Reprap Arduino Due Shield project - looking for helping eyes.

Cefiar Wrote:
-------------------------------------------------------
> bobc: I'm assuming that the 2x4 connector up in
> the top right corner is for 4 thermistors?

Yes.

> Something we might want to consider here is
> putting a 1N746A (3.3V, 0.5W Zener diode) across
> each. This will limit the voltage into the Due to
> a max of 3.3v, if the hot end voltage comes in
> contact with the thermistor. I just checked, and
> the diodes are about 4 cents AUD each. To protect
> the Zener & the Due from long-term over-voltage,
> we should probably put a 100-150mA fuse (0 ohm
> resistor style - I've seen these for about 35
> cents AUD) in series with the thermistor (between
> the Due input/pull-up resistor/filter cap end and
> the thermistor input). This way, if the thermistor
> input does short to a >3.3V source somehow, the 0
> ohm fuse will blow if the current dissipated by
> the Zener is too high, protecting the Due.

I was also thinking about how to protect the thermistor inputs, both for ESD and excess voltages. This is really not my field, so I'm happy to take advice. I found various things on the web e.g. [www.thebox.myzen.co.uk] but I like the idea of adding PTC fuse. I was thinking something like MF-MSMF030-2

As well or instead of zener, I am thinking of BAV99, which has fast response time. It seems like a good idea to add some series resistance and a capacitor as well, not sure how all this will fit on though

Last time I looked at thermocouples, the interface ics were expensive and not necessarily more accurate. However, the new MAX31855 looks cheaper and better, worth considering I think.

NoobMan Wrote:
-------------------------------------------------------
> So please do it like the arduino guys and rest of
> world did it: give fair warning - and cross your
> fingers, hope for the best.
> And thats it. If you think it through all in all,
> thats a very good solution, really

Well, now you have hit a pet peeve of mine. Compared to the professional stuff I work with, the Arduino/Reprap stuff is pretty amateur and at times laughable. Simplicity is all very well, but it translates to people blowing up boards and generally unreliable operation.

I can see how it happened, amateurs who don't really know any better cobble together stuff from standard components and hack some software and it all "sorta works". Then they spend forever trying to fix niggling little issues by stirring the pot, but never really advancing.

So I think the "Arduino way" is very amateur and for a small amount of effort can be easily improved on, I can't take "it's ok to be as bad as everything else and its impossible to improve" as very useful advice. RepRap is supposed to be about innovation, it truly will get stuck in a rut of mediocre implementations if people follow your advice.

I know it can be improved, because the products my company make have all inputs protected to +/- 24V, and we have a tight cost and size budget.

Ok, rant over

Not something good to say, so i started with a sorry line, but thats how i see it. Perhaps i was too radical, or too pictorial, but well if i had social skills i would be with some top models now instead of posting about electronics. And please, there is no need to call names and adjectives and such. I suppose we are friends here, and better than that. If am i wrong about that, tell me and i will stop posting in your thread.

The zenner you need to clamp that must have a power rating of 12-3.3v = 8.7V @ ~40 amps = 348 Watts - not 0.5w. Because the hotend power comes from psu directly and if it shorts to thermistor leg it gets full psu power, nothing standing in between.

Sry but lets not mix overvoltage with overcurrent protection (ovp vs ocp). Or just me thinks are different.

Seriously 3.3v zenner 0.5w cant clamp a 12v, ~40amps atx psu across its terminals. This was for thermistor short to hotend block with v+, so its overcurrent. Serious solution i can think of, what some psu have is electronic fuse circuit on output, when output shorts it cuts the line, that is a low side mosfet with a big heatsink the kind there is no place onto the shield. Still doable coz it has fastest reaction time, else would need impossible heatsink for any longer. And that introduces at least one junction drop, and has its caveats, like you need to put that before the ramps board, not on it.

If 12V from psu and hotend shorts though thermistor leg and comes to the board, i can not think of something else that is gonna save it completelly. Perhaps other can and just i cant.

My point was like this. You have some current you dont want, you try to 1) dissipate it or 2) sink it to gnd.

