ForumsHardware ← Soundplane - ghost notes/incorrect tracking

Does the soundplane suffer issues during the summer with warmer temperatures?
(Ive got about 29C during the day in my studio, though no direct sun)

I ask, as I play my soundplane daily, and its become noticeable over the past few weeks, its tracking seems to have really deteriorated... (and Ive spent an unhealthy amount of time trying to recalibrate/tune parameters to sort it out :( )

Ive always got some ghosts notes, and dubious track arrange the edges, but now even row 2 and 4 and the 3 outer columns seem to all be suffering... either by the tracking being inaccurate in the X/Y direction by nearly half a cell, or introducing additional touches.

I perhaps notice it more as I play in fourths, so the Y being out can really be heard when is modulating things and sometimes its bad enough to think its in the next cell, so triggers the wrong note ... I guess I'm also potentially notice the X tracking more, as I am playing un-quantized.

(of course, Im only talking about issue which are for touches > 1" apart, as per current spec)

Ive tried both the official release of the soundplane software, and the current development branch, both behave the same :(

Is there any hope of seeing a new version of the soundplane software soon, that might improved the calibration routine, or the tracking software... I know it was mentioned many months back ... perhaps this will help resolve the issues?

I really cant think of anything else to try, to gets its tracking more accurate. :(

Im really disappointed, as for 6 months its been the controller i always just turn to, it lets me relax and noodle, and just enjoy the flow - but this has turned to frustration as I hear it glitching, and I find myself again trying (in vain) to recalibrate it.

I don't know about any issues with warm temperatures. However, I guess it's possible that the combination of constant high temperatures and daily playing has caused the rubber surface of your instrument to stretch out a bit. Your report that things are going slowly more out of whack is also consistent with surface stretching.

The good news is, there is likely an easy mechanical fix for this you can do. Rest assured that we'll get your instrument working one way or another. If we can't figure out hardware or software modifications that you can do there that make it work for you, I'll take it back for repair.

First, I would like to ask for some of your time to understand what is going on, and see if my idea about stretching is right. Do you have a utility you can make movies with? Screenflow has a free trial and works well for me here.

If you can make a movie of the response from a single finger as it moves around the surface, I will be able to see if the surface is doing what it should be mechanically. To do this, you want to turn "voices" to 0. This will cause the display to show the raw, unfiltered data. Set the view mode to "calibrated" and run a finger left to right in five passes, one over the center of each row.

I made this kind of movie, just doing one pass at top and one at bottom, and it looks like this: http://madronalabs.com/media/temp/touches0.mp4

I would describe my Soundplane as working well, but you can see from this movie that the single touch spreads out more in the upper right of the Soundplane than other places. The software is able to correct for this, but if it were much more spreading, there would need to be a hardware correction to fix it.

Being a curious guy, maybe you have opened up your Soundplane already? If so, you will have noticed that there is some thin veneer applied to the basswood spacers that you see when you take the back off. These spacers hold the sensor boards against the surface when the back is attached. The veneer is applied by me when making each instrument, and sometimes I have to go through a few iterations of tweaking the veneer to get the spacing just right because each surface is a bit different.

Long story short, you can likely add some veneer strips to compensate for the stretching and fix the instrument yourself. First, let's see if the stretching is the problem.

ok, uploaded a video as you requested... here : https://www.dropbox.com/s/ocr32uct37mk5iy/SpCalibIssue.mov?dl=0

(fyi, here is a picture of the calibration
https://www.dropbox.com/s/gmhwec1wexl7dcr/Screen%20Shot%202015-07-17%20at%2016.32.31.png?dl=0)

I ran my finger over row 1 , then 5, 2,3,4 (and then over the end columns) as evenly as possible.

you can see some unevenness, and in particular around column 26 (in all rows), which is noticeably,when playing, more sensitive that any where else on the board.

the I switch to touch mode.and first play a M7 interval (carefully centred on both cells!) (so 1 up, and across), you can see the root is okay(ish) but the G is half a cell away, and this typically will randomly choose between the D and G... so its unpredictable... also of course even if it is the right note... its now got Y = 0, what I played was Y=0.5

after this i just play D and G, and you can clearly see the ghosts notes firing...
(Ive tried with higher thresh, but it doesn't make any difference unless its at about 0.022, which means you have to almost hammer notes, no subtlety.

Ive tried manually altering template, usually it calcuales around 0.3, but I've tried much lower and much higher, again which no perceptible difference in getting reliable touches.

