DIY Single axis USB controller

[jacob]

@Sclass -

Some of my early outputs looked a lot like Predator's wrist computer

What I have is a size=10 byte array called digits where each array entry shows what segments to light up. E.g. digits[2] = B11011010 and a size 4 byte array called position to pick which of the four common cathodes to sink into, e.g. position[3] = B11111011. Then I just feed a set of those into the two shift-registers to light up one number. E.g. if I want to light up the hundreds, I'll SIP feed position[3] followed by digits[(int)(velocity/100)%10)] bit-by-bit while flipping the clock pin back and forth. This sounds like what you had?

I do worry whether the Arduino will be able to keep the digits lit "enough" once I start adding more displays to the loop() cycle. (I have noticed that "small" digits like 1 are brighter than "large" digits like 8. Fortunately, my marketing department was the first to volunteer for my astronaut corps. They were as brave as the rest but rather more stupid ) It's funny that the "best programming practice" for space launches very much resembles a quant-style trading algorithm as opposed to the ghetto-spaghetti style often used for physics research.

I prefer the 1960s Starship Enterprise style and programming from scratch for KSP. Early on I used a library, but I wasn't happy not knowing what was going on under the hood. (The stitch'n'glue approach to programming just doesn't appeal to me.)

I'm thinking of getting some 5V voltmeters to use as fuel-gauges using pulse width modulation, e.g. outputting 4V on the pin would show 80% full. Just have to find some cheap ones. 5V seems rare, but 5V is what the Arduino offers. Any ideas?

I've seen a couple of OLED displays in people's KSP controllers(*), but that will have to wait for the multifunction displays for the next project, e.g. Cessna172, A10, F16, or Boeing737. Who am I kidding, it'll be the A10

(*) One potential style-cheat worthy of an OLED would be the Nav-Ball (as seen in the center bottom for the video I included above).

I did look into button matrices. The amount of wiring required is brutal though, especially given my rat's nest wiring style. I really need to figure out how to prettify my wiring. One problem with the shoebox approach is that everything is best kept on the lid lest one uses really long wires. I don't have a good solution yet. It's a mess. It might be better to compartmentalize with a bunch of small protoboards, like the one at the bottom of the shoebox, instead of feeding a single large one.

[sky]

Cake pans are pretty easy to work with.

[loutfard]

That, or cheap plastic boxes. If you like to cleanly rack mount electronics for customers, do have a look at Gainta g17081ubk series 1u rack mount cases. They're cheap and very practical.

[jacob]

I put the 20W amplifier I built many years ago into a 1U rack mount, but I think I'm in love with the cake pan idea. It's perfect Kerbal style!

To get an idea of what other people have built, see https://www.google.com/search?client=ub ... =823&dpr=1

[sky]

If you get 2 cake pans you can bolt or clamp them together to make a case.

[Sclass]

E.g. if I want to light up the hundreds, I'll SIP feed position[3] followed by digits[(int)(velocity/100)%10)] bit-by-bit while flipping the clock pin back and forth. This sounds like what you had?

se as fuel-gauges using pulse width modulation, e.g. outputting 4V on the pin would show 80% full. Just have to find some cheap ones. 5V seems rare, but 5V is what the Arduino offers. Any ideas?

I did look into button matrices. The amount of wiring required is brutal though, especially given my rat's nest wiring style. I really need to figure out how to prettify my wiring.

Ok. A lot here. Yes what you are doing is a common cathode muxxed display. This can all be done without glue logic though. Instead of using glue I used small MOSFETs (n type IIRC) on the cathode strobe lines. And optionally p type on the anodes. You can see them along the edge of the display in the image. The microcontroller handles all the strobing and latching through code and port writes. Your dimming is likely a bottleneck in the HC logic.

The PWM can be done with the AVR too but it puts the burden on the pwm port. It can all be done in code. No need for fuel gauge, you just use a stiff 5v rail and precalculated pulse durations. If you’re really tricky you put your pulse durations in your digit lookup table to get consistent brightness.

You can engineer all this yourself or just buy the ASIC. They are less than a $1. The OLED displays are super cheap too.

You can clean up your wiring using precut jumpers. Notice I mostly use the orange 0.3” ones. They cost money but you’ll be surprised how few you actually use. My $20 kit from 1997 is still full. I bought one bag of every color except orange which I bought three of. You can never have enough Orange 0.3”.

