What's the distortion adding feature of the new $$L console?

[Kev]

thanks Wayne

I don't want to hijack the thread ... this is one of those subjects that just fascinates me ... hence the stuff at TT and the in the GDIY updates

you're not the only one that say this and I'm not making any conclusion or assumptions

others have said it is only shows a fault through the electronics.
so hard to do this test with only acoustics and part of me does feel that there probably is an asymmetrical response at the eardrum and associated parts

don't hijack the thread Kev !!
genuinely fascinated
new thread required

 

[AnalogPackrat]

[quote author="Kev"][quote author="AnalogPackrat"]If you look at a typical time domain representation of a signal, you can see that there are two axes of symmetry--time and magnitude

Now take our old friend the sawtooth wave. You have to invert its magnitude and mirror it in time (in half cycle pieces) to get it to match itself. It is not possible to do away with the mirroring. The sawtooth contains only even harmonics.[/quote]

thanks AP
I too wanted to look at the waveform and look at possible techniques from the synthesiser world

but I was trying to move away from continuous repetition of waveform (cycles) and look closer to the rate of change of the amplitude
I think the non-linearity we might want is continuos and without instant change[/quote]

Well, where I was going with that train of thought (before the old noggin' lost it while I was typing) was... Take the difference between the input and the "harmonically enhanced" output. What does that difference signal exhibit w.r.t. symmetry? I'm pretty sucky at math anymore, so the pragmatist in me wants to get on with the experiments (either in solder or simulator). :grin:

AP
"the saw tooth has only evens"
I thought both saw and reverse saw contained all integer harmonics ??
they sound the same anyway
but can help hightlight a circuit that is asymmetrically clips ... sory I'm back on clip and soft clip again
perhaps I'm not getting the significance of the mirror in half cycle pieces.
...
wouldn't that make it more like a triangle wave ... lots of odd ?

Saw and reverse saw both contain only even harmonics. The only difference is the phase (or polarity) of the harmonics relative to the fundamental, I believe. In fact, simply inverting the polarity of the sawtooth makes it "reverse saw." Seems obvious this is going to sound the same to your ear.

And I can hear the difference in some instruments when they're "upside down" at the ear. Shoot me if you think I'm wrong.

once they have passed through a circuit that introduced distortion ?
or simply straight out of the recorder as pos or neg polarity ?

this was very much the content of that thread I started ... long ago in the wayback machine
it was also the subject of one of the Group DIY updates.

I'm curious about this phenomenon (manah manah), too. I know that a non-square pulse train has that oboe sound. It's an oddball wave, too, in the sense that, if you remove the DC you get tall, but narrow pulses on one polarity and short, but wide ones on the other. It's almost like one set of frequencies pushing and another set pulling, but their energy is balanced. Makes my brain hurt to think about it, but I wonder if that's a clue.

A P

 

[Kev]

[quote author="AnalogPackrat"] Saw and reverse saw both contain only even harmonics. [/quote]

the wikipedia currently tells a different story

http://en.wikipedia.org/wiki/Sawtooth_wave

or am I reading it wrong ... and I don't want to suggest that Wiki is gospel

A sawtooth wave's sound is harsh and clear and its spectrum contains both even and odd harmonics of the fundamental frequency. Because it contains all the integer harmonics, it is one of the best waveforms to use for constructing other sounds, particularly strings, using subtractive synthesis.

I've done too much reading this morning and getting a headache ... more coffee ... no more net time ... later

new thread on the absolute polarity thing

 

[AnalogPackrat]

I did some more checking and the Wiki page is correct! Saw has all harmonics, not just even. Retract! Retract!

And I still have to be here at work for three more hours in this obviously deficient state of mind....

A P

 

[AnalogPackrat]

OK, in a vain attempt to salvage my questionable academic reputation here, I have put together a simple spreadsheet to help visualize harmonic content by additive methods.

Get the spreadsheet here.

Column B is a simple index for frequency generation. Column C is the fundamental. Columns D thru M are the 10 harmonics you get to manipulate. Column O is the summation of the harmonics and is plotted in the graph window. Row 1 is the polarity setting for each of the harmonics (I did not provide for complete phase shifting of the harmonics relative to the fundamental). Row 2 is the amplitude formula (which currently contains polarity * [1 / f^2]). Row 3 is the harmonic relative frequency multiplier (1 is fundamental, 2 is second harmonic, etc.).

Manipulate Rows 1-3 to create the various textbook waves. Set the polarity to 0 for columns you don't want to use (if you want to use fewer than 10 harmonics).

