Frequency to Pitch voltage conversion?

[Lostgallifreyan]

I welcome any recommendations for a single IC to do this. It has to be very
easy to use.

I'm getting the frequency extracted from the input signal ok, I just want
to convert a linear DC voltage representing frequency to one linear with
pitch, without using discrete components and worrying about temperature
dependence. I don't mind paying more if it saves work, especially if I can
get the voltage converter to be a single 8 pin DIL IC. I'd like to cover 5
octaves, with 1V/oct out, and hopefully the input scale will also fit
within a 5 volt range. I need tuning accuracy to within 100th of a semitone
(1 cent).

 

[Tim Williams]

Well, probably any VCO will do it... I don't know about accuracy but you
might tweak the circuit keenly to fix that.

XR2206 comes to mind. MAX038 I think. One or both of them is out of
production or something. I'm sure there's a thousand better chips out
there.

Tim

 

[Franc Zabkar]

I welcome any recommendations for a single IC to do this. It has to be very
easy to use.

I'm getting the frequency extracted from the input signal ok, I just want
to convert a linear DC voltage representing frequency to one linear with
pitch, without using discrete components and worrying about temperature
dependence. I don't mind paying more if it saves work, especially if I can
get the voltage converter to be a single 8 pin DIL IC. I'd like to cover 5
octaves, with 1V/oct out, and hopefully the input scale will also fit
within a 5 volt range. I need tuning accuracy to within 100th of a semitone
(1 cent).

What about a PIC or Atmel micro with an ADC and a DAC? You could also
replace the F/V converter (?) on the input with a firmware routine.

BTW, I'm not clear what you mean by converting frequency to pitch.
These two concepts mean the same thing to me. Do you perhaps mean that
you want to convert a linear range of frequencies spanning 5 octaves
to a logarithmic scale?

For example,

C0 = 130.8Hz -> 0V
C1 = 261.6Hz -> 1V
C2 = 392.4Hz -> 2V
...
C5 = 4185.6Hz -> 5V

BTW, I'm using the data at this URL:
http://www.apronus.com/music/lessons/unit01.htm

- Franc Zabkar

 

[Lostgallifreyan]

Well, probably any VCO will do it... I don't know about accuracy but you
might tweak the circuit keenly to fix that.

XR2206 comes to mind. MAX038 I think. One or both of them is out of
production or something. I'm sure there's a thousand better chips out
there.

Tim

A VCO isn't what I need, I need DC voltage conversion, not actual pitched
AC output, as I'm feeding a voltage to an existing bit of MIDI related
hardware that will use MIDI for signalling after conversion to the right
voltage scale.

 

[Don Bowey]

A VCO isn't what I need, I need DC voltage conversion, not actual pitched
AC output, as I'm feeding a voltage to an existing bit of MIDI related
hardware that will use MIDI for signalling after conversion to the right
voltage scale.

Very interesting....

Exactly why do you not want to use a VCO for this task? What would you
rather use?

Don

 

[Lostgallifreyan]

What about a PIC or Atmel micro with an ADC and a DAC? You could also
replace the F/V converter (?) on the input with a firmware routine.

I thought of a PIC chip, but that's not the simple answer I want. I'm
hoping there's a self-contained 8-pin DIL as standardised and easy to use
as a general purpose op-amp is. I need 12 bits or more of accuracy anyway,
and I'd surely do better with a small analog IC.

BTW, I'm not clear what you mean by converting frequency to pitch.

In my first post I said "..to convert a linear DC voltage representing
frequency to one linear with pitch". The first would be V/Hertz, and the
second is V/Octave.

These two concepts mean the same thing to me. Do you perhaps mean that
you want to convert a linear range of frequencies spanning 5 octaves
to a logarithmic scale?

Yes! One way is to use matched pairs of transistors in a single package
and some fiddly construction involving a custom thermistor, the result
being bulky, inaccurate, unreliable. That concept hasn't changed since it
was first used in analog synthesizers.

What I'm looking for is a single 8 pin DIL IC that would do the same task,
as it seems likely that such a basic task has been done this way by now. I
can't find one though. I think what I need is a DC log amp with fast slew
rate, but when I look for log amps I find too many, they all seem to be
spec'd for complex RF applications, and I can't find any simple answer I
can understand. I want input and output to be positive DC values. RF has
nothing to do with my need.

