2Box Machine Gunning Question

[TangTheHump]

Kabonfaiba,

Thank you again. This is tremendously helpful! :-)

To my ears, the results with tom and hi-hat are very much the same as the snare test. As follows:

1.) Unlimited polyphony, round robin ON: best, most natural sounding
2.) No polyphony, round robin ON: good, but not as good as (1)
3.) Unlimited polyphony, round robin OFF: machine gunning is obvious
4.) No polyphony, round robin OFF: worst machine gunning of all

I'm wondering, in the case of test 2 (monophonic polyphony, round robin on), how 2Box handles this. I can see at least two scenarios (and perhaps there are more). First scenario is absolute monophonic implementation: each successive stroke truncates the decay of the previous stroke. Second scenario is duophonic implementation: to allow better audible round robin results, 1st stroke plays, 2nd stroke plays, every 3rd stroke truncates the oldest stroke, causing whatever remaining decay of every third eldest stroke to be truncated. Ultimately, my question is does the 2Box implement absolute monophonic presentation at all times when polyphony is reduced to one / zero?

Interestingly, I played these audio files for my significant other, who is neither a musician nor a detailed music listener. Without comment from me, she came to the same conclusions I did. More interestingly, I'll underline her immediate reaction when she first encountered test 3 and test 4, which was: "Ugh! Those last two sound absolutely terrible! The best sounds are from test 1." She came to this conclusion within milliseconds of hearing the tests; her reaction was immediate and emphatic. So much for the idea that casual listeners don't notice machine gunning!

[JmanWord]

The title of this thread should probably be changed to Machine Gun Question....... since the testing is just on machine gun theories in general, not the 2Box.

[Kabonfaiba]

Final results are in!

Rack Tom Polyphony Test

Hi Hat Polyphony Test

Same as before:

1) Unlimited polyphony, round robin ON.
2) No polyphony, round robin ON
3) Unlimited polyphony, round robin OFF.
4) No polyphony, round robin OFF.

The Hi-hat test exposes a new flaw; too much polyphony is a bad thing.

But why is that? We already know that VSTs dynamically control a lot of things beyond simply playing waveforms... well the test exposes why controlling the stacking of sound is just as important as letting the sound stack in the first place. There is indeed a polyphony sweet spot, as such.

[Kabonfaiba]

There's more to come guys, this is just my initial experiment to see if the presentation is correct, and additionally, if the experiment is worth continuing.

The snare is a Tama Tempesta from BFD3 from one of the processed presets, I collected 4 distinct samples of the "Half edge" articulation at 90% velocity.

I sliced the waveforms up (they are all 2 seconds long of decay - which is really very short) I manually layered these, and spaced them in my NLE, so basically there's no interference from BFD3 that could skew the results. Manual round robin is tricky, but I only had 4 samples and I didn't want to start over at that point.

I'm not surprised it sounded as bad as it did, I've heard the same effect in SD2 as well. Machine gunning with peak velocity is bound to happen, and secondly, there's no timing humanization going on here either; every hit is bang on time like an atomic clock. I gave it the worst possible test after all.

Maybe a gorilla of a drum player could replicate what I did on an Akit, it might sound surprising similar. I'm going to test an unprocessed floor tom next.

Btw I didn't notice the difference right away with lack of polyphony until I heard it without round robin, it's really obvious there - now I knew what to listen for, I go back to the first two and now hear the subtle difference it makes.

[Tommy_D]

Kabonfaiba,

1.) Unlimited polyphony, round robin ON: ignoring the one pop transients, this is the most natural sound
2.) No polyphony, round robin ON: amazingly, this still sounds pretty natural; not as good as (1) though
3.) Unlimited polyphony, round robin OFF: machine gunning is obvious
4.) No polyphony, round robin OFF: worst machine gunning of all examples

What if you took one of the stock 2Box floor tom sounds (one with a reasonable number of samples) and tried the same thing? This would eliminate the pops and save you from needing to do any sampling. With an instrument that has sustain, I'm guessing the result of limiting polyphony will be even more pronounced. At any rate, whether you have more time to contribute or not, your example seems to show both unlimited polyphony and round robin sample selection contribute to removing machine gunning. Thanks again for doing this! Really appreciated!

Agreed with your findings of the samples, though I don't feel that the first sample sounds good either. It could just be a bad sounding snare sample from the get go, but the pops and the limited round robin effect make it sound a bit wonky to my ears.

