Analog Pitch Shifter

bharat pathak wrote:
> Jerry,
> 
>     I do not have analog background. So i do not know what can
>     be achieved and what cannot. By the way are there any good
>     "analog signal processing" books out there?

That depends on what you mean. Digital signal processing is mostly 
mathematics, and it it possible to do mathematical operations with 
analog circuits. Most designers of such circuits would be happy with 
slide-rule (i.e., 1%) accuracy. Somewhat better results -- maybe 0.1% -- 
can be had with the precision op-amp circuits of analog computers.

In the other hand, modulation, demodulation, filtering, channel 
compensation and more were all done routinely with analog circuits long 
before digital processing (or indeed even solid-state electronics) was 
contemplated. Circuit-design texts cover those applications.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On May 30, 3:55 am, Nils <n.pipenbri...@cubic.org> wrote:
> bharat pathak schrieb:
>
> >     I do not have analog background. So i do not know what can
> >     be achieved and what cannot. By the way are there any good
> >     "analog signal processing" books out there?
>
...
>
> In short you can do the following mathematical operations using analog
> electronics:
>
>    - add
>    - subtract
>    - multiply
>    - tanh
>    - exp
>    - square
>
> If you add an opamp you get:
>
>    - divide
>    - log
>    - roots
>
> You can approximate other functions like sin or cos either using
> polynomials. For sin the tanh function can be directly used for small
> angles.
>
> That's enough to do quite a bit of math. All the essential stuff is
> there. Some of the functions are hard to build in practice if you need
> high precision and low noise. Temperature compensation is not easy to
> get right as well.
>
> You can do delay using BBD chips or allpass/dome filters. That'll give
> you some memory if you need it. Not exactly high precision but it works.

you need to do *varying* delay *and* some means to splice and
crossfade.  other than the rotating tape head, i dunno how that could
be done non-digitally.  also, if you don't like glitches in your pitch-
shifted output, you need to be able to analyze the input signal to
look for good splice points where there is more similarity in the
waveform.  dunno how one would do that outside of a discrete-time
context.

r b-j

Jerry Avins wrote:

> bharat pathak wrote:

>>> Is it possible to create a pitch shifter using an analog circuit? 
(snip)

>> Why do it the hard way when it can easily be done in digital domain??

> Hard or easy, you imply that it can be done. I had thought not.

It depends on what the meaning of "digital" is.

I believe that there are analog tape players that do
it using electromechanical means.  Otherwise, it shouldn't
be too hard using sampled analog electronics, such as CCD's
(I believe called Bucket Brigade devices).  The latter may
or may not be considered digital.

-- glen

On May 30, 1:28 pm, robert bristow-johnson <r...@audioimagination.com>
wrote:
> On May 30, 3:55 am, Nils <n.pipenbri...@cubic.org> wrote:
>
>
>
> > bharat pathak schrieb:
>
> > >     I do not have analog background. So i do not know what can
> > >     be achieved and what cannot. By the way are there any good
> > >     "analog signal processing" books out there?
>
> ...
>
> > In short you can do the following mathematical operations using analog
> > electronics:
>
> >    - add
> >    - subtract
> >    - multiply
> >    - tanh
> >    - exp
> >    - square
>
> > If you add an opamp you get:
>
> >    - divide
> >    - log
> >    - roots
>
> > You can approximate other functions like sin or cos either using
> > polynomials. For sin the tanh function can be directly used for small
> > angles.
>
> > That's enough to do quite a bit of math. All the essential stuff is
> > there. Some of the functions are hard to build in practice if you need
> > high precision and low noise. Temperature compensation is not easy to
> > get right as well.
>
> > You can do delay using BBD chips or allpass/dome filters. That'll give
> > you some memory if you need it. Not exactly high precision but it works.
>
> you need to do *varying* delay *and* some means to splice and
> crossfade.  other than the rotating tape head, i dunno how that could
> be done non-digitally.  also, if you don't like glitches in your pitch-
> shifted output, you need to be able to analyze the input signal to
> look for good splice points where there is more similarity in the
> waveform.  dunno how one would do that outside of a discrete-time
> context.

That's for time domain pitch shifting.

Weren't the original vocoders for frequency domain
pitch shifting analog?  A bunch of narrow-band filters +
envelope followers driving the gain controls on a bunch
of appropriately re-pitched oscillators into a summing
node.

A type of additive pitch shift can be produced by using
a slightly mistuned single-side-band receiver, although
that's probably not what the OP wants.

.


IMHO. YMMV.
--
rhn A.T nicholson d.0.t C-o-M

Ron N wrote:

   ...

> A type of additive pitch shift can be produced by using
> a slightly mistuned single-side-band receiver, although
> that's probably not what the OP wants.

That's a frequency shifter, not a pitch shifter. All frequencies are 
shifted the same amount rather than multiplied by a constant.

BTW, I consider the rotating-head pitch shifters to be mechanical, 
rather than analog. Of course, the usual analog electronics were used 
for record and playback, but nowadays that part could be done digitally.

The first commercial application of rotating-head was time compression 
of audio. A good steno can type faster than many executives talk, and 
speeded-up playback of dictation tapes made for efficient transcription. 
As far as I know, early video slo-mo was accomplished simply by slowing 
the tape on a spiral-scan recorder. It works with VHS and Betamax.

jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On Thu, 29 May 2008 19:28:10 -0500, "ryder1650" <ryder1650@yahoo.com>
wrote:

>Is it possible to create a pitch shifter using an analog circuit? I know
>this site is about DSP's but it seems to have a lot of readers, 

   And a significant amount of DSP code is simulation of analog
circuits, so posters here might naturally know about these things.
It's been over a day and I'm surprised no one else has responded with
what I describe below.

