How bad is automotive environment?

Go To Last Post
68 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi.
I need some practical hints. Actually, how bad is automotive environment?
For example, may I use direct connection of buttons/contact switches to the inputs of AVR chip, with internal pull-up resistors?
Or I need to use external, pull-up resistors with lower resistance?
Or I need optocouplers on inputs?
Is it possible to use software instead of hardware? (Some sort of filtering)
Let say, the wires from the CPU to the buttons/sensors are 1..2m long.

Regards.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The automotive environment is very hostile as far as electronics are concerned! You might be better off using separate MCUs and LIN.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

leon_heller wrote:
The automotive environment is very hostile as far as electronics are concerned! You might be better off using separate MCUs and LIN.

Leon

That is exactly what I am doing. Central CPU, with RS485 bus to the peripheral modules.
But at the end point, some devices have to be connected to the slave modules with piece of wire.
So, in the case of inputs, I am wondering about the schematic of the inputs - whether to use optocouplers or I can use some simpler interface, eventually with software filters.
I have scores of sensors, buttons, switches (on the control board, for example) and using optocouplers in all these cases may complicate the design too much.
That's why I am asking for some practical experience, if someone already have it.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'd use another controller for the panel. That's what we did on the military equipment I worked on.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Automotive environment is not only about how to connect front panel with 2m cable, but also about power supply filtering.

Some people may put another micro on the front panel, but some just put filtering on the lines. RC filter, strong pull-up, maybe schmitt trigger chip like 74HC14 to make the signal clean to AVR. Protection diodes even.

Same thing with power supply, first thing is a fuse, then some diodes to protect from wrong polarity or negative spikes, RC or LC filter to filter out noise and keep voltage stable while battery dips when starting the engine, and possibly overvoltage protection with resistor and zener diode or TVS etc. All this before your normal 7805 regulator. Best to use LDO if 6V battery system or you need AVR running when starting the engine. Even better to use 3.3V AVR voltage so it works with less voltage on battery.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

leon_heller wrote:
I'd use another controller for the panel. That's what we did on the military equipment I worked on.
Leon

Yes, but what about inputs? Let say - you have 15 buttons on the control panel (dashboard) and a CPU mounted inside shielded box, somewhere inside the dashboard. The power is perfect (with protections, filters etc.). So, what is, (in your opinion) the optimal way to connect the buttons to the CPU?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

With strong pullups and capacitors?

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ESD is a major problem in cars. Since the basic operation of a car forms a static generator and you've got a lot of carpet with static generated when people get in/out.
Most carmakers spec way in excess of what CE requires. OnSemi have specific protection diodes as do many other manufacturers. Use software to filter the inputs also - you'll most likely need to debounce anyway.
Optos are a waste of time.

Have a look at some production hardware in a modern car to get some ideas.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This document may be helpful:

Suppression of Transients in an Automotive
Environment, Harris AN9312.5

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

There is also the issue of temperature. You can drive to places north of the Arctic Circle, in the winter. Plenty cold. You can have a vehicle in places where it is very hot.

Simple things like electrolytic caps have major problems with those extremes. That is also why Atmel has "automotive grade" processors. You will notice, on checking, that the major spec difference is just temperature range.

Another automotive issue is power (voltage). Your power conditioning circuit needs to be able to tolerate this, not your processor, fortunately. There is this thing called "load dump". It happens when the vehicle is running and very little load (headlights, etc, off) and someone disconnects the battery. The alternator output can jump up to 65V for some milliseconds before finally dropping. Basically, the regulator cannot change the field current fast enough to keep it from happening (due to field inductance). Those milliseconds are enough to wipe out most regulators. You need to pick a regulator that will tolerate these voltage extremes, in addition to temperature and everything else.

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Further on Jim's comments - I often find myself brushing up on overvoltage protection on this page. It's a good read I think

http://www.maxim-ic.com/appnotes.cfm/an_pk/760

oddbudman

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

As I can read, most of your posts concerns power lines - noise on power lines, surge protection, over voltage, ESD, load dump etc. Well, I already designed the power system for my project and it have all protections and noise filters.
So, I think to try simple input schematic - on the picture. What do you think about it?

