Cheap Arduino Nano - well, you got what you paid for (and the fix)

[Rick Law]

Well, you do get what you paid for.  But like buying a cheap solder station, the first thing after receiving it is to open it to see what kind of fix it needs.  After that, it may do the job.

If you have a NanoV3 with CH340G looking like this, your Nano may also have the same problem.



Of late, I have been working on something that is kind of volt sensitive.  At first, I was using an UNO and it went pretty well.  I decided to switch to a Nano to get the components closer.  So, I switched over to an Nano V3 with FT232 RL ($6-$7).  Things went pretty well.  Minor adjustments from the UNO.

Doing something else for a while, the Nano with FT232 is now busy.  I have a pair of super cheap Nano with CH340G ($2.88 each), so I switch to one this super cheap Nano.

Well, things went a bit wild occasionally (as I said, volt sensitive stuff I was doing).  Took me a while to figure out it is wild when I use VIN but ok with USB.  I know the same power brick works, so I hooking a scope on the +5V.  The +5V was having a 160mV-170mV noise like this:



The bottom of the board looks like this:



Ah ha, a good challenge - after a bit of tracing and debugging, I see they use a thin trace from the 5V regulator to the 10uF capacitor.  Per spec sheet, it recommends a 20uF.  Looking at the trace, not only is it thin, it snakes around first to the top side and connects to the components, then back down to the bottom and finally connects to is the 10uF and as the last thing.  So, I came up with this addition:



[note: the bright stuff on the regular pin is actually reflected light]
- The ground trace turns right before the AMS1117's Vout tab and the trace goes under the AMS1117.  It is scraped at an appropriate point to provide an addition solder point for ground.
- A 10uF tantalum (spec wants tantalum) is connected from the AMS1117 tab (Vout) to that newly created solder point for ground.

The end result: With the same power brick and environment, the +5V noise now is down to 11mV, exactly the same as the Nano with FT232 and UNO version.



Well, I love this.  Not exactly something a professional should do as time is money, but for me a hobbyist, this is part of the fun.   I like finding and solving problems.  This Nano now works like the FT232RL version of the Nano I got (for this application anyhow).

By the way, since I had a pair of the Nano, I use the loose board for the pictures.  The board that is running is soldered to other things and difficult to photo.

Rick

[Rick Law]

Did you move the existing 10uF cap or add another one to bring it up to 20uF?
Could a single 20uF cap in the existing position have been equally suitable? Or was the new cap positioned there because the long thin trace was the problem?

I think the thin trace is part of the problem.  Shorting the tab to the existing 10uF gave me some relief but did not solve the problem. 

The datasheet suggest a 20uF, so I left the other 10uF there.

Since I replicated the solution for both my Nano, I suppose it is not a bad cap with way below capacity.  The chance of getting two bad caps at the same spot is low.

Your question gets me thinking: Then again, it is a cheap board - the whole thing may be using below grade capacitors so the on-board cap could be the problem as well.  Perhaps replacing the cap may have done the job.

[Mechatrommer]

Well, you do get what you paid for.  But like buying a cheap solder station, the first thing after receiving it is to open it to see what kind of fix it needs.  After that, it may do the job.

ditto! Recently i got a rc receiver module shorted between 2ch output pin.luckily its nondestructive upon turning on.this is not noname china, but got brandname china. China made is not suited for the unskilled on the field. Better lookout for review 1st b4 buying. If i get hit and unskilled in fixing it,i sure mourn like a dying whore.

[Rick Law]

The proper way to do this is to use low ESR caps, not simply use bigger caps. Per USB specifications, the MAX cap size for a USB device is 10uF if there is no in rush current limitations, despite most USB device won't meet this standard. However, for uses such as a Arduino, I think a 4.7uF polymer and several 10/100nF MLCC are more than enough.

The data sheet for AMS1117 recommends 22uF tantalum.

