PNOscan sensors

It's been six years since I installed my PNOscan II.  I don't know if any of these photos will be of assistance to you, but here is a photo-album that I made while things were 'in process':  https://goo.gl/photos/mZ6UywoFz79DXAMr9

Also, while I can't get to the underside of the keys easily, here is a link to an image of the Sustain pedal board:  https://photos.app.goo.gl/3Zg3AXvfhnM6MQxo6

The LED laser is part D14, glowing red in the photo.  I presume that the key sensors use the same type of device, though perhaps not the exact same unit.  I cannot easily see identifiers on it - they may be on the underside of the laser.

Are you considering building your own? If so, consider the CNY70 sensors as described at https://pianoclack.com/forum/d/107-cybr...ano-action and https://pianoclack.com/forum/d/163-buil...the-action and https://github.com/jkominek/piano-conversion and related

Nice discussing this topic with you, Gruust.

Even if the CNY70 data sheet does not say anything about time resolution, both the Cybrid and the piano-conversion have been able to demonstrate that the time resolution is fast enough for playing. For distance, I am not sure what the commenters you saw said, but as far as I am concerned, if properly utilized it can be used to measure distance. Cybrid does so with a (double, IIRC) comparator. The piano-conversion does so with ADC's. So from an electronic point of view I think the CNY70 is at least more than adequate and perhaps even very good.
Regarding the size, I admit it is on the large size, even though I'd not call it "rather bulky" as you say. I wish it was a bit thinner and a bit smaller so that one had a bit more ease of assembling and aligning the parts, however it's not too big compared to the pitch of a regular sized keyboard (if you want to make keyboard sensors) or to the pitch of the hammers (if you want to make hammer or shank sensors). Now, if your interest is in a reduced size keyboard (see paskpiano.org if you don't know what I'm talking about) then the CNY is certainly too big.

Alternatively, there is the Everlight EAITRCA6 which is half or less in each horizontal direction (so you could cram four of them in the space of one CNY70) and 1/4 of the thickness. And its data sheet does mention the 10-20 microseconds mark of time resolution, so perhaps that's already what you have in mind? Or are you thinking of something else? Please let me know! FWIW, Jay also is interested in exploring such a sensor, see https://github.com/jkominek/piano-conversion/issues/44

Last, but very importantly, how can you doubt that 10 microseconds is not enough? Microseconds, not milliseconds! So the sensor is capable or responding to a change of position of the key 0.01 milliseconds after the key has moved. Given that all PTQ player are quite content with 2-3 milliseconds and many even with some slower response rate, I wonder if I am missing something in your reasoning. Do note that Cybrid has a sequential scan of the keypresses (and releases) but **does not** wait for the key press to complete. So in this design the time resolution is more about the frequency of the sequential scan than the time resolution of the sensor. Of course you could duplicate all the electronics for each key and this is (partially) what piano-conversion does, with one individual ADC channel per sensor (two channels per key, or 176 channels total), but at some point you'll have to serialize anyway, since MIDI is a serial, not parallel protocol. Well, in theory you could give each key its own sound generator, but at that point you'll lose things like sympathetic resonances, so I don't think that's what you have in mind. Please clarify.

Thanks!

I agree, quite nice photos and description. Thanks for sharing it @dklein!!

Are you the same person as jjYBdx4IL on github, who asked very similar question here: https://github.com/ekumanov/cybrid/issues/4 ? Feel free to ignore this question if you prefer to stay anonymous, I was just curious since it sounds an unlikely coincidence.

In my opinion, your suggested test is kind of moot. PNOScan might report reliably the same MIDI velocity, but the hammer velocities will hardly be exactly the same, and hence Cybrid or piano-conversion will not necessarily report the exact same MIDI velocity -- because of escapement and regulation. In fact 3-4 different piano technicians told me that on acoustic pianos only with the composite WNG action pianists can expect a "perfect" response from one note to the next. One of such technician (who actually was skeptic about WNG action and bought it in the first place only because the Renner one he wanted was not in stock at the time) has posted a few videos about it, including one in which separate pianists (much more advanced than I) explain this "feeling" better than I could, see the second part of https://www.youtube.com/watch?v=M5OfbIw_9dI (go forward in the video to hear the pianists, the beginning is of interest only if you care about the action itself, which you might not)

It's part curiosity, part trying to figure out what good options there are if my ordered MP11SE device shouldn't meet my expectations - because there wouldn't really be much of an existing alternative then. NV10 is way above my price range. Acoustics with PNOscan are probably above my price target, too, and their actions are probably just a service-heavy part that I wouldn't use beyond silent playing anyways (so why accept the wear and tear on it?). Grand Feel with optical sensors? Sounds like a very decent solution to me.