C0 = 130.8Hz -> 0V
C1 = 261.6Hz -> 1V
C2 = 392.4Hz -> 2V
...
C5 = 4185.6Hz -> 5V

BTW, I'm using the data at this URL:
http://www.apronus.com/music/lessons/unit01.htm

- Franc Zabkar

Your table of values is about right, that's similar to the range of control
I want. Setting base, scale and offset is easy, the bit I need is the
exponential conversion, made extremely easy to use, and also accurate to
within 100th of a semitone for musical signalling. I'd also like it to be a
dedicated general purpose part, not some part of a more complex (and
expensive) special purpose IC.

Thanks for the input. If anyone else wants to pile in, feel free. I'd
even welcome the Aussie Sledger, I don't care what he says to me if his
input is helpful and accurate.

 

[Lostgallifreyan]

Very interesting....

Exactly why do you not want to use a VCO for this task? What would
you rather use?

Don

Since you ask...

I want to use the Doepfer R2M 'Ribbon To MIDI controller'. It's wonderful,
can take a voltage linear with pitch, into the input meant for its ribbon
control, the 'manual' that it's built for.

I can make a nice 50% duty cycle AC with some bandpass on audio input from
monophonic source, partially cleaned with a Schmitt trigger based on
CA4140, followed by a 4046 PLL to get a really clean square that tracks my
signal like a ninja on speed.

So I can easily get a nice V/Hz signal from the PLL, and easily get fine
pitch contol over many octaves with MIDI via the R2M. (The R2M even solves
the problem of handling the offrange signal when noise or inappropriate
signal is presnt on input)

The only bottleneck is the simple and accurate conversion of voltage to the
V/Octave wanted by the R2M.

 

[Andrew M]

I thought of a PIC chip, but that's not the simple answer I want. I'm
hoping there's a self-contained 8-pin DIL as standardised and easy to use
as a general purpose op-amp is. I need 12 bits or more of accuracy anyway,
and I'd surely do better with a small analog IC.


In my first post I said "..to convert a linear DC voltage representing SNIP

Your table of values is about right, that's similar to the range of control
I want. Setting base, scale and offset is easy, the bit I need is the
exponential conversion, made extremely easy to use, and also accurate to
within 100th of a semitone for musical signalling. I'd also like it to be a
dedicated general purpose part, not some part of a more complex (and
expensive) special purpose IC.

Thanks for the input. If anyone else wants to pile in, feel free. I'd
even welcome the Aussie Sledger, I don't care what he says to me if his
input is helpful and accurate.

TI have a log amp I believe. 100th of a semitone over 5 octaves
requires a very accurate system...

look at www.ti.com and search for logarithmic.

The small micro doing frequency detection and conversion to DC is
about the most accurate, flexible, integrated system for this app,
IMHO

-Andrew M

 

[Tim Williams]

The only bottleneck is the simple and accurate conversion of voltage to
the
V/Octave wanted by the R2M.

Ohhh, then you need a transfer function!

You can use an op-amp and a nonlinear element, for instance, FET (x^2),
diode junction (e^x), or using diodes as switches, a piecewise function
composed of different slopes. Also the inverse (sqrt(x), ln x) by switching
voltage for current in the nonlinear bit.

Similar circuits are used to convert triangle waves into sine waves, by
progressively reducing gain on the peaks, softening the shape.

Tim

 

[Lostgallifreyan]

Ohhh, then you need a transfer function!

You can use an op-amp and a nonlinear element, for instance, FET
(x^2), diode junction (e^x), or using diodes as switches, a piecewise
function composed of different slopes. Also the inverse (sqrt(x), ln
x) by switching voltage for current in the nonlinear bit.

Similar circuits are used to convert triangle waves into sine waves,
by progressively reducing gain on the peaks, softening the shape.

Tim

Did I say 'simple', and 'accurate'? It needs to be accurate, as Andrew M
said, to get to within 100th of a semitone over 5 octaves. The 'standard'
method uses the nonlinearity of one of a pair of matched transistors in a
single package, but it's neither easy or accurately controllable. These are
all problems which I am certain are best solved on a single chip substrate
as are the accurate designs of op-amps and such, and I'm hoping someone
here can name such an IC, preferably an 8-pin DIL for easiest use.

