Hybrid Carillon in the making

AlexanderBunt Just a whole lot of if else statements won't work I think (far too slow) and makes a true mess of the code of course.

This isn't necessarily true. For instance, when you want fast digital read on the Teensy, they advise that digitalFastRead() works only with a pin parameter that is a constant, e.g: digitalReadFast(10) instead of looping some variable and doing digitalReadFast(i). That is why my code looks so ugly with no for loops. As a Java dev who is used to writing nice code that goes through merciless reviews, I couldn't look at what I had to write for the Cybrid 😃

I'm not sure that applies to the analog read though.

As to advices for writing the code, I can give you a very simple example in how to loop through multiple CNY70-s by using the Arduino based (high-level) language, however I'm afraid you may need lower-level C-stuff that would directly use the underlying microcontroller and I'm of no use here since I've never been good with low-level programming...

AlexanderBunt I don't have any useful Arduino experience. In general you'll want to arrange for something which can, on a regular schedule, just read all of the relevant ADC channels and store the values. Then once you've grabbed them, do whatever checking you want.

Controlled timing can be more important than "fast". If you just check things as fast as you can in a loop, then any conditional code execution will make some iterations of the loop take more or less time than other iterations which might cause you difficulty timing things. Depending on how you implement things.

AlexanderBunt Just a whole lot of if else statements won't work I think (far too slow) and makes a true mess of the code of course.

This is being compiled down to assembly to run on a microcontroller, if the task fundamentally requires a bunch of "if else statements" then you're going to end up with a bunch of compares and conditional jumps one way or the other.

The standards for what constitutes messy firmware differ, IMO, from the standards for high level code running on luxuriously overpowered desktop/server processors which just need to respond fast enough for a human.

JayKominek The standards for what constitutes messy firmware differ, IMO, from the standards for high level code running on luxuriously overpowered desktop/server processors which just need to respond fast enough for a human.

In the world of high-level languages like Java, what is most important is easy maintainability of the code and self-explanatory code. Even a ternary operator is undesirable when it will take 10 additional seconds for someone else to read it 😀 I'm so used to these requirements, I had real difficulties understanding low-level code examples for the Teensy and I just gave up. It's a bit of a shame when I call myself a developer but that's the reality 😳

CyberGene I can give you a very simple example in how to loop through multiple CNY70-s by using the Arduino based (high-level) language

Well, that would be nice I think. I have never done any low-level coding, so that won't be my strongest point either. But maybe the if-else structure wont be that bad then. My only concern: if you make a statement like this:

1. if first key pressed -> do this;
2. else if second key pressed -> do this;
3. else if third key is pressed -> do this;
4. and so on until key 27.

Will be key 27 much slower than key 1…?

So, CyberGene is you can cive that example? Would be nice.

CyberGene Yeah for me the same. Used to write in languages as C# and JavaScript/React, so this stuff is very new for me.

JayKominek In general you'll want to arrange for something which can, on a regular schedule, just read all of the relevant ADC channels and store the values.

Exactly, but maybe just a loop over all values will be the best (and ugly) thing to do… I don't know.

AlexanderBunt Will be key 27 much slower than key 1…?

Assuming you have a while (true) loop around each ifs, no, each key will be the same. However as Jay said there may be latency on how long it takes to do the entire loop in case what you do into the "ifs" are time consuming. If you are just setting shared variables for another thread to act on (e.g. send MIDI packets), I think this loop will be fine.

And remember: premature optimization is the root of all evil, so just try it, see what happens, gain experience and change what it doesn't perform as hope.

PS: I have not done any Arduino either.

RIP Yeah doing that now. Trying and see what will happen. Thanks for the help.

Will keep you posted.

RIP And remember: premature optimization is the root of all evil, so just try it, see what happens

Wise words, as always...

@JayKominek

If I want to use your sensor board, where do I have to place which resistor and how does the CNY70 sensor be placed on the board?

For example this board:

I think the led resistor should be on Em1 - Coll1, and the phototransistor on An1 - Cath1. But am I right?

AlexanderBunt If I want to use your sensor board,

Have you tried to use KiCad for the purpose? All the information you are asking should be there.

Start https://docs.kicad.org/5.1/en/getting_started_in_kicad/getting_started_in_kicad.html if you don't know what I'm talking about.

PS: sorry if I sound too blunt, I'm just trying to protect Jay from too many questions (including my own 🤣)

AlexanderBunt That appears to be my 2020-offset-1x board. It doesn't have any resistors on it. The labeled pads you're seeing are test points.

The LEDs in my design are driven by a constant current source, so there aren't any resistors used to control LED current. Similarly, the phototransistor collector is attached directly to +3.3V, and the emitter feeds into a transimpedance amplifier. So there's a gain resistor for that, but it is located adjacent to the relevant op amp.

The orientation of the sensor is easy enough, though. The markings on one side of the CNY70 should be facing left in that image, away from the U1 silkscreen.

My document here https://github.com/jkominek/piano-conversion/wiki/Analog-Stage-Theory-of-Operation explains how my system powers the CNY70 LEDs, and then gets/makes use of the signal from the phototransistor.

