Off Topic Banter

[mctom]

Funny how I have discovered interest in even lower layer than Assembly, for the sake of pure efficiency, while others pick languages even more detached from the hardware.
"Memory is cheap", right?

[rooted]

Funny how I have discovered interest in even lower layer than Assembly, for the sake of pure efficiency, while others pick languages even more detached from the hardware.
"Memory is cheap", right?

The only layer lower than assembly would be binary in general computing so I assume you are referring to FPGA, I wouldn't call that lower level since it's abstract?

[mctom]

The only layer lower than assembly would be binary in general computing so I assume you are referring to FPGA, I wouldn't call that lower level since it's abstract?

Lower level than assembly, strictly speaking, would be machine code (binary you must have been referring to). I got my hands dirty even in that, when the ASM op code documentation had an error and something didn't work as they said.

To me FPGA is even lower level than machine code, because it's all about putting logic gates together to perform tasks. The same logic gates that are the building blocks of all CPUs, so building a CPU on FPGA is possible and commonly done.
What's lower level than the power to define your own opcodes in your own CPU?

[mad_ady]

What's lower level than the power to define your own opcodes in your own CPU?

Manipulating the electromagnetic field Magneto style, to get the outcome you want!
Offtopic - could Magneto only influence the magnetic field (B) or also the electric field (E)? Electric field seems more useful, imho...

[mctom]

Manipulating the electromagnetic field Magneto style, to get the outcome you want!

I prefer stuff that works automatically, and not flipping bits by hand.
My superpower is creating logic gates out of npn transistors! That must be the lowest level I am handy with. Home brewing transistors, meh, not so much.

Offtopic - could Magneto only influence the magnetic field (B) or also the electric field (E)? Electric field seems more useful, imho...

Finally something offtopic.
Well, every time you change a magnetic field, you create a perpendicular electric field, and vice versa. That's how light propagates.
So it doesn't matter which field is controlled directly.

[rooted]

What's lower level than the power to define your own opcodes in your own CPU?

It's the method (the development environment) in which you create the CPU which cause me to form this opinion.

But I do see your point.

I think the creation of real hardware is the lowest level, virtually is definitely cool though

[mctom]

There's nothing virtual about what I'm doing with FPGAs, as far as I can tell..
I program an actual physical chip and test my hardware live, never used any simulator.
And the chip contains actual logic gates that I merely describe how shall be connected with each other.
Of course the point of all that is to design a bunch of logic modules I need for my hardware project

True, Verilog is slightly higher level than drawing logic gates on a piece of paper. It automatically creates logic networks from C-style boolean and algebraic expressions. But still requires manual design of Moore / Mealy state machines just like in the 90s.
On the other hand, iCE40 chips have universal logic gates, which are 4-bit LUTs. So each gate can perform any logic operation with 4 binary arguments. So the design techniques from the 80s that I know (and taught dozens of students) do not apply anymore.

I do keep an eye on the results of code synthesis and inspect the chip layout to see if it does a better job than I would manually. There is a nice online chip layout viewer, that frankly is more stable than GUI version of nextpnr. ;P

And here is a clock divider that I finally transformed into a standalone module today, for reference.

Code: Select all

───────┬────────────────────────────────────────────────────────────────
       │ File: clock.v
       │ Size: 462 B
───────┼────────────────────────────────────────────────────────────────
   1   │ // A generic clock divider
   2   │ // f_out = f_in / divider
   3   │ // Output clock is post-divided by 2 for equal on and off time
   4   │ // Thus the actual divider is rounded down to nearest even number.
   5   │ module clkdiv(
   6   │     input wire clk_in,
   7   │     output reg clk_out
   8   │ );
   9   │ parameter divider = 1000;
  10   │ 
  11   │ reg [$clog2((divider/2)+1):0] clkdiv;
  12   │ 
  13   │ always@(posedge clk_in)
  14   │ begin
  15   │     if (clkdiv >= (divider/2)) begin
  16   │         clkdiv <= 0;
  17   │         clk_out <= ~clk_out;
  18   │     end else begin
  19   │         clkdiv <= clkdiv + 1;
  20   │     end
  21   │ end
  22   │ 
  23   │ endmodule
───────┴────────────────────────────────────────────────────────────────

[hominoid]

What's lower level than the power to define your own opcodes in your own CPU?