Ive watched the videos of you opening.. but looks easy enough, but not tried myself as wouldn't really know what to adjust or how to make it better :)

oops realised that was the dev version... anyway same on production version,
this is with 1.2.5, just some general movements, showing edge issues, and also how its also not predicable with intervals.
https://www.dropbox.com/s/kkh009ykcd80u0d/SpCalib2.mov?dl=0

(I play intervals like this, as not enough room to do horizontal, and this formation prevents having to play adjacent cells ... and I don't like playing in row 1/5 which I find has less Y range, and can be more unpredictable.)

The response looks pretty good overall, I wouldn't say that any major stretching has gone on. On the right side, I would say there is some spreading of the touch that exceeds what it would have left the shop with.

If you want to intervene mechanically, adding a bit of material to the rightmost spacer bar will hold the sensor more tightly together, and should reduce the spreading. What I use is a wood veneer with an adhesive back, because I have lots of this material. Anything incompressible will work equally well. thin cardboard with double-stick tape or removable spray mount to hold it in place, for example. Or, I can send you some veneer.

I would start with around 0.5mm more material over the entire bar. Then if the feel is better overall but too tight at the top and bottom, you could remove some of the material at the top and bottom. Or if it improves things but not enough, you can add more.

You can either calibrate after each change, or you can clear the calibration with "use defaults" in the app and use that as a baseline for noticing changes. In any case, proceed methodically and record the response before and after. You could use a weight of some kind on the surface as a calibration standard.

But first,

why don't you try some variations in how you are normalizing it. I do find it a bit weird that in your normalize picture there are some areas of almost zero value in the middle. This may result from pressing down harder in the middle when you normalize. Try "use defaults" and see, is the center more accurate without normalization? If so, either use the defaults or try normalizing again being sure to press very evenly. You can use an phone or a CD case for this!

When you are looking at the accuracy of touch position, look at only one touch at a time. This will give you repeatable results. Any inaccuracy in chords that is not present for single touches, is the fault of the touch tracker software not the normalization or hardware.

Also,

You could just wait and accept the ghost notes into your life until the next software update. I am sure that they will be improved by the next touch tracker. Meanwhile the thing with ghost notes is that they are tiny, so when using direct envelope control via z they may not even be heard.

I'm eager to work on the new touch tracker, but meanwhile I also need to release another plugin. So I hope some of these tips give you something to chew on, and I'll thank you in advance for your patience.

Interesting... particularly the comment about the heat.

For what its worth I returned to my soundplane after a few months hiatus and found the tracking to be much further off than I remember. Pitch tracking was okay but the inconsistency in pressure was surprising. I tried several rounds of calibration but the end result generally seemed less consistent than just clearing out all calibration.

...all my recent experimentation was during a string of very hot days (comparatively).

...the surface of the soundplane also seemed to have more vertical travel than I remember.

If you want to intervene mechanically, adding a bit of material to the rightmost spacer bar will hold the sensor more tightly together, and should reduce the spreading

Im a little confused about this, Ive watched your disassembly video (a few times), but don't see any spacers... are you saying that you thing the two sensors boards could be separating slightly?

Im sure i will see if I open up, but kind of want to check what I need before I dive in :)

try normalizing again being sure to press very evenly. You can use an phone or a CD case for this!

this also confuses me a bit, really the difference between the first step (with palm) and second.. are they calibrating different things? It would I think be useful to know a little more about this.

as Ive mentioned before, I do wonder if an 'editor' for the calibrated data would be useful,
perhaps were we could select regions, and increase/decrease/smooth out the difference, and retry... I know it could be fiddly and perhaps would need good 'explanations' to stop fumbling around in the dark.

but one of the problems Ive had in the past is, I have an issue, so recalibrate, it fixes one area of the board, but another gets worst. (so sometimes I have lived with a bad area of the board, getting a worst compromise)

You could just wait and accept the ghost notes into your life until the next software update. I am sure that they will be improved by the next touch tracker. Meanwhile the thing with ghost notes is that they are tiny, so when using direct envelope control via z they may not even be heard.

ghosts notes, true to some extent,it is more the tracking thats my main issue. and in some ways I have lived with it till now (its always existed to some extent), but its just got to a point where its pretty hard to play... it not very gratifying playing an instrument that doesn't have consistent/repeatable behaviour, which is pretty fundamental for any instrument (imho)

I'm eager to work on the new touch tracker, but meanwhile I also need to release another plugin.

Yeah, I recognise this, bills to be paid and all that :) same for all of us, and of course there is only so much one man can do at a time. so fully appreciate priorities etc.