Is this being created on a keyboard, or some sort of computer program? Or is this actually being played on the kit with sticks and human hands?

[TangTheHump]

Kabonfaiba,

That's awesome! Thank you! I agree, a floor tom (or rack tom... something with some sustain) is probably a better choice to hear machine gunning. But, your snare example works pretty well. To my ear, both polyphony and round robin sample selection affect the outcome. Rating the examples, my ears hear:

1.) Unlimited polyphony, round robin ON: ignoring the one pop transients, this is the most natural sound
2.) No polyphony, round robin ON: amazingly, this still sounds pretty natural; not as good as (1) though
3.) Unlimited polyphony, round robin OFF: machine gunning is obvious
4.) No polyphony, round robin OFF: worst machine gunning of all examples

What if you took one of the stock 2Box floor tom sounds (one with a reasonable number of samples) and tried the same thing? This would eliminate the pops and save you from needing to do any sampling. With an instrument that has sustain, I'm guessing the result of limiting polyphony will be even more pronounced. At any rate, whether you have more time to contribute or not, your example seems to show both unlimited polyphony and round robin sample selection contribute to removing machine gunning. Thanks again for doing this! Really appreciated!

[Kabonfaiba]

It's been a long while but I finally found some time to do this!

I'm happy with the experiment, but not happy with my choice of snare or the amount of round robin samples. (I only captured 4)
Also, the hit velocity was close to the maximum, which means the snap is overkill and unrealistic given the speed at which the snare is being hit.
I think a floor tom would be a wiser choice next time, but anyhow better than nothing right?

MP3 files to download:

Snare 1 Polyphoney test (no compression)

Snare 1 Polyphoney test (with compression)

I thought it would be wise given how aggressive the snap (pop) is, that compression would help make the decay a bit louder and more noticeable during the test.

You will hear 4 tests of machine gunning. What I am doing to limit the polyphony (to monophony) is cutting the sample off when the next is triggered. Effectively not allowing the samples to stack.

In the order they are heard;

1) Unlimited polyphony, round robin ON.
2) No polyphony, round robin ON
3) Unlimited polyphony, round robin OFF.
4) No polyphony, round robin OFF.

[TangTheHump]

Kabonfaiba and Chris K,

Kabonfaiba wrote:
Tang, are you interested in a round robin-esque collection of samples? or testing with 1 sample over and over? Obviously the latter is going to machine gun like no tomorrow.

Both, actually - (1) a reasonable size round robin grouping such that one would think no machine gunning would occur, and (2) a single sample over and over. You'd think the latter would machine gun like crazy, but I've got a multitimbral Roland keyboard that allows one to allocate polyphony per part. So I allocate one note of polyphony to a part. Now, I set the part to a patch that has only a single sample layer throughout the entire velocity range (0 to 127). Guess what happens? I start to hear what sounds like machine gunning. But, it's not quite the same as with a drum pad due to the fact I can't repeat the notes fast enough on a keyboard controller. With a drum pad, playing double and triple stroke rolls creates a very rapid succession of notes.

So back to this monophonic, single sample keyboard setup... as I increase the polyphony allotment, the machine gunning goes away entirely. If I play so rapidly that all notes of polyphony overlap and thus successive notes are truncated, machine gunning returns. However, as long as there is enough polyphony to cover a rapid sequence of notes with all notes sounding to full duration, there is no machine gunning. Remember again, this is a monophonic, single sample keyboard setup and it's still possible to eliminate machine gunning by only adjusting polyphony.

Consider the old drum machines, like the Linn Drum. I had one. The thing machine gunned like crazy. And, it only allocated one note of polyphony per pad! All pads were, in essence, monophonic. This is another example that makes me think polyphony is the culprit. Now of course, we want no machine gunning and realistic dynamic response, but first I'm curious how to get rid of the machine gunning. Maybe, as I used to think and Chris still does, lack of velocity layers is a significant contributing factor, too. I honestly don't know anymore which is why I'm curious to test some assumptions and theories.

Kabonfaiba , if you're willing to do some testing with the equipment you've got, that would be great. Also, yes, if you've got the ability to show what happens to the waveforms as machine gunning occurs, that would be very interesting to analyze. I was hoping to find someone who had drum pad controllers to enable the repeating of notes at the speed a drummer typically plays Indeed, there's a world of difference between how a keyboard player and drummer play. Maybe someone else will join in with this test.