Manipulating the electromagnetic field Magneto style, to get the outcome you want!
Offtopic - could Magneto only influence the magnetic field (B) or also the electric field (E)? Electric field seems more useful, imho...

I and others believe they are one in the same thing but that is yet to be proven by any unified law.

[mad_ady]

Ah, now I see @mctom's verbosity in his answers comes from a teaching background!

And that verilog code stirred up some long lost neurons!

By the way, isn't there some verilog blocks library with everything you need? Seems to me you're reinventing the wheel!

[mctom]

I think that was another way round. I tutored students to keep in touch with anybody that wants to talk about electronics. And also, ladies

opencores.org hosts multum of generic modules that are free to use, if I adopt their license or whatever.
I wrote the clock divider as the first exercise, not too complicated either. Same with UART transmitter, and I was surprised it took some 30 lines of code or so.

The problem with generic modules is they are not written with actual hardware in mind. Simple stuff like clock divider would work on any FPGA, but more complex things will require porting.
I think that's going to happen with I2C module at some point. That one has open collector ports and I don't expect that to work out of a box.

[mad_ady]

Grr... I hate unrepairable hardware...
My wife's Fitbit Versa smartwatch broke before our holiday. When put in the charger it gets very hot and doesn't turn on. A quick google search shows it's a common symptom for a dead battery. The watch is bearly 4 years old, so - I should be happy it lasted this long.
My wife likes it, so, I'll be replacing the battery.
But, getting inside is challenging and local phone repair shops do only Samsung and Apple watches... I forsee deep cuts in my fingers in the immediate future: https://www.youtube.com/watch?v=6t_h53SDlSM. Now, I need to locate a razor blade...
Assuming all goes well, what should I use to glue back the screen?

[joerg]

Assuming all goes well, what should I use to glue back the screen?

Please not the 'Jack of all trades' hot glue.

[mctom]

Oof... If I had to do this, before any attempt I'd tell the owner the repair is more likely to fail. Just look how many things can go wrong over there. :/
But still it is worth trying, of course.

For glueing screen back, I have no idea. Personally I'd try a thin double-sided adhesive tape. I'd take a piece bigger than a screen and stick it to something very flat like glass. Then use a scalpel to cut the shape of the original tape that I think is in there.

[joerg]

That would be the correct type of tape, but with a modern acrylate adhesive.

[mctom]

I don't know about chemistry unfortunately, but now when I think about it, would be much better to buy a flat sheet of such dual side adhesive material.
The point is it shouldn't be too strong. And a tiny bit will be necessary for a battery too, so it won't rattle during use.

I wouldn't rule out reusing the original one, though, if you won't allow too much dust sticking to it in the meantime.

I don't have much experience with Kinder Surprise technology, sadly. I much prefer nuts and bolts over plastic latches and glue..

[joerg]

Too bad, I have done so much with my pc last days, I lost the bash history from older than last 2000 commands. Normally I do backup my history from time to time, but haven't done last half year.

[mctom]

I use bash history very rarely for anything, because it's always a mess. I often use midnight commander which seems to be completely detached from bash history, or handles it in some weird way. I'm trying not to rely on that.

Anyway, perhaps it's a high time to set up automatic daily backups of everything?

[rooted]

I use bash history extensively, with this added to bashrc you can type part of any former command and complete it with up/down (or scroll through previous commands starting with whatever letters you typed)

Code: Select all

if [[ $- == *i* ]]
then
   bind '"\e[A": history-search-backward'
   bind '"\e[B": history-search-forward'
fi

[brad]

Too bad, I have done so much with my pc last days, I lost the bash history from older than last 2000 commands. Normally I do backup my history from time to time, but haven't done last half year.

bash history is my goto, I backup home directories every 12 hours and have copies going back many years (has been very handy)

I generally use either CTRL R for reverse search or "history | grep <part command>"

@rooted thanks for your suggestion, will have to use this as a more elegant solution

[rooted]

I also used history and grep for many years, I knew there must be a better solution

I mean I still use it if I can only remember some small part of the command.

[mctom]

So I put my hands on atlc, a transmission line simulator that's unmaintained for some 20 years. I couldn't get it to compile, it probably has something against modern gcc. And the development has been halted right after multi-threading feature has been disabled. Oh well, at least GNU parallel saves the day.