I understand one of tantalum's failure mod is dead-shot and not a good choice for power supply. But I figure, stock caps are tantalum, and all my others are using tantalum, so I went 10UF tantalum.  I left the stock alone (10uF for 1117 and 10uF at USB).  I expected their actual would be below the 10uF nominal.  The one I measured was 6-7uF, so... combining in total, I am probably over 22uF by 10-20%.

Did you move the existing 10uF cap or add another one to bring it up to 20uF?
Could a single 20uF cap in the existing position have been equally suitable? Or was the new cap positioned there because the long thin trace was the problem?

I think the thin trace is part of the problem.  Shorting the tab to the existing 10uF gave me some relief but did not solve the problem. 

The datasheet suggest a 20uF, so I left the other 10uF there.

Since I replicated the solution for both my Nano, I suppose it is not a bad cap with way below capacity.  The chance of getting two bad caps at the same spot is low.

Your question gets me thinking: Then again, it is a cheap board - the whole thing may be using below grade capacitors so the on-board cap could be the problem as well.  Perhaps replacing the cap may have done the job.

Your post got me thinking something I rejected earlier - that perhaps their whole stock of caps were bad - after all, these are cheap boards and could be the "rejects".  So my early tests of swapping caps between them would not have shown anything.

So, as a learning exercise, I re-tried a few different fix and this time with better notes.

1. Added bypass wire to stock 10uF (6-7uF measured).  Partial results, 40-60mV plus noise.
2. Replacing the stock 10uF with new 10uF but using thin trace.  Partial result (lowered noise to 20-60mV)
3. Added bypass wire directly from Vout to new 10uF at stock capacitor replacement (good result like the solution I posted.)

So, the under-performing cap and the snaking thin trace where both part of the problem.  A good cap can make up the difference with low current but I think at higher current the story may be different since the trace is thin.  Add a by-pass wire to the new capacity would make it much better but ugly.

Had I started with just replacing the cap, I would probably found it good enough but unhappy that it is not as clean as my other Nanos.  I like the solution I have now - the capacitor is near the asm1117 regulator as it should -- rather than snaking around the block.  Good learning experience and pleasing result - performs like my other Nanos.

Rick

[westfw]

It's interesting that with such a reputation for copying, there is still such variation.  Here's my Baite Arduino Nano 3.0 w ch340g.
It's got nice beefy looking ceramic caps right next to the regulator, where they belong.  (Baite seems to be one of the actual manufacturers of a lot of the low-cost boards, and they seem to have a pretty good reputation.)

[firepower]

I like the Meduino Nano it has switchable 5V or 3V3 supplies via 2 x LM1117 on the back side with FDTI chip.



http://www.rcgroups.com/forums/showatt.php?attachmentid=4980409

[DmitryL]

The proper way to do this is to use low ESR caps, not simply use bigger caps. Per USB specifications, the MAX cap size for a USB device is 10uF if there is no in rush current limitations, despite most USB device won't meet this standard. However, for uses such as a Arduino, I think a 4.7uF polymer and several 10/100nF MLCC are more than enough.

Yes, using low ESR capacitors on LDO output is a proper way to turn them to oscillators
AFAIK Datasheets mention "tantalum capacitors" for a reason: they have higher ESR that MLCC ones. Datasheets from _decent_ LDO regulators manufacturers specifically mention that using low-ESR capacitors on output is a bad thing.

[Rick Law]

It's interesting that with such a reputation for copying, there is still such variation...

I think (pure speculation) they gone pass that at least for these stuff.

The technology is too well know.  There is not much saving to merely replicate someone's board.  Laying it out by a new College grad or college student cannot take much longer than a day.

China goes ON and OFF about IP and not consistent in application of the law.  Perhaps these manufacturers prefer to have the insurance on NOT getting caught the day the law is ON.  So, instead of copying a PCB, geta student part-timer happy to have the experience and lay out and new one for cheap.

