Spontaneous left/right movements in everything

dankcushions

if you search for ibuffalo ghost inputs you will see this is a very common problem. i don't get it when i use my controllers via a powered usb hub, but that solution doesn't seem to work for many others.

nazo

Two things: first, as I said, it's not doing it in Drastic which is pretty big. Second, on my PC it doesn't do it in anything either -- and, again, just to be clear, it doesn't just not do sudden "ghost" movements, but there is no wobble at all, not even by as much as a single pixel. Given that it didn't do it before and it's not doing it in Drastic, I think this may actually be related to something that was updated. I really don't want to have to reinstall everything though...

simonster

Hi @nazo

You said your power supply was over 1 Amp, that is very low for a Pi3 setup. You should be looking at a 5.2V 2.5Amp power supply as a minimum if you have USB peripherals attached. I'm amazed the Pi isn't showing a lightning bolt constantly. Power supplies have been known to trigger these sorts of issues in the past. Is there another power supply you could try?

nazo

Uhm. Over doesn't mean under. It's not under 1A. 5MA is over 1A for example (note the capital...) I was just saying that it was well within a minimum necessary range. I simply didn't feel like unplugging it to read the exact specifications since I was doing stuff on the system when I wrote that and I already knew it was well over anything the system could need in any realistic scenario.

BTW, 2.5A is not necessary under any reasonable use. That's if you're going all out plugging in a lot of high powered devices. It would actually be pretty hard to reach that amount through normal usage. Mine isn't even "normal use." It's just a gamepad and a keyboard. Both utilize the low power USB profiles and likely are closer to 10mA than 100mA (nevermind anything like 500mA.) For the record though, it's a 2.1A power supply that seems to be able to pretty reliably actually provide high currents on the few things that can actually use it. If anything it's being underutilized right now as would a 2.5A adapter (I'll check the voltage later.)

dankcushions

drastic: may have its own way of coping with/ignoring single frame inputs. games often do. emulationstation, kodi and retroarch (rgui) do not - 1 single frame of input is all that is needed.

windows will likely exhibit the same behaviour if you recreate the scenario - e.g. use emulationstation in windows and see if there's movement after an idle period. i believe someone did this test but who knows what thread that was in.

everything you have said so far matches exactly with the 100s of instances of ibuffalo's doing ghost inputs.

Capeman

Phone and tablet chargers are not very good power supplies for micro computers. Charging a phone doesn't necessarily require precise regulated power, many of them are not manufactured with those standards in mind, especially cheap chinese chargers and mid to low range android chargers.

Connect a phone charger to a multimeter and you may notice spikes and drops in voltage, especially when a current load is being drawn.

The pi 3 especially needs a stable and constant supply of power to run properly, especially under processor load. I would recommend switching to the official raspberry pi foundation's regulated power supply and then troubleshooting from there.

nazo

Then why didn't it have the problem before? And no, Windows doesn't have this problem. Besides the fact that it never once moved even a single pixel from the center while I watched it, I actually came up with a really simple mechanism of testing: I set JoyToKey to send letters for each direction (LRUD) and set both deadzones completely down to zero with a textbox highlighted. Nothing. If it picked up even the tiniest movement it would show something.

I'm not saying ghosting is impossible, but signs point to it being something else. Possibly software even. I would like to diagnose this instead of simply writing it off and refusing to try to fix it. (Also, when I search this it sounds like people are getting different results from what I'm seeing including large amounts of ghosting over very short periods of time instead of one random input over large intervals.)

EDIT: I will not respond to "any power supply that isn't the official overpriced one is automatically the cause of any problem you might ever have" style posts. It is a good power supply.

simonster

Because the power usage of a Raspberry Pi is very much linked to activity, it is quite plausible to have different power draw in different installations. Even simple things such as the SD card read/write increase the power load massively. Equally, each emulator actually has its own effect on power draw of the Pi when running, increased higher by using shaders in Retroarch, especially those that apply geometry such as barrel distortion.

I still don't understand why people don't buy the official power adapter, and use a mobile phone adapter expecting it to work. That's like buying a new gaming computer without a PSU because you've got an old 250W PSU in your garage and then wondering why your NVidia graphics card isn't working properly.

The official power adapter is priced at around £8 from stockists in the UK. If don't think this is overpriced, especially compared to the plethora of tablet PSUs which unless you buy a Chinese knock-off one (really, don't do that) are going to set you back far more money.

