Xbee Series 1, Parallax 32400 USB adapter, Roomba 3.3V regulator, XBee fries

tonofsteel

I posted this on the Digi forum, but I thought that maybe using the USB adapter in this way maybe Parallax people might have an idea?:

I am trying to get a serial link going to a Roomba. I got a 3.3V regulator: LD1086V33 and confirmed that 3.3V is coming out of the regulator properly.

I am using the Parallax 32400 board to carry the XBee. I have used this to communicate between a STM8S Discovery board and a computer, both using the carrier. The STM8S discovery XBee was powered from the 3.3V on the discovery board and the computer used only the USB power. I have used this setup for awhile with no problems.

With the regulator circuit connected I plug in the DIN connector to the Roomba and after a couple connect/disconnect cycles the XBee no longer works. Only the Parallax board power LED lights up.

The XBee appears to get burned out from something in this regulator circuit coming from the Roomba, but the Parallax board comes through unharmed.

I am looking for the next step to take. If when plugging/unplugging the circuit would the regulator have a voltage spike that would cause damage to the XBee? The XBee must be very sensitive to something since all other electronics powered off of this 3.3V does not have any problems, only the XBee stops working.

Edit: I tried to reset, nothing happens. I have X-CTU and the serial port shows up fine, but it will not communicate with the XBee.

I read some other peoples projects that used regulators and there was no trouble.

Is it the regulator that I used? I did not see any crazy voltages on the multimeter I was using when connecting/disconnecting. I will look with a scope, but I don't understand why the XBee would fry (2 of them total now) and the parallax board not.

I notice that the Xbee's are now little heaters and the only function they serve is to get hot. I have connected some microcontrollers and sensor boards to the 3.3V and subjected them to the same connect/disconnect punishment and they do not fry. Why do the XBee's have to fry?

WTF?

Edit 2: Additional info: I did not have any serial comm connected, only the 3.3V from the regulator to power the boards. I know the Roomba is 5V and the Xbee is 3.3, the level translator isn't finished so I was just powering the board from the regulator.

I tried every reset/brick fix I could find, and none work.

I see this guy:
http://blog.lucaseckels.com/2009/04/01/xbee-adapter-for-roomba/

used a 317T regulator, so I will add that to the next order as well and start the next design from that. It still does not make ANY sense why the XBee's are getting fried while anything else connected to that EXACT SAME 3.3V works with no trouble.........

Edit 3: Would using a 3.6V zener to clip any high voltage transients help? My current plan is to install a switch between the XBee and regulator so that I can turn the XBee on only after the regulator has stabalized, and off before disconnecting power to the regulator. If this solves the problem then the regulator is doing something strange. This is not an acceptable long term solution. If the problem is isolated to the regulator throwing a strange voltage somewhere then I am hoping using a zener to clip these voltages will be enough to protect the XBee.

The only other thing I can thing of at this moment is that with the filter caps on the regulator the voltage increases/decreases at a slower rate and maybe this transient voltage pattern causes the XBee to fry? Someone at digi needs to read a book on robust product design. (Whatever I am doing may be an edge case, but other products are working fine and do not fail) It is SO frustrating to run into a problem that noone else has had, and have no clear direction on how to solve it. My next digi-key order is going to be XBee modules and regulators. And at the rate I am going a small fortune will need to be spend on XBees.

Edit 4: My final conclusion at this point is that when the DIN connector to the regulator is pulled from the Roomba, the voltage transients from 3.3V to zero. It is slow enough that I can watch the progression on the multi meter. In the steady state with everything connected the XBee will happily show its power light and flash the LED on board. All hell breaks loose only on disconnect/reconnect. I am assuming the board self-initializes on power up, and this initialization would only happen once an acceptable voltage has been reached. On disconnect the internal circuitry tries to compensate for the reducing voltage to the point of frying itself (drawing too much current at some point?)

The other regulators don't have this problem because either they somehow switch to zero quickly or are small enough that by being small they fall to zero quickly.

