Welcome to the Piano World Piano Forums Over 3 million posts about pianos, digital pianos, and all types of keyboard instruments. Over 100,000 members from around the world.
Join the World's Largest Community of Piano Lovers
(it's free)
It's Fun to Play the Piano ... Please Pass It On!
Cybergene: do you think it would be feasible to have two teensys sending 2 streams of MIDI signals along the same MIDI channel number (merged in a USB hub)? One of the teensys would scan the hammer optical sensor and only sends note-on events, while the other teensy scans the key movement and only sends note-off events... Even if possible, I believe it would be easy to run into cancellation problems between the two streams and very fast trills would end up with a lot of silent notes or a few strangely longer ones...
This is easy. They can even be just two separate MIDI interface connected to the computer sending at the same time and I believe most VST-s can respond to multiple interfaces at the same time (maybe even on different MIDI channels but I believe the two interfaces can both send on the same channel). One of the boards will send only note-on velocity events, whereas the other will measure only the release velocity.
From mechanical point of view I think you can assume dampers are hard-linked to the keys. Although that's not exactly true and there might be some inertial differences, in most cases, including leggiero it would work. The problem with leggiero is rather one of the VST itself. On a real piano leggiero means slightly tap the key so that the hammer has high enough momentum to hit the string, without pressing the key fully. If you do that without the damper pedal pressed, it will have the effect of very muted notes because the dampers may not even be lifted, or even if lifted they will return to the string back. I'm not sure that's well modeled even in Pianoteq because it would require a MIDI support where the software knows the depth of each key at any moment + the hammer velocity hit (just for a moment imagine that you accelerate the hammer strongly without even lifting the damper off the string: there will still be a specific muted note produced and no piano currently emulates this). I believe only AvantGrand pianos right now send such data in their MIDI stream. Note also that this is different than note-off velocity. It's a continuous position of the key (hence damper) which you can then use to calculate a release velocity of the damper at the middle of the key travel where that happens.
Originally Posted by xooorx
Yes. Velocity is being measured here as time taken to cover a fixed distance. That 5 bits wouldn't be the resolution of the measurements: It's just something that lets you set up some start and finish posts so you can make some measurements. If they're in roughly the correct places, at roughly the correct distance apart, then the key will work and measurements can be taken. The accuracy of the timing measurement is the important thing.
I completely agree. You can measure velocity between any two points. This won't hamper the precision of measurement because you will use the correct distance factor when calculating the velocity.
The only mechanical consideration that needs to be taken into account in such a scheme is that the velocity measurement needs to be performed above the escapement and that's usually just 1-2-3mm of distance before the stop rail. This is needed to ensure that the hammer is detached from the key and free flying. Otherwise you risk measuring wrong velocities. Imagine for example if your first point of measurement is when hammer is 25mm before the stop rail. You can activate this point by slowly pressing the key halfway and so the measurement is started, you can wait a lot there with your finger and then throw the hammer quickly towards the string: the velocity cannot be calculated properly due to wrong duration. Also, you need to be within the last 1-2mm of hammer travel to allow for fast repetitions near the bottom of the key, without full release. So, 5-bits can be enough to position the two points within the last mm or can be too wide. So, it needs to be experimented.
I'm not sure that's well modeled even in Pianoteq because it would require a MIDI support where the software knows the depth of each key at any moment + the hammer velocity hit (just for a moment imagine that you accelerate the hammer strongly without even lifting the damper off the string: there will still be a specific muted note produced and no piano currently emulates this). I believe only AvantGrand pianos right now send such data in their MIDI stream. Note also that this is different than note-off velocity. It's a continuous position of the key (hence damper) which you can then use to calculate a release velocity of the damper at the middle of the key travel where that happens.
Again I was probably being naive about the sound generation capabilities of current software, and was assuming that leggiero would be the consequence of fast/high midi note-on occurring simultaneously with very slow/low midi note-off, translating into something akin to the damped harmonics in guitar playing (to be frank, never tested it in pianoteq... will do one of these days).
You spiked my curiosity about the yamaha midi implementation, but could not find in the reference manual any mention about key/damper position. In fact, event the note-off is ambiguous because page 3 of the N1x MIDI reference states that note-off ranges from 0 to 127, but in page 16 note-off becomes fixed at v=64.