I am just 'finished' my first version of the Hybrid Carillon. I must say, apart from the power, which I don't know how to get right, it works very well… I used one Teensy 3.5, with 4 16x multiplexers. So I have 64 sensors put through just 4 analog inputs on the Teensy. I know, this seems to be far to slow, but its not. Or at least, not mentionable…

The only problem I am facing now, is the power. For now I use breadboards for the connections. This is of course a temporal solution and I want to change this to custom made PCB's. But for the power I use two 'breadboard 3.3v/5v' power supplies for 48 sensors. These offer 700Ma. And I know 1400Ma can not be enough for 48 sensors, because the sensors each need 50Ma for as far as I know…

But which solution should I use for this problem? Maybe you guys have a clue…? @JayKominek @RIP @CyberGene

Hi AlexanderBunt , great news, well done!

Perhaps look at something like this:

https://www.soundimports.eu/en/mean-well-rs-100-24.html

That'll supply plenty of nice and stable power, yet outputs 24V DC. So some conversion/regulation electronics are needed to give you the 5V and/or 3.3V, whichever one you need.

Maybe there are other power supplies that would give you the voltages right away, like a computer ATX power supply with modified cabling. I personally like beQuiet! computer power supplies because they're decent and… drumroll… quiet.

Cheers and happy building!

HZ

AlexanderBunt Congrats for not giving up, feel free to share your progress video 🙂

As to power, I use an iPad charger which is 2.1 A and is ripple free (well, almost but better than any other power I tried). Take in mind that changing the USB cable that you use for power may change the resistance and hence the voltage calibration of the CNY70 sensors 🙂 Ask me how I know 😀

AlexanderBunt The only problem I am facing now, is the power. For now I use breadboards for the connections. This is of course a temporal solution and I want to change this to custom made PCB's. But for the power I use two 'breadboard 3.3v/5v' power supplies for 48 sensors. These offer 700Ma. And I know 1400Ma can not be enough for 48 sensors, because the sensors each need 50Ma for as far as I know…

Are you using optical sensors, like the CNY70? Do you want reliable dynamic control? If so, stability of the power supply to the LED is essential. If the power fluctuates, the light output fluctuate, and so the apparent distance that the phototransistor "thinks" the hammer is at fluctuates, and so the actual dynamics (and perhaps even timing) fluctuates. The stability is essential. All commercially available power supplies that I know of are not stable enough if you want to be accurate, even though you might live with their limitations as @CyberGene has done.

See https://github.com/jkominek/piano-conversion/wiki/LED-Power-Board for a simple solution (which has also a PCB already designed at https://github.com/jkominek/piano-conversion/tree/master/hardware/led-power and you may be able to order it already assembled from a PCB company.

PS: I assume you mean 700mA and 1400mA etc. Case is very important: A=Amper=current. The lowercase a seems to imply acceleration. Lowercase m = 1/1000. Uppercase M=1000, so a million more than you (most likely) meant

AlexanderBunt because the sensors each need 50Ma for as far as I know…

I assume you stuck with the CNY70s.

Their IR LEDs do not consume 50mA. Rather, they're current devices. They consume approximately 1.25V. (That does not mean they should be attached to an ideal 1.25V source!) And the 50mA value comes from the "Absolute Maximum Ratings" section. You want to give them less than that. I settled on 20mA.

I believe you were using the design where you put a resistor in series with the LED to control the current. That value, along with the forward voltage of the LED, will determine how much power is consumed by each of the LEDs.

Mean Well is a perfectly good power supply manufacturer, but I'd buy supplies much closer to what you actually need, rather than fiddling with trying to drop down from +24VDC to something small. (Anything which can do that without burning up will introduce its own noise and problems.) They manufacture a number of supplies which directly output +5VDC. Maybe even +3.3VDC, though I haven't checked their catalog for that. I suggest fully enclosed supplies.

Ideally, you'd want to have designed it to use a constant current supply for the LEDs, and chained them together in groups of ₂₄. Without knowing exactly how you've done the physical wiring, I can't say if it is possible to go back to that at this point.

CyberGene As to power, I use an iPad charger which is 2.1 A and is ripple free

🤣

RIP you may be able to order it already assembled from a PCB company.

It has a number of through-hole parts, I doubt any of the inexpensive places will assemble the whole thing.

JayKominek

CyberGene As to power, I use an iPad charger which is 2.1 A and is ripple free

🤣

I know why you gave it the LOL 😀

But let’s say this, among the power supplies one can usually find at home, the iPad one is probably the best. I even found some website that cracked it open and analyzed it against many similar chargers and praised it for being designed way better than the rest and having more advanced circuitry to provide less ripple. I need to find the link. Of course there are probably better power supplies out there 🙂

Yes, typically 5V on a series of one 220ohm resistor and a 1.25V led. There remains 5-1.25=3.75V… then the current is 3.75/220=0.017=17mA.

Okay wauw thanks for the many answers! I will make a photo of the wiring when I am at my brothers house (thats where it's standing…), probably tonight. Don't panic when you see the absurd amount of wires haha.

JayKominek Mean Well is a perfectly good power supply manufacturer

3.3v power supply

Probably this…?