Back to the issue at hand, if you open an SSH prompt and run a jstest, whilst using the Pi can you see the ghost inputs in the console?

Capeman

@nazo said in Spontaneous left/right movements in everything:

EDIT: I will not respond to "any power supply that isn't the official overpriced one is automatically the cause of any problem you might ever have" style posts. It is a good power supply.

You'll catch more flies with honey than vinegar, my friend. Just remember that when you show a negative attitude toward strangers who waste their free time trying to help you.

mediamogul

@capeman

That's good advice, but I doubt he'll respond to any "vinegar that isn't overpriced honey is automatically the cause of any problem you might ever have" style posts.

nazo

Just to nip this entire derailing in the bud, I tested the power supply's output. I didn't have the patience to wire up a current meter test, but voltage is easy enough. During startup I saw it around 4.99V most of the time, dropping to 4.98V for a split second during heaviest load. Then in the menu it stayed rock-solid on 5.00V most of the time until it had been idling for quite a while at which point it jumped up to a whopping 5.01V fairly stable (still dropping to 5.00V roughly half of the time for a while before it eventually settled on 5.01V, so maybe really closer to 5.005, but my multimeter only goes to the second decimal point.) When the screen dimmed after a long while of idling it jumped up to 5.02V. That's a range of +/- 0.4%. I consider this to be within the 10% most people consider an acceptable tolerance (though my rule is 5%.) The voltage did not budge at all when it finally picked up a "ghost" input.

The power supply is neither underutilized nor insufficient. Which I pretty well already knew. This one has been quite reliable for a long time in a lot of applications. BTW, one thing I think you missed is that it's doing this in the menus. That means while idling. So even if the power supply were unable to keep up during heavy use it wouldn't have the same problem idling in a menu.

Actually, I've had a chance to test it on another RPi3 using a filesystem image. No "ghosting" on the other. I will be comparing the hardware differences between them next.

As for the vinegar, I stated already it wasn't that and I wanted to diagnose the other possibilities. At the very minimum people could have humored me and looked into the other possibilities somewhat instead of repeatedly insisting it was 100% guaranteed to be that. BTW, flies like balsamic vinegar much more than honey.

Darksavior

Beat me to it. I was going to suggest testing different retropie builds. That would prove an update introduced the ghosting.

Capeman

@nazo Actually by testing on another Pi3, you may have inadvertently stumbled onto something. There are 2 different manufacturing facilities that produce raspberry pi boards. One is in the UK and the other is out of China. There have been many discussions about the differences in the quality and it's sometimes almost obvious to tell the difference when you see a board from each facility side by side, I have definitely run across both. Might be worth checking out your boards and seeing if they differ in components (easy to spot with the large parts like USB and hdmi connectors). If the ghosting is related to the board that would be good to know in the future. Kind of an outlandish thought, but this ghosting issue has been been around for a long time and there is literally no definitive solution.

**Edited for accuracy, i got a beer or 2 in me.

nazo

Yeah, something like that may be a possibility. I will be swapping things around and see what happens momentarily. The second one is in a hard to open case though.

@darksavior said in Spontaneous left/right movements in everything:

Beat me to it. I was going to suggest testing different retropie builds.

Here's the funny thing. Same build. It was a filesystem image. I'm going to have to test the two hardware components -- the RPi 3 itself, and the DAC+ addon. The problem is, either one, if it's a 100% hardware issue, is going to be a huge problem for me. The other RPi3 isn't mine and I'm pretty broke atm. The DAC+ I consider non-optional (ask an audiophile about the difference between this and even cheap USB dongles. I have a Fiio E18K that is comparable -- surely better even -- but it's thoroughly dedicated to another task and obviously I can't buy another right now) so if it's it I need to figure out why it's doing it and somehow fix that if it is at all possible without digging into source code or something.

My suspicion is the DAC+. I believe it uses some GPIO functionality. And RetroPie explicitly supports some of the GPIO addons for retro gamepad adapters.

EDIT: The one without the ghosting says "Made in PRC" and has a darker PCB with black plastic on the ribbon connectors. The one I've been using says "Made in UK" and has a lighter PCD with white plastic on the ribbon connectors. I had considered the darker one to be inferior because the TRRS connector has some bit of oxidation or some sort of stain (but I'm not using that analog connection anyway.) I'll test the actual hardware and see what happens.

