From: Andrea TONI Subject: -- FtoV -- Date: Mon, 10 May 93 11:40:55 MET Organization: ESA-ESTEC (European Space Research and Technology Center) Keplerlaan 1 2200 AG Noordwijk (EUROPA) Phone: Int +31 1719 83606 Fax: Int +31 1719 84697 Machine: HP-UX sihp03 A.B8.05 A 9000/730 941406112 Mailer: Elm [revision: 66.33] Hi, I try to post this last weekend .. but I don't think it went on the list .. I'll try again .. Basically a FtV looks like .. Sig (f) | \ --->----| \ | | | | |OA >--->--| One Shot |--->--| LPF |---------> out (CV tension) --| / | | | | | | / | / GND the OA is an OP amp that function as comparator that is suppose to "square" the signal (sig(f) frequency) in a kind of zero-crossing fashion in which the positive front are used to trigger the O.S producing every time a pulse of Tm. The LPF is used to find the average value generated by the OS. Since this pulses are all equal in length and amplitude, the average value depends in a unique fashion by their frequency .. (that is indeed the freq of SIg(f). SO the tension going out by the LPF is proportional to the pitch of the used note and can then be used to CV and tune the VCO . this diagram can help a bit more ... Sig (f) | | ^ ^ | / \ / \ |/___\______/ ___ \_________________________________ t \ / \ / \ / \ / v v OP comparator | |--- ----- ---- | | | | | | |__|________|___|____________|__|_________________ t One Shot | Tm |-- --- --- | | | | | | |_|_________|_|______________|_|___________________ t LPF | | ___________________ | ___________________________/ |_____________________________________________ t Nice he .. to draw in ascii mode !!!!!!! arghhh I just would like to add that the length of Tm has to be smaller then the shortest Ts of sig(f) otherwise the OS wouldn't be able to keep track of sig(f). The drawing is of course an ideal situation .. but in the reality the LPF is a bit slow in catching up the quick frequency variation. The most difficult thing is of course in the accuracy and stability required. In the standard Cv/Oct two semitone are equal to a difference of 83 millivolt (!!) and since ears are able to fetch a de-tune of 1/4 of semitone .. we can assume that errors need to be in the range of 10 millivolt .. and in the analog domain you know what it means !!! (minimum noise .. thermal attentions, hi-level components .. ) But the advantage compare to a digital pitch-to-midi is the uniform functioning of the circuit .. the converter doesn't have to discriminate between a legato, vibrato or glissato .. the output tension (CV) will *dummyly* follow the frequency of sig(f). From here the name pitch-follower may seems more appropriate since the circuit doesn't detect the sig(f) but it barely make a conversion. In a pitch-to-midi device instead everything is much more complicated since the box has to discriminate semitones closer to the real pitch and to approximate to the real freq of the pitch-bend command. I have both an analog (Korg MS20) and a digital (Roland VP70) PtV(Midi) and even if the Korg looks simpler it is in practice more usable .. the digital one required a lot more attention when you feed in voice (no bending no rapid detuning etc etc ..) The Analog Devices Databooks (89/90) contain a lot of hi-level IC already suited for the purpose .. AD537 (VtoF) AD650 (FtoV and VtoF) AD652 (VtoF) AD654 (VtoF) Low cost .. ADVFC32 (FtoV and VtoF) .. and of course if you find any old databook (the new ones are as bad as digital synth compare to analog ..!!! ) you may find several application notes and proper explanations .. Enjoy !! for analogue@magnus.acs.ohio-state.edu; 10 May 93 9:42 MET (for rick@sara.nl, from sscprick@diamond.sara.nl); Cc: rick@sara.nl Subject: Pitch To Voltage converter Date: Mon, 10 May 93 09:40:38 N From: Rick Jansen > Well, you haven't told us anything specific about how you want to use the > PtV converter so you'll get some answers that won't be applicable.>> > What do you want to use it for? Sorry about that :) I'd like to use a PTV in processing existing sound, like from a record or microphone, with the Formant. I've been playing with the Envelope Follower, which delivers the amplitude (envelope) of an input signal, plus a trigger (gate)signal if the input signal is above a settable level. You can use this to drive a synthesizer. For example, you can filter an existing song so that mainly the percussion is audiable, and set the Envelope Follower trigger level so that it generates a pulse for each beat, and thus have your synth play along. It seems to me a PTV converter fits in with this type of thing very well. > What's the frequency range you want to cover? audio range: 0-20kHz > What's the input signal look like? anything, ranging from RuckZuck to Music Non Stop > How fast does the conversion have to work? not sure, but so as to cover my need > How accurate does the conversion need to be? mwah, withing 1 to 0.5% I'd be very happy with > Do you want a linear or a log output? linear will do fine. Control voltages are linear (1V/oct), the VCO's already contain a lin-log converter. > True enough, but these devices won't be too useful for electronic