I only hope this can receive some future priority, against all the calls for new versions of Aalto/ Kaivo/ new plugin/ Modular interface... which perhaps generate more revenue, than software that is given away free and has less than 100 users (albeit its part of an expensive hardware product)

Id dive into the touch tracking software myself, but frankly, when I checked it out, its obvious the amount of knowledge/experience you have built up in this area, my efforts are pretty futile by comparison... a few hours of your time, are worth hundreds on mine.

So I hope some of these tips give you something to chew on, and I'll thank you in advance for your patience.

Of course, I will try... as Ive said (repeatedly) , I love playing the soundplane, its a wonderful experience and its noticeable how much its teaching me too, so I really don't want to put it away until it cooler (3+ months here) or a newer software version is out.

anything I can do in the meantime I will definitely try, and of course try to be patience :)
(let me know about the spacer / calibration and I will look into this.)

thanks for your help
Mark

I looked at the disassembly video again, and I seem to have totally hidden the spacers, because I wasn't thinking about showing them. So your confusion is natural—sorry about that. If you look on the underside of the aluminum back you will see five wooden bars glued on. These hold the sensor layers in place and allow room for the ribbon cables.

The thicker the bars are, the more they put the sensor layers including the foam rubber "waffle" into compression. Ideally, when the instrument is at rest, everything fits tightly together but the foam is only compressed a very small amount: 0.5 mm or so.

On Monday I can get back to the shop and take some pictures of the bars if that would be helpful. Or, I can simply make a sketch of what should be going on. It's really simple.

Now with two reports (above) of changes maybe due to heat, of course it has me thinking that there is something I can do better. I tested the response of the foam over many months before shipping an instrument, but not at temps of 30C. Possibly the foam rubber will permanently deform enough under these conditions that it will be noticeable.

However, a sample size of two instruments is not a lot! So I have to proceed carefully and avoid jumping to conclusions about what might be happening. One thing I can do here is try some experiments with the rubber under compression at higher temperatures.

Every kind of compressible stuff loses some of its ability to return to its original dimensions over time. The foam rubber was chosen after testing around ten different materials and reading specs for many more. It worked well in testing but everything degrades over time and it is certain that heat accelerates this process. Luckily, there are two fixes we can do: add spacing material, or replace the foam layer entirely. I would like to get more data and try a fix with spacing but can also send you a new foam layer if that turns out to be a fix.

Mark, you have a Soundplane you bought second hand, is that correct? Can you remind me of the serial number?

I know I sent one instrument to a hot climate—I can ask the customer how it's going or maybe @rastkopravi will see this and give us some info.

Please understand that I stand behind the instruments and it bothers me a lot when any of them are not working well. Certainly my goal is to make them play consistently and accurately. I agree these are important qualities for any instrument. When I say "accept the ghost touches for now" I'm not trying to say there is no problem, but rather trying to give you another option until the time I can do work on the issue.

The new tracker is a big priority. Luckily for me, you are in a great position to test any changes and seem to have some time to do so—I'm thankful for your involvement and look forward to getting you some options to try ASAP.

this also confuses me a bit, really the difference between the first step (with palm) and second.. are they calibrating different things? It would I think be useful to know a little more about this.

The first pass gathers the normalize map. This tells the software how much each taxel (sensor junction) moves when the same amount of pressure is applied. Typically there are small variations over the middle of the surface, and big consistent changes around the edges, due to the way the instrument is held together. The normalization corrects for both of these.

An ideal way to get the normalize map would be to have a weight of around 100-200g with a flat bottom. Moving this weight around would put a consistent force everywhere on the surface. In practice, moving a palm around with a consistent pressure has given good results, and most people have hands permanently attached, so I have been recommending this instead of something more complicated. But if you are getting inconsistent normalizations, maybe try some kind of moveable weight setup.

The second pass generates a template touch. This is used to discriminate touches from ghost touches and other noise by recording the shape of data a single finger makes. it's important to move reasonable slow here! I would allow at least 10 seconds to traverse a row in one direction.

In the second pass, consistency of pressure is not required. I recommend a medium to firm pressure. If the pressure is too light then the resulting template will be quite large, and will not have a well defined effect.

Sorry, been busy trying to get Axoloti ready for its release :)

my soundplane, is #54 I guess from second batch? (yes from a guy in Norway)

Heat, agreed two SP is not much of a sample set, could be a wide variety of other things.
I guess my hypothesis is not so much about stretching, more that the rubber is warmer, and so is taking longer to return to rest state, which I think would lead to false touches, as the other touches are not properly released yet.
its hard to say if it is, but it does feel a bit more 'sluggish'... but perhaps Im imagining it :)

I tried recalibrating with a CD case to get an even pressure... and its different, but not really better.

