Trying to emulate a Minimoog saw

xav · Last edited by xav on Wed Sep 24, 2008 2:28 pm; edited 1 time in total

I did several tests with a minimoog. The sound of the saw was my first preoccupation. The oscilloscope shows a perfect saw for the G2, and a rounded one for the Minimoog, especially in it negative side. I tested the lowest note the tuner recognized, E 32'. The sound is a little bit different to my ears.
This what I did to imitate the saw of the mini, listening, looking to the oscilloscope and a multimeter:
1. I tried to tune perfectly the Minimoog to the G2 and to get the same volume.
2. I reversed the wave of the G2 since it is in the opposite phase, to compare easier the shape. (why did clavia change the sense of the saw since the G1?)
3. I tried to change the waveform with a shaper and FM. The best results were when the transformation is dynamically synced to the waveform. FM changed the evolution of the wave under 0, and the shaper rounded especially the negative side too. It is gradually modulated by the saw (before it is reversed). And I tried a overdrive that shapes the wave in a good way (visually...).
I cannot say the result is perfect, since it still lacks a little bit of high frequencies to match to the Minimoog saw, but it is nearest than the pure saw of the G2. I'll appreciate if anybody get a solution to get closer.
Here are the pictures of the analyzes and the patch. Variation is the pure saw, and 2 is transformed.

E 32' saw minimoog.png

Description:

Filesize:

137.32 KB

Viewed:

536 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

E 32' saw G2 FM, Shape, overdrive.png

Description:

Filesize:

140.65 KB

Viewed:

479 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

E 32' saw G2.png

Description:

Filesize:

155.15 KB

Viewed:

505 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

Test saw moog.pch2

Description:

Download

Filename:

Test saw moog.pch2

Filesize:

1.29 KB

Downloaded:

1579 Time(s)

iPassenger · Posted: Fri Sep 12, 2008 12:36 am Post subject:

Thanks Xav,

The waveforms certainly look similar if not the same. I found similar differences in the waveform when I studied my waldorf pulse's Saw/Squ waves etc vs the G2 (which is pratically perfect). Although the effect appeared to have a bias to the note played. In other words higher notes produced more accurate waveforms than lower ones, (could this be to do with the saturation into the filter perhaps?). In order to try emulate this I tried placing EQs on the outputs of the Oscs but with limited success.

Anyway I look forward to testing this patch out when I am back at home.

R.

kmmcdonald · Joined: Oct 08, 2005 Posts: 22 Location: USA Audio files: 1

xav:

What software are you using to provide the oscilloscope & spectrum analysis?

thanks

Keith

BobTheDog · Posted: Sat Sep 13, 2008 11:10 am Post subject:

Looks like The MultiMeter and tuner in Logic and NI Reaktor doing the scope

Derek Cook · Posted: Wed Sep 17, 2008 12:41 pm Post subject:

One nice thing on Yamaha's AN1x, PLG150-AN and AN engine in the EX5 (yes, I have them all as well as my beloved Nord G2 Engine!) is the ability to "round" the edges of the waveforms to emulate the less than perfect waveforms of classic synths.

So is it possible to create something like this on the Nord? I haven't really thought about it, but the Yamaha AN Edge parameter must be equivalent to an LPF where the cutoff is synced somehow to the VCO pitch, which sounds like what XAV is trying to do.

xav · Posted: Wed Sep 24, 2008 2:24 pm Post subject:

Sorry for the delay. Yes, Bob The Dog is right; Logic, Reaktor. But the shape is only useful to monitor what I'm doing. The best tool to see the differences was the spectrum display (especially RME's one).
I tried to do the same job with a Wersi Organ. The Flute 16' is beautiful. It looks like a sine for the first half circle, and something more chaotic. I never succeeded in making such a waveform with the G2. It would be so nice to draw a one cycle waveform and to use it as an oscillator.

Tim Kleinert · Posted: Wed Sep 24, 2008 4:09 pm Post subject:

xav · Posted: Wed Sep 24, 2008 5:22 pm Post subject:

You are right Tim (as usual). I gonna try this way. Not easy but the sound seems good... maybe more flexible chaining two seq.

Tim Kleinert · Posted: Thu Sep 25, 2008 1:54 am Post subject:

xav · Posted: Thu Sep 25, 2008 6:47 am Post subject:

Thank you Tim. Those 3 ways you mention are really interesting. Your digital manipulations and considerations are amazing.
About fine tuning, the minimoog used for those tests waas really weird. When it's perfectly tuned for the lowest note, it's extremely out of tune in the higher range. I don't know if it is normal.

iPassenger · Posted: Thu Sep 25, 2008 9:02 am Post subject:

Tim Kleinert · Posted: Thu Sep 25, 2008 9:54 am Post subject:

cheers, guys. Smile

There's no magic to all this. It's just a self-resetting counter circuit, clocking at audiorate, which is fed by a keyboard-tracking lin-expo converter. If I reset it to zero regardless of how high the overflow value is, the generated oscillations will always have the same length of samples and will therefore not be jittered by the sampling frequency. This is why the waveform always sounds clean, but the tuning is a bit skewed. (The other reason for the tuning issues is that the LevelScaler output isn't accurate enough for these kinds of applications.)

I use these kinds of counter circuits alot in my recent designs. In the voice autotuner, there are four such counters.

Oh, btw, the two gain stages (RANGE and MAKEUP gain) in the saw shaping example patch are actually superfluous, since tweaking the saturation curve gives the same form of control. So you can make it more efficient by leaving those out.

As I say, the DIY sawtooth core can drive any nonlinear processing without ever aliasing.

