JP1 Remotes

[ryanosman]

I documented a couple things, thought I would post them:

How to fix an RCU810 so it doesn't reset itself sporadically, and how to add a 'hard' backlight key.

http://www.ryanosman.com/rcu810/rcu810.htm

Enjoy, feedback welcome.

[johnsfine]

All these UEI remotes leave reset floating that way, but only the RCU810 has much of that reset problem. Are you sure that connecting the reset signal that way solves it?

I've always assumed (and seen a fair amount of symptomaitic confirmation) that the problem was the sudden dip in battery voltage as the backlight turns on. I didn't look at your backlight change carefully, but I think you're turning on the backlite while the CPU is still hibernating. THAT seems likely to fix the reset problem.

Doesn't tieing reset high your way prevent you from using JP1 with batteries in the remote?

[ryanosman]

First of all, let me set all worries aside - these hacks work. solidly. for about a week now.

My guess is that the other UEI remotes have a resistor somewhere (inside the CPU even perhaps) that pulls the RST high unless someone externally pulls it low to reset the CPU, and that this resistor somehow got overlooked with the RCU810, or at least some of the earlier RCU810s, because they still sell them at radio shack, so I can't imagine they still have this problem.

Next, your guess about sudden drops in voltage: that was definitely my first guess also, so I tacked on a couple more caps (tantalums even) at strategic places that would for sure solve the problem if that's what it was but it didn't stop the problem. Running the remote from a regulated power supply didn't help. Watching the power line on the scope showed that brownout was not the problem.

I next tried adding the caps to the crystal that have positions on the silkscreen but UEI must've later decided weren't necessary. They were right. We normally design in these 'stability' caps also, but personally I think they're only helpful in rare circumstances (extreme cold, or crappy/defective crystals)

Then I noticed the RST pad sitting there. I looked at it on a scope & found it to be very shaky & unstable. Soldering a wire to it, then touching the wire could cause it to reset, so it's obviously not even pulled high with a resistor. So that seemed like a very suspicious problem. I tied it to VDD & instantly the remote is solid as a rock. Normally I suppose one would tie it high with a resistor (that's how we normally design things), but in my personal opinion, I think a wire here is just as good. At any rate (choose either way), holding reset high makes the RCU810 completely stable. And causes no problem like the one you ask about concerning batteries... which I don't understand... I wonder if you perceive that I am suggesting to pull reset low? which would cause the thing to never work at all. Or maybe if you tied RST to V+ before the regulator which would also be bad. well, not as good anyways. maybe bad. Anyways, fear not... it works great. With batteries.

The button hack I suggest toggles the backlight exactly the same way as toggling it with the front side button. All I'm doing is adding another button in parallel with the existing one. Just that mine you can feel in the dark. Both buttons caused occasional resets before the RST mod, and both buttons cause no problems after the RST mod.

Finally one thing to consider. I think the reason we may have experienced that turning on/off the backlight seemed to occasionally coincide with the unit resetting makes total sense after watching the RST line on the scope. The slightest waver in power could send a shiver down the reset line. Pull that baby high & problem goes away.

I hope this helps.

Ryan
http://www.trinity-electronics.com

[johnsfine]

When you use JP1, the reset line is pulled low through a resistor to keep the remote's cpu from interfering with the SDA and SCL signals while the PC is controlling those signals.

If reset is pulled high inside the remote (unless it is done with a much higher KOhm resistor than used in the JP1 cable) the PC can't pull it low.

Any interference on SDA and SCL must be either minor or rare, because several people have used JP1 with JP1's reset connection completely broken. That does mean that the remote never resets at the end of an upload nor when the cable is disconnected. That means that some of what you upload won't be visible to the remote until the next battery reset.

Your tests sound very convincing regarding the reset signal vs. battery problems as the cause of the spurious resets. They contradict a lot of anecdotal eveidence (you say a regulated supply doesn't fix it. Others say just switching to brand new batteries fixes it.)

Unless you can explain something that makes the top part of this reply wrong, I think you should change the advice you're giving others so that they use a high value resistor rather than a wire.

My limited understanding of electronics says a signal must be way below the midpoint of high and low voltage before it will reliably be interpretted as low (which would argue for a very high resistance pull up resistor in case someone is still using 10K in the JP1 cable). But lots of anectdotal evidence (and a little of my own testing with an RCU810) says reset will be reliably interpreted as low (active) when it is around 70% of the CPU's supply voltage. That would imply a wider range of possible pull up resistors would all be OK. Any ideas on that topic (unless you're rejecting the whole idea that the JP1 cable should pull reset low)?

[ryanosman]

Right you are... typical me there forgetting about JP1 I read 'JP1' in your original post and saw 'remote control' sorry... like I said - I didn't understand what the heck you were getting at - not of course it makes total sense... using a wire solves the self-resetting but you do have to remove it to use the JP1. May I suggest 100K resistor - looks to work OK with JP1 and no resetting. Lower values will fight the 10K in the JP1 interface, and higher values should be fine, but the higher they are, the less they contribute to the sporadic resetting.

Thanks for pointing that out - might've been a few weeks before I tried to JP1 my remotes & figured it out.

Okay, I think the modification is ready for the public now?

[ryanosman]

Sorry... stop the presses... the 100K isn't enough to keep the unit from resetting. Now I have to redo the webpage again... I think I'm just going to pull the whole thing off for a week now. Anyways, I guess the solution now is to have a header plug that dead shorts pin 1 & 5 that you put on the jp1 header when you're not programming. So I'm going to try that for a while & if it's all good, then I'll put the page back up.