Another potential issue I could have with PNOscan is the way it's getting calibrated: AFAIK it calibrates top and bottom positions of each key. From that I infer that it guesses the sensors values for the relevant points in between, ie. between which values to measure dt etc. I'm sure that guess work will do just fine in most situations. However, I'm also thinking: why not just use a step motor to depress each key slowly and record somewhat more accurate position information? Optical solutions like the one in AvantGrands seem to avoid all of these issues by a precise fabrication of the light barrier dimensions and precise positioning on the hammer shanks - a precision that's probably not achievable with DIY. AFAIR they also use some sort of linear encoder to determine absolute key position. But I neither like the keys nor the price of the AvantGrands. And all of their tech is pretty much proprietary. You'd always be dependent on them. With PNOscan (or a DIY alternative) there is much more flexibility. Well, and I also like to tinker. Haha.

Doesn't look like it.

I collected a few strokes in a MIDI0X or similar software, wrote the log to disk and parsed it. Wouldn't recommend trying to step up the values manually.

Thanks! On properly working keys on my almost 20 years old Promega 3 it's usually +/-1 difference. Problem is, some keys (or key pairs) are always off by like 8 or 12, and I seemingly can't fix it.

High res velocity data is transferred via CC#88. You should see corresponding control messages if it's active.

That comment was not targeted at PNOscan, rather at using a digital piano action instead of an acoustic one. I assume the digital piano actions have been built in a more resilient way, generally, even if it's only because they are much simpler.

Yeah, they don't have it in their support section for download. And Google finds nothing relevant about it, really. Mysterious program. How do you get it? The parts listing also doesn't mention it.

Yeah, that tool looks really nice. So many options to screw around with the device.

Is it possible to make such adjustments per key?

1000 scans per second. That's PNOscan III. For 1900 bucks. For that money I could almost attach a separate processor to each single key. The software seems nice. But the hardware seems rather dated.

That's a valid point.Though I wouldn't entirely agree with "fine". If you assume a travel time of 0.3ms, which is a little cumbersome assumption because it comes from hammer travel time AFAIR, a basic time resolution of 1ms would mean washing out the measurement over a longer distance, implying that at high velocities the measurements cannot accurately reproduce the velocity at the precise, relevant point. Also, what makes you think that time resolution has no effect on "position" measurement? If you sum up intensity over the entire 1ms, you will end up with the aforementioned washing out/averaging out/built-in "anti-aliasing".

That would be a quite lenient way of measuring velocity. Usually you'd want the hammer (key) speed at a very specific point in space.

That's quite optimistic. But if we knew the sensor IC, we could at least guess what that time window could be - assuming they are simply doing a momentary readout and nothing else. But even that would be an assumption. And why would they tell us about the low scan rate instead of the actual precision that's way better?

Easy, because to be profitable you need to sell large volumes. How many of these they can expect to sell? Not enough.

That's called the N1X or NV10 (and you can guess why they went down a completely different route, and didn't use PNOscan). I think you are heavily overestimating the sales numbers for such products. Are there any numbers out there for these products and PNOscan etc.? I would guess that stuff like AvantGrand and Novus probably primarily exist to improve their reputation. In the overall market I doubt that these products have much of a relevance. But that's just an educated guess based on the price differences. Also, they have to provide home service for these products. That requires a world-wide network of trusted partners. Very expensive.

Also look at the Lachnit MK23. Who would buy that? There is no guarantee that the producer will be able to send you replacement parts in even a year down the road (small producers go out of business all the time). And I guess there is no service network at all. So, if you are Yamaha, you have a HUGE advantage here (which is why they can still afford to put that subpar action into their AGs). Though, if you'd put small sensors into a portable device, the need for a service network would be reduced. But MP11SE and VPC-1 aren't exactly portable, and the keys in the VPC-1 aren't really that great either.