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Diode there does not protect from negative spikes, unless you add series resistor.

Pull-up on the button side, capacitor on the AVR side, series resistor in between.

I'd omit the diode as it does not do much. If I wanted to be really sure, I'd put two diodes there after the series resistor. Then again, they need to be schottky, otherwise AVR built-in diodes take all the heat.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

GaryPeek wrote:
This document may be helpful:

Suppression of Transients in an Automotive
Environment, Harris AN9312.5


Harris Suppression Products was acquired by Littelfuse in 1999.
Here's the application note: http://www.littelfuse.com/data/e...
Here's all their Automotive application notes: http://www.littelfuse.com/design...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jepael wrote:
Diode there does not protect from negative spikes, unless you add series resistor.

The diode protects from positive spikes (unintentional connection to +12V) but maybe it is superfluous. On the other hand it rises the level of zero to 0.6V, lowering noise resistance.

OK, lets see, another variant:

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I am using the same circuit as above without the resistors, for the diodes I use 350mv schottkys. If you know the output impedance of the sensor you want to interface it with then the resistors are a good idea, but if you are building a universal type of device then either use very low value resistors or not use resistors at all. There are air flow/pressure sensors out there that have very high output impedance.

For general purpose inputs to the ADC, I plan on putting diodes + cap on the input then run the signal to an OPAMP, from the OPAMP it goes to the AVR. The OPAMP serves as a another layer of protection.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

When you used diodes or zeners for protection you need resistors. Their purpose is to limit the current. If you look at a typical I/V diagram of a diode or zener, you know why you need current limiting.

Stealing Proteus doesn't make you an engineer.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

So, the above schematic probably will work very well. But it contains 5 elements per every input needed.
Why not to use an optocoupler in order to get simpler schematic.
What you think about below schematic. Is it safe enough to use internal pull-up, if the optocoupler is inside shielded box?

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

And then transients still can kill the LED inside the opto. I have never seen opto's in automotive modules.

I usually use the circuit I attached. R3 is for inputs switching only to 12V, not required for inputs switching only to ground.

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Optos give you galvanic isolation but won't protect you from fast transients.

AlanTo - what's the use of adding an op-amp for 'extra' protection? You still need to protect the op-amp. If it fails, your product has failed and the customer doesn't care if it is the op-amp or the processor.

Philips have some nice integrated diode protection devices - BZA100 which can cut the component count.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ST also make Transil arrays, for example the ESDA6V1U1 with six of them in a SO8.

If you use resistor and capacitor arrays you can drastically cut component count and the PCB estate needed.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ArnoldB wrote:
When you used diodes or zeners for protection you need resistors. Their purpose is to limit the current. If you look at a typical I/V diagram of a diode or zener, you know why you need current limiting.

for example, you are making an analog datalogging system for TPS,MAF/MAP,intake air temperature, etc...
If you are making a product for a specific model car, and you know exactly the characteristics of the output impedance of the sensors then you can make a wise decision on how big or small the current limiting resistor can be without compromising the signal. If the device is to be a universal type of device, then you either run with a very very small current limiting resistor or no current limiting resistor at all.

In my application, I can not have any resistance between some of the output pins to the sensor, and some of the input pins can only tolerate a very small current limiting resistor or it will cause errors on the measured signal.

I am unsure of what the best approach is, certainly I will use schottkys and a ceramic cap on each input/output. I will also use some op amps as a extra layers of defense. The OPAMP will blow if the transients are powerful enough to get past the diodes, but my experience is that OPAMPs can take more punishment than the typical uC pin.

Also it is much easier to diagnose a burnt OPAMP than it is to diagnose a burnt micro due to voltage transients on the pins. Usually only the OPAMP circuit connected to the input/output is burnt, so you know exactly what sensor pin is giving you trouble. A micro with a burnt pin, will often just be dead, and you might not have much clues as to what is the issue.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If the unit failed you know the design is not bullet proof enough :P

I don't see why you cannot include a relatively big series protection resistor on an analogue input. On a generic datalogger system you want the input impedance to be as high as possible in order not to influence the signal too much.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
And then transients still can kill the LED inside the opto. I have never seen opto's in automotive modules.