The equation in the rest of the cells computes the amplitude (and polarity) weighted harmonic. It is currently using a sin. You can change this to cos in one cell and copy to all the others (being careful not to accidentally mess up the other cells).

I fiddled around a bit and found a combination of amplitude and polarity for an all even harmonic wave that "looks" right. Note that it is basically a skewed sawtooth. It ramps from zero to max in 0.188f, back to zero in 0.312f, zero to min in 0.312f, and back to zero in 0.188f. Pretty interesting, I think.

Have fun with it...

A P

 

[Kev]

thanks AP and no worries
and yes
very interesting

as soon as I have fixed my communications with the server I may have something simple to look at that looks at a single harmonic with an asymmetric in amplitude and an asymmetric in time ... err well almost depends on point of view.

just using the simulator to illustrate a point and not suggesting it helps to tell us what it might sound like

for those interested
http://www.diyfactory.com/data/discussion/20percdist.htm
but I can't seem to get it working right now ?? bare with me ...
[edit] I think it may be working now

don't read the FFT bit yet ... it's a work in progress ... yeah right :roll:
but if you do ... do do the London Police Whistle Google search for a little fun.

 

[Mendelt]

[quote author="mediatechnology"]Very well put analogpackrat.

So how about this method to generate even-order via multiplication?

Take the antilog of (log(Ein)*2).

This is squaring or Ein^2. Or as previously put, a signal AM modulating itself.

This would be a log-domain method of multiplying a signal by itself rather than using a linear analog multiplier. Second-order results?

And for those not math-challenged: Would the antilog of (log(Ein)*3) produce odd-order results?

If so, we do the math using transistors we can use either two or three junctions or a X2 or X3 gainstage.[/quote]

I'm a bit rusty on the math too (and I remember all this from my vst-plugin-programming days so i'm not sure if all this applies to analog)

As far as i remember sig^2 will generate 2nd harmonic, sig^3 will generate 3rd harmonic and so on. You can google Chebychev polynomials if you want to know more about this.

So if the transfer curve of a certain circuit is logarithmic for example you can find the amount nth harmonics that is created by that circuit by looking at the coefficient of the nth polynomial in the taylor series of the transfer curve. But I'm not really sure how to do the math to do this.

 

[recnsci]

qucky calculation gives bit like (sin(x))^3=-0.75*sin(x)+0.25*sin(3x)

Playing with analog multyplyers is dificult because of dynamic range
requirements (that is if you want to get more than 50 dB of input dynamic
range). OTOH we dont need full 4 quadrant mult for this since we dont
need pure quadratic function, we need something like x+ax^2. So we
take nice 1 quadrant mult ( like one in MAT02 datasheet), and put
DC bias at input that will provide normal operation for full dynamic
range of input signal. Get gain control on input and inverse control
on output (dual pot will give you this but dont know how much
freedom you will have with control law). This will give you one knob
control over level of second. Big problems will be noise and HF response.

Now another aproach: somewhere on passdiy.com Nelson Pass gives
his rendition of famous JLH class A amp, called PLH. In nice little pdf
he wrote there is interesting discussion on playing around bootstrap.
With single pot we change working mode from pushpull-ish (with more
odd THD signature) to singleend-ish (with more even THD signature).
Now, there is some NFB around this amp. If we decrease NFB (and
attenuate at input to compensate for higher closed loop gain)
ratio between lower and higher products in spectrum will change.
Equivalent thing with decreasing NFB. So now we have >>some<< control
over THD signature. Problem is it probably wont put more than 10%
before clipping. And, god knows how will it sound.
There is ,at least in theory, workaround to pull more than 10% from
this nice little three transistor thingie. Figure out what is gain of undistorted
part (like, measure low level incremental gain), and substract controlled
amount of undistorted signal from output of our variable THD amp. What
would be handy is if we could provide automatic gain compenstaion
in next stage to keep level of undistorted same and thus change just
relative amount of THD. Now problem: we are feeding clean and distortin
amp in parallel. I would personaly consider mandatory that both amps
clip in clean controlled way and at SAME input level. Thus there wont be
funny break points in compound transfer function. There are bunch of
possible refinements and problems with this bridged aproach, but I'm
aranting too long anyway

 

[recnsci]

There is one serious problem with Taylor, and that is requirement for
no breakups in first n diferentials for polynomial of n-th order.
This is basic reason why it is useless to look at THD spectrum of
for instance output of FWR driven by sine. Drive it by square wave
and you have just DC on output. Drive it by triangle and you have
triangle of doubled frequency. Drive it with saw and you have triangle
of same frequency. Super nice for VCO in modular synth but useless
for any conclusions of behaviour with real world signals. That is, it will
be nice FX but not sweetener.

cheerz
ypow

 

[Svart]

now that is cool. I was thinking of something like this but could never realize it as you have Wayne.