I guess I am going to start digging for any documenation from Digi on power supply design recommendations for these. I know there are smart power management IC's that will allow for detection and control of power supply conditions to stage the power up of devices on board. If I was to build a product with XBee's I would be seriously concerned about the reliability and robustness of the product when using simple voltage regulators, it took a few cycles with my setup to fry them, a regulator with more favorable conditions may appear to work, but after 100 cycles the XBee failure may show up.

The reason other 3.3V supplied boards are not failing is because they have been tested or designed for being supplied from a simple regulator and the voltage transients are handled correctly.

To prove/disprove this theory I am going to measure the dv/dt with a working XBee load connected to my current setup, and the previous setup that worked with no problems for me. I can then use a switch (dv/dt very high) and add in capacitors (dv/dt approaching my Roomba setup) and see at what point failure occurs if any. (Maybe power up sequence is really what is doing it in)

Maybe there is a doc somewhere I am going to find that outlines this mode of failure and these tests will be unnecessary.

The only other thing I can test for is as I mentioned before is over-voltage conditions on connect/disconnect. I can't see this being an issue though because all other 3.3V dev boards connected (including the parallax FTDI adapter that the Xbee is riding on) do not show any problems.

*Someone* else has to have run into this though I suspect with all the prototyping going on?

tonofsteel

Pulled out the oscilloscope and did some further testing. The waveforms attached

I am guessing this 8V peak taking 5ms to finally settle to 3.3 is the root of my problems?

Edit: Given this regulator outputs this, what are the options to control/protect from this? I am thinking:
- A zener (>1.5A since this is what the regulator is rated at) to clip anything above 4VDC?
- Some sort of power management IC that only connects the load after voltage has stabliized for a period of time?
- Add capacitance until the start-up conditions can no longer charge to 8V before settling?

I am going to try other regulators and perform the same test/measurement. I am sure others have to perform differently and may have internal circuits to prevent the above from happening. I am going to try the regulators used in these (I would like to find a linear one for testing though rather than add complexity of switching):
http://blog.lucaseckels.com/2009/04/01/xbee-adapter-for-roomba/
http://www.digi.com/support/productdetail?pid=3257

Anyone run into regulator issues like this?

Chris Savage

Should this thread be marked solved or at least unmarked as unsolved?

With high voltage spikes like that on the power supply rails you can bet the XBee Module is probably latching up and going into thermal runaway (heating up).

tonofsteel

Chris Savage wrote: »

Should this thread be marked solved or at least unmarked as unsolved?

With high voltage spikes like that on the power supply rails you can bet the XBee Module is probably latching up and going into thermal runaway (heating up).

It is solved. If you are using an xbee they do not have any tolerance at all to any over voltage. Ensure that your regulator start up does not have any glitches that will result in voltage spikes. I am sure this is not a problem with 5v to 3.3 volt but for my scenario regulating 23v down to 3.3 I ended up with 8v spike. A soft start circuit has been suggested but I am currently using a zener to clip the overvoltage. I also have obtained some switch mode converters which look promising in terms of avoiding this voltage spike and reducing power consumption.

Chris Savage

Yeah, XBee Modules do not have any protection diodes on the I/O pins either making them not tolerant to 5V signals even using a resistor.

tonofsteel

Chris Savage wrote: »

Yeah, XBee Modules do not have any protection diodes on the I/O pins either making them not tolerant to 5V signals even using a resistor.

I had it on the parallax 5v/3.3v adapter for all the above so logic was never 5v. The micro controller boards I have been using (stm8 stm32 pic32) do not fry from this startup voltage spike. The xbee would die every power up while all other electronics connected to same supply would not.

This would worry me to use the xbee in a product because if it is that sensitive how do you protect against real world emi esd etc. Protection methods will clamp/control it but they are real world devices so some well get by. I am confident the micros etc will survive but not so much the xbee.

Chris Savage

It sounds like you use your own regulator so it may be something about that regulator or the design?

tonofsteel

Chris Savage wrote: »

It sounds like you use your own regulator so it may be something about that regulator or the design?