How???
The LED in the opto have maximal current 50mA, so if the resistor is 10k, it can withstands 500V spikes on +12V line without any damage, even for long time.
Also, when the button (sensor) is not switched on, it is hard to imagine what noise can make LED to emit light.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

500V spikes are the least of your worries. ESD events are measured in terms of kV. Also with fast transients, capacitance can work for you and against you. In series, stray capacitance will couple fast transients, in parallel, capacitance will slug down the leading edge of the transient. Joe average optocoupler has capacitance between the led and opto transistor.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, I made some simulations of optocoupler variant and it behaves very good. Even on 10kV ESD (HBM), the maximal current does not exceed normal mode of operation of the optocoupler (1A for 100us). The output is not overloaded as well.
On normal operation it works without false pulses on the output even when the +12V source contains a lot of garbage - pulses to 500V/10us + some HF signal (4V/10MHz)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

But simulations do not reflect the real world ;) The GIGO principle applies.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
But simulations do not reflect the real world ;) The GIGO principle applies.

Hm. Do you have some real world experience with optocouplers in automotive (or other very noisy) environment?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

It depends on how good the simulation is!

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Hm. Do you have some real world experience with optocouplers in automotive (or other very noisy) environment?

I've made a couple of relatively simple modules for my car. And I have not used a single opto in one of them. Just plain resistors, caps and Tranzorbs in the arrangement I posted earlier.

And yes, they all still work flawlessly after a couple of years ;)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The problem with optocouplers isn't so much false output due to noise (fast transients etc) but rather acting as an effective barrier against these transients.
The HBM is fine for general application, but the auto environment is rather specific in that you have potential for more static to be generated due to people's clothing and the car fabrics as well as static generated by the movement of the car. Some carmakers will specify something like 30kV coupled directly to the electrical input. So you can use optos, but you still need to add protection in which case to optos offer very little as you don't need galvanic isolation. Also remember optos age.

JayJay's arrangement is pretty standard for ECU inputs. For user controls where exposure to ESD is much higher, circuit layout become more critical.

Ultimately, it comes down to requirements. If your customer demands a certain spec, then you've got to comply otherwise choose a standard that you think is applicable and comply with that. I dare say most of the aftermarket car goodies would just be compliant with the CE requirements (if that).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

One little thought. After some meditation, I decided, that 1mA maximal current of the protection diodes is simply not very plausible.
There are several arguments:

1. In the data sheet, there is no any explicit parameter about these protection diodes. I didn't found anything at the atmel site as well.

2. It looks like these diodes are actually parasitic diodes of the output stage MOS transistors. But in this case, the max current should be comparable with max current of the output transistors, because the area of the diodes is the same.
3. btw, there is a parameter in the datasheet:

Quote:
DC Current per I/O Pin .............. 40.0 mA

But I/O means Input and Output, not only output.

4. In PIC16 datasheet, this parameter is explicitly mentioned as:

Quote:
Input clamp current, IIK (VI < 0 or VI > VDD)........... ± 20 mA

But pics have maximal output stage current 25mA - it looks comparable with clamping current.
The technologies of AVR and PIC are probably similar, so...

So, is 1mA an urban myth, or what?

Unfortunately, I don't have laboratory just now, so if someone want to make some tests and to post the results, I will be very happy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

DC current In and Out, means the port can source or sink 40mA. From the viewpoint from the pin, current sinking means current is flowing into the pin.

Microchip apparently choose to specify the max.clamp current and Atmel didn't. Atmel did once state that 1mA is regarded safe (I believe in an email to someone here on the board).

The max. current through the protection diodes has nothing to do with maximum current a pin can sink/source. The diodes are explicitly added and are not parasitic, because the reset pin does NOT have one to VCC ;)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
DC current In and Out, means the port can source or sink 40mA. From the viewpoint from the pin, current sinking means current is flowing into the pin.

