[info] Power supply - kb1gtt

[mobyfab]

Seems like those are not depletion mosfets.

If you want to go cheap, go LDO. It's not like it's using a lot of power anyway right?

[AndreyB]

for fuses do we like F0603C2R50FWTR and F0603C3R00FWTR? 16 VDC 2.5A and 3A Current Rating.

[AndreyB]

Seems like those are not depletion mosfets.

http://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeyDN2625K4-G ?

[kb1gtt]

Sweet, I only changed to the IXT chip, because I could not find the DN2625. The DN2625 is specifically called out of by the app note I was using when I designed this circuit. See page 4 of the below. So yes that DN2625 is perfect. I see Vgs is -1.5 to -2.1 vs the IXT at -2.5 to -4.5. So I will probably need to change D1002 for a slightly higher voltage. But that should not be a big deal.
http://ww1.microchip.com/downloads/en/AppNotes/AN-D66.pdf

Hmmm, I see Rds of 3.5 ohms. That's quite high and will create heat. It's the regulator is producing 5A at 5V and 80% effeciency, the regulator will draw about 2.5A at 12V. Then 3.5 * 2.5^2 = 22 watts. That's allot of heat. However if we can use the IRF chip with an internal resistance of 0.18W, that drops the watts from 22W to 1W, and if it's the 0.55 ohms of the IXT, the watts drops from 22W to 3.5W. However as pointed out, we don't run full power. So is it reasonably to expect that if someone uses the 5A output, they also either add a heat sink, or they use the different chip which would generate less heat? I say use the low cost chip, then add an alternative note to the schematic. Also for now I say get all three chips, and I can check them all out. However for starters lets populate the DN2625.

A note about the IRF640. The graphic of the MOSFET symbol shows a depletion, but the spec for Vgs is positive, while the IXTA6N50D2 shows the depletion graphic, and specifically notes depletion mode, then it also specifies Vgs with a negative number. I'm not sure if IRF640 is depletion or not, so it may or may not work.

@motofab, feel free to compare this switching regulator against this linear http://rusefi.com/forum/viewtopic.php?f=4&t=893&p=16027&hilit=PWR_5V_linear#p16027 It's basically the same thing, but linear vs switching. Both regulators have max voltage limits, which are being extended by this depletion MOSFET. Also please keep in mind that many older dash boards drive the gauges off a 5V supply, which consumes allot more amps than we are consuming. This circuit only costs like $2 for the regulator, the high voltage suppression is what's currently expensive. Both linear and switching need this suppression.

[puff]

for fuses do we like F0603C2R50FWTR and F0603C3R00FWTR? 16 VDC 2.5A and 3A Current Rating.

Several years from now I seem to have fixed my 2007m.y. MacBook Pro power supply/IO board by replacing that tiny guy...

Still can't justify the use of a serviceble! part in such a small form-factor.

[kb1gtt]

Can you get those in low qty? Octopart seems to indicate an order of 3k is required.

With a lack of quality control in the field, we are sure to see bad practices resulting in over current situations. When this happens, do you want to replace a board or a fuse? I guess we should say if they screw it up, then you screwed it and you should buy another board from us. But that's not how we roll. We prefer to offer the option of repair instead of replace, as it is much lower cost and faster to repair. We think it's generally what people really want, and after all they can replace a failed board if they so desire. The use of fusing on the PCB is less common, and allows a flexibility not commonly offered by other PCB layouts.

Every electrical circuit should have a fuse at the battery, which should trip before the PCB fuse trips. However I'm sure someone will think they route the wire harness to close to the exhaust, which results in failed insulation that shorts to higher energy 12V which then causes a dead short on the 5V, AKA basically no fuse. I like using these fuses as 0R jumpers as it allows PCB repair instead of costly failures when you thought you ran the wires far enough away from the exhaust manifold, but apparently did not. The fuse at the battery is required and important, it's what keeps you project from catching fire and burning to a crisp. However we have an extra protection on the board which allows cost avoidance under some error conditions.

I would say we are good with the 3A fuse, and if it is not to expensive, I would suggest using them for the 0R jumpers as well. We could probably even bump those up to 5A with out to much issue. We probably want to use a 10 or 15A fuse for the GND conductors. The GND conductor has good heat sinking and especially with multiple injectors, it can normally see 5A or more.

[puff]

I meant have you considered using some larger size, more accessible fuses?