[FrankBuss]

I just got this board (actually 10 of it, very handy for small projects), EUR 2.39 if you buy one, free shipping, crazy:

http://www.ebay.de/itm/272308424805

and looks like they have changed the layout, or it is another variant. This is how the back looks like:



The output capacitor is directly connected to the voltage regulator, but still 10 uF. This is how noise looks like, when powering Vin from external 7 V:



The noise  of my scope itself when shorting the input to GND is about 16 mV, so this looks very good.

The board works in Windows and Linux, no driver installation was required (but I might have installed a driver for Windows for other boards, Linux needs no drivers). Just select "Tools->Board->Arduino Nano" and it works. I tested it with an application which reads the value of a 10k pot and prints it to the serial port and the value was very stable, +/-1. Of course, the usual blinky application worked, too.

I can recommend this. If you add a breadboard, jumper wires, some components like resistors and LEDs, and an USB cable, it is still less than EUR 10, cheaper than a full size Arduino, but good for beginners. And if you pay  EUR 2.57, the header pins are soldered, but this is too expensive

http://www.ebay.de/itm/171068597664

[bingo600]

I like the Meduino Nano it has switchable 5V or 3V3 supplies via 2 x LM1117 on the back side with FDTI chip.

It costs $17 

I'd can get 3 x 5v and  3 x 3.3v for the same price.
No chance i'll get one of these , and what do they do wrt. 16MHz vs 8MHz clocking on the 2 voltages ?

Or i could move a solder bridge (and add a cheap usb-ttl) adapter
http://www.ebay.com/itm/Pro-Mini-Enhancement-3-3V-or-5V-adjustable-16MHz-MEGA328P-Arduino-compatible/230795578198


/Bingo

[macboy]

The issue of badly implemented decoupling on LDOs is not limited to cheap Arduino clones. I have a Sparkfun version of the Bus Pirate (which is OSHW), and on theirs, the power supply can break out into oscillations in a similar way, causing all sorts of fun-to-debug issues. This was a $35 device, not $3.

[tszaboo]

It's interesting that with such a reputation for copying, there is still such variation.  Here's my Baite Arduino Nano 3.0 w ch340g.
It's got nice beefy looking ceramic caps right next to the regulator, where they belong.  (Baite seems to be one of the actual manufacturers of a lot of the low-cost boards, and they seem to have a pretty good reputation.)

Funny story, that is also bad. Regular 1117 type regulators should not be used with ceramic capacitors, because they require a minimum ESR. So that is wrong also.

The issue of badly implemented decoupling on LDOs is not limited to cheap Arduino clones. I have a Sparkfun version of the Bus Pirate (which is OSHW), and on theirs, the power supply can break out into oscillations in a similar way, causing all sorts of fun-to-debug issues. This was a $35 device, not $3.

I dont trust any OSHW project, because these makers/DIYers hackers have trouble connecting devices with a single pin the proper way. I have so many "OMG what have you done" moment when I open schematics.

Bus pirate: MIC5205 with 1 uF ceramic capacitor, datasheet:
" Ultra-low-ESR capacitors can cause a low amplitude oscillation of the output and/or underdamped transient response. " Congratulations, you managed to screw up the hardware three times!, but as a very important part, at least, you have pingvin supported. Good job.
Ah yes, and this is version 3.5 already. surely, you need like 8 different version for a hardware with 3 different ICs on it...

[rrinker]

 Made me look at a batch of cheap ones I got - the kind the come without the pin headers soldered on. Pretty cheap on Amazon. Mine look like Frank's, pretty much identical. I could just as soon scrape off the 4V regulator circuit though as all my projects power them from a 5V source, so I don't expect any issue. Working well so far in the relatively harsh environment of a breadboard with all sorts of noisy peripherals (ie, RC servos) hanging off them.

[Seekonk]

None of my old NANO boards looked like that layout. Then, the one I got in the mail today has that problem.  Something new to check out.