Audio examples of 'artifacts' or 'zippering'?

[RL]

In other external gears you have not artifacts exclusively because the manufacturer take care of this aspect and introduce in hardware or in software a smoothing circuit able to render continuous the behavior of instruments.

How can you know that? I made a check on a Roland XP50 and there are big steps when you using MIDI pitch bend changes.
To avoid steps on wide pitch bend ranges (two octaves) you need a glide hardware circuit or a special CPU - it's not so easy to implement the special algorithm in an 8bit micro controller.
RL

[gianluca13_2]

Hi Rudi,

You ready know my thought. I don’t’ discuss about the difficulty of such algorithm. Although I don’t think it is so difficult to implement also in an 8 bit microcontroller. But the problem is another. If you simply look at the number of post in the forum related to this aspect you agree a solution must be found and quickly. Want you make a test? Simply try to record a phrase with pitch wheel modulation movement in a P.C. based MIDI sequencer directly from Voyager. No external master keyboard. When you play at record time the Voyager respond very smooth because the pitch wheel information come directly by the inside system. When you play back your trace in the MIDI In of Voyager you can hear clearly the stepping of frequency shift. I never heard similar difference with other Synth. Also in little synth very less expensive. Sure Rudi here we have a problem and the problem must be solved in same manner.

P.S. Did you seen in a recent thread some other people lamented the problem of zero return of pitch wheel leaving out of tune the Voyager. Do you remember? I signaled the midi message do not return correctly to zero position? Some one suggest to him to recalibrate the hardware. How many time hi will be obligate to repeat this operation in future? A potentiometer have normally tolerance of 10% 20% and worst in linearity. You can’t be confident in a mechanic positioning of zero return. Here also you must think a software solution. Normally modulation wheel are translated in midi position via data table with dead zone to avoid unreliable end stroke and zero position. Did you take care of this aspect?



Cheers,

Gianluca

[Jazzpunk]

http://danling.com/HotLinks/misc/RME%26Triton.wav

What you hear are a minor third (3 semitones) glissando (using glide/portamento) followed by a pitchbend of the same frequency. Then both the glide and bend are repeated an octave higher. The problem seems to be only with the RME (or a Voyager controlled with a different controller).

Thanks for taking the time to post that audio example.

[Gonga]

I hope someday all this will lead to a fix.

[alisa 1387]

To avoid steps on wide pitch bend ranges (two octaves) you need a glide hardware circuit or a special CPU - it's not so easy to implement the special algorithm in an 8bit micro controller.

Dear Rudi

A simple difference-based algorithm could do the job fine - you even wont need any differential calculus. So what the f is so SPECIAL about this algorithm? (and why not just putting a condenser in the CV path - it would smooth the CV as well)

[alisa 1387]

We really need a fix. This is not a special feature, it´s absolutely essential.

[gianluca13_2]

Hi Alisa,

You don’t know how frustrating I feel me. I spent 3000 euros for bay an instruments that is in my dreams from a lot of years. And now I discover it is affected by a lot of little bugs. And many of them will require no more than one hour of my everyday job to be fixed. I make a lot of effort to signal to Rudi Linhard the nature and behavior of this bugs for have back nothing or worst an answer like this:

“No, there are hundreds of professional musicans without problems.
I can't reproduce your problem. Best Regards, Rudi”

this is the answer at my question:

“Hi Rudi, Have you some news related to pitch wheel artifact problems? The impossibility to drive my Voyger via a MIDI sequencer cause me some problem. Practically I can’t record midi traces for Voyager but only audio traces and this is a big limitation. Cheers, Gianluca”

I think a Minimoog is from ever a machine for solos where the expressivities of pitch modulation is one of the most important and fascinating aspect. Now seems to me this feature should be implemented in a perfect way specially related to MIDI possibility of Voyager.
Comparing the Minimoog to an XP50 and say:

“I made a check on a Roland XP50 and there are big steps when you using MIDI pitch bend changes”

seems to me a surreal and incredible answer. Demonstrating how superficially unappreciated are some aspect of the Minimoog. Considering this answer come from a person, a specialist of Lintronics leave me totally staggered.