Or, you can just keep doing what you’re doing. It looks great and does the job.



ETA - I like this solution to simplifying buttons.

https://www.ti.com/lit/ds/symlink/tca84 ... e.com%252F

Sorry, misread your post about the shift register. You’ve basically built your own serial LED controller. I looked up the 74HC logic and it can source enough current to light up segments. What you may be seeing is a bottleneck in your 5V rail coming from what appears to be the USB from your PC. The port should be able to source some decent current but check to see what path it takes through the Arduino. You may be current limited.

ETAx2 - so I didn’t read your idea to use a serializer to read in button states. This is possible no doubt. You’ll use some cpu overhead to scan the data coming in. But the same can be said for the i2c stuff that I mentioned. The advantage is it easy to implement an i2c interrupt to stop your main program and deal with the asynchronous button io. But, the way you say is doable.

Same with the deserializing for LED. I mentioned using an ASIC that does all of this while doing current balancing. Turns out there are these vendors on eBay selling little boards with four digit seven segment displays with serial to seven segment display driver on board. Adafruit sells their own that actually uses an AVR controller stuck to the back to do this as well. These all sell for $10-20. (Wait, you said that too and I just didn’t read it)

So there are many ways to solve this. I’ve always tried to use fewer chips. Lower part counts, fewer points of failure, smaller designs, simpler more robust PCBs. But this is just my design philosophy. I think that’s why I like directly driving the LEDs straight from the uC ports using MOSFETs.

[jacob]

@Sclass - I actually have a box of jumpers. You can see me using some of them in the picture. However, the jumpers I have now are not what I had in the 1990s. These are an absolute pain to put in without bending them (I find myself using pliers, ARGH ). It's probably because I'm using cheap-ass supplies. Could be the protoboard, could be the jumpers. Everything is smaller and thinner these days, even the resistors Also, I think I should take my two protoboards and put them next to each other. At least half of my wiring problem is ending up fitting everything onto a "single wide" one rather than a "double wide".

Also, I think maybe my shoebox prototyping could be improved by using longer wires (instead of just long enough as per my frugal nature) and putting Dupont connectors on them? Currently, I've ghettoed something similar by soldering single-strand wires at the end of the multi-strand leads and covering the soldering with electric tape. You can see them on a couple of the button connections (yellow leads, black tape).

A potential problem is ending up with a pound of wires inside the shoebox. However, I don't see any other way to still be able to open/close the lid.
See last pic on https://pangalactictech.com/kerbal-spac ... ontroller/

My design strategy is basically limited by my sourcing strategy in that I don't have everything at hand and I don't know how to go in and read catalogs to find what I need because my electronics knowledge contains a large white map space of unknown-knowns: I build something only to later realize that I can buy a much cheaper solution if I only knew it was there. So for me it basically looks like: 0) Read Dummies book. 1) This example might work/be useful. 2) Add components to shopping list. 3) If shopping list >$35, order from amazon. 4) Build and learn. 5) Realize that this could have been done cheaper and better with a ready-made solution.

Doing it the hard way is still cheap tuition money though.

PS: I highly suspect that the main bottleneck is the game's output frequency on the computer side.
PPS: I found a couple of 5V voltmeters on eBay and ordered a couple via chinapost. I figure I can use analogWrite() and wire the voltmeter in parallel with a sufficiently large capacitor to get rid of any jiggling due to PWM or overly slow updates if necessary.

[Sclass]

My design strategy is basically limited by my sourcing strategy in that I don't have everything at hand

Doing it the hard way is still cheap tuition money though.

PS: I highly suspect that the main bottleneck is the game's output frequency on the computer side.
PPS: I found a couple of 5V voltmeters on eBay and ordered a couple via chinapost. I figure I can use analogWrite() and wire the voltmeter in parallel with a sufficiently large capacitor to get rid of any jiggling due to PWM or overly slow updates if necessary.

Just checked around. No kidding all the good 3M stuff is no longer made. Breadboarding has kind of gone the way of through hole components.

Sounds like you’re learning a lot. It’s a fun place to be.