I can't agree. The exact phrase is "per I/O pin" - the subject here is "I/O pin" - that is pin that can be configured as input or as output.
Actually atmel explicitly uses the words "source" and "sink" for currents you mentioned above.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I don't know of anywhere where atmel has specified the parameters of the protection diodes. They do specify values of input voltages that would preclude turning those diodes on.

The sink/source current ratings refer to the output mosfets, NOT the protection diodes.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

dbvanhorn wrote:
They do specify values of input voltages that would preclude turning those diodes on.

But if these diodes are not intended to be turned on, why they are putted there?
Every micro controller is intended to work with little or no additional hardware. That is why the input protection should be good enough. 1mA is too little for any real world application. It can't be true.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Actually Atmel states 1mA in this application note (page 4, second sentence).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
Actually Atmel states 1mA in this application note (page 4, second sentence).

Yes, but this is not very reliable source. In this note, 1mA is simply reasonably small current to get from the signal source on regular basis (on every period of the input signal).

If I use these clamping diodes as a protection, I will not care how the source (ESD for example) will be loaded.

Also, these numbers 1MOhm 1mA 1000V looks too round for me - i.e. random.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Maybe YOU don't care how much you 'load' the source but the diode will. If the current is anything bigger then a couple of mA you will damage the diode, and likely the input structure of the pin too. And automotive transients can carry large quite some energy, and the diodes are not designed to handle that.

1000V/1MOhm=1mA, does not look really random to me.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
and the diodes are not designed to handle that.

Or maybe they are, but we don't know.
I definitely will make some tests, but I can do that only after few months.
Someone curious enough to make them earlier? ;)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

They are not. If they were capable of dissipating any significant amount of energy, Atmel would boast about it in big letters with lots of marketing fluff. Big diodes take up a lot die space. They are sized to minimum size. Bigger dies means the IC will be more expensive.

If you dump enough energy through them, it will couple onto the VCC rail, possibly influencing other stuff on that rail other then the MCU.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

johnfound wrote:

But if these diodes are not intended to be turned on, why they are putted there?

For extra protection. Like airbags in cars. Not intended to be turned on in any normal use, but still put there just in case for extra protection.

Diodes are there for ESD and overvoltage protection and can take all the heat they are specified to take under normal use.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If the max. current of the diodes is 1mA the max power is: 0.6V*0.001A = 0.0006W = 600uW - ???
Output transistors of AVR are 2 times more powerful than those of PIC controllers (40mA vs 25mA output current) , but the protection diodes are 20 times weaker (1mA vs 20mA).
It still looks strange and doubtful for me.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Why is it strange? Apparently Microchip decided to use bigger clamps, like Atmel decided to beef up the output transistors.

Why would you want to run more current through them anyway? All it takes is a little bigger resistor to limit the current to a safe value. You need that resistor anyway otherwise the current is basically unbounded once the voltage is above the threshold voltage of the diode. Do you know how a V/I curve of a typical diode looks like?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
All it takes is a little bigger resistor to limit the current to a safe value.

Not "little bigger", but 20 to 40 times bigger. This resistor with input capacitance limits the signal frequency. There are cases, this can be important.
Also, if the protection diodes can withstand reasonably big current, you can omit external protection diodes and use only current limiting resistor in many cases.
I like to make optimal design, where every value is set reasonably, not "just in case".

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

IIRC you are going to interface to switches, so speed is a non-issue anyway. And that PICs withstand more, that's a non-issue too because you are using an AVR.

I don't understand why you are having such problem with adding external diodes really... unless you are designing a toy that will be made in the millions where every tenth of a cent counts and where reliability is not really of any importance. In many other industries, just one failure of your design can mean that whatever you saved in component costs is negated. And possible many times over.

A lot of things are here for 'just in case'. Like internal protection diodes...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
And that PICs withstand more, that's a non-issue too because you are using an AVR.