But be careful the stepping and zippering on pitch wheel is not the only problem related to this device. Try to enter the MIDI data to a MIDI monitor like MIDIOX that show the entire structure of MIDI messages. You discover that when the pitch wheel return to zero from low side or up side never reach the zero message 0x4000 but always 0x4020, 0x4013 from up side or 0x3F60, 0x3F50 from low side. And the result is with a large amount of pitch wheel modulation an hearable out of tune. Seem to you something of acceptable in an instrument 3000 euros expensive?

From my point of view it’ a shame.

You are right:

“We really need a fix. This is not a special feature, it´s absolutely essential”

Cheers,

Gianluca

[RL]

Hi,
don't worry - I'm working on a fix.

@gianluca13_2:
please do NOT make private emails public.

[till]

... zero message 0x4000 but always 0x4020 ...

First a little calculation to know how this effects the pitch:

(So you use hex values based on 16 bit. In fact Midi uses 14 bit wide pitchbend)

The pitchbend zero return is off by less then 0.2% here (0x4000 / 0x4020 * 100)
So if the pitchbend range is about a fifth (=700 cents), then the offset is 700*0.2/100 = 1.4 cent.
For an A = 440 Hz this is a offset of 0.366 Hz.
And for a bass note A at 110 Hz only a 1/4 of it = 0.0915 Hz. You need good ears to hear this off pitch when a second oscillator is also active in the sound.
I got an offset with a max of 0x0020 here on my Voyager. And it wasn't yet recalibrated in its 4 years I own it. I only readjusted the pitchbend center when the OS 2.something was released and asked to do so.
So I assume this tolerance as typical.

Well I think we should shoot all guitar players and strings. And we should get a DSP or at least a DCO Voyager

Some dead zone would improve the situation for sure. But also would change the analog feeling a little bit.

...I spent 3000 euros for bay an instruments that is in my dreams from a lot of years. And now I discover it is affected by a lot of little bugs. ...

There is no perfect synth at all. Even in this price range. And also the Minimoog got some offset in this range due to the hardware pitchbend centering. And nobody I know ever complained about it on a normal working Minimoog. And there is no dead center on the Minimoog pitchwheel without a special update as far as I know.

[gianluca13_2]

Thank you for answer Till


OK there is a little mistake in my indication. The problem come from a correct representation of converted values of MIDIOX monitor. Really I have two byte = 0x40, 0x20 to represent. But in the philosophy of MIDI protocol (each byte 7 bit) this mind to a bin value = 100 0000 010 0000 = 0x2020. Ok? This must be compared to the ideal center position 0x2000 Ok? We have now a 0.389% of error. I correctly signaled this problem come out with large amount of pitch wheel Ok?
Now try with two octave of amount you have 2400 cent and then 2400*0.389/100 = 9.3%. Seems to me not necessary calculate the meaning in Hz of this difference because ours ear appreciate mostly percentage variation but if you like at 440 Hz this is 40.92 Hz more. And this is for a movement of one side (center - up – center) if you consider a double movement (center – up center – down – center) you must double this error. Obviously this is true and more evident with large amount of pitch wheel modulation. But be careful a solution of this problem is from my point of view something so easy to do that don’t merit the time we spend for debate. I’m a electronic engineer I make hardware and software from a lot of years and some time I produce devices for my music so I know very well what we are talking about. Sure Till there is nothing of difficult to create a translation table with dead zone and nothing change in the analog feeling of Voyager. His analog feeling come from a sound unique and not from some errors in the software. I’m not surprise by the fact so few person signaled this problem. I think a lot of persons use the Voyager without MIDI connection or with MIDI used just for a program changing. I quite sure because I’m the only person to signal an other bug very evident related to pitch wheel still present in version 3.3 and solved in 3.4. If you use a the same time the pitch wheel of master keyboard and the pitch wheel of Voyager after a movement in both side the pitch wheel in de Voyager stop completely to work. And this is very evident but not one signaled this problem. But now this problem is fixed.