If I can't get a single tiny analog IC to solve this, I'll have to consider
the answer that Andrew M gave, as I already have a 'ninja on speed' fast
tracking squarewave that would be ideal input for a frequency detector that
fed a digital conversion to analog output in V/Octave scale. Again, if
anyone can recommend an existing IC I can buy and use as easily as using a
standard op-amp, please name it.

Please understand that I am NOT into reinventing the wheel, or solving this
as a fun excercise alone. This is surely a problem that 40 years of IC
manufacturing have solved cleanly! All I need is for someone to name the
IC. (I've tried to find it, I really have, and already posted about why
that search fails to get a simple answer.)

 

[Lostgallifreyan]

hmm, maybe a Frequency to Voltage Chip?
http://www.national.com/ds/LM/LM2907.pdf
it has examples.

I've got one of those. It won't do though. If it was a pitch to voltage
converter, that would work though. People have said that pitch and
frequency are the same thing, but that's just overlooking the MOST
important difference between them. Those chips you linked to will output
the V/Hz signal, not the V/Octave signal I need.

or are we talking about a digital/serial
stream of coded data to be translated into
an Analog output ?

Not exactly, but one option is to use digital code to count the frequency,
make the log conversion to values representing pitch, then output via DAC.
I still think that's a sledgehammer to crack a nut if I have to reinvent
what should be a standard item by now.

 

[Jamie]

A VCO isn't what I need, I need DC voltage conversion, not actual pitched
AC output, as I'm feeding a voltage to an existing bit of MIDI related
hardware that will use MIDI for signalling after conversion to the right
voltage scale.

hmm, maybe a Frequency to Voltage Chip?
http://www.national.com/ds/LM/LM2907.pdf
it has examples.

or are we talking about a digital/serial
stream of coded data to be translated into
an Analog output ?

 

[Lostgallifreyan]

wish you would clear up what you need.
they make standard components for industrial use

And I seriously and devoutly wish that people would recognise that a
frequency to voltage converter is NOT the same as a pitch to voltage
converter!

I've repeated this crucial difference as clearly as I can in so many posts
that it amazes me that anyone can still miss this point.

 

[Lostgallifreyan]

You listed a nice set of industrial voltage to frequency and frequency to
voltage converters but missed the need for log conversion, and missed that
the musical scale, pitch, isn't going to be catered for by general purpose
tachometer gear, which is what the IC's you linked to are mainly for.

I appreciate the help, but I'd like it better if people read what I've said
rather than leaving me having to write more to steer people to the point.

The point is in the thread title, damn it! >

 

[Jamie]

I've got one of those. It won't do though. If it was a pitch to voltage
converter, that would work though. People have said that pitch and
frequency are the same thing, but that's just overlooking the MOST
important difference between them. Those chips you linked to will output
the V/Hz signal, not the V/Octave signal I need.




Not exactly, but one option is to use digital code to count the frequency,
make the log conversion to values representing pitch, then output via DAC.
I still think that's a sledgehammer to crack a nut if I have to reinvent
what should be a standard item by now.

wish you would clear up what you need.
they make standard components for industrial use
to do both Freq to Voltage and Voltage to freq.
http://www.analog.com/en/subCat/0,2879,760%5F794%5F0%5F%5F0%5F,00.html
http://www.chipcatalog.com/Datasheet/70F11C3A7D3666C29CCB90AFF88D3933.htm
http://www.chipcatalog.com/Cat/85.htm

 

[Dave Platt]

You listed a nice set of industrial voltage to frequency and frequency to
voltage converters but missed the need for log conversion, and missed that
the musical scale, pitch, isn't going to be catered for by general purpose
tachometer gear, which is what the IC's you linked to are mainly for.

I appreciate the help, but I'd like it better if people read what I've said
rather than leaving me having to write more to steer people to the point.

The point is in the thread title, damn it! >

What I remember reading in the past, is that "pitch" as interpreted by
the human ear/brain system is not a simple thing at all. The
perceived pitch of a signal depends not only on the frequency of the
signal's strongest component, but also on the presence and amplitude
of various harmonics.

I think I recall reading that the perceived pitch can change if you
change the overall loudness of the sound, even if the relative
amplitudes of the sound's constituents do not change at all.

Based on this, I believe that you're unlikely to find an electronic
circuit - let alone a simple and inexpensive one - which answers the
question "What is the pitch of this sound" all that accurately over a
broad range of sounds and loudness.