1) You can not dissipate that 348W in realistic terms.
2) Yes one protect one pin with for that would need to redirect the current path so it doesnt go into the pin, but redirect ... where? Then it will sink some other path and something else will have to burn, thermistors leg, pcb tracks, and thats not really an overall protection is it.

So board doesnt get saved, not much. Something will burn, something will fail.

Sorry for that and please dont make it look like it is my fault.

lol

To be correct further about 1) i said in text you could attempt that trying to make sure you deal with it for at least 100-500ms or some time period that will surely blow a fuse, so can be underrated. But still like probably a power device in a pakage of at least to-220, and a hefty heatsink that will make sure the junction temperature doesnt reach 175 sooner than the ~100-500ms or that certain time period. To ensure the fuse will blow first, before the power device will fail. Still as big as an electronic fuse for that rating, not much different, except it doesnt imply a junction drop, thats only real difference between clamp + series fuse or electronic fuse which is say in its simplest a low side mosfet with a gate controlled so that can raise faster than its drain raises compared to sink.

But thats still big, and again not enough place on the board. And to do something like that, i would rather do it at board's power input, (like i did with mine), rather than making it further in the board. After all, probably more boards fried because of reverse voltage at input than boards fried coz of thermistors shorting to v+.

And just to clarify even further, i am all in for protection. And all in for extra features. Actually I did already protected my endstop pins with diodes, because of how old enstops were - coz i think mixups can happen. So again im all in for that. But on thermistors adc pins, nothing i could think to help them. Its adc feed to its own logic voltage so no 5v worries. And in case of that short to psu, wont help either. Its perhaps not impossible to protect those adc pins from 12v full psu power, could be done, perhaps not on the shield itself but nothing is impossible, but however i think its well over the hand - too much trouble for its worh. At least for me. Ofc anybody else can differ, and actually should, i think thats the beauty of things here.

I will end with first thing i said: sry to "defuse" you guys about that. And i will take my leave of absence now.

Edited 4 time(s). Last edit at 04/27/2013 08:36AM by NoobMan.

NoobMan Wrote:
-------------------------------------------------------
> So board doesnt get saved, not much. Something
> will burn, something will fail.

Obviously, that is the point of a fuse. Its a sacrificial component, a cheap thing designed to protect the expensive things.

A PTC "fuse" works by increasing resistance according to current, so it won't be conducting 40A. Dissipating 348W is obviously absurd, that is a straw man. Tbh, I don't think you have really grasped what the goals and methods are here.

Overvoltage, ESD protection etc are standard stuff for electronics designers, any commercial product is required to have them for CE and safety standards. I'm not an electronics designer, but I work with them. We have guys who spend all day in the lab running tests, including blowing things up, eventually they pass and we can put the CE sticker on them. Well, that is quite a simplification but that's what it looks like

Bottom line, i see choices as:
*without introducing a junction means hard clamping, hard/impossible/not woth it, imo
*introduce a series junction, means that changes resistor divider behavior, changes thermistor tables and math, and NOT in a good way: makes them less reliable and etc, makes calibration mandatory, and this is an aspect that is already poor about reprap. Pullup should of been 560ohms not 4k7, thats the sort of change this area needs, but extra junction i dont like. I am more conservative perhaps but this is so you see where im coming from, and why. Thermistor part is weak, too much to explain newcomers hard to take in and i did it quite nr of time right on this forum. If it gets any worse wont help reprap cause at all. And risk further aggravation for many, just for a eventuality mishap, i dont think its worth it.
Well this is just me and my opinion ofc.
This time im really gone

aha, i see.
Classic flamable fuse gets its wire red hot and it melts. Thats slow and the ic will burn before that happens.
Ptc fuses arent really fuses, the name is just confusing. The current passing through has to change molecular structure inside, thats even slower. Ptc usually work at 150% rating for 2 hours or 150% rating for 30 mins, thats their role, fit a place where a fuse cannot be allowed to blow, like your airplane or life keeping devices. But in reaction time not woth the name of fuse.
Now you see these arent the things fast enough for this. Only a schottky diode fast reverse time of 5ns is fast to do it, or other complicated electronic fuse circuits perhaps or hardcore clamps. These have a chance at least. But classic fuse or ptc, i dont thik of them as really options for uC or any IC protection.