music > work (I'm guessing that's the context of Rick's request). For instance, if > there's any grunge or noise in the input signal, then the output voltage > will be way off. There are far better methods like using a phase locked > loop or digitally counting the time between pulses. > for some instruments such as guitars, building a reliable > frequency counter is not easy. I wouldn't be surprised if you first need to filter the input signal, to get rid of most of the unwanted frequencies, but I'm confident an FTV converter will be a nice toy. > For guitars the signal waveform is very ragged and constantly > changing (string harmonics are almost never integral multiples of the > fundamental and the harmonic conent changes drastically depending on how > you play and what note you're playing), the waveform level is all over the > place, and guitarists generally do not play very cleanly. Also there's a> > need for the conversion to take place in a cycle or two. Certainly > difficult. Hmmm... Are you referring here to six strings and one PTV converter, Don? Or the better scheme of a PTV converter for each string? Thanks! Rick Jansen -- rick@sara.nl She's a module and she's looking good Date: Tue, 11 May 93 00:23:48 PDT From: till@lucid.com (Don Tillman) Subject: Pitch To Voltage converter Date: Mon, 10 May 93 09:40:38 N From: Rick Jansen > Well, you haven't told us anything specific about how you want to use the > PtV converter so you'll get some answers that won't be applicable.>> > What do you want to use it for? Sorry about that :) I'd like to use a PTV in processing existing sound, like from a record or microphone, with the Formant. I've been playing with the Envelope Follower, which delivers the amplitude (envelope) of an input signal, plus a trigger (gate)signal if the input signal is above a settable level. You can use this to drive a synthesizer. For example, you can filter an existing song so that mainly the percussion is audiable, and set the Envelope Follower trigger level so that it generates a pulse for each beat, and thus have your synth play along. It seems to me a PTV converter fits in with this type of thing very well. I'm looking for something very specific about the type of signal you want this puppy to work on, but when you use a maximally vague phrase like "processing existing sound", it's real difficult to be helpful. > What's the frequency range you want to cover? audio range: 0-20kHz Say what? What kind of melody are you trying to track? > What's the input signal look like? anything, ranging from RuckZuck to Music Non Stop I have no idea what a RuckZuck or a Music Non Stop is. Here's the issue: If the input signal is a perfect textbook waveform then the monostable circuit mentioned by others will work fine. If the waveform is changing amplitude and frequency content, like you'd get out of a real musical instrument, than things get more difficult. If the harmonics are non-integral multiples of the fundamental (like with any string instrument) or there's background noise, than things are very difficult. If there's more than one note happening at the same time, only heroic digital efforts have a chance of working. > How fast does the conversion have to work? not sure, but so as to cover my need (Wow, you really do hate to get specific!) Here's an example: Guitarists tend to complain about the 15 or so mSec conversion time of most guitar-midi converters. But guitarists tend to whine a lot anyway. > Do you want a linear or a log output? linear will do fine. Control voltages are linear (1V/oct), the VCO's already contain a lin-log converter. 1V/octave is a log control voltage. 1V/1kHz is a linear control voltage. VCOs generally contain log->linear converters (ie., an exponential circuit). Almost all synths are log control. The monstable circuit ideas presented so far are linear control, so they wouldn't work too well. > How accurate does the conversion need to be? mwah, withing 1 to 0.5% I'd be very happy with In your 0-20Khz linear example, .5% would be an accuracy of 100 Hz. (Yikes, this is one reason linear control isn't used very much.) > For guitars the signal waveform is very ragged and constantly > changing (string harmonics are almost never integral multiples of the > fundamental and the harmonic conent changes drastically depending on how > you play and what note you're playing), the waveform level is all over the > place, and guitarists generally do not play very cleanly. Also there's a> > need for the conversion to take place in a cycle or two. Certainly > difficult. Hmmm... Are you referring here to six strings and one PTV converter, Don? Or the better scheme of a PTV converter for each string? One PTV converter can convert one pitch to a voltage; it doesn't make a heck of a lot of difference if you have six of them running in parallel or not. I'm gonna venture a guess and suggest that a phase locked loop circuit using a log control VCO and a leading edge sensing phase detector will probably be your best bet. (You could probably even use the VCO from your synth and the phase detector from a 4046 CMOS PLL chip.) -- Don