When I mentioned bottlenecks I meant the current sourcing ability of your HC chips. Roughly speaking you’ll want 10mA to each segment of the LED. Sometimes more for daylight viewable models. If you are strobing common cathode 7segment displays you can have up to 70mA being sunk through your chips. This isn’t a problem but you likely will see dimming when “8” vs. “1”. While each pin of your cmos can easily source 10mA it simply cannot sink 70mA on one line.

You can see the tiny MOSFETs around my display. Those are the sinking devices on the cathodes. I use the uC to source current to the individual anodes. The conventional way to do this back in the day was to use individual bcd to 7seg decoders with cathodes to ground. But that is a lot of chips and wiring. I like the minimalist approach shown on my breadboard. Fewer chips. More ISR code use to control display.

I think your design can use some mosfet drivers on the cathode side. This didn’t occur to me till just now.

Not sure about the voltmeter approach. Conventionally we try to maintain a controlled voltage via a stiff supply and pwm to control current. You kind of want to measure current not filtered voltage. This is super easy with an o scope. (Peak current X duty cycle) will give you your controlled current.

I’d be looking for one of those $25 oscilloscopes before trying to use a voltmeter as an AC power meter with filtering to get a handle on your current flows.

[jacob]

Here's an LCD test of a suborbital launch, now with pretty jumper cables:

This takes up six digital outputs using a library, so I have no idea what's going on. Apparently, it's possible to use both 4 (here) and 8 data inputs. I wonder what the difference is.

It was a pain to debug due to what can only be described as a "don't eat the yellow snow"-bug---the kind that is impossible to know in advance but which you only do once. Hmm... I merged an LCD tutorial with my Simpit interface w/o realizing that pin1 (d1) also happens to be TX for the serial port used by Simpit. This created weird spurious line breaks and characters on the display as the game's msg handler was connected to the LCD's RS pin which determines where to write the data. Once I realized the problem (thanks internet), it was a simple matter of using another pin.

I'm also aware the pin d13 is related to the internal LED, so better be careful there too. To me this double use design seems like a bug rather than a feature, but what do I know ...

Using the Mega2560, I've also run into this problem: https://www.reddit.com/r/KerbalControll ... ssible_in/ This is an absolute pain since debugging the controller requires restarting the entire game which takes about 10 minutes on my computer Strangely, this is not a problem for the Leonardo. Anyhoo, the thread mentions opening a second serial port on the Mega via UART. This sounds like some dark magic much like "don't touch D1 which doubles up as TX" that exceeds the dummy manuals I'm reading.

[Sclass]

Yeah it’s been a long time since I used a module like that one. It looks like you have a lower nibble and higher nibble written in two steps. My recollection lcd controllers have the option to write the lower nibble than the higher nibble after pulling one of the control lines. Cannot remember. There’s a register select likely to pick the instruction or data byte. Then there’s the enable to read out or write the buffer. Then there is a read write line to toggle the direction of the bus.

Looks like you use the register select, the R/W direction bit is statically tied to write only and a register enable E bit prolly to swap upper and lower nibble. And you have 4 data bits wired. That’s six.

This is a leftover from the 80s when it was easier to lay out more parallel traces on a board than handle serial data in the uC. Some of the old chips had 512 words of code space so sending two nibbles wasted a couple of lines of code which were more precious than 4 circuit board traces. The fewer the wires the more complex the protocol.

The LED stuff is likely backlight control (which it looks like you’ve already current limited with a pot) and contrast/viewing angle control.

Stepping on lines and registers is kind of the risk in embedded programming. You don’t have an OS to protect you from writing over used resources. You have to be careful when you’re coding so close to the HW. Not too familiar with the Leonardo uC but it may share some vital resources between serial interfaces.

Looking good!

[jacob]

PPS: I found a couple of 5V voltmeters on eBay and ordered a couple via chinapost. I figure I can use analogWrite() and wire the voltmeter in parallel with a sufficiently large capacitor to get rid of any jiggling due to PWM or overly slow updates if necessary.

The voltmeters arrived.

I had to do some science to make them work. Mainly, I had to figure out what their internal resistance was. Here's "the science":

Code: Select all

1.1V measured in the middle of two 10ks with 5V across
5V = (10k + (1/10k + 1/R)^-1)   *I
1.1V = (1.10k+1/R)^1 *I
=>
3.9V = 10k*I
=>
I=0.39mA
So ... 
(5V-10k*0.39mA)=(1/10k + 1/R)^-1)*I = 1.1V
=> 
1.1V/0.39mA = (1/10k+1/R)^-1
=>
R = 3928 Ohms

This is very low! The typical resistance of a voltmeter, as far as I know, is on the order of megaOhms. I suppose I can thank my physics [background] for knowing where to look and how to find out even if I had no clue about the eventual answer.