 

[Svart]

I'm not sure that marketing has much to do with reality these days. I assume and hope that $$L would not fall victim to the B*hringer ways of hyping a product only to find that it *barely* has a remote relation to the actual design, ala "tube warmth".

So with keeping under the assumption that $$L stands for truth, I assume that the quality of their "variable harmonic drive" is high.

So in short, I believe they could but I think they would have refined it much more and likely found a way to extract the harmonics completely.

I think we are on the right track though.

 

[recnsci]

Unfortunately, that circuit will give you odds only. For 2x gain between
log and anti log you will have sgn(Vin)*(Vin)^2 which is not the same
as (Vin)^2.
Output of Log amp is confined to +-0.6V. That means that in worst case
second opamp will try to put 1.8V across antilog diode. Ouch

Take two nicely matched quad NPN arrays, all diode connected.
Put three diodes in Log amp feedback and conect remaining diode
to OA output for antilogging . Do the same thing with other quad
array for opposite polarity. Make sure output diodes are antilogging
voltage from their corresponding arrays so you dont have temp
drift and transistor parameters missmatch dont screw you.
Now, current trough output antilog diodes will be poportionall to
cubic rooth of input voltage. Put this contraption inside of FB loop
of another OA in inverting configuration so that antilog diodes are
feeding virtual earth of other OA. This will provide on paper
cubic function.
In reality, low level performance is affected
by the fact for low signal levels log amp is not loging actually. And you
need one hell of a precision OA cus even small Vos and Ibias of log
amp will screw things up. And you will actually want one tightly matched
octal array. And you will have dynamic range issues. And when
you solve all that, stability of whole contraption will be bitch.

cheers
ypow

 

[recnsci]

Sorry, slight error. You will need one diode around log amp and three
diodes on its output, not the other way around.

 

[Kev]

just in case some missed the link
http://www.diyfactory.com/data/discussion/20percdist.htm

two quite different resultant waveforms produced by the same distorting component and have the same THD and the same power spectrum but only differ in the phase of the distorting component

does anyone recall the Headwize saturation diodes+resistor in the feedback idea ?
the pre-emphasis de-emphasis was added later I think

does the SSL product have a mix wet/dry feature ?

 

[ruairioflaherty]

This is very cool. Last night at my electronics night course we were studying how a square wave is made up of sine waves ala fourier and when I got home you guys were discussing this here. Unfortunately I don't know enough to help but I'm watching from the sidelines with great interest. I'll build and test in the trenches whatever you all come up with.

Big thanks to Analog Packrat for the excel doc - it visually demonstrates everything I am learning,

carry on :grin:
Ruairi

 

[Svart]

oh wow that sounds awesome! imagine cymbals with a hint of that!

Wayne you are totally on to something here.

 

[AnalogPackrat]

Sorry I went AWOL on this thread. Had a 2 hour video conference call to Korea last evening after my post. Work sucks sometimes. Anyway, I did hack my spreadsheet really quickly to verify that, other than a gross amplitude difference, yes, sin^2 gives 2nd and sin^3 gives 3rd. Interesting.

Wayne--that sample is really nice. This could definitely be a useful tool to build. Thanks for the effort and for sharing that. When's the PCB going to show up in the black market? :grin:

Kev--did a quick browse-by on that page you linked. Interesting. I've always been curious about the perception of the total waveform when its contributing components are phase shifted. Might have to whip up a little program to generate some sample wav files...hmmm.

Ruairi--glad you found the spreadsheet useful. Way back in the dark days of 1-2-3 and DOS I used to fiddle around with FM synthesis the same way.

A P (spreadsheet addict)

 

[evan38109]

I was just listening to mediatechnology's demo file when my roommate walked in and said, "Ooo, what is that? It sounds really rich."

- E

 

[ruairioflaherty]

For beginners like me who had never seen a "log amp" here is a link with a little background

http://www.play-hookey.com/analog/logarithmic_amplifier.html

cheers,
Ruairi

 

[Svart]

interesting, makes you wonder if you could use this to create an EQ of sorts..