Anyway, the point of this exercise was to see if differential pairs could be routed on cheap 2-layer PCBs, such as from JLCPCB or OSH Park. Incidentally both use FR4 substrate with Er = 4.5, so I had to do the analysis only once.
The secret is to add ground copper pours around the differential pair, so it's coupled to them rather than to ground plane on adjacent layer. In fact simulation suggests that neighboring layer is so far away (1.6mm) it doesn't matter to differential pair anymore.

Well, the drawback is that electric fields extent above the board, so it's imperative that there's nothing in or near them and disrupting them..

Here's a gist to recreate my experiments:
https://gist.github.com/tomek-szczesny/ ... 0d08258b47

And attached is render of energy field in a cross section of a differential pair, surrounded by ground pours in close proximity. Funny how energy is actually transferred around the wires, and never through the wires..

[L67GS]

mctom, is punk popular in Poland? We haven't really had a punk scene (barring some inevitable underground somewhere) since the 1980's. I still listen to DK and the Sex pistols occasionally.

[mctom]

Punk had its peak in the 80s and 90s in western port city of Szczecin, as sailors were the most significant importers of western culture from behind the iron curtain.
In 2000s more mainstream pop punk bands appeared, but they all sucked, especially because of no effort towards quality lyrics.
Most punk teenagers from the 2000s era listened to 1980s music, in essence. That was considered "true punk rock".
I didn't hang out with them, because I had and followed my own opinions...

I'm not sure what's going on on the scene right now, I'm old enough to stick to personal favorites from now on. But given that "kids these days" don't pick up real hobbies anymore I'll be surprised to see any good bands in 2030s.

There are just a few Polish bands that I like, and none of them is punk. Mostly Behemoth, and stoner rock groups like Spaceslug, Red Scalp.
From what we loosely and collectively call "punk" these days, I enjoy Bad Religion, NoFX, Rancid, The Offspring, The Interrupters, Frenzal Rhomb, Vandals, Green Day.
And in Ska branch I like Op Ivy, Streetlight Manifesto.
Some consider Psychobilly a branch of Punk, there's a few good bands too, but hailing from the 80s and 90s.
And contemporary, global "folk punk" scene has some two good bands, but the [online] community is mostly a bunch of crackhead beggars. Seriously, almost all posts on folk punk groups is someone asking for money.

[mctom]

Speaking of the differetial pairs again, I was wondering why the industry doesn't make them that way - on a single layer, edge coupled and surrounded by ground. What's the catch?
Maybe the fact the fields extend over the board, and are not contained within it. EMI problems.
Or maybe I am wrong and it won't work like I think it would.

Or maybe I am wrong and in fact this is exactly what the industry does already...

This is my 1Gb switch board, clearly 2 layers and obviously a ton of differential Ethernet signals. And look at that - routed close to each other, and separated by thin ground paths, with multiple vias to a ground plane on the bottom side.
Huh. Guess I got my answer, then.

[mad_ady]

Why is the gigabit router sitting naked on your table? Preparing to get probed?

[mctom]

It's hanging on my cork board with other modules, SBCs and stuff. I bought it for a cluster build inside that military terminal I got some time ago, but that has to wait until some affordable compute modules hit the market. Right now it's near impossible to get any at all.
N2+CM, pretty please?

Spot the Odroid!

[mad_ady]

How nicely oranized your project is! Congrats on the idea!

(xu4q in the bottom right!)

Apparently your project is going to use a mobile SIM and 7-segment displays? That's way more than just a laptop in a fancy case...

[mctom]

Aaand you're wrong, that is XU4 with some random passive heatsink fitted, as I wanted to silence it for grandpa videophone project. The project's completed and dead at the same time, as Grandpa's TV is dying and will be replaced with smart TV...

But yeah, there are some laptop parts over there. Three backlight inverters, my 10A buck converter prototype, FPD-Link ("LVDS") driver, micro-USB battery charger, FT2232, speaker amp... The display's here (that was supposed to be a tachometer for my motorcycle!), alas no input devices in sight.

