Power supply - Sergey89

Hardware inside and outside of the ECU
Post Reply
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Power supply - Sergey89

Post by Sergey89 »

Image
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

All in all, not a bad regulator layout. However it's very inefficient, and you'll have to ensure you remove those BTU's to maintain a large temperature range. If you are running 1A on the 5V with a 16V battery, you'll be dissipating (16-5) * 1 = 11 watts of energy. That's 11 watts wasted to provide 5 watts of usable power for a total energy dissipated from the ECU of 16 watts. If your thermal resistance of the junction to heat sink is 5C/W, and your maximum regulator junction temperature is 120C, your maximum heat sink temperature will be 120-(11W * 5C/W)=65C. Getting a heat sink to dissipate that can be tricky, especially if you have hot components around. Perhaps the installer mounted it to the fire wall and the exhaust or engine are making the firewall a bit warm. Even if you have an hot day it can be close to 50C, and you'd have to pump that heat from the heat sink with only a 15C delta. This low delta in temperature causes your heat sink to get really larger. So I would suggest a more efficient regulator design, as it will make it much easier to get an automotive qualified design.

Look here http://daecu.googlecode.com/svn/Hardware/trunk/KICAD_Project_TRK-MPC5634_P3-P4-ETPU_IO_proto/TRK-MPC5634_ETPU_IO-board-sch_RA.pdf at Q19, I suggest that instead of your D19. It will drop far less voltage and less heat and will provide the reverse protection you are looking for.

Also I do not recommend that either of these circuits be used with clamping IO. These regulators will pull you up to 5V but will not pull you down to 5V. If there is energy dumped into the 5V rail external from the 5V regulator, you can cause the rail to exceed 5V. Most chips have a top side limit of around .3 to .7 V above the VCC rail. If you go beyond that limit, you typically cause damage to the MCU port or the MCU ESD diodes. One commonly overlooked external source of energy is the clamping diodes. If you short 12V to the ECU input, these clamping diodes will dump some mA's into the 5V rail. This is why I used the op-amp U26 on my above posted schematic. I can choose my rail voltage via P20, then the op-amp will push and pull to maintain that voltage on the rail. I used this op-amp buffered rail voltage for my clamp protected IO, as it will prevent excessive voltage to the MCU.
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

Also I do not recommend that either of these circuits be used with clamping IO. These regulators will pull you up to 5V but will not pull you down to 5V. If there is energy dumped into the 5V rail external from the 5V regulator, you can cause the rail to exceed 5V. Most chips have a top side limit of around .3 to .7 V above the VCC rail. If you go beyond that limit, you typically cause damage to the MCU port or the MCU ESD diodes. One commonly overlooked external source of energy is the clamping diodes. If you short 12V to the ECU input, these clamping diodes will dump some mA's into the 5V rail. This is why I used the op-amp U26 on my above posted schematic. I can choose my rail voltage via P20, then the op-amp will push and pull to maintain that voltage on the rail. I used this op-amp buffered rail voltage for my clamp protected IO, as it will prevent excessive voltage to the MCU.
+5v_clamp is used to supply external clamping diodes and i also use additional resistors before internal clamping diodes.

Image
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

I see 5.87504V in the area that I think is where the MCU pin would need to stay under 5.3V. Am I reading this correctly?

I once posted a hacked together wiki about AN/Digi protection it's found here https://code.google.com/p/daecu/wiki/AN_and_DIGI_Protection Sorry it's very un-polished, however the core data is captured there. I have also tested that circuit on a physical board and it works well. I shorted all 4 inputs to 18V, each pin was below 5.3V and the 5V rail was solid at 5V. I let it sit that way for half an hour.
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

I see 5.87504V in the area that I think is where the MCU pin would need to stay under 5.3V. Am I reading this correctly?
Yes. But I also use an additional resistor between the IC pin and clamping diodes.

Image

This solution allows to limit current through internal protective diodes at 1 mA.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

I believe the MCU pin will be high impedance, so 0mA flow through the R9 resistor, and therefor no voltage drop. If you have the same voltage on both sides of that resistor, isn't that still to high for the MCU pin?
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

Many IC can be used with a voltage greater then the supply voltage + forward voltage of clamping diode with current limiting.
erich
Posts: 21
Joined: Tue Sep 03, 2013 4:55 am

Re: Power supply - Sergey89

Post by erich »

User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

Typically a 6V transorber will dump significant current at 4.5 to 5V, which can skew an analog signal. Do you have a specific transorber you use?