But you are not sure PIC withstands more. You just believe in the myth about this 1mA.
I don't believe in anything without proof. Do you ever blow the AVR, because of 2mA (or 5, or 10mA) current through the protection diodes???
I am sure you are not. :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I trust the application note written by Atmel engineers and I don't think the imposed 1mA limit is a myth. I assume they know what they are writing and regard the Atmel documentation as authoritative. After all, they are the designers. And the number sounds reasonable. What do YOU think they can take?

But it's even better to design like there are no protection diodes present at all because they are basically unspecified. Basically... sound engineering practices, due diligence etc etc

And indeed I have not tested how much abuse the diodes can take. I don't need to either, I use external protection.

I don't know what will happen with overcurrent, maybe just the diode blows (might turn into a short, maybe open circuit) so you either have an output that won't go low anymore and short circuits the power rail to ground when set to ground, or you loose the protection function and it will blow up other stuff with the next transient.

For a one-off product just putting on those diodes will be less effort than looking for the absolute limits, that probably vary from device to device anyway.

It's YOUR design, and if it's goes kaputt YOU are in trouble, not us ;)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

In other application notes (AVR240, AVR245) Atmel engineers suggests using single resistor 1k or 470ohm for ESD protection of the inputs.
Here is quote from AVR245:

Quote:
ESD protection resistors are typically in the order of 1 kohm, or less.

Now, what will be the current through the protection diodes on ESD?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

That depends on what kind of ESD it is.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Diodes can witstand much higer short transient currents than continious currents.
At least this is true for IR LEDs, so this might be true for regular silicium diodes too.
For many IR remote controls designers take advantage of this fact and send a short pulsating current through the IR LEDs that are much higher than the specified continious forward current for the LEDs.

Example of IR LED: http://www.vishay.com/docs/81001...

Quote:
Forward current IF 100 mA
Surge forward current tp ≤ 100 µs IFSM 2A

So even though this diode can only withstand 100 mA continious, it can handle 2A surge current if tp ≤ 100 µs - a factor 20 higher.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If you inject enough current into the protection diode, you get what is called 'latchup'. This was common on early generation cmos devices and mainly cured on later generation devices. Nevertheless, iject enough current and you'll get problems. This is why prudent engineers don't rely on the internal diodes.

http://www.piclist.com/images/ed...

With an external diode, you don't have these problems.

As to how much the internal diodes cope with an ESD event depends on the actual ESD event. The HBM is a popular spec but you should design to the expected environment. CE requirements differ for industrial and commercial applications in regards to fast transients. Obviously in an industrial setting bigger motors etc are expected, thus larger transients. Similarly for an automotive environment, carmakers will usually have their own spec - much worse than what the HBM is. There's most likely a SAE spec available for a price that outlines current automotive practice.

That's ESD covered, then you have EMC. You can expect to have radio transmitters installed in a vehicle be it a mobile phone, two way radio setup or CB radio. Again, SAE specs should cover expected signal strengths.

You need to decide what you are going to comply with. This depends on the application of the device - for something like a GPS device where the failure or inoperation causes only annoyance then you can go for a less stringent spec but if your device might cause injury ( say a window controller) or something that can disable the engine etc then you would want to go with a spec that is in line with current auto practice.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Kartman wrote:
If you inject enough current into the protection diode, you get what is called 'latchup'.

I thought it too - that the latch-up is the reason for so little current claimed by Atmel. But, then I found this very interesting document: http://www.ninenine.com.tw/pdf/A...

This document says that AVR are much more robust than other manufacturers MCU, including the latch-up effects.
Also, I tried to emulate ESD event on the input with diode protection. The result is, that on claimed 7000V HBM for ATTiny2313 the current through diodes is 5A for 150ns(50%). If I used low current diodes, even shotky, the voltage on the input raises above 6..7V or even 11V (for BAT54) because of the internal resistance of the diode.
In order to be able to fix the input on 5.5V it needed 0.5A or 1A diodes. BTW, there is little difference between shotki and regular diodes.
I will post the model file if someone wants to play a little.

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

johnfound wrote:
I will post the model file if someone wants to play a little.