However, even though these voltmeters are no good for circuit-testing, they'll still serve my purpose drawing a safe 1.25mA max from the Arduino which was my main worry.

I set it up the Arduino with pulse-width-modulation (a map() to analogWrite() ) on a low-pass filter (the wiring with a capacitor). No smoke appeared although I did close my eyes before I switched it on. It's a habit of mine ... founded on the belief that my eyelids double up as safety glasses?!?

Now I have to figure out what to "meter" for the launch. I'm thinking Fuel and Atmospheric Pressure. Since I bought two meters, I took the other one apart. (It's easily reassembled.) The plate is "printed metal". I want to replace this plate. The ghetto way would be to tack spray a printout from Inkscape on top. Maybe there's a nicer way, like remaking a new plate. This would require figuring out how to cut super-thin metal as well as printing on it. Perhaps the CNC router will solve this?

ETA: In the final installation I forgot to include both the resistor and the capacitor. This works fine and in some ways better. The internal resistance will draw a max of 1.25mA, which is negligible as each pin on the pro micro can handle 40mA (<200mA combined).

[jacob]

Something similar is possible in reverse. Instead of individual wires from each button, I can send 8 of them into a 74HC165 and have only 3 wires go back to the Arduino. This is not done yet, but should be similar.

This is done now. Nothing exciting to show though. I need hook up another 74HC165 in serial, because I do have more than 8 buttons.

I used: https://www.makerguides.com/interfacing ... h-arduino/ ... except I used two-pin buttons, so I rewired them all using pulldown resistors.

Add: 16 done. I'm running out of components and jumpers, so some buttons remain unconnected and the 6 inputs on the second 165 are connected to ground. Leaving them floating didn't work. (Also those buttons aren't the greatest.)

[jacob]

I bought four MAX7219 controlled 7-segment displays. This created another potential problem.

Apparently, when connecting them in series, it is a good idea to shunt and remove the SMD diode (D1) which is just to protect the board from accidentally switching 5V and ground. Also put a 100nF capacitor between Vcc and GND to reduce noise (the two holes below D1 will do that.)

I spent around 10 minutes trying and retrying soldering the pins on. GND was the worst!, what an exercise in frustration! As a good^H^H^H^Hbad workman, I blame my tools. My rusty old telephone pole of a soldering iron from freecycle has seen better days. So has my eye-sight. Also, maybe I should figure out what a flux pen is? Not really something we used when I learned to solder 35 years ago.

An alternative would be to solder in multistrand wire, which I think will "take" better, to daisychain the next 3 displays. And wire the power in parallel from an external supply. This should allow me to keep the diodes w/o suffering the foolish "safety" voltage drop. Also, I can wire in the capacitors off of the board.

Alternatively, I could go spend some money on tools, since this kind of problem may occur again. Like, when I have to transfer this growing mass of wires and chips to a perfboard.

[rref]

I spent around 10 minutes trying and retrying soldering the pins on. GND was the worst!, what an exercise in frustration! As a good^H^H^H^Hbad workman, I blame my tools. My rusty old telephone pole of a soldering iron from freecycle has seen better days.

The tip is oxidized which prevents the solder from clinging to it which again reduces heat transfer to the lead and pad.
Clean the tip [edit: Not the one pictured above. It is done. Get a new one.] on brass wool before soldering something, then apply a protective layer of flux core solder to the tip after having soldered something.
Use as low a soldering temperature as possible* and/or turn off the soldering iron as soon as possible to increase tip longevity.
Chose a tip that is approximately the size of the component you intend to solder.*

* A trade-off with a lot of other parameters such as large ground planes and other unwanted heat sinks.

[jacob]

@rref -

Rather than throw money at it, I figured I'm going to attempt making my own replacement tip. It's an M4 screw tip. I have some 4mm brass(*) rod, so I'm going to put a piece on a drill press and try to file it pointy; then cut some threads into the other end.

(*) I know copper is better, but brass is what I have.

