Analog tuner idea (try #2..no show)

[Magnus Danielson]

>>>>> "S" == Synthfool  <Synthfool at aol.com> writes:

 S> (forgive me if this gets posted twice, I didn't see a copy from DIY after 12
 S> hours)

 S> Hello all,

 S> A while back I felt it necessary to build myself a box to help tune
 S> synthesizers.

Hey, arn't they allways in tune ;)

 S> The unit I built is very simple- a six position switch that selects taps on a
 S> voltage divider of resistors to give 0,1,2,3,4,5 volts out thru a 1/4" jack.
 S> This parallels to two mini banana plugs for my voltmeter, so I can monitor
 S> the actual output voltage. There is also an octave trim and a ten turn
 S> "offset" pot that adjusts the output voltage up or down about 100mv. 
 S> There is a voltage regulator, but no output buffering and as such the offset
 S> control is needed to compensate for varying vco input impedances and loads.

 S> While my unit is very usable, it obviously still suffers some drawbacks.
 S> Having just ordered some .1% Holco resistors (from mouser) with plans to
 S> build "Tuner #2", I thought I'd bounce some ideas back and forth with you
 S> guys before I embark on my new version. Possibly we can come up with a nice
 S> design and do a PCB layout and all.
 S> It is a lot easier and more accurate to use something like this than a
 S> standard 1v/oct keyboard, IMHO.

 S> Here are some of my thoughts so far-

 S> 1. A dual rotary 8 pos switch (mechanical or electrical) that selects a xtal
 S> divided audio tone along with the  output voltage. I would probably just base
 S> my unit in the key of C, but I can see the possibility of doing a "key"
 S> selector too. While the voltage would be easy to offset, I haven't explored
 S> how difficult the frequency divider part is.

Getting a set of A4's is trivial using simple divide by 2 stacks like
a CMOS 4024 or something. Getting other tones is quite possible, but
there is drawbacks. For instance will top-octave-divide down cursuits
do this by haveing a high tone and fixed divisions to get all twelve
tones, but the drawback is that they will be sligthly off...
Getting them more exact using this technology needs greater dividers
and this calls for higher common frequency...

A good technique to solve this problem is to use an phase-accumulator
runing at a fixed known frequency. The phase accumulator is nothing
but a addition cursuit and a output latch. The output is being feed
backwards to one of the inputs and the other input is being constantly
fed with a fixed number. This constant will be linearly related to the
output frequency.

	                 c
	fout = fconst * ---
	                2^n

where
	fconst	is the addition clock frequency
	fout	is the output frequency
	n	is the number of bits in the accumulator
	c	is the addition constant

To show the power of this technique I migth add that a 24-bit
accumulator is enougth to take all MIDI notes (0-127) at any bend
uppward (0-99 cents) within +/-0.025 cents of limit. You can safely
allow a cheap 8-bit CPU running a small fixed or timed loop with the
calculations.

You will not need to use a divide down cursuit, but if one has an
external divide down cursuit one can use the high accuracy in the high
frequency range and just pick it down. There is a of course allways
the problem that while the frequency exactness is high at high
frequency is the momentary phase error may be higher due to simple
aliasing. The simple cure for this is either to increase the clock or
tap a lower frequency. If you allow for longer accumulator you may
gain low frequency exactness. In a 8 bit CPU can an 8 bit extra
exactness be just a few extra instructions per cycle... which migth
call for a sligthly lower loop clock, but then again it may be worth it.

 S> 2. A thought based on the Digisound Cal-tuner that uses a ring modulator to
 S> beat an internally divided tone against the vco under tune in order to
 S> exaggerate the beating effect. This in conjunction with LED's possibly as a
 S> tuning indictor using comparators.

Why not use an FM detector and allow a tap-in of an multimeter?

An audio-visual combination is a good thing, neigther of them will
survive on it's own I think...

 S> 3. Multiple 1/4", 1/8" and banana output jacks.

Good thinking. Throw in some RCAs and XLRs while you are at it :)

 S> 4. A momentary "ground" switch to facilitate back and forth tuning when doing
 S> vco scale adjustments.

Are you sure you necessary want to tune toward 0 V ?
I'd prefer an A/B setting if I where to go all the way.

 S> 5. Sine wave converter for higher reference frequencys and square waves for
 S> low freqs.
 S> I find tuning vcos to sinewaves is easier, but at very low freqs the sines
 S> tend to get lost.
 S> Possibly an inverted vcf that will increase the harmonic content of the
 S> tuner's ref freq as the ref voltage goes down.

You have poorer pitch judgement down there... but the overtones helps
you out.

 S> 6. A jumper with an alligator clip as a ground. This just seems to be a good
 S> idea.

Not sure why you want it... but why not, could be usefull.

 S> 7. A built-in 4.5 or 5 digit DMM and amp/speaker would be nice too.

Neat, there's plenty of things one can come up with once you have one
of those aboard... but then, this is what they call creeping features.

 S> Any ideas?

I hope some poped out above... 

I would use some voltage reference (there's fairly cheap 10V and 5V
onces), use sligthly trimmable divider cursuits or multiplying DACs to
get the appropriate voltage. If one uses a multiplying DAC some care
has to be taken so that there is even steps... this can be done either
by having the upper reference a little higher (say at 10.24 V or 5.12 V) 
or possibly moving the lower sligthly below zero (say at -0.24 V or
-0.12 V). This way will the steps fit nicely to a logarithmic scale.

With some care built in to both the referece voltage and reference
frequency generators and the measuring of the missmatch one could also
be able to measure (and possibly trim) the "nonlinearness" or rather
deviation from the logarithmic curve. Good measurement of that can be
usefull in many ways for discovering temperature drift, power drift
(some cursuits have dependece from the power) and continental drift
(couldn't resist that one :).

For instance, the ASM-1 VCOs have still some dependence on the
supplied power for it's tuning, some minor changes including hooking
it to a voltage reference will fix that. Also, this VCO is a good
example of a VCO having demands for large ranges of sweeps (about
1:1000000 is certainly not bad).

If one builds this the rigth way we can add modules until we got
ourselfs a Audio Precision killer :)

And yeat again I couln't resist babbling on... how do you all surivive?

Cheers,
Magnus