I am not sure what exactly fuse that was. It was a fast fuse with a 'P' sign on it. 3A in 0603. It was cheap that I didn't bother bying 10 or even 20 of them not to keep change in my pockets.

But that was Moscow. The most irritating part is that it was 2 hours to get to their office and another 3 hours to get back home), and their courier delivery was ridiculosly expensive..

[AndreyB]

http://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeyF0603C3R00FWTR 2,990 Can Ship Immediately Minimum: 1

[mobyfab]

Here's my power supply:


Basically I use an input diode, a resettable fuse, a TVS diode and a pretty tough automotive LDO.
In theory you only need the LDO since it's got protections everywhere. As long as your input cap is rated 60V+ and has no polarity (unlike mine).
It worked perfectly for me so far.

@kb1gtt I had no idea you needed so much power, so it might not be that useful.
You might want to go synchronous. (ie: LM25117-Q1)

Hope that helps anyway.

[kb1gtt]

So far we have not needed the power. We typically consume 0.125A, with an inrush around 0.30A. Allot of that is the LCD screen, and diagnostics LED's We allow extra amps to be used by external gauges, or external devices which may draw more amps than what we need. However that has not happened yet. I designed it to be a bit on the large side such that if we needed power, we had it available, and we would not need to redesign that section. The total cost is around $2 to $5 in low qty and less if we increase qty. We figured that for a couple bucks the flexibility would be handy.

In your design, I don't see an RF choke. Do you have a current probe? Can you measure the current ripple on the 12V power line? Do you know how it survives against EMC issues? In Frankenso R0.5, I'm adding an RF choke to help prevent noise from escaping up the power leads.

Your board can operate with less than 90mA? I plan for an ambient of 85C and a board up to 105C. I see this fuse trips at 90mA at 85C and wold trip at much less ma's at 105C. Do you expect a different ambient temperature? I guess if it trips, I see it still allows 14V / 2.6 ohms = 3.8A dead short to GND. Am I reading that correctly that tripped ohms is the max ohms of 2.6 ohms? If it allows 3.8A, I guess it really doesn't matter if it's tripped all the time. My eyeball guesses the traces could survive a 4A short for a decent amount of time. If the PTC allows 3.8A, the BAT165 would fail. So I guess you have a replaceable fuse. I see the 100V surge voltage would conduct about 80V / 2.6 ohms = 31A. I'm thinking your traces would fail if this is a sustained surge, but your BAT165 would likely fail long before the trace fails. Does the BAT165 fail short or open? Often they short then open, so I'm not sure if the trace would fail or the diode. I think this design has about the same surge suppression and EMI protection as the Frankenso R0.4 layout.

Have you tried doing a dead short of your 5V to GND like if someone screwed up the wiring on the TPS? Does it really allow 3.8A? Or does the PTC have more ohms when it trips?

I would like to compare costs at some point. Currently we are at about $2 for the switching regulator, and then another $2 for the N-MOSFET depletion mode over voltage protection.

[mobyfab]

There's no RF choke, I might add one later on. I'm not using an SMPS so I don't expect a lot of current ripple. Especially that I have a big output cap that smooths things out.
I had a look with a scope and voltage ripple is around a couple mV on a lab PSU IIRC. I assume the long wires and battery of a vehicle will help as well.

The board operates with 50mA when everything is turned on. (leds included)
2.6 ohms is the resistance when reset 1h after a trip. (basically the maximum resistance while not tripping)
It then slowly decreases.
R 1max = Maximum resistance of device at 20°C measured one hour after tripping or reflow soldering of 260°C for 20 sec.
When tripped it's in the Mohm range.
I did short tests on all rails and nothing smoked

I don't expect more than 50C.
If the input voltage minus the diode forward drop gets above 16.7-18.5V, the TVS (rated for 15V continuous) will short, which will trip the fuse.
If current gets above 0.1A-0.4A, fuse trips.
Works in pretty much all cases.

The LDO is around $0.25, bat165 $0.1, TVS $0.15, fuse $0.4 in low quantities. You can probably get them cheaper in the US. So around $0.9 total.

Please keep in mind I am far from being an expert in electronics design

[kb1gtt]

I'm attempting to follow the design criteria posted here. http://rusefi.com/wiki/index.php?title=Manual:Hardware which notes an ISO spec for automotive grade power supplies. The MAXIM app note is easier to read and includes the key specs for the voltage surge and low voltage dips during cold cranking. This includes repetitive 100V peak voltage spikes on your 12V line. Also I follow AEC Q100 Grade 3 for the temperature limits, as it was the easiest spec to comply with for automotive environments.

