Xhip vs Analog Hardware

[Mutant]

Elo everyone.

So... Today i found a nice set of raw oscillator samples from the new excellent Moog Sub37.
http://dsiforum.com/viewtopic.php?t=6118

As always i got curious if i could get somewhere close to that kind of saw wave with Xhip.
And this time i can say that i got a PERFECT match by human ears standard (not digitally bit perfect of course).
Made the test in 96Khz SR.
In this clip first half is Xhip and second half is Sub37.
https://app.box.com/s/xyf8t0lhlukaais4pmsz9bgfg4b8ng5r

And in this clip Xhip is on the left and Sub37 on the right.
https://app.box.com/s/5fu0dnt92svyzulq7qoss3ctxbwpqmyh

The preset:
https://app.box.com/s/tdk5al1c15girv2yhujowr91c9lurrar
As you can see, there is nothing out of ordinary here, just matched the pitch and filters and the loudness, no rocket science.
Took just few minutes.

Also to be fair i need to say that i used ReaEQ with sub half decibel adjustments there and there to make it a perfect match while looking at the spectrum on maximum zoom in Voxengo SPAN - though it was indistinguishably close even before the EQ.

I kind of can't believe it myself.
These are MOOG oscillators that everyone says they are "Warm and rich sounding Moog saws you can't imitate with anything else." and "You have to have a Moog to get that Moog sound.".
Did i just get a 1500€ sound for 0€ on a VSTi that is not even an emulation of Moog ? WTF ? Hahaha !

Of course just the oscillator doesn't make a synth and if i wanted to match the filter i would probably not get so close, but still... it is a perfect match of a Moog oscillator on a freebie VSTi.




ps. To avoid opening a new thread every time i want to match something with Xhip, i will just post it here in the future, thats why the general "Analog Hardware" in the thread title.

ps2. 40 views and only 4 downloads of the audio files each and 1 download of the preset... hmm maybe not enough drama here, or maybe people just don't care as much as i do.

[Mutant]

Look on top right for the download button if you want to analyze the audio in your DAW.

[aciddose]

Try the pulse wave. The ramp waveform in any synthesizer is going to be identical unless you're dealing with rare cases in which the buffer saturates or so on, such as is possible in the MS- synthesizers given the right ("wrong") FET although unlikely if not impossible (need at least -4 VGS) in the factory configuration.

This is because the FET source is supplied by the raw, unregulated +18v output from the PSU capacitor before the regulator while the peak level of the SCR ramp generator (oscillator core) should be approximately 14.4v - 15v depending upon drop voltage and regulators, temperature and so on.

In order for the FET to saturate requires the positive offset generated by the buffer (opposite direction of a NPN buffer = drop voltage) to be in excess of the extra 2.5 to 3 volts provided to the buffer source. In other words accounting for all the issues the FET VGS must lead to an offset of at least 4 volts, saturating the buffer.



Beyond that effect which is the only major effect I'm aware that exists in any practical circuit, there is also the issue of the reset pulse and the way current flows to reset the ramp voltage after reaching its peak. In the korg MS- and various other oscillator circuits using an SCR relaxation oscillator core the reset is as close to instant as possible.

In other circuits such as the moog (as in minimoog) oscillators utilizing a schmitt trigger, the length of the pulse and the current flowing depends upon the hysteresis circuit as well as the active element under control by the reset pulse. At high frequencies (>5k) in some circuits (e.g. roland SH- series) a resistor acts to limit the reset current in higher ratio to the complete cycle length which helps to counter-balance non-linearity generated elsewhere in the exponential current source among other sources. This can make the current flow with an RC curve rather than linearly for the reset portion of the waveform.

Still, in all those cases of which I am aware none of the effect applies to the audible band <20k other than very small high frequency differences similar to for example a fraction of a decibel of high-cut, not at all audible.



Now, the important bit. The pulse circuit, sync circuits and response under modulation are not so simple. Sync circuits in many cases are, and while the effects of modulation due to longer sync pulses can change in software vs. electronic circuits these effects are rare. Primarily the waveshaping method used to generate the pulse and triangle waveforms makes up the most significant audible difference. At low frequencies for example the minimoog oscillator waveshapers will produce audibly muted (high-cut) pulse waves and significant leakage of the reset pulse in the triangle wave, adding additional high-frequency harmonics making it "brighter" than a pure triangle. A ramp waveform shaped into a triangle is always imperfect. A pure triangle can only be generated by a different sort of oscillator core, a triangular core rather than a ramp core.



