how stable is padauk mcu ?

[czorgormez]

hello everyone

i am following padauk and cheap mcu topics for a while  and we need a ultra super giga cheap micro at high quantity, around 400k per year. so every cent is important. in the past i developed several codes for elan (em78p153 series) used in high qty. (more than millions) 
and didn't have any problem. they all worked as expected and still working fine after 2-3 years 24/7.   but elan is a little expensive anymore. so i started looking for cheaper alternative. 14 pin 1k otp rom 40 bytes ram is more than enough.  Padauk PMS152-S14  seems perfect match.
it is also pin compatible with elan (no hw change needed) in same package, probably pin compatible with some microchip mcu in same package too.       

so i wonder if anyone test padauk in long term, not in a toy with high volume ? i have to decide and order around 100k mcu next month so any idea or advice could help.

ps: it is used in satellite stb signal switch boxes (called as diseqc multiswitch). every box uses 24-32 chips inside. so even small batch (5000 qty) needs around 100k mcu.     

thanks...

[Yansi]

Jeezus crist what do you need 32 microcontrollers for in a damn SAT/CATV splitter?

[czorgormez]

it is an active device with rf switch matrix not a splitter.

a satellite receiver sends a data signal over lnb cable when  channel is changed (diseqc protocol) so multiswitch knows correct polarity for that channel and mcu decodes this information and arranges pin diode (rf switches) matrix according to the data.
some models has 16 input (every satellite has 4 output 950 to 2150 mhz) and 32 outputs (satellite receiver/stb) 32 output model needs 32 seperate microcontroller for isolation. also decoding multiple signals and controlling switch matrix (there is 512 active switches) by one or two mcu becomes super comlex.

[Scrts]

I worked for a company that did and still does multiswitches. Nobody ever used 32 microcontrollers...
Here's 5x32 multiswitch with one micro:

[czorgormez]

that depends on the structure, i worked with several multiswitch companies and all of them using seperate micro for every output also many chinese low cost multiswitches are same.
i am sure because i reverse engineered many of them. the device on the your photo may be an expensive model (comparing to the china made-low cost units) 
so different markets has different needs. i am eager to know details of your devices if you like to share. 

[Yansi]

Awwww... it is really a matrix switcher?  I gotta have somewhere a likely 32channel one laying somewhere.  I didn't really understand the purpose of the device, just seen the metric craptons of diode switches in there.

But as far s I remember, there was 1 MCU at maximum in it. I can look it up a take a picture.

I am pretty sure you do not need 32 MCUs. 

[brucehoult]

I wish there was a way to follow a thread without having to post to it first. But since I have to ..

Is it really common for electronic components that work on arrival and survive a burn-in to fail in the field, if they are used within specs?

[OwO]

I wish there was a way to follow a thread without having to post to it first. But since I have to ..

Is it really common for electronic components that work on arrival and survive a burn-in to fail in the field, if they are used within specs?

It depends on how you define "common" and what kind of reliability you are targeting. But for a very rough ballpark approximation you can use the 10:1 rule: for every 10 defects that an elaborate burn-in test catches there will be one that it doesn't catch. So if you are aiming for 3 nines of reliability (1 defect per 1000 units delivered) you will need no more than 1% fault rate during burn-in. This all assumes shit is correctly designed of course, as you can easily have time-bomb failure modes (e.g. electromigration, thermal cycling, metallurgic issues, etc).

[Nusa]

I wish there was a way to follow a thread without having to post to it first.

I believe that's what the "Notify" button is for.

[I wanted a rude username]

I believe that's what the "Notify" button is for.

I think I overused mine, as it doesn't seem to notify reliably any more.

Also nice would be an option to unnotify a thread once you've posted in it! (There may be a Simple Machines Forum hack for this but it's not clear if it works.)

[Monkeh]

I believe that's what the "Notify" button is for.

I think I overused mine, as it doesn't seem to notify reliably any more.

Also nice would be an option to unnotify a thread once you've posted in it! (There may be a Simple Machines Forum hack for this but it's not clear if it works.)

Try looking at the same button..

[brucehoult]

I wish there was a way to follow a thread without having to post to it first.

I believe that's what the "Notify" button is for.

That's something different. I tried it on another topic and it doesn't show up in the "Show new replies to your posts" that I usually check -- it sent me an email that got buried in hundreds of other emails.

[I wanted a rude username]

Try looking at the same button..

Notify? Ineffectual if you've posted.