Linear regulators are basically the same as switch mode regulators internally. See this video which explains it a bit more.


About wire EMC issues, I'll assume that we are dealing with frequencies below a couple kHz, so Standing Wave Ratio (SWR) is not a problem, and wire length is not causing RF issues due to length or propagation delays. According to power stream, I get the below ohms for 18 awg and 14 awg wire.
14awg is 2.525 ohms / 1000ft --> 0.01515 ohms / 6ft.
18awg is 6.385 ohms / 1000ft --> 0.03831 ohms / 6ft.
http://www.powerstream.com/Wire_Size.htm

For a 14 awg wire 10mV of ripple would be 0.10V / 0.01515 ohms = 0.66A ripple.
For a 18 awg wire 10mV of ripple would be 0.10V / 0.03831 ohms = 2.61A ripple.

Radio waves are primarily based on current and your antenna length. If you are measuring small ripple on a wire, you are likely radiating allot of RF. For the Frankenso R0.4 design, I could not measure any ripple on the 12V side of the regulator. which means it was below 1mV. But even then we had one person who had a noticeable noise in his radio, and required additional RF filtering. I would suggest the addition of an RF choke in your design.

Also we have one member who appears to test automotive devices, and he's noted that the ISO 100V peak is a repetitive load dump simulation. It's not just a one time issues like when a head light blows. I think it was repetitive when you had certain injector and ignition failures. AKA one injector or ignition failure will still let the engine operate.

[mobyfab]

In fact I expect the switch to generate more ripple on SMPSes than LDOs.

On bikes most of these wires are around 24AWG. (they are much shorter than in cars)
Do you have a link or information to calculate the required RF choke?

Thanks

[kb1gtt]

These two pages are kind of handy, for helping make predictions. But at the end of the day, you have to test and measure.
http://rusefi.com/forum/viewtopic.php?f=4&t=394&start=124
http://rusefi.com/forum/viewtopic.php?f=13&t=1180

[mobyfab]

Thanks!

This looks like it's adequate:
http://www.vishay.com/docs/34158/icm1206.pdf

[Rhinoman]

Linear regulators are basically the same as switch mode regulators internally. See this video which explains it a bit more.

I got half way through the video and then gave up. The fundamental issue that he is missing is that the linear regulator has an op-amp that drives the pass-transistor in its linear region, hence the term linear regulator. The control loop is very slow and the transistor never turns on or off quickly (if at all) and hence doesn't generate RF - an LM2940 has a bandwidth of around 10kHz, it easy to demonstrate the linear operation with a current probe on the front end. I'm sure that you could force it to switch with a fast switching input waveform but if you're putting that into a linear regulator then its time to go back to school; automotive transients, although harsh aren't fast.
The switcher has a comparator that drives the pass-transistor very quickly to turn it hard on and hard off, typically operating at hundreds of kHz with switching edges in the MHz region, the EMC issues are made worse by pumping short current pulses into the inductor which can generate additional ringing in the tens or hundreds of MHz regions and by the snappy response of the schottky diode. These current pulse are usually drawn from the battery and hence travel the full length of the wiring, pre and post filters are usually required and the PCB layout is critical. If the point of the video is to teach that switchers are as easy to use as linears in a product then its just plain wrong.

Incidentally if you look at a datasheet for an LM2940, the datasheet shows the feedback on the positive side and a PNP pass transistor, I doubt if any LDOs use an NPN because it would need a charge pump, also a switcher typically uses a FET.

[Rhinoman]

Also we have one member who appears to test automotive devices, and he's noted that the ISO 100V peak is a repetitive load dump simulation. It's not just a one time issues like when a head light blows. I think it was repetitive when you had certain injector and ignition failures. AKA one injector or ignition failure will still let the engine operate.

I designed ECUs for an OEM, not the biggest manufacturer but with production volumes of around 500k per annum. The ECUs were fully type approved so they were designed to meet ISO specs or higher and I spent months in various test chambers while those tests were run.

Load dump is a very specific set of circumstances were you get a battery disconnect with a discharged battery, this is explained in the ISO spec, it doesn't occur from a blown headlight bulb or a failed injector, those situations are no different from turning off the lights or ignition and are covered by a different test (Pulse 1).