Yes PTCs don't limit the current, other than the fact they have some small resistance. They have to heat up and then their resistance gets higher but that takes time. I have also known them to catch fire.

A small series resistor after the pullup can be used to limit the current without affecting the reading. Clamping it to 3.3V without affecting the reading is an issue though. Better to use a lower full scale ADC value like 1.5V so there is headroom for clamping.

---

[www.hydraraptor.blogspot.com]

NoobMan Wrote:
-------------------------------------------------------
> Now you see these arent the things fast enough for
> this. Only a schottky diode fast reverse time of
> 5ns is fast to do it, or other complicated
> electronic fuse circuits perhaps or hardcore
> clamps. These have a chance at least. But classic
> fuse or ptc, i dont thik of them as really options
> for uC or any IC protection.

You can get PTCs that trip in ms, so it is clear you have absolutely no idea what you are talking about, making your advice quite useless.

If you don't think the design is worth improving, I would ask you to respect your pledge to not contribute further. You also might consider it helpful to readers to fix your spelling and grammar before posting, your posts are a mess.

Ok, here is a proposed design I am toying with, there may be better component choices.

The zener is a fast acting type at 3.6V. The "fuse" is a PTC with trip time 0.2s at 290mA, hold current is 120mA. Cost per input easily less than a $1.

To be clear on the goals, the circuit should protect against transient ESD spikes without damage. In case of short to 12V, the circuit should protect against damage to Arduino at the cost of damage to the protection circuit.

I will run the design past my colleagues on Monday, they seem to find these little challenges entertaining

Edited 2 time(s). Last edit at 04/28/2013 06:40AM by bobc.

A 3.6V Zener will probably leak too much at 3.3V compared to a 100K thermistor so low temp readings will be inaccurate.

If the thermistor is shorted to 12V D502 will conduct and put 12V on the 3.3V rail destroying the MCU.

Can't see the point of the PTC. 290mA into a CMOS input for 0.2s will blow the crap out it.

---

[www.hydraraptor.blogspot.com]

This is maybe what I would do.



I would look to see if I could increase the series resistor enough to not need the PTC without the ADC input leakage current creating a significant error. Don't have time to look at the data sheet right now.

Edited 2 time(s). Last edit at 04/28/2013 08:00AM by nophead.

---

[www.hydraraptor.blogspot.com]

Silicon is ("mostly") EM speed on a microscopic waffer. If thats to be protected, somesthing faster would be awesome, but we dont have that yet so we are limited at using something that is at least equally fast and try do our best. So by necesity silicon is required to protect silicon. Nothing else works. If there are fuses and other devices those will be subordinated, meaning silicon will trigger and control them. If there is a fuse it only works coz there is silicon somewhere that buys it time to blow it on purpose. But the fuse alone cant work. Ptc or others even less so. Because all these are based on effects of the current passing through, that means current has to pass through first for the effect to take place. So by definition of their causality effect, they are too slow. Miliseconds range, tens or hundreads are for fuses. The silicon time domain is in nanoseconds scale. That means in the same time a flameable fuse blows, IC may blow a thousand times.

Its "mostly" EM and not fully so, only mostly because of parasitic capacitance associated to silicon junction and also dissipation requirements. This is why processors had to go for lower voltages, now are stuck to smth like 4-5 Ghz freq and we have multicores instead of single core. If we had technology to make switches with no parasitic effects and zero losses then we would all have truly speed of light processors that would be single core and would never need to be upgraded.