Maybe it's finally time for me to put this note-off debacle to rest... If you feel that at the current state of software implementation, having more detailed note-off does not make a significant difference in playing dynamics, I will truly trust your judgement because you have proven many times to possess a most sensitive touch and ear, and have shown that probably no one understands more how the mechanical side of the piano translates into MIDI mapping and how MIDI interfaces with sound generation.
Honestly, I don't think note-off is such a big deal. I have a very fine implemented damper modeling on the AvantGrand where slow release of the key would gradually dampen the note but I can't find any practical difference with the Cybrid that lacks any note-off behavior besides fixed value.
As to the Yamaha implementation: I'm not sure what their MIDI chart is but we have discussed on the N1X thread that key depth is being transmitted as aftertouch in a very chatty manner and I can confirm that.
Honestly, I don't think note-off is such a big deal. I have a very fine implemented damper modeling on the AvantGrand where slow release of the key would gradually dampen the note but I can't find any practical difference with the Cybrid that lacks any note-off behavior besides fixed value.
I think note off is for wind instruments (and maybe synths) ... not for pianos.
I think note off is for wind instruments (and maybe synths) ... not for pianos.
Now that is something I really don't understand... In very slow pieces I assume that fine control of key depth during key return would be extensively used for harmonics (because a good player would be able to let the damper barely touch the strings). That's in my essence as a guitarrist... At the piano I am currently at the chimpanzee level of key finesse, so it's not that I am missing control... It's only about knowing that you should practice with the system that allows maximal control, so that it is there for you to overlook but also aim to...
The thing is, some effects may not be apparent when you try to test them, however they are cumulated in your aural memory through the subtle differences. I can tell you for sure I don’t hear a difference when playing a single note with and without the una-corda pedal, however when I play some passage with many notes it’s very obvious. Maybe the note-off behavior has a similar and even subtler unconscious effect. I wouldn’t discard it just because I don’t see a huge difference between it and Garritan CFX with the fixed behavior.
I think note off is for wind instruments (and maybe synths) ... not for pianos.
Now that is something I really don't understand... In very slow pieces I assume that fine control of key depth during key return would be extensively used for harmonics (because a good player would be able to let the damper barely touch the strings). That's in my essence as a guitarrist... At the piano I am currently at the chimpanzee level of key finesse, so it's not that I am missing control... It's only about knowing that you should practice with the system that allows maximal control, so that it is there for you to overlook but also aim to...
The position, shape, and material of the piano damper make guitar-like stopped harmonics all but impossible.
I don't find great differences between my Kawai CL35 and the MP11, the first sends fixed note off velocities, whereas the MP11 sends different values, I don't remember if is uses the whole range from 1 to 127 o something narrower. Only in staccatissimo may be there is some diference, since I never release slowly keys on purpose.
The only mechanical consideration that needs to be taken into account in such a scheme is that the velocity measurement needs to be performed above the escapement and that's usually just 1-2-3mm of distance before the stop rail.
Does the action itself have any trim pots (erm...) regulating screws that could adjust this zone?
The only mechanical consideration that needs to be taken into account in such a scheme is that the velocity measurement needs to be performed above the escapement and that's usually just 1-2-3mm of distance before the stop rail.
Does the action itself have any trim pots (erm...) regulating screws that could adjust this zone?
Yes, thats the standard regulation performed by piano technicians and tuners. I followed the same steps as described in Reblitz’s book.
The only mechanical consideration that needs to be taken into account in such a scheme is that the velocity measurement needs to be performed above the escapement and that's usually just 1-2-3mm of distance before the stop rail.
Does the action itself have any trim pots (erm...) regulating screws that could adjust this zone?
Yes, thats the standard regulation performed by piano technicians and tuners. I followed the same steps as described in Reblitz’s book.
I mean that 1-2-3mm in particular, if you can make fine adjustments to its width and/or position without causing problems elsewhere. Together with a screw adjustment of the sensor position like you mentioned, plus 5 bits of DAC control of the comparator levels, *plus* (it turns out) 2 bits of control of the voltage reference used by the DACs...
I mean that 1-2-3mm in particular, if you can make fine adjustments to its width and/or position without causing problems elsewhere. Together with a screw adjustment of the sensor position like you mentioned, plus 5 bits of DAC control of the comparator levels, *plus* (it turns out) 2 bits of control of the voltage reference used by the DACs...
