Announcement

Collapse
No announcement yet.

Official Users Topic: Roland TD-50

Collapse
This is a sticky topic.
X
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Hey Peter,

    Originally posted by Peter Warren View Post

    I quoted Cobalt sky but it is actually your post. Where did you get this information? This is incorrect for the VST's that I use.
    you said so?

    Originally posted by Peter Warren View Post
    My understanding having been a beta tester for some VST companies and opened up sample library wav files, is that most use 6 or 7 velocity levels
    and feature space analysis clearly supports my point, I think: https://www.vdrums.com/forum/general...24#post1153324

    Furthermore I doubt that there is any human out there who can reproducible strike at 128 velocity levels.

    kr

    Timm

    Edit: To have that clear, I´m not speaking of the number of actual files on harddisk. With sample I mean a sample file produced by sampling an actual drummer.
    Last edited by DrStackVanHay; 06-19-17, 12:25 PM.

    Comment


    • Originally posted by DrStackVanHay View Post
      Hey Peter,



      you said so?



      and feature space analysis clearly supports my point, I think: https://www.vdrums.com/forum/general...24#post1153324

      Furthermore I doubt that there is any human out there who can reproducible strike at 128 velocity levels.

      kr

      Timm

      Edit: To have that clear, I´m not speaking of the number of actual files on harddisk. With sample I mean a sample file produced by sampling an actual drummer.
      My understanding from being a beta tester is that they are all individual hits. I have opened a file of hi hat samples and seen 1.7 gig for a single hi hat. I can easily ask to confirm. Some libraries may have used other methods.

      Comment


      • Originally posted by DrStackVanHay View Post
        Hey Peter,



        you said so?



        and feature space analysis clearly supports my point, I think: https://www.vdrums.com/forum/general...24#post1153324

        Furthermore I doubt that there is any human out there who can reproducible strike at 128 velocity levels.

        kr

        Timm

        Edit: To have that clear, I´m not speaking of the number of actual files on harddisk. With sample I mean a sample file produced by sampling an actual drummer.
        I did hear back from a developer and he said volume changes and envelopes only. He was not aware of any other company doing anything other than using a variety of samples. He pointed out that any changes including pitch degrade the sound of the sample. I know that many companies use lossless data compression to speed up loading and reduce the strain on the CPU.

        Keep in mind that most sample companies record multiple hits at 6 to 7 velocity levels so there is a pool of variation at each volume. That is why buzz rolls sound so good on VST's.

        As Alan said, a combination of both methods [Roland and VST] would be ideal. I am sure that will be the case eventually.

        Comment


        • Originally posted by Peter Warren View Post

          I did hear back from a developer and he said volume changes and envelopes only. He was not aware of any other company doing anything other than using a variety of samples. He pointed out that any changes including pitch degrade the sound of the sample. I know that many companies use lossless data compression to speed up loading and reduce the strain on the CPU.

          Keep in mind that most sample companies record multiple hits at 6 to 7 velocity levels so there is a pool of variation at each volume. That is why buzz rolls sound so good on VST's.

          As Alan said, a combination of both methods [Roland and VST] would be ideal. I am sure that will be the case eventually.
          Hey Peter,

          sorry, please could you elaborate for me: You say multiple hits at 6 to 7 velocity levels what exactly does that lead to?

          multiple (lets say 5) * 7 * 127 = 4445 recorded samples for snare center hit?

          kr

          Timm

          Comment


          • Originally posted by DrStackVanHay View Post

            Hey Peter,

            sorry, please could you elaborate for me: You say multiple hits at 6 to 7 velocity levels what exactly does that lead to?

            multiple (lets say 5) * 7 * 127 = 4445 recorded samples for snare center hit?

            kr

            Timm
            For example,120 to 127 may be very loud hit,110 to 119 the next level down,etc. I am not a programmer but my understanding is that certain number ranges use the same volume samples. I have found looking at midi data that the first 50% of volume range does not get used as much as the 2nd.

            BFD used to be an exception in that they just had samples lined up from quiet to loud. The problem was that when you hit 127 you were playing the same sample . They have some type of round robin now in BFD3.

