Bitmask or dithering in a patch?

Modularmind

I've spent some time the last couple of weeks trying to create replicas of two amazing "bitcrushers" - OTO Biscuit and WMD Geiger Counter. I'm almost finished with the Biscuit patch, but I'm stuck on the Geiger Counter. It has a "mask" mode, which is described like this in the manual:

"Bits/Mask. This switch controls how the Bit Depth knob works. In Bits
mode, the Bit Depth knob reduces the resolution of the signal. In Mask
mode, the signal is filtered through a number (0 to 255). This mode can be
used to reduce noise and add gain. In pre wave table mode, it can turn off
small ranges of the wave table, creating altered harmonic content from the
tables."

I don't know much about bitmasking. Is it a sort of dithering, and is that what is being described here? I'm using the ADC and DAC modules to convert the signal to 8-bits. Is there a way to "filter the signal through a number (0 to 255)" with the modules in the G2? What does "filter the signal thru a number" actually mean?

I'm really out of my depth here, since I know so little about these things. But I want to learn from someone who knows more, and I think it can make for a very nice distortion patch. Just running the signal thru a custom wavetable is nice in it's own right.

Any thoughts?

/Mm

speak_onion · Posted: Wed Aug 24, 2011 11:48 am Post subject:

I'm not sure, and that text doesn't specify exactly what's going on, but I think what the masking does is to force one or more bits to be either 0 or 1 (low or high), regardless of the input signal.

ian-s · Posted: Wed Aug 24, 2011 12:48 pm Post subject:

speak_onion is correct.

bit masking involves a logical AND operation. So if you mask an eight bit value with 0xF0, you will only get the most significant 4 bits (the lsbits will be zero). 0xFE would zero only the lsb.

You could patch it with a constant + ADC and 4 dual AND gates I think.

The term is common in assembly language programming.

Roland Kuit · Posted: Wed Aug 24, 2011 2:01 pm Post subject:

Funny, today I was toying with an osc and came up with this. The other patches where the result.

Roland Kuit Muxulator.jpg

Description:

Filesize:

82.45 KB

Viewed:

19364 Time(s)

Muxolator.pch2

Description:

Download (listen)

Filename:

Muxolator.pch2

Filesize:

1.2 KB

Downloaded:

4099 Time(s)

Roland Kuit · Posted: Wed Aug 24, 2011 2:02 pm Post subject:

audio result:

Roland Kuit Audio yo bit.jpg

Description:

Filesize:

166.77 KB

Viewed:

19362 Time(s)

Audio yo bit.pch2

Description:

Download (listen)

Filename:

Audio yo bit.pch2

Filesize:

2.18 KB

Downloaded:

4142 Time(s)

Roland Kuit · Posted: Wed Aug 24, 2011 2:10 pm Post subject:

and with sequencer, pattern gen:

Roland Kuit Audio-wavetable II.jpg

Description:

Filesize:

155.44 KB

Viewed:

19359 Time(s)

Roland Kuit Audio-Wavetable.jpg

Description:

Filesize:

142.6 KB

Viewed:

19320 Time(s)

Audio-wavetable.pch2

Description:

Download (listen)

Filename:

Audio-wavetable.pch2

Filesize:

2.1 KB

Downloaded:

4166 Time(s)

Audio-wavetable II.pch2

Description:

Download (listen)

Filename:

Audio-wavetable II.pch2

Filesize:

2.22 KB

Downloaded:

4233 Time(s)

Modularmind · Posted: Fri Aug 26, 2011 2:06 am Post subject:

@Roland: That last audio-wavetable patch looks similar to the one I use in my geiger counter patch. I've copied and modified a waveshaper patch by Fozzie (user here) where the signal is split into positive and negative parts and each part is treated by a dedicated wavetable. I haven't tried to see how much difference that makes.

@Ian: In my Biscuit patch you can turn off the bits one by one. If you turn them off from least significant to most significant it sounds like regular bit reduction (like with the digitizer module). If I understand you and speak_onion correctly instead of just turning them off I replace them with a constant value high or low, and that then is essentially bitmasking? So for instance if I want 7 bits, not regularly reduced but masked, I should replace the least significant bit with a constant value? How do I know if it should be high or low? Or have I got it all wrong?

I’m at work at the moment. I’ll try to post some pictures of example patches once I’m back home.

Thanks so much for every ones input!

/Mm

Roland Kuit · Posted: Fri Aug 26, 2011 2:47 pm Post subject:

@Modularmind, I ill look into it. The pat. gen. can be used as a kind of formant osc too.

Modularmind · Posted: Mon Aug 29, 2011 1:53 pm Post subject:

Ok, so this is a test patch where I've tried (without really knowing what I'm doing) to use the advice from Ian and speak_onion. Anyway, I've put a logic gate between all the bits from the ADC module and the DAC module. All gates are set to AND except for D7 which has it's gate set to OR (it could just as well have been bypassed I guess...). A constant module converted to binary through a ADC module is feeding the first input of every gate. The other input is fed by the sound input (converted to binary with an ADC). Reducing the value of the constant module reduces the bit resolution. It's sounds like the Digitizer but with more room for fine adjustment of the bit reduction. So if I've done this right bit masking is just one way of approaching bit reduction. How does that compare then to "regular" bit reduction (if there is such a thing)? Does bit masking turn off the bits following a pattern, rather than turning them off from low to high (D0, then D1, then D2...)? This is a mystery to me...

A strange thing I found when I compared this way of masking the bits with the bit reduction of the Digitizer module is that the ADC to DAC modules with all bits going through sounds just like the Digitizer module when it's set to Bits: 10 (not 8 as I would've expected). The Digitizer is much noisier at 8 bits. Another thing is that the headroom seems higher in the Digitizer module. The ADC to DAC clips earlier with the same input volume. Wonder what this all means??

This feels like being on an adventure.... Cool

Input, please!!

/Mm

Bitmask vs Digitizer.jpg

Description:

Filesize:

56.52 KB

Viewed:

499 Time(s)

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

Bitmask vs Digitzer.pch2

Description:

Download (listen)

Filename:

Bitmask vs Digitzer.pch2

Filesize:

1.39 KB

Downloaded:

3900 Time(s)

speak_onion · Posted: Tue Aug 30, 2011 1:00 pm Post subject:

Doesn't the biscuit offer a dedicated button for each bit, when doing the bit masking? That might be a more flexible way to implement the masking than a knob. That way, you could turn off a more significant bit, while allowing a less significant bit to continue operating normally.

grimley · Posted: Tue Aug 30, 2011 6:11 pm Post subject:

Modularmind · Posted: Wed Aug 31, 2011 12:53 am Post subject:

Modularmind · Posted: Fri Nov 11, 2011 3:19 am Post subject:

My quest to understand bit masking and how that relates to "regular" bit reduction (like with the digitizer module) and also the concept of dithering is not completed. But some of it has become a bit(!) clearer. First of all I have realized that dithering is a concept very different from bit masking and bit reduction. From what I've gathered (I found a really good PDF from Ozone on this topic on the net) dithering is used as a technique to hide or mask quantizations noise/errors that occurs when reducing bit resolution. Bit masking on the other hand seems to be a way of reducing the resolution in finer increments than with regular bit truncation/redcution. The result is a smoother bit reduction.

I tried my best to replicate the Geiger Counter, but haven't been pleased with the result. I'm not pleased with the Bicuit patch either. But when I took the best of both patches and put them together and made some adjustments, the result was much more satisfying. I've posted the patch here: http://electro-music.com/forum/post-344557.html#344557

I want to say thank you to everyone that helped out with this!

/Mm