I guess I'm still a bit vague on what I should use for extra spacing material, don't really have any spacer that I can think of, and not sure where to get some, but will have a look around... (perhaps I can order something online?)

one thing has struck me, that may be an 'easy temporary fix' for the touch tracker.
Its pretty obvious that the ghosts notes are usually just one cell from the actual touch, but this should be impossible as a new touch should not start within 1 cell range... its suppose to inhibit close touches... so this seems like it might actually be a small bug.

ok, it wont solve the issue of the tracking being out, but it might help reduce ghost notes, and also (in my experience) sometimes fixing such bugs can reveal other issues.

54: from second run, not first. OK that's good information about timing. The spacing material is the same in both cases.

If you have time to mess with spacing, send me an email and I can send you some spacer material. It's sounding more and more like maybe a new touch detector is going to be the best fix for you, however. Luckily you can try and even build dev versions, so it shouldn't be too long until you have something to test.

The situation with the ghost notes and the existing tracker is a little complicated. Your idea is reasonable but doesn't take into account that touches can start out in one place and get moved frame by frame in an attempt to converge on the best solution. So even setting a high inhibit threshold, a touch can start out two keys away from the existing touches and then get sucked in, so to speak. The new tracker uses a different approach and will not have this problem.

email sent... thanks.
yeah, I suspect better TT/Calibration will help, but it may be spacing also plays a role.

touch tracker 'bug' -yeah I fully recognise why its doing it, the 'sucking in' is a really noticeable phenomena, but I think even that should be 'curtailed' to some distance (perhaps closer than the normal 'new touch'?) because this sucking in, if done too close will start generating ghosts notes as well.

I think the rule is 'reasonably' firm, you cannot have 2 notes that are closer than N, because once you get this, chaos follows pretty quickly, regardless of the original cause.
(e.g even if I slide two notes close together the TT will soon start having issues).

perhaps its possible, that perhaps N may be slightly different for different scenarios, e.g. sliding together (touch age?) or very new touches, or perhaps even touch thresholds.

but I look at it this way, I can get the SP into a position, where with one sustained touch, can be playing two (sustained) notes on adjacent cells... for me this is breaking a precondition... id be firing code asserts :)

you can see it in my video - https://www.dropbox.com/s/kkh009ykcd80u0d/SpCalib2.mov?dl=0 @ 0:50
ok, this is edge of the board, but I get this in the centre of the board too.

anyway, I only said an 'easy fix' as I thought it might be doing an additional inhibit check, but I do recognise the TT code is complex, and there are lots of 'use cases' so improving one thing, may make something else worst/stop working... so just an idea, one that you are in a much better position to judge, it could well be time much better spent on the new TT.

So, so good. Randy, I tip my hat to you, sir! There should be an award for elite instrument makers, a few have started to spring up here in London but SP wins (hands down) for beauty and awesomeness and people saying, "what the hell is that?!".. Which we like!

So just a quickie: The extreme sides notes of my plane tend to offset touches inwards, playing ghost notes on the next row or note (row 1 and 5 triggers ghosts on 2 and 4).

I've had a good play every day this month and a hefty fiddle with the lowpass and thresholds. Have you any advice on calibrating this effect out? Should I play around the edges more for a while? Thanks for your hard work in these puppies!

Hi Andrew,

The issue with the top and bottom edges is one I have experienced too. I tend to compensate by attacking the notes harder at top and bottom and saving those slow onsets for the middle rows. A higher threshold will help, but of course that's not always what we want.

To a large extent it's really down to the way the instrument is put together. The way the frame is tensioned causes the actual force of the touch to be spread out more near the top and bottom.

There is also an effect where a very nearby hand or finger causes what, for brevity, i'll call a negative pressure. You can see this by looking at the calibrated data without any touches active. XY mode may be the best way. If you press with a pencil or something, you can experiment with what happens when you keep your hand away. By approaching more perpendicular to the surface and not keeping your hand resting near the edge, the effect can be mostly avoided.

This is definitely on my list of things I'm trying to improve in the new software.

Haha! Thats so good to know that The Master also experiences this! I quite like the challenge of techniquing out the quirks of a new instrument. Its like learning a violacello;). And I have also been attacking he edges with a little more velocity than other notes.

I am loving spending time with the SP (which I have yet to name), and creating SP specific Kyma instruments. An edge-offset could be handy..