You *might* be able to develop something using an ADC, a digital
signal processor, and a microcontroller, if you could find and encode
a sufficiently-accurate model of the ear/brain system. It's entirely
possible that someone has done this as a research project, but I've
never heard of a commercial product of this sort.

 

[Lostgallifreyan]

[email protected] (Dave Platt) wrote in

What I remember reading in the past, is that "pitch" as interpreted by
the human ear/brain system is not a simple thing at all. The
perceived pitch of a signal depends not only on the frequency of the
signal's strongest component, but also on the presence and amplitude
of various harmonics.

I think I recall reading that the perceived pitch can change if you
change the overall loudness of the sound, even if the relative
amplitudes of the sound's constituents do not change at all.

Based on this, I believe that you're unlikely to find an electronic
circuit - let alone a simple and inexpensive one - which answers the
question "What is the pitch of this sound" all that accurately over a
broad range of sounds and loudness.

You *might* be able to develop something using an ADC, a digital
signal processor, and a microcontroller, if you could find and encode
a sufficiently-accurate model of the ear/brain system. It's entirely
possible that someone has done this as a research project, but I've
never heard of a commercial product of this sort.

All it needs is a base 2 logarithm The details of pitch are as you say
though, but that will be down to the musician to handle. But to get that
far, I'll need the frequency-related voltage (or the frequency itself)
converted accurately to voltage that is a base 2 log such that octaves
increment linearly with voltage. Also, I want to avoid thermal drift and
unwanted complexity in construction and obtaining specialised parts.

Specifically, the details of pitch being hard to quantify are related to
trying to get pure intervals that relate either to the octaves, or to equal
temepered scales, but all I want is the octaves. That solves two problems:

1. It becomes compatible with stardard synthesiser architecture, as
developed over 40 years.

2. It puts all the finer points into the hands of the musician, where they
belong.

 

[flexten]

Didn't do any searching for you but these words may help: Compressor,
Expandor, Log Amplifier, Antilog Amplifier, Differential Log Amplifier,
Log Comverter, Data Compressor, Log Ratio Circuit.
The basic circuit looks like 4 op-amps so there must be some one chip
soln around.
Best

 

[Jamie]

You listed a nice set of industrial voltage to frequency and frequency to
voltage converters but missed the need for log conversion, and missed that
the musical scale, pitch, isn't going to be catered for by general purpose
tachometer gear, which is what the IC's you linked to are mainly for.

I appreciate the help, but I'd like it better if people read what I've said
rather than leaving me having to write more to steer people to the point.

The point is in the thread title, damn it! >

Hmm.
well, then you must be referring to sampling a block of
audio looking for a domain of frequencies ? and i will say again!
if you don't clearing specify what your looking for you are
going to get answers that you don't want.
i think what you are looking for can be done via computer
programming in windows using the sound card with some DFT coding.!
more likely to be FFT which is just a fast version of DFT. there are
also FFT code frags around for Mpu's but you will need a large code size
and memory size for that process if you want an accurate one.

so basically your looking to analyze a range of frequencies and
determine certain aspects of it ? if so i think you should look into
doing some Sound card programming. it has the ADC converter to PCM bits
that you will need to perform the math.
there are also tons of DSP references out there.

 

[Lostgallifreyan]

well, then you must be referring to sampling a block of
audio looking for a domain of frequencies ? and i will say again!
if you don't clearing specify what your looking for you are
going to get answers that you don't want.
i think what you are looking for can be done via computer
programming in windows using the sound card with some DFT coding.!
more likely to be FFT which is just a fast version of DFT. there are
also FFT code frags around for Mpu's but you will need a large code
size and memory size for that process if you want an accurate one.

so basically your looking to analyze a range of frequencies and
determine certain aspects of it ? if so i think you should look into
doing some Sound card programming. it has the ADC converter to PCM
bits that you will need to perform the math.
there are also tons of DSP references out there.

I'm not trying to analyse a complex signal. I keep saying that I only want
a conversion of voltage scales.

All I'm looking for is a single IC to make a base 2 log conversion of
frequency to pitch. V/Hz, to V/Octave. I don't mind if it takes its input
as a train of regular pulses or as a voltage derived from a train of pulses
(as produced by the frequency to voltage converters I'm already aware of,
if the voltage output of the PLL I'm using doesn't help). This IC should
preferably be an 8-pin DIL, very easy to use.