Post once in TEA, be cursed with incessant notifications forever. 

[Monkeh]

Try looking at the same button..

Notify? Ineffectual if you've posted.

Post once in TEA, be cursed with incessant notifications forever. 

Look closer at it.. you may notice two additional characters.

[I wanted a rude username]

It obviously won't be Unnotify unless I've previously clicked on Notify.

Maybe you have the "Turn notification on when you post or reply to a topic" profile option enabled.

[Monkeh]

It obviously won't be Unnotify unless I've previously clicked on Notify.

Maybe you have the "Turn notification on when you post or reply to a topic" profile option enabled.

I do, yes. You wanted to unnotify - that would be how.

[Mechatrommer]

I am pretty sure you do not need 32 MCUs. 

maybe a $50 mcu... 32x $1 mcu is still cheaper, and SW development can be simplified greatly if they are doing repeatable/same task. isolation can also be achieved as already mentioned by OP, and he is already aware of single mcu system and notified us that using single mcu will become super complex for him, or more expensive... since nobody answered spot on to what he asked, i guess nobody have done the test. maybe his best bet is to buy few batch and stress test them himself. i cant make any usefull advice since i'm clueless to what he's building and i dont own padauk programmer/mcu, i'm just an eager viewer with the same interest to know if padauk mcu is reliable or not? not how many should i use them on a board. cheers.

[OwO]

Well if you somehow solve the isolation issue, the figure of merit of the MCU would still be GPIOs per dollar. Large MCUs get expensive very fast but don't offer proportionally more GPIOs. PMS152-S14 has 11 available IOs (after subtracting one for the command interface) and costs 5 cents, so a giant MCU in a 144-pin package would need to cost less than $0.65 to compete. However a long chain of 74hc595 might be another good option to look into.

[brucehoult]

Well if you somehow solve the isolation issue, the figure of merit of the MCU would still be GPIOs per dollar. Large MCUs get expensive very fast but don't offer proportionally more GPIOs. PMS152-S14 has 11 available IOs (after subtracting one for the command interface) and costs 5 cents, so a giant MCU in a 144-pin package would need to cost less than $0.65 to compete. However a long chain of 74hc595 might be another good option to look into.

Shift registers certainly work to expand I/O if you don't need super high switching rates, and I've done this myself, especially for things such as driving 7 segment displays, but you're only getting 8 outputs for each $0.30 or $0.40, so you'd be better off chaining together those PMS152-S14s with most of them simply emulating a shift register :-) Or use a slightly smarter protocol such as sending a 6 or 8 bit bitnumber to set or reset in binary and each PMS152-S14 absorbs the command if the bit number is 0..7, otherwise passes it on with 8 subtracted from it... depends whether you usualyl want to change everything at once, or just one at a time.

[OwO]

74HC595D is ~2.5 cents on LCSC so it is slightly cheaper than the padauk solution But the routing and control software may be a bit more difficult.

[czorgormez]

as i explained above our hardware is ready and proven for years. it is 4 layer and 2 of them is microstrip transmission lines/couplers some planar inductors etc. other 2 layer is a groundplane for that microstrip line. that gorundplanes are mandatory to have high rf isolation above 40db or more. ten 2 ghz signals coming in and switches out to the 24-32 outputs.

so there are very little space left for routing digital tracks/power lines etc. and remember there is 240 to 320 switches in that board. so using one micro sure it is possible but i have to route digital signals all over the board. it may be done in 6 layer board but that would be expensive bcs we use non standart stackup. that microstrip lines need more track to ground space in substrate to compansate manufacturing tolarences. also you should consider that boards are manufacturing in low cost factories not specialized for rf boards. the challenge is making it cheap, reliable and enough isolation- output signal level. otherwise it is super easy to do it many other ways.

for the similar device mentioned above with 9 inputs and 24 outputs selling price is more than 400 euros. dont know the wholesale price or bom cost but dont think it is under 100 euros. our bom cost is for the same model is only 18 usd, you do the math !

we used elan low cost microcontrollers for years without a problem. sure sometimes they fail in programming step but they never failed after. and the numbers are big i mean millions of them. not in a toy working in a device 7/24. i never saw burned-erased-stopped elan mcu in this period exculuding water damage, accicentaly supplied by high voltage etc. but i am sure every mcu would fail in same scenario.  they are very immune to static and other noises.

so padauk is an lottery i never wanted to have. sure ill test them in different harsh conditions but it is still possible to miss something.