trying to make sense of it all...

Brian

[Chunk Systems]

Originally Posted by OriginalCrash

But what about my FreqBox in hard sync mode? It tracks very well

Originally Posted by OriginalCrash

Same with what I've read of the Boss OC-2. So how could units like these track effectively but a pitch-to-CV converter present difficulty?

Neither of the units you've mentioned convert the frequency of the incoming signal to a voltage or number and then generate an output signal from that voltage or number. Take the FreqBox for example. You can read on the manufacturer's website about how this works, but it involves simply restarting the output oscillator when a new input cycle is detected. That's a different thing entirely from trying to convert an input frequency to CV.

Traditional octavers also do not convert the input frequency to a voltage. Inside all of these things you will find a divide-by-two counter implemented in some kind of digital logic which simply produces an output edge on every second input edge. If some of these edges are due to harmonics then it still works, you'll just get harmonics in the output signal too.

So that's why neither of these kinds of boxes do frequency -> pitch conversion per se.

[OriginalCrash]

But they still have to determine the pitch (which is what I believe you said is the time consuming part), don't they? For example, the FreqBox... audio triggered VCO. Initially, the VCO is set to whatever frequency the knob or expression pedal is set at. When you play, it triggers the unit to "play" a waveform at that frequency with the dynamics of your playing. If hard sync mode is on, it resets the initial frequency of the VCO to that of what you're playing. So first it has to determine the pitch of the input signal, then it forces the output frequency of the VCO (not sure if it does this by changing the control voltage feeding into it or some other way). So you've got a two-step process... determine pitch, do something with it. Which step causes the tracking problems in a pitch-to-CV converter?

Brian

[Chunk Systems]

Originally Posted by Chunk Systems

So that's why neither of these kinds of boxes do frequency -> pitch conversion per se.

Sorry, I meant to say pitch->CV conversion.

[Chunk Systems]

Originally Posted by OriginalCrash

But they still have to determine the pitch (which is what I believe you said is the time consuming part), don't they?

No. They don't "determine the pitch" in any meaningful sense. They just do a particular thing when a particular event occurs at their input.

[OriginalCrash]

Hmm... maybe this is getting too technical for me...? I just don't see how it can reset the frequency of the oscillator to the frequency of the input signal without "knowing" what that frequency is...

Brian

[Chunk Systems]

Originally Posted by OriginalCrash

When you play, it triggers the unit to "play" a waveform at that frequency with the dynamics of your playing. If hard sync mode is on, it resets the initial frequency of the VCO to that of what you're playing.

I'm not sure that's true. It doesn't say anything on their website about changing the frequency of the output oscillator based on the input frequency, it just says the output oscillator is reset each time an edge occurs. This would naturally give the output sound a strong component at the frequency of the input and probably sounds pretty cool but it's not the same thing as converting the input frequency to a voltage.

[OriginalCrash]

Now that I read it phrased that way, it may be making more sense to me. So you're saying that it's not actually resetting the output frequency of the VCO to that of the input signal... it's just making it start over at a rate equal to the input frequency... Allowing it to do so simply whenever a "peak" occurs in input waveform instead of having to determine what the frequency of that signal is. Here's a link to the manual; it's got a more detailed explanation of hard sync and a quick illustration of how the two signal are interacting on page 13. Basically the VCO is still trying to output the frequency it's been set at, but it keeps getting reset based on the incoming frequency, right?

Brian