I wrote in another post that Roland could significantly improve the sound of their modules simply by eliminating machine gunning. I believe this to be true. Same for Yamaha. The actual modules sounds aren't that bad, but the machine gunning truly kills any sense of realism. Get rid of machine gunning and you're a big step up the ladder. Or, at least, this is another of my theories.

Thanks to both of you for replying and getting in on this.

[Chris K]

Angr77,



On the 2Box module, I believe there is a parameter that allows one to set a pad / instrument to monophonic (meaning it is allocated a single note of polyphony). I don't think one can do specific polyphony allotments per pad (such as assigning 6 notes of polyphony to a given instrument), but the monophonic setting does the trick for this test.

On the TD-30, there is no way to allocate polyphony. In fact, my hypothesis, based on what my hands and ears tell me, is Roland hard codes polyphony allotments based on the type of instrument. One would think the TD-30 would use all available polyphony and divide it up, dynamically, as needed, but this doesn't seem to be the case. I believe this is why certain instruments (such as cowbell) always machine gun, even when no other instruments are playing. It's because the TD-30 hard codes the polyphony of a cowbell at 1. Seemingly, the TD-30 doesn't have enough processing power and/or polyphony to allow all instruments fully dynamic polyphony. (i.e. Fully dynamic polyphony = all instruments draw as many simultaneous notes as they need from the total reservoir of polyphony.) Thus, Roland makes hard coded assignments based on where they think the polyphony is most needed, but this fails because one never knows what instruments a drummer will choose to play a given pattern.

A cowbell, due to its short duration, seems like an instrument that wouldn't need much polyphony. But, start playing fast patterns and voila. All of a sudden, the decays of each cowbell note start to overlap and with insufficient polyphony, the result is truncated notes, producing machine gunning. That's the theory I'm hoping someone will test with their 2Box module - cause the 2Box to machine gun by forcing limited polyphony. Note, I'm specifically not talking about number of samples and velocity layers. It's polyphony I want to test. E-drum folks (myself included) commonly think insufficient velocity layers is the main symptom causing machine gunning. However, I'm beginning to think this isn't the case. I hypothesize restricted polyphony is the root cause. Thus, I'd like someone interested to test this as currently I don't have the right equipment to do this testing myself. :-)

You can have any machine gunning with 1wav sample and 128 voice polyphony on any instrument categories, in the past Keyboard had 1 PCM wav accross the Keyboard, for Piano, acoustic guitar, bass or any instrument, which caused repetitive sound and lack of variation on strike and unnatural sound, and extremely "machine gunning or repetitive sound" with faster playing without reaching 128 notes polyphonic, since you can heard the problems with normal playing, with time they have added low wav sound layers to reduce the artifact on repetitive sound and added 16 multi trimbral on a keyboard range on hardware to variate the sound. Latly Korg Kronos have a 2Gig piano with multi-layers wav which sound very professional..

You can have machine gunning on snare and tom with lower fill tempo, it's more the same wav sound is repeated it self which caused this, you can take 128 voice polyphony and can get easily machine gun on a computer, until the sound wav or attack don't change, forget it, you will still heard the same sound repeated one after one which make machine gunning, the ears recognize it easily it's the same as, play a Film and put the 1.JPG repeated one after one, you wont like it because it's the same 1.JPG is playing, same for sound, play 1. wav each time, your ears will get bored, even if there is 3 wav lower velocity wav it will still not sound right.

Take Any keyboard or any others instrument it's the same problem until there is no variation on the sound on each strike, it's a no go...VST Keyboard, Piano, Guitar, Bass, Orchestra Symphonic, Choir, Sound, Strings etc. beat hand down any Keyboard out there most of them, on any companies, Roland, Korg, Yamaha, problem caused, lack of variation on the sound because of the Memory required, in any case polyphonic are never exceeded. Now Korg have partnered with VST Ivory Sound which have one the best piano out there.

VST have set the bar high vs Typical hardware instrument, unless the hardware can receive VST, they will still sound much better in details and quality.

[Kabonfaiba]

I'm getting intrigued by this too... I may be able to help you Tang.

I don't have 2box to try this with, but I think I can recreate this experiment with my video editing software in a way that would simulate polyphony. I could show the results visually as well as an audio file.

I need some time to do this properly, but I'll try it and post the results. (mp3 file and picture of waveforms)

Tang, are you interested in a round robin-esque collection of samples? or testing with 1 sample over and over? Obviously the latter is going to machine gun like no tomorrow.