IMHO, my thoughts are zenner alone may be too weak, so i would rather use it in classical fashion in series with 100k resistor to gnd, so reverse biased zener will drive the base of a power transistor / igbt either way probably darlington config for gain speed, and would have the collector/emitter would do the effective clamping. Sry cant draw, im away and posting from phone. The cheap transistor / expensive igbt choice would probably be tradeoff speed vs leakage. And probably a to-220 package because less footprint. And that would have wide track, direct and isolated gnd path to gnd input as in priority star gnd routing, not a 16 mil track with 2 microvias. And also 2 fuses, one on each thermistor leg coz afterall can short either one.

Or, better to bring in a game changer. Perhaps can go with a high side driver with some current control and an inhibit signal that can be used to turn it off instead.

But still, honestly i wouldnt do either. Too many complications, too many components, increases the shield too much. And for what, just for the eventuality of a short.

It doesn't need to be that complicated because a small series resistor limits the current to something that can easily be handled by a zener diode long enough to open a PTC, or maybe indefinitely. Since it is after the pullup it doesn't affect the reading significantly.

Often a series resistor is all you need as the MCU inputs will have clamping diodes to VDD so as long as you limit the current they survive. And when speed isn't critical a small cap to ground behind the resistor gives ESD protection as well.

---

[www.hydraraptor.blogspot.com]

i am trying to learn what i can from this thread, i am trying to understand how your circuit works nophead, though its late here and i am tired.

its just that i think i have seen more than a few threads with things that have gone wrong that could be prevented, i think that some things have come to the point where we need to think more about stuff like this.

i gotta laugh at my self some days i built 2 working repraps, but i am still amazed when i get LEDs to blink. also kinda ironic considering from what i remember i was one of the earlier people to post about arduino mega and the idea to make a board for it. i was glad to see that much better people than me got it to where it is.

---

[mike-mack.blogspot.com]

I see how it works, but if you're comparing it to the circuit from bobc, be aware that it's backwards. In nop's diagram, the Arduino is on the left labelled MPU, and the thermistor is on the right labelled "Therm". In bobc's diagram, the thermistor input is on the left, and the Arduino input is on the right labelled "Therm3"

With nop's design, if the thermistor input goes to the +V rail , you will get current flowing through the 4k7 to the 3.3v rail , which doesn't kill anything as the resistor has (+V - 3.3v) across it, so the appropriate amount of current flows through the 4k7 for that voltage. The current then passes through the 22R (which limits any current to a reasonable max value), plus the PTC (which, once it actually trips, reduces it even further).

After that, you have two diodes. One reversed biased (in parallel with the cap), and one forward biased (to the zener). The reverse biased diode acts to protect if someone connects a -ve voltage to the thermistor somehow. The 1k/3.3V zener produces a nice stable 3.3v rail (independent of the the Arduino regulator, which in turn is less likely to be fried). If the voltage gets over 3.3v + the voltage bias of the forward biased diode, it conducts and effectively clamps the rail at that combined voltage. The fact that the forward biased diode has to conduct means that the zener's inbuilt current passing at voltages lower than its trip voltage is effectively nullified.

The cap performs the same function as in the traditional circuit, which is just to smooth out the input voltage. By placing the cap after the two diodes, the voltage that appears across the cap should be limited. This stops any stray voltages on the thermistor input charging the cap up, which then would discharge into the Arduino, even if the voltage is removed from the thermistor input.

In normal operation, the voltage going through the voltage divider created by the 4.7k resistor and the thermistor is always less than 3.3v. The 22R and PTC aren't part of the voltage divider, so the voltage at the voltage divider should be pretty much unchanged at the Arduino input. How much it will actually change by would be determined by how the Arduino ADC is internally wired.

It might be worth dropping the zener to a slightly lower voltage (eg: 3v), due to the added difference caused by the forward biased diode.

Note: IMO you could probably use one zener/resistor for all the inputs, which reduces the component count slightly. If someone shorts all their thermistors to the hot end +V line at once, then they're probably well out of the league of our ability to offer protection.

PS: Regarding the change of the voltage divider from 4.7k: IMO 560R is too low a resistance, while 1k is reasonable. Thermistor tables for 1k are already in firmwares like Marlin, so it makes sense to use the same value. From memory, 1k also worked out to produce more reliable results from 180 up to 260 Deg C, while 4.7k seemed to have a nice sweet spot from 50-150 (which is perfect for a heated bed). Can't remember whose blog post it was, though I suspect it may have been nophead.