I like this buck regulator. The ripple is a bit high, however if it's set to something like 6V, then a LDO with a high ripple rejection like MAX1857 is used, you can get a nice clean automotive grade supply.
Welcome to the friendlier side of internet crazy :)
erich
Posts: 21
Joined: Tue Sep 03, 2013 4:55 am

Re: Power supply - Sergey89

Post by erich »

That's why I used a 3.3v rated voltage transorb
http://www.semtech.com/apps/product.php?pn=SRDA3.3-4
I'm not worried at all about ripple. There's alread a LDO on the discovery board that the analog stuff references. I also added 470uf caps on the 12v, 5v, and my own 3.3v LDO that I use for pull-ups.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

How is this situation deal with "To return to a non-conducting state, the current through the device must fall below the ISB (approximately <50mA) and the voltage must fall below the VSB (normally 2.8 volts for a 3.3V device). If a 3.3V TVS is connected to 3.3V DC source, it will never fall below the snap-back voltage of 2.8V and will therefore stay in a conducting state."

The datasheet seems a bit light on the details. If you have a valid 3.3V I would expect that after a transient event, the diode will remain in a low impedance state, however I don't know what that impedance would be. If you have a current limiting resistor, and the protection diodes impedance is low enough, the diode would drop the voltage at the MCU pin for a period of time. Once the protection diode(s) unlatch then it would bump back up to normal voltages. However with out knowing the diodes impedance, I don't know what current limiting resistor value is required. I guess I can ballpark the impedance at 2.8V/.05A=56 ohms during that condition. The 2.8V is a min, so if we do the same for 3.3V/.05A= 66 ohms. Hmmm, also if I follow the curve, it shows the impedance increases to around 1megOhm for the punch through, so I'm not sure we can rely on that low of an impedance. With a valid (and solid) 3.3V signal on the harness pin, the current limiting resistor would drop to at least 2.8V from 3.3V and would have to limit conduct at least 50mA, so it would have to be at least (3.3V-2.8V)/.05A ohm= 10 ohms. Any lower resistance would likely not survive the snapback. A this point higher ohm current limiting resistors would be unknown if they would work or not.

If I assume a hot battery at say 18V and the IO pin has been shorted to it, we would need to limit such that this chip can sustain a continuous current. However I only see a curve up to 1S. So I don't know what we can sustain for a continuous current. Hmmm, how to limit such that the diodes don't go wrong. I'm not sure.

Hmmm, it would still need the low side shotkey as they reverse breakdown is 3.5V, the STM32F4 can only handle .3V, so the reverse polarity still needs to be protected against negative voltage spikes.

Hmmm, I would also expect this to be noisy as hell when conducting. Basically if a design that can handle being connected to a continuous DC supply, this chip will conduct, then it will probably pull down the MCU pin voltage, then it will stop conducting, however if you have a DC supply, it will start to conduct again. Seeing that the conduction really turns on around 1uS to 10uS, I would expect a splatter of RF around 1MHz to 100kHz, which could couple into other circuits. My experience with negative resistance diodes includes GUNN diodes which are used to make GHz signals in radar equipment. So it's possible we'll see much higher frequencies as the TVS goes into this negative resistance characteristic.

Hmmm, also the STM32F4 is 3.6V not 3.3. I guess we would have to make sure we limit the upper limit to 3.3V instead of 3.6V.

I see they also have 5V version of this chip.

To me it seems like this chip is well suited for short term ESD events. However I don't think it's so good for general purpose pin protection, as it can self latch into a conducting state where it would skew an analog signal, and it has non-specified continuous ratings. With out specs for continuous operation we are left to guess and hope for the best.
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

kb1gtt, I have a question about your power supply circuit.
Screen Shot 2013-12-11 at 9.49.55.png
Screen Shot 2013-12-11 at 9.49.55.png (17.77 KiB) Viewed 19387 times
The maximum G-S voltage of a typical mosfet is no more than +/- 20 V, but the TVS clamping voltage in the automotive application is more than that value.
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

What do you think about this solution?

Image
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

Oops, I really should use MMBZ18VALT1G (18V) instead of GSOT24C-GS08 (24V). You are correct that 20V g/s is lower than the 24V clamp I was using and should be clamped to 20V or less. Note the 18V will really be closer to 19 as I'm using both diodes back to back. I'm not crazy about adding extra parts to do the clamp, I would tend to prefer the 19V clamp. I like that yours will work up to that 72V with that switcher chip I used on that IO board. So my schematic is hard clamped to 19V, while yours will work for higher voltages. I think both are good designs with their own pro's and con's.

I did a QUCS simulation to check out several things and look at wave forms and such. I might suggest a 1k or 5k for R3 as the 10k only conducts current that rival leakage currents, and I don't think you want to rely on the conduction of most diodes when you are only using the leakage currents.

Image
Source found here http://code.google.com/p/daecu/source/browse/#svn%2FHardware%2Ftrunk%2Fsimulations%2Finput_power_protection
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

ISO 7637-2 consists information about test pulses for the automotive power supply system. http://www.empek.com.cn/uploadfile/download/uploadfile/201304/20130411045252975.pdf

Below you can see the parameters of the test pulse 5.
Screen Shot 2013-12-13 at 9.19.18.png
Screen Shot 2013-12-13 at 9.19.18.png (36.74 KiB) Viewed 20011 times
According to this information the TVS diode must handle the large current transients during the pulse.