>so maybe
>someone can help me out. 

   It depends on what you call 'analog'. Fully electronic pitch
shifting was originally done in the 1970's, I'm pretty sure using a
bucket brigade device (BBD - they're digital sampling-wise but analog
amplitude-wise) and a VCO for its clock input and a variable-slope
"triangle" waveform for the voltage control input of the VCO. By using
two of these with the waveforms driving one of the VCO's inverted, one
can switch the inputs and outputs between the two BBD's and get a
continuous, if crude, pitch shift - no fancy stuff like zero crossing
detection I saw mentioned in some other comp.dsp thread.

This page has a very short description of such a pitch shifter:
http://www.geocities.com/scaprile/fxa.html


>
>Thanks
>

   (It looks like some attribution(s) was lost in quoting, but all I
did was hit reply, please don't blame me...)

On Fri, 30 May 2008 13:28:53 -0700 (PDT), robert bristow-johnson
<rbj@audioimagination.com> wrote:

>On May 30, 3:55 am, Nils <n.pipenbri...@cubic.org> wrote:
>> bharat pathak schrieb:
>>
>> >     I do not have analog background. So i do not know what can
>> >     be achieved and what cannot. By the way are there any good
>> >     "analog signal processing" books out there?

>> ...

>> You can do delay using BBD chips or allpass/dome filters. That'll give
>> you some memory if you need it. Not exactly high precision but it works.
>
>you need to do *varying* delay 

   BBD's with ramp-waveform-driven VCO's for clock inputs do exactly
that (see my other post).


>*and* some means to splice and
>crossfade.  

   One can use r-c networks driving voltage-controlled amplifiers to
fade in and out between the two BBD outputs, so the sound isn't so
glitchy in switching between them.

>other than the rotating tape head, i dunno how that could
>be done non-digitally.  

   Well, now you know...

>also, if you don't like glitches in your pitch-
>shifted output, you need to be able to analyze the input signal to
>look for good splice points where there is more similarity in the
>waveform.  dunno how one would do that outside of a discrete-time
>context.

   That's indeed harder to do in analog electronics. But it should be
possible to delay switching between the BBD's until the next, say,
positive-going zero crossing comes along (this might need one or more
flip flops and a few logic gates, so it's no longer purely analog...).
Of course a zero crossing in one direction is not neccesarily the same
part of a repeated waveform as a different occurrence of the same
polarity zero crossing, but doing that might sound better than
switching at an arbitrary time.

>
>r b-j

glen herrmannsfeldt wrote:
> Jerry Avins wrote:
> 
>> bharat pathak wrote:
> 
>>>> Is it possible to create a pitch shifter using an analog circuit? 
> (snip)
> 
>>> Why do it the hard way when it can easily be done in digital domain??
> 
>> Hard or easy, you imply that it can be done. I had thought not.
> 
> It depends on what the meaning of "digital" is.
> 
> I believe that there are analog tape players that do
> it using electromechanical means.  Otherwise, it shouldn't
> be too hard using sampled analog electronics, such as CCD's
> (I believe called Bucket Brigade devices).  The latter may
> or may not be considered digital.

Anything is considered digital when it looks good in the marketing 
brochures. Many things have been called digital, merely because there is 
a chip in the system with a part number starting 74... :-)

Regards,
Steve

Steve Underwood wrote:
> glen herrmannsfeldt wrote:
>> Jerry Avins wrote:
>>
>>> bharat pathak wrote:
>>
>>>>> Is it possible to create a pitch shifter using an analog circuit? 
>> (snip)
>>
>>>> Why do it the hard way when it can easily be done in digital domain??
>>
>>> Hard or easy, you imply that it can be done. I had thought not.
>>
>> It depends on what the meaning of "digital" is.
>>
>> I believe that there are analog tape players that do
>> it using electromechanical means.  Otherwise, it shouldn't
>> be too hard using sampled analog electronics, such as CCD's
>> (I believe called Bucket Brigade devices).  The latter may
>> or may not be considered digital.
> 
> Anything is considered digital when it looks good in the marketing 
> brochures. Many things have been called digital, merely because there is 
> a chip in the system with a part number starting 74... :-)

One can even buy digital alkaline cells.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On May 31, 12:06 am, Ben Bradley <ben_u_brad...@etcmail.com> wrote:
>    (It looks like some attribution(s) was lost in quoting, but all I
> did was hit reply, please don't blame me...)

no blame necessary.

> >other than the rotating tape head, i dunno how that could
> >be done non-digitally.
>
>    Well, now you know...

okay, i'll buy that.  now for the CCD device that just had its output
faded to zero, then you need to reset its delay.  is there any
(digital) logic in doing that?

> >also, if you don't like glitches in your pitch-
> >shifted output, you need to be able to analyze the input signal to
> >look for good splice points where there is more similarity in the
> >waveform.  dunno how one would do that outside of a discrete-time
> >context.
>
>    That's indeed harder to do in analog electronics. But it should be
> possible to delay switching between the BBD's until the next, say,
> positive-going zero crossing comes along (this might need one or more
> flip flops and a few logic gates, so it's no longer purely analog...).

yeah, that's sorta how i felt about it.

> Of course a zero crossing in one direction is not neccesarily the same
> part of a repeated waveform as a different occurrence of the same
> polarity zero crossing, but doing that might sound better than
> switching at an arbitrary time.

not a particularly good pitch detector.   i wonder how to do an analog
AMDF or ASDF circuit?  maybe you could have a circuit for each lag.
then you need a way to pick the minimum lag.

i dunno.  an intriguing concept (pure analog pitch shifter) that i
hadn't thought too much about.  it looks to me that Vladimir has the
simplest plausible concept of it.

r b-j