PPS: As mentioned in my last post, I'm not really happy with PTC's either as, like nop says, they can do all sorts of weird things, including catch fire if you try and pull too much current through them. I've seen this first hand about 3 times now, which is about 4 times more than I ever wanted to.

Edited 1 time(s). Last edit at 04/29/2013 07:46AM by Cefiar.

The ptc and 22 ohms resistor are part of divider, and thats the problem, like anything else that introduces a voltage drop, it limits the lower point of adc reads, and because thermistor is lower side, lower voltages are where temperatures are read. Thats why i didnt like to change that.

Let me make a scenario. Ppls copy paste tables, someone copies a not shifted table, that means system gets a maximum value after which all higher temps wont produce higher readings coz voltage cant decrease any lower. For examples this with a non-shifted table will read a max temp of say 250C and when temp will be 350, it will still be read as 250C. Why am i only one thinking thats dangerous. Take as example last guy posting about themistor tables right on this forum, he just resorted at trying different tables in firmware and chose one that seemed to fit. Next one doing that will get a roof temp that masks higher temps. May get more confusion than protection.

The lowering of pullup value is to be credited to Gen7 Research page, there explains why its better, mostly better for higher temperatures. Pullup change also changes the 22R drop, and if x2 for each themistor leg, thats even more voltage drop. Idd not much for 4k7, but gains with lower value, 1k or 560 pullup.

Im all in for adding features, i made reprap.org/wiki/RDB that has much of what i think is good, but its diy version and different approach. Still, does have other features even if discussable. Why not get something like that instead. Sry for typos and etc, posting from phone. Done speaking my mind now so cheers.

Err, do you understand how a voltage divider works?

Get 2 resistors and make a voltage divider. At the centre point of the divider measure the voltage with a multimeter. Now do the same, but put a 22R in between the centre point and the multimeter. The voltage will be as near to the same as it makes no difference.

This is due to the fact that the multimeter should have a fairly large impedance.

Looking at the Atmel 640/1280/2560 docs, there is a whole section on the circuitry around the analog inputs: Section 26.6.1 Analog Input Circuitry (from [www.atmel.com] ). From a brief look over the SAM3X datasheets, it uses the same store-and-hold mechanism, and therefore should have similar input circuitry.

The way the circuitry works is to connect an internal capacitor to the input through an internal series resistor. The time it takes the capacitor (which on the Mega is about 14pF) to charge is the minimum sample time.

From the Atmel 640/1280/2560 docs:

Quote

The analog input circuitry for single ended channels is illustrated in Figure 26-8. An analog source applied to ADCn is subjected to the pin capacitance and input leakage of that pin, regard- less of whether that channel is selected as input for the ADC. When the channel is selected, the source must drive the S/H capacitor through the series resistance (combined resistance in the input path).

The ADC is optimized for analog signals with an output impedance of approximately 10kΩ or less. If such a source is used, the sampling time will be negligible. If a source with higher impedance is used, the sampling time will depend on how long time the source needs to charge the S/H capacitor, which can vary widely. The user is recommended to only use low impedant sources with slowly varying signals, since this minimizes the required charge transfer to the S/H capacitor."

So by adding a series resistor in there, the only thing you really reduce is the time it takes to charge the internal capacitor. Since the internal resistance between the input pin and the capacitor is between 1k-100k, 22R and a PTC should not noticeably increase the time it takes to charge the S/H cap much at all.

The main issue I suspect will be the standard input leakage of the pin, which I can't see looking over the docs for either the Atmel 640/1280/2560 (Mega) or the SAM3X8E (Due). It is something that I think would need to be tried to see if it has any effect.

Edited 1 time(s). Last edit at 04/29/2013 07:50PM by Cefiar.

The S/H cap will get charged from the 10uF so the impedance of the input network has no effect on its charging time.