This method seems good. Short 1&5, unit is stable after pressing buttons repeatedly for a minute, remove jumper, press light button 5 times & it resets.

[johnsfine]

I'm quite surprised that 100K was too high. I'm still pretty sure the other models just float that signal without problems. I can't imagine what might be wrong with that trace on the board and/or that input to the chip that would need lower than 100K to make it stable.

Your idea of an extra header used when not using JP1 is good solution and minimizes changes to the remote, espicially for anyone who already added the JP1 header before deciding to address the reset problem (avoids needing to take the remote apart again). But I think a fixed pull up resistor would be more convenient operationally. If 100K is too high, I think the internal pull up should be 10K and anyone using it should make sure they have a 1K resistor in the reset signal in the JP1 cable. That would cause a 0.55 milliamp drain on 6 volt batteries while the JP1 cable is connected. I don't think that's too serious for normal JP1 use.

[unclemiltie]

The web page mentioned at the beginning doesn't appear to exist anymore. I've got two RCU810's that I like but they do occasionally (and for no reason) reset. It doesn't seem to be battery related.

If there's a fix that someone has figured out, I'd be really happy to hear about it.


thanks

[johnsfine]

It sounded like ryanosman's fix (either version) would work. Either permanently on the PCB or in a JP1 header without cable, connect a pull up resistor to the reset signal.

As mentioned above, if you use the permanent version, make sure the reset resistor in the JP1 cable is a much lower value (for 10K pull up, use 1K in the cable).

Before ryanosman found that approach, other people solved the problem by using more stable (voltage) batteries. Regualr alkaline batteries may only be stable enough when they're very new. Changing batteries when they are only 20% or 30% drained would probably fix the problem, but has obvious disadvantage.

You might try so NiMH rechargeable batteries. They're more expensive than regular rechargeable (much more expensive than disposable) batteries. But they have voltage characteristics that will probably prevent the RCU810 reseting for a much larger fraction of their charge life, and after that you just recharge them anyway.

[ryanosman]

I tried 10K & it still wasn't 100% stable, so I did the wire jumper connected to a header thing. It has been working 100% ever since. I just haven't had time to take pictures & put them back up on the site.

And BTW, I have always used nimh batteries, and I also observed the sporadic reset with the remote hooked to a nice big stable power supply, so I doubt that the problem is battery related.

Personally, my guess is that something inside the remote (either firmware or hardware related - probably hardware - possibly less than perfectly manufactured CPUs) causes the reset line to go to zero occasionally, and pulling it hard high keeps that from happening, and using anything less than a 0k jumper doesn't keep it high hard enough.

But I'm not really interested in debating about this stuff again.

If anyone is interested, I fixed my remotes doing this, so I would suggest anyone having this problem do it to theirs.

Best wishes.

[johnsfine]

It's hard to imagine any explanation other than a serious manufacturing defect in the CPU that would explain needing a lower resistence pull up than 10K, and of course you're right that the extra noise (reaching the reset signal) resulting from weak batteries is not large enough to explain needing so much pull up.

I think there are enough user reports to conclude that there exist plenty of RCU810's that just plain don't have this problem (reset floats reliably high, like most unconnected inputs usually do). I think there is also good evidence from user reports that some RCU810's are sensitive to the noise induced by weak batteries and work just fine with brand new batteries. You've made clear that there is either a wider range or a second cause.

Your extra jumper, added for ordinary use and removed for use of the JP1 cable, seems to be the only solid answer.

[unclemiltie]

ryanosman:

can you please describe the temporary jumper that you put into your RCU810's to make it more stable?

Are we pulling the reset line on the JP1 header up to VCC? is that on the JP1 header? which pins?

I can probably make some 6-pin connectors with a jumper in it so that I can pull it out when I use the JP1 cable.

jonsfine:

I've had this problem on two of the three RCU810's that I have, the newest one is the one that seems to be most stable. I've used both NiMH rechargeables and a variety of alkaline batteries in them and don't find any correlation between the batteries and the random reset.

I like the remote layout so I'd like to keep it but it's a pain in the neck with the resets.


Thanks

[ryanosman]

Jumper pin 1 to pin 5 - one is reset, the other is vcc - your problems will be gone. If you use a removeable plug, then you can take it off to program your rcu using the jp1 interface.

[unclemiltie]

I've been reading the S3C8 processor manual to learn how the extenders work and came across something interesting here.

They recommend that users of the processor set the "watchdog timer" to keep devices from getting too out of control. What this does is that you set the timer and it starts to count down, if it gets to zero it will pull a hard reset on the processor. The object here is to have the software periodically come in and reset the counter so that it never gets to zero.

If the software goes haywire and doesn't get there in time, the device will reset.

now, if the RCU810 has this timing loop somewhat close to the edge, or if the performance slows down with battery voltage, the watchdog could be going off and pulling RESET down to restart the processor. With a pull-up resistor as suggested earlier, the RESET by the watchdog would still do the trick. But with the RESET pin tied high with the jumper it may not make it.

The jumper in both of my RCU810's has made them rock solid for a month and it's no big deal to pull them out when I want to program the remote, so problem solved. But reading the book gave me this thought that the watchdog is what's causing this

[johnsfine]

Sorry but I think your theory is full of holes.
1) They aren't using the watchdog timer at all.
2) Low battery doesn't slow performance (before more serious problems are visible).
3) Some people report the problem even with good batteries.
4) As you report, the jumper on the floating reset line (which would have no effect on the watchdog) fixes the problem.