            Comment


            • Originally posted by Peter Warren View Post
              ...Where I think the TD50 needs improvement is in the memory size. I can hear very clear machine gunning on the toms...
              Interesting, because the only way I'm able to get full-on machine gunning effect on the TD-50 is by playing outside the sensitivity range set for a given pad/trigger. But that would be operator error rather than the TD-50 causing the scatter gun blast. ; )

              As I noted elsewhere, you can get something that sounds close to machine gunning on acoustic drums if you practice playing that way (as some metal and hardcore drummers make a point of doing, in fact). If you play that way (within the set trigger sensitivity range) you get a similar result on the '50. But otherwise, I have found it to be a non-issue.

              As with the Ascent track posted by Bruce a page or so back, in fact.
              TD-50KV extended kit with KD-A22 kick, DW pedals/stands. SPD-30 and SPD-SX. TD-30 for additional triggers & layering. Muse Receptor 2+ Pro w/ SD v2.4.4.

              Comment


              • Originally posted by Peter Warren View Post
                For example,120 to 127 may be very loud hit,110 to 119 the next level down,etc. I am not a programmer but my understanding is that certain number ranges use the same volume samples. I have found looking at midi data that the first 50% of volume range does not get used as much as the 2nd.
                I believe that's a fairly accurate description of the 'calculated siblings approach that Timm referred to earlier. A smaller number of hits are used (scaled) to generate the volume increments.

                Otherwise the only way you can make it work with only 650 or so samples for each drum is to have little or no positional variation (every hit in exactly the same place) or have no round robin (only one hit at each position for each level). To generate a separate wav for a round robin of each zone hit at each velocity you need about 5,000 wav files per instrument (and for long decay instruments the wavs can easily be 20-30s each).

                Originally posted by Peter Warren View Post
                BFD used to be an exception in that they just had samples lined up from quiet to loud. The problem was that when you hit 127 you were playing the same sample . They have some type of round robin now in BFD3.
                Again, if repeated hits are registering 127, then that generally means that the player is playing outside the trigger sensitivity range, and has not been calibrated appropriately. I don't really think we should blame either a VST or a module (though I know people often do) for that. ; )
                Last edited by CobaltSky; 06-19-17, 06:01 PM.
                TD-50KV extended kit with KD-A22 kick, DW pedals/stands. SPD-30 and SPD-SX. TD-30 for additional triggers & layering. Muse Receptor 2+ Pro w/ SD v2.4.4.

                Comment


                • Originally posted by CobaltSky View Post

                  I believe that's a fairly accurate description of the 'calculated siblings approach that Timm referred to earlier. A smaller number of hits are used (scaled) to generate the volume increments.

                  Otherwise the only way you can make it work with only 650 or so samples for each drum is to have little or no positional variation (every hit in exactly the same place) or have no round robin (only one hit at each position for each level). To generate a separate wav for a round robin of each zone hit at each velocity you need about 5,000 wav files per instrument (and for long decay instruments the wavs can easily be 20-30s each).



                  Again, if repeated hits are registering 127, then that generally means that the player is playing outside the trigger sensitivity range, and has not been calibrated appropriately. I don't really think we should blame either a VST or a module (though I know people often do) for that. ; )
                  The developer varified that they record every hit. Every drum is different,but they could have 6 volume levels with 20 hits at each level on a snare. They may have less on the rimshot. Then there are 6 mic channels so it is a lot of files.

                  The Jarrah 16 tom in the TD50 machine guns in my opinion. Try making a recording of it with all effects off,especially the transient designer.

                  Machine gunning is usually not a problem with 1/8 notes and quarter notes. It becomes much more obvious with 16th notes.

                  Comment


                  • Originally posted by Peter Warren View Post
                    The developer varified that they record every hit. Every drum is different,but they could have 6 volume levels with 20 hits at each level on a snare. They may have less on the rimshot. Then there are 6 mic channels so it is a lot of files.
                    That would be a minimum of 91,440 files per drum/cym. At an optimistic average of 250KB per sample, that would be getting up to around 350GB of raw data per kit. However even on compressed install format, most of the VST libraries I've encountered are less than 1/10th of that size (and most of the EZX libs, for instance, are 1GB or less).