I showed the scope traces and asked this question on a few other forums and asked others I know. I tried several LDO regulators from several manufacturers and they all showed the same response. No one ever looked into this or notices LDO startup behavior because if you connect virtually anything other than an xbee you would not have trouble thus never hook up a scope. Also I am regulating down from 23v so most people also are doing the same circuit but with 12v or lower so the same ldo will not spike as much or have the same response.

We went through the data sheets and my design and could not find anything that was out of spec or wrong. From what I gather most ldo designs are put together from the data sheet, connected to power and if the output is regulating you connect the rest, I heard almost no one will connect a scope to check the startup like in my captures just for the fun of it if it appears to be working correctly.

It is not a problem with the circuit or regulator since if you went to the lab and tried it you would see the same thing, however most robustly designed products would not Smile out from a single 1ms spike. The discussions all headed towards using a soft start since the xbee is sensitive.

No one that I know of was willing to connect up the same circuit in there lab, it would be interesting to know if other real world testing from some one else could provide some further insight. I would really like to know either what I am doing wrong or if this is just how a non soft start ldo works.

Chris Savage

I didn't catch the 23V input voltage. I can take a look at the regulators we use and see what we typically experience at startup, but I don't expect it to be quite that high.

tonofsteel

Chris Savage wrote: »

I didn't catch the 23V input voltage. I can take a look at the regulators we use and see what we typically experience at startup, but I don't expect it to be quite that high.

If you could try this out and see what happens for you and any conclusions you may have it would be greatly appreciated. I have heard a lot of speculative answers but nothing where anyone replicated the problem.

One conclusion was that the regulator cannot handle the high slew rate of connecting it directly to a 23V source. Since it is a NiMH battery it would not be "soft" like a wall wart or some other lower powered "pre-regulator" and will hammer the full voltage where as other supplies would have to react to the load being connected and the transient response would be different enough to not cause this spike to occur. I am currently using a LM317 and the Linear LT3080 has been suggested as a superior part that may react differently.

Tracy Allen

I'd be very surprised to see a transient like that at start-up from a regulator, no matter what the input voltage so long as it is within spec. And I do quite often look at things like that on the 'scope. I'd really want to know if there is an endemic problem of that sort. Nevertheless, in your case, or if I saw it in one of my own circuits I'd immediately suspect a circuit design issue in choice of capacitors, or in the layout, or even more likely some effect of Murphy's Law, a cold solder joint on a ground lead or something of the sort. It is particularly puzzling since you tried several regulators from different manufacturers with the same results.

Do you have a photo and/or schematic of the circuit that you could share?

Chris Savage

I agree with Tracy. It is unusual, though I hadn't tried such a high input voltage before, so I thought I would see how our Board of Education Regulator handled various input voltages at the output. This board has a National Semiconductor LM2940-5.0 5V regulator eith a 1000uF capacitor on the output. We use 1000uF because the board has servo headers, but the minimum output capacitor required for this regulator to be stable is 22uF. Anyway, here's my scope shots at 1ms / 1V per division. The first screenshot is with 12V externally switched into the regulator. I chose 12V because this is the highest power supply we sell to plug into the board. My bench supply doesn't go up to 23V, however I did try 20V and that is shown in the second image. These screenshots are pretty much what I expected to see.

P.S. - The reason the trace is not at 0V initially is the 1000uF capacitor holds a charge for a long time.

Chris Savage

I should have read your original message in more detail. I see you're using the 1086 regulator which I would recommend against. We had/have a board that uses them and initially when it was designed there were no known issues, however later we had to rework the boards to solve a terrible oscillation issue and we don't use that regulator in new designs anymore. :nerd: As I recall not only does the minimum capacitance have to met on that regulator, but you really need Tantalum caps for ESR considerations.

tonofsteel

Chris Savage wrote: »

I should have read your original message in more detail. I see you're using the 1086 regulator which I would recommend against. We had/have a board that uses them and initially when it was designed there were no known issues, however later we had to rework the boards to solve a terrible oscillation issue and we don't use that regulator in new designs anymore. :nerd: As I recall not only does the minimum capacitance have to met on that regulator, but you really need Tantalum caps for ESR considerations.