[TangTheHump]

Angr77,

Angr77 wrote:
How can this polyphony approach be tested on a 2Box or a TD30?

On the 2Box module, I believe there is a parameter that allows one to set a pad / instrument to monophonic (meaning it is allocated a single note of polyphony). I don't think one can do specific polyphony allotments per pad (such as assigning 6 notes of polyphony to a given instrument), but the monophonic setting does the trick for this test.

On the TD-30, there is no way to allocate polyphony. In fact, my hypothesis, based on what my hands and ears tell me, is Roland hard codes polyphony allotments based on the type of instrument. One would think the TD-30 would use all available polyphony and divide it up, dynamically, as needed, but this doesn't seem to be the case. I believe this is why certain instruments (such as cowbell) always machine gun, even when no other instruments are playing. It's because the TD-30 hard codes the polyphony of a cowbell at 1. Seemingly, the TD-30 doesn't have enough processing power and/or polyphony to allow all instruments fully dynamic polyphony. (i.e. Fully dynamic polyphony = all instruments draw as many simultaneous notes as they need from the total reservoir of polyphony.) Thus, Roland makes hard coded assignments based on where they think the polyphony is most needed, but this fails because one never knows what instruments a drummer will choose to play a given pattern.

A cowbell, due to its short duration, seems like an instrument that wouldn't need much polyphony. But, start playing fast patterns and voila. All of a sudden, the decays of each cowbell note start to overlap and with insufficient polyphony, the result is truncated notes, producing machine gunning. That's the theory I'm hoping someone will test with their 2Box module - cause the 2Box to machine gun by forcing limited polyphony. Note, I'm specifically not talking about number of samples and velocity layers. It's polyphony I want to test. E-drum folks (myself included) commonly think insufficient velocity layers is the main symptom causing machine gunning. However, I'm beginning to think this isn't the case. I hypothesize restricted polyphony is the root cause. Thus, I'd like someone interested to test this as currently I don't have the right equipment to do this testing myself. :-)

[angr77]

How can this polyphony approach be tested on a 2Box or a TD30?

Angr77

[TangTheHump]

yzf125,

Thanks for the advice re contacting Jman. I'll give it a shot.

Anyone else here with a 2Box module interested enough to try the test?

[yzf125]

I would PM Jman as he is a very active mod over at the 2box forum http://www.2box-forum.com/index.php?board=1.0

[TangTheHump]

I'm wondering if a 2Box module owner would do me a favour. On the 2Box module, I believe one can alter the polyphony for a given instrument. I'm wondering if someone would be willing to set the polyphony for a given instrument to monophonic (or single voice), meaning each time that instrument is struck, the module always terminates the current sounding voice for that instrument and starts the voice again at whatever the new velocity level is. What I'd like to get is a recording of singles and doubles at different tempos so I can hear if this creates the type of machine gunning I hear in the TD-30 and other drum modules. This is purely an academic request, although the result might form part of a letter I send to Roland re suggested improvements for their drum modules.

I'm specifically curious about polyphony as opposed to velocity layers because, based on what I've seen in the TD-30, the TD-30 seems to hard code polyphony allotments based on instrument and not on how many actual voices of polyphony the module currently has available. Thus, certain sounds (like cowbell and certain toms) always machine gun as it seems their polyphony allotment is hard coded at 1. Other instruments (other tom voices, etc.) seem to have a higher polyphony allotment and thus these instruments still machine gun, but less so. And yet again, a scan few instruments in the TD-30 seem to have no polyphony restrictions other than the total polyphony capability of the module. Those instruments almost never machine gun.

So yeah, I'm wanting to test this theory with the 2Box module because, unlike the TD-30, I believe the 2Box module allows the user some control over polyphony.

My theory dates back to my use of synthesizers... units like the Fairlight CMI, Emmulator II, Emmulator III, Rhodes Chroma, Oberheim Matrix 12, and Kurzweil K series (K2500, K2600, etc.). These are all multitimbral synthesizers that give the user full control over polyphony allotments. When one allocates too little polyphony to a given sound, the result is very much like machine gunning. Indeed, one can hard set the velocity at a given level (so the same sound or sample layer always plays) and remove machine gunning simply by increasing polyphony allotment. So my theory, in regard to drum modules, is machine gunning is more a function of polyphony allotment than velocity layer restrictions.

Is there a 2Box owner out there willing to try this experiment and post the audio results here? Thanks in advance for everyone's time and consideration.