Maybe you should mention this is an LTspice simulation.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You don't know what diodes Atmel uses and what other elements they include that are not published. There are lot of diodes types so a simulation is at best just a guess.

What exactly are you trying to prove?

That document, to me, is a lot of marketing fluff. It does not tell HOW long the devices stand the test.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Since OnSemi et al make specific diodes for ESD protection, one would conclude this is a specific application that is best served with a specific product.

So how would one design a circuit with the bare minumum components to meet a ESD spec?

Use metalised buttons/enclosure
Make the buttons etc thick enough so they provide a degree of insulation
add some capacitance to the inputs
add some resistance to the inputs
add specific protection diodes
add spark gaps on the pcb

You'll most likely see evidence of one or more of these techniques in commercial gear.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jayjay1974 wrote:
What exactly are you trying to prove?

I am KISS principle maniac. I believe in the words of Antoine de Saint Exupéry: "It seems that perfection is reached not when there is nothing left to add, but when there is nothing left to take away".

So, I am seeking a way to design things as simple as possible without sacrificing robustness and reliability.
In particular, now I make a redesign of whole electric and electronic equipment of my car, so that is the reason I am seeking for the simplest way to design automotive electronics.

Kartman wrote:
So how would one design a circuit with the bare minumum components to meet a ESD spec?

Well, if the quoted above document says the truth, you don't need external components - AVR chips like ATTiny2313 should withstand 7000V HBM and 500V MM even if you simply route the pins to the "outer world".

BTW, the ESD protection diodes by OnSemi are zener diodes, not clamping. Also, its max. DC current is far more than 1mA in order to withstand peak currents of several amperes for 100ns.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

But I think 500V MM is not all that much in a car.

As I said before, I made a few automotive modules myself, and one connects to the switches that turn on the interior lights. Those switches are very crude devices and they switch with lots of sparking.

I would apply the KISS principle at a higher level instead of at componentlevel... If you design on the bleeding edge of the specifications, relying on marketing fluff and not a single bit of overdimensioning or additional bulletproofing you might risk all kinds of problems that could be really hard to debug. I wonder what is ultimately simpler then, just adding the extra bulletproofing components (like everybody does in the industry) or spending lots of time looking for easily prevented problems.

Maybe you have to spend $10 on protection components. I don't know how much you value your personal time, say it's $10/hour I think the choice is easy as fixing problems for sure takes more then one hour.

I don't say it's impossible to only have a single resistor and abuse the internal clamping diode to interface to 12V signals and relying on them for ESD/EFT/EMI; but most of us think that an extra diode and a cap per signal improves reliability tremendously and as such is fairly cheap insurance.

What's left to say... just try it :D And report back ;)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
now I make a redesign of whole electric and electronic equipment of my car,

Redoing the display, or actually redoing the engine control, braking, etc.

If your display fails in a car it is not the end of the world. If your cruise control, engine control, or braking suddenly develop a mind of their own, and do not function as anticipated, it could be fatal to you, or others on the road. It suddenly becomes a Mission Critical interface. Over protection and redundancy are then key components to the design.

JC

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
BTW, the ESD protection diodes by OnSemi are zener diodes, not clamping. Also, its max. DC current is far more than 1mA in order to withstand peak currents of several amperes for 100ns.

How does one differentiate between a zener and a clamping diode?

John, If you're happy with using the HBM by all means fly with it. For CE compliance the test is done to exposed surfaces, whereas one carmaker's requirement is 30kV to the input conductor - not via insulation! There's a world of difference between these two ESD specs. Personally, I'd seek out the SAE spec and comply with that - at least you can show due diligence if anyone asks the question. I know for my equipment I comply with CE but there are particular hazards in the application that CE doesn't address so I protect against them.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This is what I use with great results on one of my projects... The LC filter provides 17dB attenuation with low voltage drop at 120Hz/5A, 600W TVS diode clamps fast transients at 18v, its protected against reverse polarity, and the input is fused (not shown) I have a 20mm fan in there to keep things cool, project has more power electronics, could be ignored...