Thank you Chris for building up a test circuit and Tracy for your suggestions. The circuit is the same as what is in the data sheet for the 317. I don't have the circuit handy to take a picture but it is a typical breadboard version with jumper wires (longer than they need to be).

One thing I notice is the voltage slew rate on your version vs. what I have. Mine goes straight up, where yours has a much slower rise time. I don't remember all the different tests I did with changing out capacitors but I know I added/removed different amounts to see if there were any changes. I did not have anywhere near 1000uF though.

At this point I think I am going to tear everything apart and start from scratch with a new breadboard and new parts and trim the wires to a more appropriate length. I am going to try with a bench power supply vs. a battery and see if the response is different. I did not have any issues with regulators when connecting them to a power supply, and most projects use either a bench or wall wart supply that already drops the voltage to 5 or 9 volts. Only when I use a battery do I notice this. I am going to add a very small resistance(s) to see how changing this rise time might affect the circuit.

Is suspecting a battery vs power supply as being a potential cause nonsense? Do you have any hobby battery packs handy that you could connect to see if they respond any differently?

The 1086 was the first regulator that I tried and I swapped it out. Good to know to avoid this part from your experience with it. The LM317 is used for the current revision I have on the board. I think I did not have any caps that met the ESR requirements at the time. I am going to have to put the LM2940 in my next parts order to see if it works differently than the current regulators I have.

Chris Savage

The LM317 is a variable voltage regulator so typically you would use a resistor on the GND/ADJ pin to adjust the voltage output. While you can do this with some other regulators, the 1086 was not designed to be used in this manner and that could be a contributing factor.

As for low slew rate on my board, I am using a 1000uF capacitor. At power-up the capacitor is almost a short circuit on the bus while it charges initially. At 1000uF it stands to reason it would take longer to come up. I could do the same test with the minimum 22uF capacitance and I would expect the slew rate to be higher and the rail to come up to voltage much faster.

On many regulators Aluminum Electrolytic caps work just fine in place of Tantalum caps. Just don't use ceramic caps on the output, and make sure the capacitor is the minimum recommended output capacitance and type, especially for ESR.

Yeah, sorry about the 1086, that one I will never use again.

Tracy Allen

Robert Pease wrote and nice little book, Troubleshooting Analog Circuits, which is chock full of advice and anecdotes. Bob designed or had a hand in many of the voltage regulators sold by National Semiconductor, and he was known as the "king of the bandgap". Anyway, his book has quite a lot to say about voltage regulators.

Choice of capacitors. The output of voltage regulators has what amounts to an inductive output, that is, its output impedance increases with frequency and that is accompanied by phase shift. A low ESR capacitor at the output can and will resonate with that inductance. At worst a bad choice causes wild oscillations and overshoots. (As in an overshoot to 8 volts when turning on) In a more benign form it causes a predictable increase in noise level in the output at the resonant frequency. A (slightly) higher ESR is necessary to dampen the oscillations. The book has a whole appendix on the output inductance and choice of capacitor for the venerable LM317. More modern designs arebetter at handling low ESR capacitors, still though, ceramics with their own voltage dependence can cause weird oscillations.

One anomaly Bob mentions in the book can happen when a voltage regulator is driven by an inadequately filtered transformer supply, and then the output of the regulator suddenly drives a short circuit to the point of foldback current limiting. The shutdown of current demand causes the transformer to generate a spike of voltage and poof goes the regulator. The solution was more input capacitance. That does not have anything to do with your present issue, just a reminder to be alert.

An RC battery pack would probably have very low output impedance, but the voltage regulator should be fine with it, and it would only indirectly qualify as a "potential cause of the nonsense".

Chris Savage

I can attest to the choice of capacitor being a huge factor in modern LDO regulators. You could get by without on on old-school regulators like the 7805 under ideal circumstances. But the modern LDO regulators always have a minimum output capacitance value. And that specification often includes a certain recommended type of capacitor for ESR reasons. Sometimes when you don't meet the recommended specs you experience unintended results.