Capeman

PRC = Peoples Republic of China. Which strikes me as odd, because i always thought the pi's with the black ribbon cable connectors looked to be made to higher quality standards than the white connector versions. I would have expected the UK to be the better product, but the versions with the white connectors sometimes have crooked connectors that just look cheaper to me.

I just checked, the one i use, which was ordered from Element14, appears to be the PRC one also. From what I've read is the ones from E14 were better built.

Strikes me as odd that the china version would be superior, but, i guess every major tech company including apple and samsung manufacture in china, so i guess the quality can be pretty good if the price is right.

nazo

I'm not seeing many physical differences between them to compare. I actually said I thought that one initially looked inferior because it actually had oxidation on the TRRS connector which striked me as meaning not using a very good alloy (typically they're at least nickel plated these days.) However, I don't know what exactly it is because it doesn't really look exactly like normal rust. It's almost more of an inconsistent brown stain. In fact, it sort of looks like the coffee stains on my coffee press. The metal on the TRRS connector on the UK board is clean and shiny.

Anyway, after I'm done with this test I'll post images. Interestingly enough with the DAC+ plugged into the PRC board I'm not getting ghosting, but, I forgot to switch memory cards, so it's not actually initialized or being used at the moment and the system is using the built-in sound. Honestly, there are several variables and it takes a long time to test since all I can do is just let it sit there a while and see what happens, so this is a bit frustrating. What I can say though is that before when I ran jstest --event I saw a lot more coming up before I switched things around, so right now it seems to be consistent with something in this current setup being "right" and something on the previous setup being "wrong."

nazo

Well... You may be right. I'll be... So here's the thing. I remember now that I actually did switch the two boards a while back. I was having troubles with the power connector on the PRC board being looser (though both seem to be pretty bad and prone to momentary power loss from just a very minor bump of the cord, the UK one was a bit better in that respect and thus I went with it.) I didn't think it made any difference since they keep their's velcroed to the back of the TV and I have to keep mine loose and closer to my speaker amplifier. I performed another test, swapping the memory card and leaving the DAC+ in, running everything from the PRC board and zero ghosting -- unlike what I had with the exact same configuration on the UK board. It just never even occurred to me that there could be actual differences between the boards.

So I guess it must have something to do with the USB handling on the UK board. In fact, this might explain how a powered USB hub could make a difference for anyone if maybe it's dropping power too low or something (perhaps it doesn't like devices that simply don't use much. As I said, this gamepad likely uses exceptionally little power. Less than the wireless keyboard/mouse remote thing surely. One of these days I need to rig up an in-line USB current meter. Some day.) EDIT: As far as quality control questions go, I have to say, the UK board looks like they left bare copper exposed to me. Note how dark many things like the run connectors are. They're obviously not gold-plated as with the PRC board, but if it is copper it has oxidized a fair bit (which perhaps could mean it's affecting the actual lines further down if it has spread.) If it's something else I've never really seen a metal of quite that particular color before personally.

I'll do some more testing and give it more time, but so far I should have seen quite a few events show up in jstest by now.

BTW, I got curious and dipped a q-tip in alcohol. Most of the brown stuff came right off with no real scrubbing. Clearly not oxidation. I just never bothered with it before since I consider the analog output on the RPi to be too awful to use in anything less than an emergency.

EDIT: Now the UK board isn't showing the ghosting either. I guess there is a variable here I may be missing. I may have to test both over a MUCH longer period of time...

PS. Before anyone starts, no I'm not overclocking. The heatsinks are just there because I like all of my hardware to last as long as possible. If I still have a use for this RPi3 in ten years, I want it 100% working. Neither has ever been overclocked and if Raspberry Pi themselves checked them the flag will not have been changed (or does it still do a flag? Eh, if it did it would still say zero or whatever.)

nazo

Well, I'm really having a lot of troubles pinning this down. For a really long time the PRC one produced zero ghost inputs. Then both stopped producing ghost inputs. Now both do both with and without the DAC+ addon. I finally found another gamepad to test with (one I hate, unfortunately) and am running jstest with it. So far zero ghost inputs with it. Meanwhile, now that I know about this jstest, I've rebooted into Linux on my PC and am running jstest on it after installing the requisite joystick package. So far zero ghosting on the PC. I went ahead and plugged in the other iBuffalo and have jstest in another "window" (Byobu) running and it too shows zero ghost inputs. No events at all showing up at all yet except both showed a center a few moments after I plugged them in (which is weird because "no movement" isn't supposed to be an event, yet the value was exactly zero -- aka center.) If I'm assuming it really IS the iBuffalo controller despite all the inconsistencies like the fact it never did it before then it should be producing the exact same result on my PC. Or even just similar results. But no, it's showing zero problems on the PC.