[mctom]

But hear me out, it took me 15 years but I finally got around to release a demo of my crappy high school band!

https://www.youtube.com/watch?v=ywGLy-bhpdw

[mad_ady]

Aaand you're wrong, that is XU4 with some random passive heatsink fitted

Any xu4 with a passive heatsink becomes a xu4q in my book!

[brad]

An impressive tape plotter (powered by ODROID)
https://www.youtube.com/watch?v=js4_p1S9vIM


An Arduino controls the mechanical parts and an ODROID-C2 (probably) handles user-interface on a TFT LCD.

[rooted]

It's very interesting, I'm surprised by the quality of the label.

[mad_ady]

That is amazing!

[mctom]

I was genuinely pleased that the CD with MRI scan of my cat got a viewer executables for both Windows and Linux. It's a Java app, but still a very nice touch.

[mad_ady]

Yup, my MRI scans were also cross-platform (might be for mac lovers, though)! But they didn't contain cat pics...

[joerg]

I use bash history extensively, with this added to bashrc you can type part of any former command and complete it with up/down (or scroll through previous commands starting with whatever letters you typed)

Code: Select all

if [[ $- == *i* ]]
then
   bind '"\e[A": history-search-backward'
   bind '"\e[B": history-search-forward'
fi

Some days now and I can say that was one of the most helpful hints for me. Thank you, @rooted!

[rooted]

I use bash history extensively, with this added to bashrc you can type part of any former command and complete it with up/down (or scroll through previous commands starting with whatever letters you typed)

Code: Select all

if [[ $- == *i* ]]
then
   bind '"\e[A": history-search-backward'
   bind '"\e[B": history-search-forward'
fi

Some days now and I can say that was one of the most helpful hints for me. Thank you, @rooted!

I'm glad it is useful, you're welcome.

[mad_ady]

Oof... If I had to do this, before any attempt I'd tell the owner the repair is more likely to fail. Just look how many things can go wrong over there. :/
But still it is worth trying, of course.

For glueing screen back, I have no idea. Personally I'd try a thin double-sided adhesive tape. I'd take a piece bigger than a screen and stick it to something very flat like glass. Then use a scalpel to cut the shape of the original tape that I think is in there.

The replacement battery arrived and I replaced it! It was tricky to fit all the stuff back into the case, and I was left over with a small screw that I couldn't fit back because the things wouldn't align... Put the ribbons back and... I could hear a whining noise and the thing got hot really quickly (it did this before changing the battery, when it died) No screen activity, no leds on the back side... Oh, man... I wanted to prove you wrong! (And also save up on buying a new one)... But at least I got the experience points... And had fun (in a masochistic way)...

Here are more details, for those that are into pain: http://adrianpopagh.blogspot.com/2022/0 ... versa.html

Any suggestions for a smartwatch that has better battery replacements?

[mctom]

Sorry to hear the patient's dead, but I agree it was worth trying.

Soo maybe the charging circuit is broken and overcharged the battery, so it got hot. OR anything else is broken and discharged the battery too fast and too deep. That could be tested by measuring the old battery's voltage.
But either way, it's borked I'm afraid.
In the latter case, the voltage converter could have gone into a hiccup mode - so the converter stops for a while if overcurrent condition is detected. That means its frequency will be lower than usual, that's why it went audible. It generates one or two frequencies, though. More peaks were detected because the sound source didn't emit pure sine wave, so there are a few frequency components associated with them.
But then again, if the whining was a hiccup mode trying to protect the device, it did a really poor job - so that's either a design flaw nobody cared about, or it was the power converter itself that got damaged.

In any case, RIP smartwatch.
NOW you can take out the display and figure out how to drive it through SPI or whatever! Screw that 7-segment display

EDIT: And for product recommendations, I dunno.. It took me two years of owning my smartphone to figure out BT doesn't work, so I don't BT at all. My "smartband" I bought for myself was given to my dad.
But if it comes to repair-ability, how about pinetime? At least it's open source..

[mad_ady]

I didn't know PINE went into smartwatches as well . Looks good if I were to use it, but a downgrade for my wife... Thanks for the suggestion, though!

[mctom]

Too bad it doesn't have e-ink, such a wasted potential!
...To build a wearable device that has to generate six auxiliary voltages to refresh it

[rooted]

While not a smart watch I use an Honor Band 6 which is inexpensive, does what I need it to do and has great battery life (around 10 days).