                    Originally posted by Peter Warren View Post
                    The Jarrah 16 tom in the TD50 machine guns in my opinion. Try making a recording of it with all effects off,especially the transient designer.
                    Interesting. I'll check that out. To date, as I said, most probs I've seen seem to be caused by sub-optimal trigger settings. But maybe there's an issue with that particular instrument on the TD-50.

                    Originally posted by Peter Warren View Post
                    Machine gunning is usually not a problem with 1/8 notes and quarter notes. It becomes much more obvious with 16th notes.
                    Agreed. But it's also player, style and technique dependent IMO.
                    Last edited by CobaltSky; 06-19-17, 06:58 PM.
                    TD-50KV extended kit with KD-A22 kick, DW pedals/stands. SPD-30 and SPD-SX. TD-30 for additional triggers & layering. Muse Receptor 2+ Pro w/ SD v2.4.4.

                    Comment


                    • Originally posted by CobaltSky View Post

                      That would be a minimum of 91,440 files per drum/cym. At an optimistic average of 250KB per sample, that would be getting up to around 350GB of raw data per kit. However even on compressed install format, most of the VST libraries I've encountered are less than 1/10th of that size (and most of the EZX libs, for instance, are 1GB or less).



                      Interesting. I'll check that out. To date, as I said, most probs I've seen seem to be caused by sub-optimal trigger settings. But maybe there's an issue with that particular instrument on the TD-50.



                      Agreed. But it's also player, style and technique dependent IMO.
                      Where are you coming up with those numbers?
                      Snare:
                      Head 120 samples[6 velocity level 20 hits per level]
                      Rim 60 samples
                      rim click/cross stick 40 samples
                      x 6 mics =1320 wav files

                      On average,I have seen 100 samples for snare head and 50 for toms.

                      Comment


                      • Originally posted by Peter Warren View Post

                        Where are you coming up with those numbers?
                        Snare:
                        Head 120 samples[6 velocity level 20 hits per level]
                        Rim 60 samples
                        rim click/cross stick 40 samples
                        x 6 mics =1320 wav files

                        On average,I have seen 100 samples for snare head and 50 for toms.
                        Yes, that's about what I see as well. Which lines up precisely with the point Timm was making:

                        Originally posted by DrStackVanHay View Post
                        ...There are e.g 127 velocity samples available but only a few are real samples, the others are calculated siblings of those...
                        The calculated siblings are produced to generate 127 velocity levels from a much smaller number of actual recorded strokes, and that was a point you were disputing:

                        Originally posted by Peter Warren View Post
                        Where are you getting this information about VSTs? It is incorrect for all the ones that I have. Toontrack,FXpansion and Slate record every sample. I have had beta versions from FXpansion where I can open the folder and play every wav file.
                        In order to not be relying on calculated siblings, yet have varied hits - let's say strikes at 6 different positions (including rim) with 20 multiples of each strike each recorded through six mics at every one of 127 levels:

                        127 x 6 x 20 x 6 = 91,440

                        and regarding the accumulation of sample file sizes:

                        91,440 samples per instrument X 250 KB average sample size = 22.86 GB per instrument (and for a whole kit, 22.86 GB x 14 instruments =~ 320 GB). That would be a conservative estimate of the ballpark library size to accommodate a different/original recording for every sample, which is not what we see (with any current VST I've encountered to date).

                        The difference is illustrated where in your calculations you are allowing for 6 velocity levels instead of 127. That's about right and matches what I am seeing also. And it matches what Timm said about calculated siblings, where level and timbral changes (and/or blending) are applied to candidates from the original set of recorded hits at each of 6 levels (approx. - the actual number varies from one VST to the next) to generate the 'illusion' of 127 levels, where only a fraction of the 127 stroke levels for each position of each instrument represents an actual/original recording (the other 121 or so are faked/simulated/modelled).
                        Last edited by CobaltSky; 06-20-17, 01:01 AM.
                        TD-50KV extended kit with KD-A22 kick, DW pedals/stands. SPD-30 and SPD-SX. TD-30 for additional triggers & layering. Muse Receptor 2+ Pro w/ SD v2.4.4.