Hi xooorx... I am not entirely sure if I am understanding you correctly, but regulation of the hammer positioning should not be used as a method for decreasing comparator levels. Although the piano action can be regulated by many millimeters along several of its joints, there is an optimal point of calibration for each key that depends on the action of (at least) the neighboring keys. Changing action regulation in order to gain sensor efficacy is not the best strategy: piano regulation should be optimal and only then sensor adjustment should be performed to accommodate any loss in sensor sensitivity, and not the other way around...
To all DIYers out there with an interest in Cybrid:
I am definitely interested in reproducing the Cybrid prototype. I have been in contact with Cybergene and contacted the chinese supplier that produced his PCBs; since I intend to have them provide the fully assembled PCBs (ie, with all electronic components soldered in and ready to operate) this has called for a lot of additional information because cybergene only order the PCBs and done all the soldering by himself (my estimate is around 6000 solder points!!! for the entire project - not happening for me...), the assembly process requires a bit more information and I have been working with them for the last two weeks (daily emails with specs, alternative bill of materials, additions to the PCB files - awesome client service).
The Cybrid project requires three type of PCBs: - 88 PCBs for the individual optical sensors (one per key) - 18 PCBs for the group note modules (each processes the information from 5 sensors/key/notes) - 1 PCB for the MIDI USB processor (in addition it will be necessary some hardware for positioning the PCBs within the piano, the Teensy MIDI/USB microprocessor, and some cabling to link everything together).
I have now a preliminary quote for the group note modules (only ask for that PCB since it is the more complex and therefore all future problems would be identified and worked out). I say "preliminary quote" because that are a few technicalities not fully set and because one of the electric components (the variable resistors or trimpots, 15 per module PCB) are still quoted as purchased from a major manufacturer instead of being purchased from an alternative chinese supplier (cybergene also used cheaper chinese trimpots without problems). That change alone (still under evaluation) will reduce the overall cost of components by more than 50%.
My current quote now is of 354 USD for 20 module PCBs, or 1154 for 100 PCBs (ie, 231 USD per set of 20). If the change in trimpots supplier is possible, I estimate the price to become ~280 USD for 20 PCBs and ~850 USD for 100PCBs (ie, 170 USD per set of 20 PCBs).
Based on this, I estimate that the cost of 100 sensor PCBs and 2 MIDI PCBs (always a good idea to produce a few spares) to be in the range of 100 to 200 USD.
Thus, if 4 more PW members want to get together with me in this order, the major electronic components of the prototype can be ours for less than 300 USD per user.
Anyone interested please send me a PM and we will work from there (and if it has more expertise in electronics than me, that would be highly valuable to double check the specs...).
$300 sounds like a great price to me (it includes the parts, right?).
I also feel a bit more responsible, so I will try to summarize what we have already exchanged in PM-s and put it in GitHub, as well as regulation considerations, frame mounting and crucially my own procedure I devised for aural velocity trimpot calibration which isn’t very difficult and yields very good results.
Yes, it includes all electronics in the PCBs except the teensy microprocessor (I am estimating for the remaining 2 types of PCBs since I was only quoted for the "module-note" PCB), but the 300 USD is in case there are 5 persons interested (ie, a grand order of 100 PCBs of the module, 950 PCBs of the sensors, and 10 PCBs of the MIDI-USB-teensy). If no one else is interested, the cost for me alone will jump to roughly 600 USD (plus import taxes, for sure).
I am not confortable with sharing the quote details in this post, but will gladly do it by PM.
I think those (or similar) videos were posted on my initial research thread. It most probably uses PnoScan or similar commercial silent system with sensors under the keys.
This thread is awesome. Reading through it today. So I also have been prototyping some ideas, after seeing CyberGene's inspiring keyboard. So cool! Great work.
I was going to type it up into this thread but decided a video might be more fun. So I collected video snippets I've taken and narrated the work so far and direction. Alot of these ideas are inter-related to what I am reading here.
This thread is awesome. Reading through it today. So I also have been prototyping some ideas, after seeing CyberGene's inspiring keyboard. So cool! Great work.
I was going to type it up into this thread but decided a video might be more fun. So I collected video snippets I've taken and narrated the work so far and direction. Alot of these ideas are inter-related to what I am reading here.
I really enjoyed your video introduction Greg. There can never be enough project detail.