The load dump test is usually 6 or 7 repetitions although I had one job where 100 repetitions was specified, that was for a large diesel engine that was being used on generator sets that had no battery, a battery on a trolley was used to start it and that was then taken away so there was a load dump on every start.

[Rhinoman]

Here's my power supply:
Basically I use an input diode, a resettable fuse, a TVS diode and a pretty tough automotive LDO.
In theory you only need the LDO since it's got protections everywhere. As long as your input cap is rated 60V+ and has no polarity (unlike mine).
It worked perfectly for me so far.

I think your TVS is too small, on a load dump you will get 100V through something like 1 ohm, which is a lot of current, the input diode or TVS will fail short circuit which will may blow the fuse, desolder itself or blow the tracks. The polyfuse will probably be very slow to blow and I'm a bit suspicious about its voltage rating too, there is another potential explosion there. Which LDO is it?
Have you examined the original ECU? if so what protection does it use?

[Rhinoman]

About wire EMC issues, I'll assume that we are dealing with frequencies below a couple kHz, so Standing Wave Ratio (SWR) is not a problem, and wire length is not causing RF issues due to length or propagation delays. According to power stream, I get the below ohms for 18 awg and 14 awg wire.

Up to 2Ghz and wire length can be critical - I once flew to Detroit, shortened a cable by 4" and then flew home again. The ECU had already passed stand alone tests but the harness on the vehicle was different. Previously I had increased the gate resistors on the injector high side drivers because of a radiated emissions issue at 20MHz. The tests don't go down to a couple of kHz, I don't recall the lowest frequency, its about 25kHz or 50kHz

[mobyfab]

Here's my power supply:
Basically I use an input diode, a resettable fuse, a TVS diode and a pretty tough automotive LDO.
In theory you only need the LDO since it's got protections everywhere. As long as your input cap is rated 60V+ and has no polarity (unlike mine).
It worked perfectly for me so far.

I think your TVS is too small, on a load dump you will get 100V through something like 1 ohm, which is a lot of current, the input diode or TVS will fail short circuit which will may blow the fuse, desolder itself or blow the tracks. The polyfuse will probably be very slow to blow and I'm a bit suspicious about its voltage rating too, there is another potential explosion there. Which LDO is it?
Have you examined the original ECU? if so what protection does it use?

LM2931.
The polyfuse is only rated to 24V, the TVS is rated to 15V continuous.
I actually expect the TVS to switch before it gets too high, or the transient to be short enough.
Which original ECU are you talking about? This is not an ECU in my case. The OEM one is potted so... (it's a motorcycle)

Thanks

[kb1gtt]

but with production volumes of around 500k per annum.

I could see how an extra $1 could be very important in that situation. I can also see how you could justify resources for full testing and validation. I don't have quality controls to rely such that I can ensure each board meets specification. I also I don't have full validation testing to ensure the design is operating properly. I agree I'm probably over-designing it a bit, but in low qty I see it as less important.

Thanks for the clarification about the load dumps. I keep forgetting it's the battery disconnect, not the head lights. Do you happen to know what surging you get from a head light or fuel injector? I believe that kind of surging is mostly caused by the alternator overshooting.

About EMC, I might be using a wrong standard here. At work I have to deal with 461 RE102 which goes up to 18GHz, as well as CS114, but for commercial stuff I expect CISPR 11 which ranges from about 20MHz to 1GHz. Is that a wrong standard? Is there a more appropriate standard for automotive applications? The switching regulator is in the KHz range, so I expect it to be naturally quite at the higher frequencies. I should probably add a section to my hardware design guide noted at the below link. I currently do not suggest an EMC specification.
http://rusefi.com/wiki/index.php?title=Manual:Hardware

Per page 55 (marked 49) found here https://snebulos.mit.edu/projects/reference/MIL-STD/MIL-STD-461-CrossReference.pdf see this snippet. It provides a vague overview of these different standards.

Capture.PNG (78.78 KiB) Viewed 17262 times

I understand that many LDO's will do the switching thing instead of operating in the linear range of the transistor to operate in the saturated state. This helps increase efficiency. I also understand that your upstream and down stream caps are required to help prevent your transistor from oscillating. Different LDO's will are more susceptible to oscillation that others. Also keep in mind even the older lower bandwidth linear regulators, are a high gain device. Any high gain device can easily oscillate when you do not expect it. If you expecting oscillation, like with a switcher, then from a functional standpoint you are all set as you have designed for this oscillation. You might not meet EMC, but functionally you should be OK by design.