But here's a brain fart: A workaround for me, would be to create a larger zone out of the two lower rows givng a two-row Y position and leaving the top three to the default 4ths. That could be epic! Thanks for responding, R.

@andrewbird... I think we all have similar problems.

mine had recently improved slightly, I suspect partly due to reduced ambient temperatures... but I'm still getting stray touches. but I think, we all know that these are due to limitations within the touch tracker.

(unfortunately, due to family/work pressures Ive not had time to take apart the soundplane yet to see if mechanical changes can help)

Its a bit frustrating, some days/times the Soundplane feels like the perfect instrument, you feel in touch with it... but then this is spoilt by it false touches/glitching and the moment is 'lost'... it would be a dream to have it behave perfectly all the time :)

(I'll point out, for me, the ghosts notes/glitch tend to make midi mode pretty pointless, if your on the edge of the board, or close multiple touches... as it causes gate triggers, its not so bad with T3D/Aalto/Kaivo, as I avoid gates... its why I pretty much only use Aalto/Kaivo with the soundplane which is a real shame!)

Randy, is your plan to give the touch tracker some 'quality' time after Virta's release?

This is what Ive been impressed with on Geerts development on the Linnstrument, they have not added many features, but over 6 months they have consistently been improving response, consistency and 'feel', and Roger has clearly stated this is his #1 priority, to ensure the feel is 100% , as why most users buy these instruments.

this is no 'criticism', I completely understand the competing time pressures, more just a plea to give the soundplane software some priority, over perhaps the more 'lucrative' plugin development... but improved consistency in the touch detection would be enormous for soundplane users.
(to be clear I only want consistent to the 1"/non-adjacent cell, and over the whole board... Id much prefer this over any attempt to 'add features' e.g adjacent cell detection.)

anyway, fingers crossed...

"I know I sent one instrument to a hot climate—I can ask the customer how it's going or maybe @rastkopravi will see this and give us some info."

Hi,

yes I live in a rather extreme of temperatures 40 Celsius in April and then like rain for four months with humidity 100 :-) Burma.

But I keep my instruments in a air conditioned room and I have a dehumidifier in the same room.

My Soundplane works really well.
I took it for concerts here and in Vietnam outdoors https://vimeo.com/127172158 and everything was ok . Great actually :-)

It was stuck in the Myanamr customs for a month and it was either humid or hot in that storage but again all was well.

sorry I could not help more and I send you greetings from the rainy Yangon

rastko

Randy, is your plan to give the touch tracker some 'quality' time after Virta's release?

Definitely.

this is no 'criticism', I completely understand the competing time pressures, more just a plea to give the soundplane software some priority, over perhaps the more 'lucrative' plugin development...

It's not a matter of choosing what might be more lucrative. Plugins are what keep the business afloat. If I don't release a new plugin or two each year there is no more Madrona Labs, simple as that.

Mark, I am sorry the pace of my development is too slow for you—please know that I have not forgotten about your desire for better responsiveness over MIDI. And please be reminded that I'm just one person.

hey, I'll be buying verta too ( have to have all ML products ;) ) and completely understand new products (and upgrades, aalto 2.0? ) pays the bills, we all need to eat and make a living
(and as developers we all need to do new things too, to keep fresh!)

and I also completely understand maintaining/improving existing software is hard given operating system changes etc (I just spent a day fixing EigenD for el capitan),
then there are feature requests etc... its never ending (I know been there too!)

as i said, no criticism intended... just a request.

perhaps, soundplane software 2.0, could be a paid upgrade? that could help fund its development? ( though i suspect time is more the issue, but its a thought)

@rastkopravi , cool music and video :)

interesting i dont have air conditioning, and actualities a very dry (low humidity) climate here... but id guess high humidity would be more a problem than low.

I will say, Im not 100% sure if my issues are common, perhaps its also partly comes down to type of pieces played or technique? unfortunately there is little feedback from other soundplane owners (either one way or the other)

anyway, it has improved, so I'm back to playing the soundplane each day, so thats cool.

... and Im looking forward to virta :)

I would be doing the wrong thing to charge for the Soundplane software. I think that it has been a work in progress for some time, and so I prefer to keep it free. But thanks for helping me think outside the box of my habits.

You are not alone in the issues you are finding with precision, especially over MIDI. Everybody has a different reaction to them—some people focus on the technical details and try to iron out the problems, while others kind of probe out the limitations of the instrument, put those in a corner, and get on with music making. Both approaches are valid—your feedback is helping me improve both software and hardware.