At this point I'm pretty well at a loss. Working with the assumption that it is the gamepad until I can figure out some consistency here, what gamepad similar to this IS good without a bad d-pad? None of that 8-way crap especially. (And I'm definitely looking for something similar to this. For me the SNES gamepad always used to be one of the most comfortable and this one seems to be close enough because it feels just great. Not looking for a PS1/PS2 style controller and absolutely will not use a Xbox 1/360/"One" controller, period.) I'm not overflowing with cash, but at the very minimum I need to tell them what to get and unless I can verify this, I guess I shouldn't tell them to get the iBuffalo.

edmaul69

@nazo so i didnt read all of this book but from things i did read. Cellphone chargers do not maintain a constant voltage. The way rapid charge works is it increases and decreases voltage to prevent exploding batteries. The raspberry pi 3's system requirements is listed as 5.1v 2.5a so a cellphone charger will never maintain that. I have only used the white connector versions. I have not had issues as long as i am not trying to power a hard drive from the same power supply. As far as the loose usb connector, the cable can make all the difference. Some are looser than others. Try other cables. A lot of the power supplies made specifically for the pi have a tighter fit.

nazo

@edmaul69 said in Spontaneous left/right movements in everything:

@nazo so i didnt read all of this book but from things i did read.

You should have. This has been covered.

@edmaul69 said in Spontaneous left/right movements in everything:

Cellphone chargers do not maintain a constant voltage. The way rapid charge works is it increases and decreases voltage to prevent exploding batteries.

You took the term "charger" too literally. It's a USB power supply. It outputs a constant 5V. For the record, lithium ion chargers output a constant voltage and vary the current.

@edmaul69 said in Spontaneous left/right movements in everything:

The raspberry pi 3's system requirements is listed as 5.1v 2.5a

5.0V +/- at least 4% tolerance (much more than 0.4,) probably 5% or more up to 2.5A - https://www.raspberrypi.org/help/faqs/#power
Going by official statements from the people who made the thing, it likely uses close to 1A with the way I'm using it. Perhaps peaking around 1.5A or so. This is rounding up. The 4.98V minimum and 5.02 maximum voltages it produced are far within the official specifications by 10 times or more.

EDIT: Ok, I'm going to make the power supply people really happy by finally answering the big question. Or actually unhappy because you are not going to like the results... I have been meaning to rig up a USB current test connection for some time but just never really got around to it. So I guess this was a good occasion to finally stop putting it off, more to indulge my own curiosity about that gamepad than anything else. Anyway, sitting idle in the system menu (where the ghost inputs are most noticeable) it's using about 0.425 amps. Yes, less than 500mA. When I let it sit for a good while (around the sort of time range when ghosting shows up) it drops to around 0.325A (I guess there must be some power saving implemented in the hardware after all, even if only minimally so.) I told you my use case was less than they were discussing. Now, that's not a real test of course. So I fired up the PS1 emulator and ran a couple of games. The absolute highest peak value I could get it up to for a split second was 0.515, but it wouldn't hold that for more than half a second or so. About 99% of the time in PS1 gaming it was holding about 0.482 or so (it wavered a lot going down to 0.47 something and up to 0.49 something, but hovering most around that area.) I don't guess I have anything that makes extensive use of the GPU handy. I thought maybe Kodi could push things a bit more since it uses the CPU and the VPU for video playback (at least it has some minimal acceleration capabilities) but curiously enough it actually went down to about 0.385 -- lower than the main menu was before it idled for a while. I think I don't have a single power supply in my entire house that can't handle my use case scenario.

Now, I couldn't help but to indulge my curiosity about this gamepad. How much does it use? Is it somehow using a lot and pushing USB controllers hard? My PC can output a higher current than the RPi can (it specifically chooses to violate USB standards in that respect so it can charge even high powered phones.) Nope. 0.002A idle usage. Peak usage if I'm mashing buttons as hard and fast as I can can get up to 0.004A for a split second, but most wild button mashing results in 0.003A usage. So, if the problem is in any way power related it's not using enough power since input devices often tend to be closer to 10mA rather than 2mA.

I will not in any way whatsoever address further discussion of the power supply. (I shouldn't have responded to this.)