Cheers,

Gianluca

[till]

...
So if the pitchbend range is about a fifth (=700 cents), then the offset is 700*0.2/100 = 1.4 cent.
For an A = 440 Hz this is a offset of 0.366 Hz
...

...
Now try with two octave of amount you have 2400 cent and then 2400*0.389/100 = 9.3% ...

See the error in your calculation:
the result is not percent but (musical) cent.
Just write down your calculation with the units after the numbers (as we were trained in the early school days)
So the result is 9.3 cent
One cent = 1/100 of a small second interval like from note A to A sharp.

Therefor the error in pitch in your 2 octave pitchbend example is not 40.92 Hz !!!
(this would be about a whole step and would be noticed by everybody not being deaf)

100 cent = frequency shift by the factor of the (12th root of 2) = about 1.05946
(the 12th root of 2 is the formular for a small second interval in our well tempered scale in western music.
The 12th root of (2 by the power of 12) is a whole octave = factor 2)

440 Hz (note A) * 1.05946 (factor for 100 cent) = 466.16376 Hz (note A sharp)
1 cent (in this interval) = (466.16376 Hz - 440 Hz) / 100 = 0.26164 Hz

And so your example's 9.3 cent offset equals 0.26164 Hz * 9.3 = 2.43325 Hz (and not 40.92 Hz !!!)

And this is not so much. Of cause you can here a 2.43325 Hz beating to a reference frequency without the offset. But you will like not notice this in a multiple track recording at all (unless you are a really trained musician).

... I’m a electronic engineer I make hardware and software from a lot of years ...

And I am just a master carpenter, but I know the difference of percent and musical cents

Sorry to all here that get borred by those calculations

[gianluca13_2]

Ok Till you are right,

I make a sacred error but the problem remain. I don’t know if are my hear to much trained or not but the out of tune it’s hearable. I whose in studio some evening ago and in a precise moment I must renounce to record the Minimoog in MIDI way recording it directly in audio to avoid this problem.
I whose very high in scale and probably this make more evident the problem but this whose the situation.
But regardless of subjectivity of detuning perception why the machine must send an imperfect message? This is incomprehensible for me. It’s so easy to avoid this problem from software point why we must accept a similar situation?

In any case thank you for the lesson and my compliments you are a master carpenter with an amazing mathematical and musical preparation.
I would have a lot of carpenters like you in my factory.

Cheers,

Gianluca

[Carey M]

***edited***read the thread properly, listened to the audio example***edited***

This zippering/stepping via MIDI pitch bend is a bit of a bummer... I hated the way Sunsyn stepped when using pitch bend. Not even my MKS-50 or Matrix-1000 exhibit any stepping (eighties analogue modules), regardless of the controller used. Of course, one way around this could be using a simple Midi->CV device, that smoothly converts MIDI pitch bend data to CV... A bit of an bass-ackwards solution though...

- CM

[Amos]

gianluca, before your calculations did you confirm that the pitch wheel is calibrated to return to zero? You can check this from the Master Menu; select Software Version, press Cursor and press Enter. This is assuming you use a keyboard Voyager and not the rack.

I'll suggest to add a small first-order lowpass filter to the pitch-bend CV, after the digital to analog conversion... perhaps this would help with the smoothness of incoming pitch-bend CC messages. If they are sent at full bit depth and at a fast rate... if not these conditions to begin with, you won't get smooth results anyway.

[till]

A full 14 bit pitch bender move from to to bottom would need:

- 2 by the power of 14 (number of bits) = 16384 messages for full pitch bend range

- and each pitch bend message is 3 byte
1 status byte, 1 byte for the least significant 7 bits, 1 byte for the most significant 7 bits

- 16384 different values * 3 byte * 8 bits per byte = 393216 bit

- Midi speed is 31250 baud = 31250 bits per second

- so a full sweep at full 14 bit would take 393216/31250 = 12.583 sec
(but only if there is not a single other midi data transmitted at the same time. And no Midi clock)

So data reduction is need. Or play the pitch bender reeeelllllyyyy sssssllllooooowwwww.

And a real hardware pot used for pitch bending on an analogue synth is also not perfect smooth.