Just did the maths again, and you can improve the filter performance by lowering C3 to ~1800uF, to 32dB @ 120Hz attenuation...

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Unixbeatch - you'll find most auto apps will avoid using electros - the temperature means they don't last too long. Also there's the physical size which means you've got to tag them down somehow otherwise they'll vibrate to pieces Captain. That's not to say you can't use them in your own apps. There's also the issue of disipation in D3 in event of a load dump - hopefully the fuse will save the day - try feeding in 22V and see what happens. I found out the hard way.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You are right about the electros, I just didn't have much choice here. In the actual app I chose an electro rated 10K hours @ 125C, they are expensive, but cheaper than the much larger inductor I would need for lesser capacitance. It is a through hole device in a radial package, and I locked it to the board with a large drop of 5-minutes epoxy.

The TVS device is rated for 600W peaks up to 10ms. Following this rating, it should be ok even if the voltage should reach 100V with a 5A load attached. BUT I guess there are cars that are more noisy than mine, or owners who are missing some strenght in their lower arms when time comes to tighten the battery connectors... It is trivial to add a few more of these 0.10$ devices in parallel to scale with the need.

Now I'm not saying this will pass certification or that it is foolproof, but it has been installed in an old Oldsmobile Eighty Eight for bout 4 months now and it works like a charm. The filtered 12v source rarely goes over 14v.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have first hand experience with AVRs in automotive applications. I'm interfacing to the factory steering wheel buttons and engine sensors. I treat each input individually. Rule of thumb is a series impedance as big as possible to limit current. In cases like a sensor where this is not an option, then a series ferrite bead will become high impedance at high frequencies - which is what a transient event is.

In a few cases, I have series PTC fuse (<100mA) with TVS diode. I also use PTC and TVS to guard against load dumps and reverse voltage.

I agree with your approach of less is better, but, you could spend more money than you save when you consider validation time.

With regard to spark gaps, I have used this once before with good success. But, be aware that each spark leaves a little carbon at the place of the arc, and the protection degrades with the number of sparks. You can't use solder mask on a spark gap, and it is better if the traces are gold.

Temperature is a big concern. Stay away from the internal RC oscillator if you have critical timing like serial communications. The frequency will vary widely based on temperature.

If you're laying out a PCB, I recommend the series impedance, TVS, cap, and pullup. You can adjust the values as appropriate. You can change the series R with series bead or a PTC and you can leave the other parts off if unnecessary.

Good Luck!

official AVR Consultant
www.veruslogic.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

What about outputs?

I'm currently working on a project that may need to splice into some of the car's internal signal lines (mostly for stuff like turning the interior lights on and off) and I'm at a loss for how this should be approached. Would simply using some buffer IC to interface to the car's circuit boards be wise?

From what I understand, in most cases the modules in a car are simply looking for a high/low signal or a specific voltage for input.

Where possible, I plan on using the car's own CAN and LIN buses to send commands, but I expect this will not be possible in all cases.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Naturally I have a few outputs on my automotive modules, I normally use automotive graded Smart fets, intelligent FETs and what other marketing names they have. But my car predates CAN or LIN, and is purely relay and switch based so that's easy :)

SmartFETs are fully protected e.g transients, overheating, short circuits, reverse battery. Some allow you detect open-load and such with a seperate status output. They are expensive though, and generally difficult to get.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'm curious why the 78M05 instead of an actual automotive rated part.

I've had reliability problems with the lower current versions of the 7805 over the years, from all manufacturers. Nothing really catastrophic, but a noticably higher infant mortality.

I didn't see a fuse, to deal with a shorted D3. Most power diodes fail shorted.

Good job on including D1, many people miss that.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I use a 7805 simply because I had them lying around, and the 5v load is very light. I agree they are crap but they are very cheap, and in this case the casing of the project itself is constructed like a big heatsink, with fins, and has mounting rails and brackets to hold the regulator against it. I have no real fear for the regulator.

The input is fused, but as noted before fuse does not appear on this schematic snippet.