                        Comment


                        • Originally posted by CobaltSky View Post

                          Yes, that's about what I see as well. Which lines up precisely with the point Timm was making:



                          The calculated siblings are produced to generate 127 velocity levels from a much smaller number of actual recorded strokes, and that was a point you were disputing:



                          In order to not be relying on calculated siblings, yet have varied hits - let's say strikes at 6 different positions (including rim) with 20 multiples of each strike each recorded through six mics at every one of 127 levels:

                          127 x 6 x 20 x 6 = 91,440

                          and regarding the accumulation of sample file sizes:

                          91,440 samples per instrument X 250 KB average sample size = 22.86 GB per instrument (and for a whole kit, 22.86 GB x 14 instruments =~ 320 GB). That would be a conservative estimate of the ballpark library size to accommodate a different/original recording for every sample, which is not what we see (with any current VST I've encountered to date).

                          The difference is illustrated where in your calculations you are allowing for 6 velocity levels instead of 127. That's about right and matches what I am seeing also. And it matches what Timm said about calculated siblings, where level and timbral changes (and/or blending) are applied to candidates from the original set of recorded hits at each of 6 levels (approx. - the actual number varies from one VST to the next) to generate the 'illusion' of 127 levels, where only a fraction of the 127 stroke levels for each position of each instrument represents an actual/original recording (the other 121 or so are faked/simulated/modelled).
                          This is incorrect. I don't know why this is being repeated. Vst's only do head,edge,rim and cross stick at the most. They choose ranges to place the samples in. There are no fake samples. I asked a developer. I have seen folders before they are copy protected from Fxpansion as well.
                          Out of 127,it is divided up into 4 to 9 volume ranges. Each range has a round robin pool of samples. Note that those samples can vary in volume when played even if they are recorded at the same approx volume.
                          The Mimic thread has an actual programmer that would maybe explain this.

                          Comment


                          • Originally posted by Peter Warren View Post

                            This is incorrect. I don't know why this is being repeated. Vst's only do head,edge,rim and cross stick at the most. They choose ranges to place the samples in. There are no fake samples. I asked a developer. I have seen folders before they are copy protected from Fxpansion as well.
                            Out of 127,it is divided up into 4 to 9 volume ranges. Each range has a round robin pool of samples. Note that those samples can vary in volume when played even if they are recorded at the same approx volume.
                            The Mimic thread has an actual programmer that would maybe explain this.
                            There's obviously a communication disconnect here. You keep stating "this is incorrect" but then corroborating it with the facts and details you yourself cite.

                            While it may be true that there are no fake samples *on disk* (depending on the VST), out of the 127 velocity levels that are reproduced in response to a hit, a majority of them are imputed (interpolated/calculated from 'adjacent' siblings).

                            As you've said, "out of 127, it is divided up into 4 to 9 volume ranges". Which means that a pool of samples for a given velocity level are being played back in modified form to cover up to 20 or more levels of the 127-step velocity range of the MIDI data stream. Which is exactly what I understood to be Timm's original point, and that you have been disputing.
                            TD-50KV extended kit with KD-A22 kick, DW pedals/stands. SPD-30 and SPD-SX. TD-30 for additional triggers & layering. Muse Receptor 2+ Pro w/ SD v2.4.4.

                            Comment


                            • Hey,

                              Originally posted by CobaltSky View Post
                              There's obviously a communication disconnect here. [...]
                              As you've said, "out of 127, it is divided up into 4 to 9 volume ranges". Which means that a pool of samples for a given velocity level are being played back in modified form to cover up to 20 or more levels of the 127-step velocity range of the MIDI data stream. Which is exactly what I understood to be Timm's original point, and that you have been disputing.
                              Yes indeed!

                              Superior Drummer (e.g.) covers all 127 velocity levels. If it is using only a few 4,5,6, or 9 or whatever actual recorded samples to do so, which Peter now repeatedly confirmed and which can clearly been seen in my analysis posted some time ago, there are calculations done on those „fundamental" samples to do so. Thats simple, plain and obvious.

                              kr

                              Timm

                              Comment


                              • Hello there;

                                I've tried to read whole 64 page before I ask this question, but as far as I see none of you've asked it or replied (:
                                I want to know if there is a way to copy the kit's instrument to another kit's instrument. I mean, for example; I edited my ride cymbal on Kit#1. But i want to copy the exact same ride to the kit#2 without changing any other of its pads.


                                Many thanks for your assistance.

                                Comment

                                Working...
                                X