GND was definitely the tricky one to solder on account of having a very large area on the PCB.

Any recommendations for flux? I've never used it before. I saw one that looked like a needle.

(I looked at my roll of tin and it says flux=1.8%)

[jacob]

I want to replace this plate. The ghetto way would be to tack spray a printout from Inkscape on top.

This might work:
https://www.instructables.com/Heatless- ... CB-Making/

[rref]

Rather than throw money at it, I figured I'm going to attempt making my own replacement tip. It's an M4 screw tip. I have some 4mm brass(*) rod, so I'm going to put a piece on a drill press and try to file it pointy; then cut some threads into the other end.

It is solderable so why not. Find some scrap PCB and practice desoldering and resoldering components. Scrap PCBs are also great for learning how far you can push it wrt mechanical stress (fingerspitzgefühl) and heating before the components break or fry, the PCB delaminates, the PCB reaches its glass transition temperature and warps etc.

GND was definitely the tricky one to solder on account of having a very large area on the PCB.

One way of bodging this would be to use a hairdryer (any heat source) to raise the overall temperature of the PCB (and the ground plane) before and during the soldering process. Just keep the PCB temperature below the melting temperature of the most fragile components (and the melting point of your solder unless you want SMD components flying everywhere). Again practicing on scrap PCBs is a good way to reach "kennen" faster.

Any recommendations for flux? I've never used it before. I saw one that looked like a needle.

A Chemtronics rosin or no clean flux dispensing pen would be a good starting point. Just apply it on the component leads and solder pads before soldering.

Quality of solder tin can vary wildly. Tin-lead (Sn63/Pb37) solder with flux cores have a melting point around 180 degrees Celsius and is easier to work with than lead-free solder. Hopefully Sclass will drop by with some brand recommendations.

[jacob]

First, I spent $20 at Harborfreight for a metric tap&die set. Then I spent 4 hours figuring out how to cut a proper thread in a 4mm brass rod that I already had from my clock-making WOG.

Answer:

• Brute ape force - no
• 45 deg chamfer to start - no
• 45 chamfer and taper and ape force - no, threads increasingly destroyed, but getting there
• 45 chamfer and taper and 1/2 turn forth followed by 1/4 turn back (trained neanderthal) - also no
• 45 chamfer and taper and 1/2 turn forth pushing down hard followed by 1/4 turn back without force - yes!

Physically, the cheap die comprises 3 (120 degree angle) x 4 (sets of) microplanes that carve out a thread deeper and deeper. Turning forwards carves increasingly deeper. Turning backwards clears the debris. Then go forward again. Understand and respect the physics!

The new tip is better than the previous tip. And good enough. I have a new one set up to go with a finer point, but I might just not bother with it. Still ... the "making it from scratch and seeing it work"-feeling is extremely satisfying!!

This morning I spent a few hours fiddling with Ledcontrol.h. First attempt worked out of the box. A couple more hours of derpy programming and it was possible to print any decimal number whether positive or negative. (The first three numbers should somewhat sum to the fourth +/- rounding!). Given the struggles with this according to google, I was positively surprised how few [hardware] problems I had. Maybe I'm better at soldering than I thought.

[Sclass]

Sounds fun. I have that exact iron for plastic welding. I use a nail (with head) as a tip.

The displays look nice. Well done. Red acetate works really well as a lens. It makes the unlit segments less visible. You can get it cheap from the kids craft section. My favorite source is printer transparencies. You can make your desired shade of red and put some text on as well with a color printer. Print borders. I had good results with a color laserjet printed mirrored so you can flip it over to protect the soft side.

@rref I generally stick with Kester products for availability and consistency. Lead free silver solders need hotter tips to flow properly. I use exclusively lead tin because it’s cheaper, easier and more reliable.

[jacob]

Red acetate works really well as a lens. It makes the unlit segments less visible.

The camera does weird things with the light, like making the segment looks half-way orange. The display looks much better in reality. When it's running, all segments light up equally in a very bright red (better than the DIY LED-controller above).

I avoided having to desolder the diodes by feeding each segment individually with 5V. (Hence four red wires instead of one.) I've noticed that a lot of online complainers use dupont connectors. That might be the source of their digital noise. My daisy-chain is soldered all the way through.

And it really helps to solder before the first cup of coffee, not after.