For example:

Us = 87 V
Ri = 0.5 Ohm
Vcl = 30 V

I = (87-30) / 0.5 = 114 A
P = 114 * 30 = 3.42 kW

So the peak power of the TVS diode should be more than 3 kW.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

Blah, that's far worse of a spike than I was planning for. My plan was just a guess and was based on if you exceeded 20V for more than around 5 to 10 mS, it would blow an upstream fuse. I'll have to re-evaluate each design after I get to absorb that spec. Which will take a couple days. I'm also sure it will result in more QUCS simulations. So stay tuned.
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

This is a preview of a new design
Attachments
power_supply.pdf
(61.52 KiB) Downloaded 3345 times
User avatar
mobyfab
Posts: 139
Joined: Tue Oct 29, 2013 10:09 am
Location: Versailles, France

Re: Power supply - Sergey89

Post by mobyfab »

I would use something like that: http://www.ti.com/product/tps54239e

Compared to an LDO you just need to add a tiny inductor. (3x3mm)
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

Has the "LM267X Made Simple (Version 6.0)" software been used to do some kind of verification? I see that software is noted in the datasheet for the switcher.

Upon a quick look, my primary concern would be ripple rejection. Do you know what kind of ripple to expect on the 6V? Also do you know what kind of ripple rejection to expect from the linear's? I know that these linear's often pass the ripple right on through, which can be a pain. I see those ferrites, but I also see a switching frequency of 260kHz. I would expect that to be below the ferrite cutoff frequencies, but above acceptable usable tolerances. I'm wondering what the expected ripple is at those points. Perhaps a simulation can help shed some light on the expected ripple, as that can be hard to predict via data sheets and pencil techniques.

I see page 2 of 3, are there other pages that can be shared? Often I need to see what other components are being connected to know what kind of tolerances and ripple are acceptable. Even if only a draft, it helps me understand what the requirements would be.

I've seen several "automotive" switchers that generate 2A. I guess that kind of current is useful for driving dashboard gauges directly. It might be worth considering a more powerful switcher if you plan for to drive dashboard devices at some point.
Welcome to the friendlier side of internet crazy :)
User avatar
Sergey89
contributor
contributor
Posts: 839
Joined: Wed Sep 25, 2013 5:30 pm
Location: Russia, Velikiy Novgorod

Re: Power supply - Sergey89

Post by Sergey89 »

kb1gtt wrote:Has the "LM267X Made Simple (Version 6.0)" software been used to do some kind of verification? I see that software is noted in the datasheet for the switcher.
No. I have a problem with installation of this software and plan to test it later.
Upon a quick look, my primary concern would be ripple rejection. Do you know what kind of ripple to expect on the 6V? Also do you know what kind of ripple rejection to expect from the linear's? I know that these linear's often pass the ripple right on through, which can be a pain. I see those ferrites, but I also see a switching frequency of 260kHz. I would expect that to be below the ferrite cutoff frequencies, but above acceptable usable tolerances. I'm wondering what the expected ripple is at those points. Perhaps a simulation can help shed some light on the expected ripple, as that can be hard to predict via data sheets and pencil techniques.
Maybe I will build this circuit on the PCB and do the test.

This figure is not promise anything good at 260 kHz.
Screen Shot 2013-12-18 at 14.56.29.png
Screen Shot 2013-12-18 at 14.56.29.png (31.31 KiB) Viewed 19931 times
I see page 2 of 3, are there other pages that can be shared? Often I need to see what other components are being connected to know what kind of tolerances and ripple are acceptable. Even if only a draft, it helps me understand what the requirements would be.
A bit later I will publish all schemes.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: Power supply - Sergey89

Post by kb1gtt »

I finally got a chance to read through ISO 7637, here's my summary of the major concerns to keep in mind when making a design.

Pg 10 includes a suggested wiring network used to simulate the vehicles wiring when the vehicle. It includes a typical inductive and capacitance properties of ha wire harness.
Pg 12 specifies a power supply with resistant only properties of .01 ohm or less impedance under 400Hz.
Pg 14 specifies -V transients caused by inductive loads. It notes a peak of -100V with a total time of up to 2mS.
Pg 15 specifies +V transients caused by parallel load switching. This spike can get up to 37V for .5mS and happens regularly.
Pg 16 specifies a DC motor disconnect. It's decay can take up to 2 seconds.
Pg 21 specifies an alternator, or battery disconnect. Spikes up to +87V with a maximum time of 400mS.

There's lots of other good information in there, but those are the high lights I see in terms of making sure you uphold a good design.
Welcome to the friendlier side of internet crazy :)
Post Reply