Another significant contribution is interaction along the PSU rails between multiple oscillators. This can lead to oscillators in a modular system syncing without being connected when their frequencies get close to an integer fraction of one-another.

This is due to the sync pulse leading to a large variation in supply current drawn from the PSU. It can generate a pulse large enough to put any other oscillators on the same supply line in sync. The more out-of-phase the oscillators are the larger the pulse required to sync them. This effect is known as "soft sync".

Along the same line (literally) variations in current from other sources such as LFOs and envelopes can bleed into the oscillator cores and create frequency or sync threshold modulation. This can lead to higher frequencies becoming more "dirty" as they become more susceptible to interference from supply variations.

The most common case you'll see this effect is in a synthesizer with a poorly regulated supply, lacking bypass capacitors on opamps and other locations in which current is varied in combination with specific sorts of oscillator cores which are more susceptible to these variations in supply current. You'll often get 60hz (or 50hz) ripple and pink noise as the primary sources of noise on the supply lines.

These effects can be minimized by proper bypass of various points in the circuits. Both at the source of the variable current (to smooth the current variation itself) as well as via further filtering and regulation of supply lines to sensitive modules like oscillator cores and audio amplifiers.



These PSU related effects are less likely to be experienced in modern, well designed equipment. The modern sub-37 is nothing like the minimoog in this respect. The minimoog itself with old out-of-spec capacitors will likely have a far less regulated supply and poor filtering/bypassing of modules with either highly varied current demands or sensitivity to power supply variations.

[Mutant]

That's a lot of info. I'll have to read it few more times to understand it better.

[aciddose]

Do you have any analogs to play with? I suspect you can get the MS-20 to sync and get most of the other effects out of it with some careful tuning. (For example, throw it in the freezer for an hour )

For the more extreme stuff though you need to either do mods or build your own circuits. Of course going out to buy very specific synthesizers just to hear a particular effect is crazy, at least in my opinion with the prices where they are.

I remember the MS-10 (the 20 has the same circuit, and others) drove me nuts with the effect on the triangle wave. The middle part of the triangle flips around / wraps due to the reset pulse being mixed in. They attempt to filter it but it becomes obvious at very low frequencies like 60hz.

If you want something like a sine to back up the bass in a preset, you can forget about using the triangle in the minimoog, MS- or various other synthesizers. They all have this high frequency pulse at various levels and it is really a pain. Sometimes it is okay because the high frequency component from other waveforms (pulse, etc) are much higher anyway. In those cases where you want your sine at 0db and pulse at -12db though, the triangle isn't pure enough to provide anywhere close to the same timbre you'd expect.

It can be nice if you go at it more freely, "go with the flow". It does have the capability to sound awesome. Trying to match a signal generated with pure waveforms though you'll find it simply isn't possible on an analog.

Anyway yes, a lot of info, my post covers some chunk of the complexities of analog oscillators and waveshapers. People don't really separate the two although once you have a lot of experience with analogs and modulars you're forced to. A waveshaper circuit is really an isolated component. It often requires a ramp input with specific properties, but it is otherwise entirely independent from the oscillator core.

Some oscillator modules for modular synthesizers come with only a ramp output for example. You can then get various waveshaping modules separately to produce pulses, triangles, sines and other waveforms.

When people say these things online like "the oscillator from X sounds fat and analog!" you always have to remember "oscillator" is something very generic. They're probably referring to something very specific and very complex and it is simply way, way beyond their ability to express that in a technical way.

They may mean "the pulse waveform from X sounds fat and analog while set at exactly this width and played between C1 and C3 while driven at max mix level into the filter with minimal resonance".

[psmacmur]

The PAiA FatMan is a perfect example of an analog with soft sync, and a well-known mod I couldn't be bothered to install that adds the bypass capacitor you're talking about. Also, it sounds FAT and ANALOG *all the time*, maybe because its filter is only low pass? Well, depending what Fat means; maybe when it syncs, not so much? When sync'd, though, it "cuts through mixes" and stuff. pew pew pew

[Mutant]

Do you have any analogs to play with?

Do i really need them if i have Xhip and some parametric EQ ?

When people say these things online like "the oscillator from X sounds fat and analog!" you always have to remember "oscillator" is something very generic. They're probably referring to something very specific and very complex and it is simply way, way beyond their ability to express that in a technical way.

Yep.
Unless they say it while playing 1 note, 1 oscillator on some synth review video on Youtube.

Like for example here with Prophet 6 saw (but with a chord):
https://youtu.be/kGkM0GLMVh0?t=33
"That is a rich, full sounding saw tooth wave" and i agree.
Which btw. i downloaded the video, extracted the audio and matched few days ago too, just the effects werent as spectacular as with the Sub37, so i did not feel the need to post it here.