Is Bobc's circuit really the opposite way to mine? I read it the other way and drew mine in what I thought was the same direction, hence why it is backwards.

If we assume input leakage is less than 10uA, which CMOS MCU pins usually are. Then to make less than 1 LSB error on a 10 bit number the series resistor should drop less than 3.3V/1024 at 10uA, so I think the resistor can be up to about 320R. So 300R 1/4W could be used negate the need for the PTC and will handle being shorted to 12V.

Edited 2 time(s). Last edit at 05/02/2013 09:36AM by nophead.

---

[www.hydraraptor.blogspot.com]

nop: Actually you seem to be right. They are the same way, in which case bobc's wouldn't actually protect much as IMO the PTC is on the wrong side (it should be in-line with the 22R in his diagram).

Yup Cefiar you were right about divider, i overlooked schematic too fast and messed up with some things in my head. Sry for that and you were right about it.

Hey bobc,

On the topic of protection, it would be worthwhile adding diodes on the FET outputs too, such as described in the following thread: [forums.reprap.org]

Hi !

Are you still looking for helping hands?
I developed the easy electronics and the graphics LCD shield for Sanguino & Melzi.
I have a bit experience in designing analog & digital PCBs. I my job i'm designing RF electronics for e.g. for cordless phones which sometimes sell in millions.

What design tools are you using, is there somewhere a feature list of tht what you want to build available?

---

Grüße / Regards

STB

______________________________________________________________

Basics about MOSFETs

nophead Wrote:
-------------------------------------------------------
> This is maybe what I would do.
>
> [forums.reprap.org]
> lename=clamp.jpg
>
> I would look to see if I could increase the series
> resistor enough to not need the PTC without the
> ADC input leakage current creating a significant
> error. Don't have time to look at the data sheet
> right now.

It can be done with a nuch simpler circuit.
The input of the ADC behaves like a capacitor with some pico farads. No static current is flowing into this pin.
In this case, taken into account that the thermistor signals time constant is in the range of some 100ms, a simple series resistor of ~22kOhms followed by ~470nF (against noise) is enough to protect the input against voltages up to 30V. Clamping is done by the internal ESD protection diodes.

---

Grüße / Regards

STB

______________________________________________________________

Basics about MOSFETs

All inputs have some static leakage current, even if is nA.. It should be on the datasheet but others have reported it isn't.

---

[www.hydraraptor.blogspot.com]

It's in the datasheet SAM3X, SAM3A Series page 1420. Atmel specs. +/- 500nA which gives an error of approx 10mV with 22k series resistor.

Some thoughts on MOSFET switching. Gate drivers for the MOSFETS are mandatory, driving the gate with 3.3V results in bad Rds_on behaviour.
You should think about driving the gates with 12V. MOSFETS in this case achieve significantly lower Rds_on and are cheaper compared to the 5V (logic) FETs.

---

Grüße / Regards

STB

______________________________________________________________

Basics about MOSFETs

10mV will give about 5C error at 250C.

---

[www.hydraraptor.blogspot.com]

Catching up...

Cefiar Wrote:
-------------------------------------------------------
> nop: Actually you seem to be right. They are the
> same way, in which case bobc's wouldn't actually
> protect much as IMO the PTC is on the wrong side
> (it should be in-line with the 22R in his
> diagram).

Sorry, that was a bit confusing, my circuit is opposite way, thermistor on left, MPU on the right. I use the convention input on left, output on right, but should have labelled it better.

Quote
STB
It can be done with a nuch simpler circuit.
The input of the ADC behaves like a capacitor with some pico farads. No static current is flowing into this pin.
In this case, taken into account that the thermistor signals time constant is in the range of some 100ms, a simple series resistor of ~22kOhms followed by ~470nF (against noise) is enough to protect the input against voltages up to 30V. Clamping is done by the internal ESD protection diodes

That certainly does sound simpler.

Quote
Cefiar
On the topic of protection, it would be worthwhile adding diodes on the FET outputs too, such as described in the following thread: [forums.reprap.org]

That is a good idea, I will add them.