[Rhinoman]

LM2931.
The polyfuse is only rated to 24V, the TVS is rated to 15V continuous.
I actually expect the TVS to switch before it gets too high, or the transient to be short enough.
Which original ECU are you talking about? This is not an ECU in my case. The OEM one is potted so... (it's a motorcycle)
Thanks

At 100V load dump through a 1 ohm source you would be trying to push approx. 80 amps through that TVs which would fry it pretty quickly. The TVS should only be to protect the regulator so you just need to ensure that it clamps below 60V. The load dump pulse is an exponential rise and fall over a maximum of 450ms so the clamping voltage makes a massive difference to the power dissipation.
I've dealt with a couple of motorcycle units - usually a Dremel with a brass brush takes care of the potting.

[Rhinoman]

About EMC, I might be using a wrong standard here. At work I have to deal with 461 RE102 which goes up to 18GHz, as well as CS114, but for commercial stuff I expect CISPR 11 which ranges from about 20MHz to 1GHz. Is that a wrong standard? Is there a more appropriate standard for automotive applications? The switching regulator is in the KHz range, so I expect it to be naturally quite at the higher frequencies.

CISPR12
CISPR25
ISO7637
ISO 11451
ISO 11452

The fundamental frequency of the switcher may be in the kHz region but it will be capable of generating noise well into the MHz region.

[mobyfab]

LM2931.
The polyfuse is only rated to 24V, the TVS is rated to 15V continuous.
I actually expect the TVS to switch before it gets too high, or the transient to be short enough.
Which original ECU are you talking about? This is not an ECU in my case. The OEM one is potted so... (it's a motorcycle)
Thanks

At 100V load dump through a 1 ohm source you would be trying to push approx. 80 amps through that TVs which would fry it pretty quickly. The TVS should only be to protect the regulator so you just need to ensure that it clamps below 60V. The load dump pulse is an exponential rise and fall over a maximum of 450ms so the clamping voltage makes a massive difference to the power dissipation.
I've dealt with a couple of motorcycle units - usually a Dremel with a brass brush takes care of the potting.

Indeed that would not be good.
I guess I could replace it with a 10-20 ohm resistor since I only need 100 mA tops.
The TVS can be replaced by a 30-45V one.
The only other thing connected to the input rail (behind the BAT165) is a LIN transceiver that's protected against 100V transients. (TJA1020 or equivalent)

[kb1gtt]

I've seen people also put in resistor / inductor choke, such that they simply filter out higher frequency signals, such that the transients voltage spikes are blocked. However with the spike duration around 450ms, that might be a bit hard to accomplish for this application.

I'll see if I can find some of the key parts of those specs to use as a guide for EMC. Perhaps I can find a freely published equivalent mill spec. One of the issues is that every different part of the world has different levels of caring if you radiate RF.

[AndreyB]

Sweet,... The DN2625 is specifically called out of by the app note .. So I will probably need to change D1002 for a slightly higher voltage. But that should not be a big deal.

please be more specific - placing an order for BOM.

[kb1gtt]

The current design uses a 14V D1002, with Q1002 having a max Vgs of -2.5 to -4.5, so the max output voltage is between 16.5 and 18.5.

The DN2625 has a Vgs of -1.5 to -2.1, which means we want D1002 to be around 19V - 2.1 = 16.9. I see the 1N5247B is 17V. Which is a max output voltage of 18.5 to 19.1.

So if we change to the DN2625, we would also want to change D1002 to 1N5247B

[AndreyB]

D1001 is BZT52C7V5-7-F which could probably have a 1NXXXX designation. Something like 1N5343BRLG would probably be good enough.

Something like 1N5343B would be through-hole, so we need a combined 1206/through-hole package for it probably? At the moment 1206 only.

[kb1gtt]

What is the desire for a 1NXXX designation? Thru hole is generally expensive to manufacture. I was thinking this was a question for someone who had a pile of 1NXXX diodes who wanted to use that instead of SMT diodes that they did not already have. Pretty much any 7.5V zener will work for this.

Could we clip the leads to make it looks like this

Capture.PNG (15.52 KiB) Viewed 17133 times

Looks like the oblong issue as well. Those fuses are slots.

[kb1gtt]

The 1N-series designation system was established in the 60's. I believe that all 1NXXXX didoes will be thru hole. Also I believe that the above Vishay diode is the same as a 1NXXX diode. However it does not have axial leads.