[aciddose]

The PAiA FatMan is a perfect example of an analog with soft sync, and a well-known mod I couldn't be bothered to install that adds the bypass capacitor you're talking about. Also, it sounds FAT and ANALOG *all the time*, maybe because its filter is only low pass? Well, depending what Fat means; maybe when it syncs, not so much? When sync'd, though, it "cuts through mixes" and stuff. pew pew pew

The effect I mentioned actually happens without any sync at all, it's rather sort of like an electronic "bug"

You can reproduce it almost perfectly using soft-sync if you have the correct sort of oscillator though.

In the schmitt trigger sort, you'd input the negative sync pulse to the reference level for the comparator. This lowers the level at which the oscillator's internal sync circuit kicks in. You can then control the amount by scaling the pulse.

The same works in the SCR-based oscillator cores. It requires some slight differences in the circuit and I'm not so certain it is possible in the MS- implementation at all. I use a variation of the circuit with a PNP buffer for the reference voltage though which makes the voltage accurate (no diode drop) and also allows you to input sync pulses.

[aciddose]

Unless they say it while playing 1 note, 1 oscillator on some synth review video Youtube.

Ah... This is one reason I cry myself to sleep at night wishing I wouldn't need to live on this planet anymore.

Reminds me of "this one goes to eleven" from spinal tap. There is no limit to the range of bullshit subjective nonsense musicians will produce to explain something when they don't understand it. Worst of all, these situations in which there is no difference to begin with and they invent nonsense to explain the difference they "hear".



Oh well... You know what they say, to each their own, whatever floats your boat and so on.

[xoxos]

achievement award, asset to the community.

[Mutant]

You forgot the sarcasm emote.

[chk071]

Who are "they"? Not that i don't think to myself pretty frequently that "they" are stupid too though. Makes me feel smarter. Heh.

[aciddose]

One really common effect in analog waveshapers is actually that the ramp input to the pulse shaper is high-pass filtered. This means the "square" is actually more like a 51% pulse due to the curve this creates on the ramp. The point it crosses zero is then slightly further ahead than usual, giving the zero volts (typically center) position on the pulse-width knob a little positive offset.

This is easy to cure. The synthesizer merely requires an offset to correct this with a trim-pot. Of course which sane designer would ever bother to spend fifty cents on such a thing? No one, that's who.

The consequence is that "square" is actually 51% or so, easy to achieve in software.

The first thing I do in software is set the pulse to 51%. It makes every filter sweep sound fatter. Why? ...because I'll tell you why : the even harmonics and their interaction with the filter oscillation, especially in non-linear filters.

The more difficult part is that the gain of the shaper is less than infinity. The gain is often high enough so as to make it needless to care, however in certain circuits (SH-) the waveshaper's gain is 20 to 50. This means low frequency ramp inputs will produce a pulse with a much softer edge on one side.

Easy in purpose-made software (the waveform edge is actually a ramp) however all but impossible to get right if not coded to achieve this in the first place.

[Mutant]

@Aciddose: Is it strange or is it not strange that i had to close the 24dB low pass filter on Xhip quite a lot (8.14 Khz) to match the spectrum curve of the Sub37 saw ?
I would assume that the OP at the DSI forum sampled his Moog with the filter fully open, so there has to be some other force at play to make it "mellow" like that, maybe some of the rare electronics "magic" you talked about earlier.

[aciddose]

Nope, for it to be filtered requires it be filtered.

A pure ramp waveform has harmonics at 1/N.

This means:

1. 0
2. -6
3. -9
4. -12
5. -14
6. -15.5
7. -17
8. -18
9. -19
10. -20

From there we have the usual well-known rule in decibels:

• 10^1 = -20
• 10^2 = -40
• 10^3 = -60

-20db each time you add a zero. 1, 10, 100, 1000, 10000.

You should see the amplitude of harmonics falls off very slowly.

At even 100hz, you only get 200 harmonics by 20khz. This is only -46db.

The 1/N fall-off is the fall-off of a discontinuity. Any infinitely fine impulse will fall at most at this rate. The impulse (edge) harmonics will fall off faster only once it is filtered.

It would be very unusual for the oscillator core to produce this effect at a frequency like 100hz. The consequence would be that at higher frequencies (~5k) oscillation would stop entirely. Remember that the minimum filter slope is 6.02db/octave, which is actually just rounded off. The slope is actually 20db/decade. Decade = adding a zero, 1 -> 10.