I'm having problems with the accuracy eBay bought RTC's. The best I could get, a DS3231 seems to run slow about 1-2 seconds a day and the DS1302 is even worse than that.
Anyone experiencing similar issues or just a bad batch (I have different ones bough at different times last year)?
One or two seconds a day is not THAT bad for a stand-alone battery-operated clock.
Assuming jremington's calculations are correct, 10-20ppm is about 5 times worse than the spec. So, yeah... Not bad for something from a random eBay supplier...
If it's constantly off, you can put a correction factor in software and add one or two seconds every day, etc.
....Your computer keeps "perfect" time because it's set periodically to match an atomic clock somewhere on the Internet. Same thing with your cell phone. Clocks powered by the AC power line are "perfectly accurate" because the power company keeps the line-frequency very accurate, and even though it's not perfect, they make sure the long-term average frequency is nearly-perfect (so the number of AC cycles in a year is probably correct down to the cycle, assuming there are no power outages).
Battery clocks (motorized or digital) and mechanical clocks don't keep perfect time and have to be re-set once in awhile.
1 second /86400 seconds (per day) = 11.6 parts per million
Yes, many others have had similar experiences. Fake and resold "reject" ICs are extremely common. You might check whether there is a functioning temperature register in the "DS3231".
I paid $9 USD for a DS3234 IC from digikey. You can buy a whole board on ebay for $0.99 and $1.98 shipping. How does that work?
Out of the box mine was 0.5 parts per million fast at room temperature. That's about 50ms/day. By tuning the aging register I was able to reduce the error by a factor of ten.
I paid $9 USD for a DS3234 IC from digikey. You can buy a whole board on ebay for $0.99 and $1.98 shipping. How does that work?
Out of the box mine was 0.5 parts per million fast at room temperature. That's about 50ms/day. By tuning the aging register I was able to reduce the error by a factor of ten.
Thank you. This gives a great comparison.
I don't buy the boards, generally speaking I get the IC and solder it to my own development boards, so I probably get it even cheaper than that, making a huge difference for the list price. Its just a chip, no need for an external crystal as the DS1307.
Regarding external sync on PC/Phone clocks... I've just been checking the DCF-77 radio clock receiver which is just a bit more expensive than a genuine DS2332 and would perhaps make more sense where accuracy is a must.
jboyton:
Very interesting reading. Thanks for that.
But it still doesn't quite answer my question. If I buy 100 cheap ebay chinese boards how many are likely to have counterfeit ICs?
I don't think anyone can give an answer to that question. I can advise on my experience:
I generally get the cheap stuff just to prototype and on my final boards I source components from several suppliers depending how critical they operation is.
Reputable Chinese suppliers such as "UTSOURCE" as just as reliable as Farnell, mouser or digikey without the added nonsense of minimum order qty.
For others most times they are just reselling and some times they have no idea if their products are faulty. I had all of this happening:
Completely fake chip, not working (This is rare)
Some failed components in a batch
Some components don't perform to specs
Used/recycled components
Reliability issues (For example, some SMT ATmega328's work fine but die after programming 30-50 times.
Generally speaking if you're buying shift registers, resistors, relays, led's, etc. at a few pence each they will not be counterfeit, and postage costs may mean you can save 50% or more by ordering well in advance from china. For things such as expensive ADC's, exotic IC's, gate drivers or some kind of Mosfets that are sold at a considerably lower price nearly 80% are knock offs. For example a IRF540 rebranded as a more expensive mosfet or a recycled one, which has been used before. I only get these from reputable sellers.
Cheaper mosfets such as IRF540/IRFZ44 are generally fine
I've found connectors to vary in quality, but generally the price difference more than makes it up. For example, if I know a given connector will need to be inserted/removed several times or being in a 'wet' environment I get the genuine thing. Otherwise the Chinese versions are fine.
DVDdoug:
Clocks powered by the AC power line are "perfectly accurate" because the power company keeps the line-frequency very accurate, and even though it's not perfect, they make sure the long-term average frequency is nearly-perfect (so the number of AC cycles in a year is probably correct down to the cycle, assuming there are no power outages).
Meaning: the hours and minutes are correct (except if there's a power outage), but the seconds might be off slightly. I've seen this myself with a pair of electromechanical clocks I have (I believe they drifted seven seconds within a day or two, then they corrected themselves).
However, some AC-powered clocks (cheap Chinese digital clocks, like you might find at a department store) keep lousy time. Apparently, they rely on their own, internal timing circuit which keeps lousy time. Some cheap Chinese digital clocks, though, do rely on the AC frequency to keep time, and thus are as accurate as an electromechanical clock would be. The only way to find out which kind you have is to plug it in, set the time, and wait.
If I buy 100 cheap ebay chinese boards how many are likely to have counterfeit ICs?
I bought 3 of the chinese DS3231 boards from eBay for about $3.00 each, and checked them against a GPS synchronized time pulse.
All three seem to be genuine DS3231 chips, but out of spec, with room temperature drifts of 3-4 ppm. I assume that these were simply reject parts.
I was able to adjust the drift with the aging register, but have not subjected them to further tests to see if the temperature compensation is actually working.
link=msg=2424803:
Some cheap Chinese digital clocks, though, do rely on the AC frequency to keep time, and thus are as accurate as an electromechanical clock would be.
I can see how an electromechanical clock (with a synchronous motor) can maintain net accuracy (over a 24H period), but a digital one relying on an AC signal surely is prone to errors caused by noise?
casemod:
I can see how an electromechanical clock (with a synchronous motor) can maintain net accuracy (over a 24H period), but a digital one relying on an AC signal surely is prone to errors caused by noise?
I'd say near 100%.
I only buy from reputable sources with traceability - digikey, mouser, avnet, arrow - and a couple parts elsewhere, MAX7219 from taydaelectronics and voltage regulators from dipmicro.
I go by the # of complaints we see about integrated circuits bought on e-bay.
Also, it doesn't seem possible to sell chips for so much less than the distributors, including shipping, when the distributors buy entire reels & boxes of parts for resale, unless illegitimate chips are being purchased for resale on ebay.
I have checked on some of the parts, the MAX7219 for instance, with the OEM, and was told the distributor who sells the chips for 1/3 the price of Digikey, is not an authorized reseller. Pictures of the DIP clearly show the part is not the same as a part obtained direct from Maxim. Pictures of delidded parts also show the part is different. The counterfeit parts have been shown to work in realworld circtuits tho.
Even the US military can be duped into buying fake parts. So surelly there is a risk buying from chinese sellers, but they are not the only source of fakes...
Taking these RTC's as an example. They cost 800% more on farnell than on eBay, so I assume from the beginning I'm not buying the genuine product. But for most stuff it's OK. As long as its compatible (physically and electrically) and accuracy is still within aceptable levels, this is not an issue to me, in fact it gives me more options to choose according to what I'm doing.
As a Hobbyst I'dd say most of my stuff doesnt perform any critical tasks. Same as running an atmel at 16MHz from a 3.3V supply. Surelly is not guaranteed, but I never had any issues with it and in fact I find it more reliable than using level translators, when prototyping (before a good PCB is designed) where noise and poor connections are common.
So all in one. Do they sell fakes? Yes! But that's not necessarilly a bad thing as long as there is a market for them. It all depends on the requirements. But if you're looking to buy the real item at 10% of the retail price expecting it to perform equally well. Well... it will not!