@xav: sounds like the tracking needs to be calibrated on your mini.

cheers again, Smile

tim

iPassenger · Posted: Thu Sep 25, 2008 12:39 pm Post subject:

yeah i got that tim Smile

... The tuning issues are a bit weird, i don't understand why those values on the mixer make a reasonably accurate tuning but they do.. so I'm not going to worry about it.

Been messing with my own multiple osc synth based on the no alias saw, using the saw to derive a pulse (comparator), tri (wavewrap) and then sine (shaper).

Probably being a bit thick here but what does non-linear processing mean?
If I use say, a standard sine osc to control a crossfader, is the sine going to introduce an element of jitter into the crossfade? And your conter circuit avoids this issue by being unaffected by sampling freq? Is that what your getting at?

Cheers

R.

Tim Kleinert · Posted: Thu Sep 25, 2008 1:02 pm Post subject:

xav · Posted: Thu Sep 25, 2008 2:19 pm Post subject:

[quote="tim"]cheers, guys. Smile

It's just a self-resetting counter circuit, clocking at audiorate, which is fed by a keyboard-tracking lin-expo converter. If I reset it to zero regardless of how high the overflow value is, the generated oscillations will always have the same length of samples and will therefore not be jittered by the sampling frequency.

I can just imagine what you say, like I was dreaming while reading Jean-Paul Sartre's books in philosophy... but it's sure I don't understand how you designed a oscillator with a mixer, a compare to level, and a value switch. And why this DIY is it aliasing free? But I'm sure you explained that very well... It will take time for me to get it.
Any way, thank you for sharing.

Tim Kleinert · Posted: Thu Sep 25, 2008 2:26 pm Post subject:

@iPassenger:
Oh, I just saw that I didn't really answer your question. So I will try herewith.

Let's say, we want to have an oscillator tuned to C4. This equals to about 261.63Hz . Within a digital system clocking at 96000Hz, this would equal 96000/261.63=366.93 samples. As you can see, it isn't 366 samples and also not 367 samples, but something inbetween. In other words, our digital system cannot faithfully represent this frequency with a constant amount of samples. In tech-speak: The oscillator frequency is non-fractional to the sampling frequency.

There are two solutions: Either we agree on a constant 366 or 367 samples and accept that our oscillator will be slightly out of tune. (My DIY design does exactly this). Or: We use a non-fractional counter, which will produce many oscillations of 367 samples, with an oscillation of 366 samples thrown in there once in a while in order to compensate (so that it averages out in the long run).

However, this causes a periodic jerk in the oscillation, as the wave cycles will be mostly 367 samples, but then 366 samples thrown in between now and then. This jerk is called jitter, and it causes unwanted periodic sidebands. In tech-speak:The oscillation is jittered by the sampling frequency.

The artefacts caused by these "jerks" are strongest on steep waveform flanks. Pulse waves have two of these, sawtooths one, that's why these waveforms are more prone to producing artefacts than eg. sines. Simply put, oscillator anti-aliasing smoothes these flanks so that these jerks become less intrusive.

But if we stick to a fixed amount of samples, we're out of trouble. Smile

Hope this clarifies.

cheers,
tim

@xav:
The mixer is in 100% feedback. When I apply something at its input, it gradually builds up internally. The comparator detects when the buildup has maxed out and resets the circuit to zero by interrupting the feedback loop (the value switch). And so the whole thing starts from the beginning again. Rise.... -reset-rise.... -reset... voilà votre oscillateur de saw clown

xav · Posted: Thu Sep 25, 2008 4:26 pm Post subject:

Ah yes, I understood. Thank you very much.
The next step will be to appreciate the logic of your keyboard-tracking lin-expo converter. I presume this conversion has an arithmetic explanation I'm not able to understand.
But with this auto-reset mode, you are sure to be synced with the sample frequency, isn't it? I think about that amazing 5,4 sec delay designed by Blue Hell that used a sample frequency divider to alow a delay module to process alternatively looped samples and new ones. That was first time I realized the PCM limitation could be controlled and even be used in a creative way.

iPassenger · Posted: Fri Sep 26, 2008 1:50 am Post subject:

Cheers tim, fab explanations.

So i was already doing this non-linear wot-sit. Smile

Your explanation for the cause of the sample freq jitter is great too, to hell with tuning, tight waveforms are better, particularly for osc sync etc.

I presume this kind of osc would prove difficult or impossible on the G1 due to the lack of consistent calculation order?

Presumably these side bands caused by the jitter don't sound horrendous on their own when just a clean osc sound is used but once combined as part of modulation signal etc... the results become increasingly easy to hear, as the artifacts affect the process.

Tim Kleinert · Posted: Fri Sep 26, 2008 1:54 am Post subject:

Tim Kleinert · Posted: Fri Sep 26, 2008 2:05 am Post subject:

iPassenger · Posted: Fri Sep 26, 2008 2:54 am Post subject:

As soon as I played around with your osc module and made my own i realised how easy it would be to add sync, I used a logic OR module, one input from the osc to be synced comparator output and one from another diy osc comparator output.

It certainly sounded cleaner...

Also using this method I figured it would be much easier to build the sync reset points to the right place and what not. Though I didn't experiment with it for long as I started building the rest of the synth to go round it. I'm ashamed to admit that I used standard filter and env modules. *hangs head in shame*.

xav · Posted: Fri Sep 26, 2008 7:58 am Post subject:

Thank you Tim. You help me very much.
So one problem of tuning is the lack of precision of the Lev scaler module. For a 6db/slope, it makes more than doubling the amplitude. Maybe one can find a way to improve that...?
The other is your choice to control the oscillator by a multiple of samples (the lowest possible fraction of the clock). As soon as you control a DIY oscillator that way, it won't be jittered by sampling frequency, and won't be perfectly tuned.