3 rail power supply chip

[abecedarian]

Up to 60v input source compatible.
Configurable 'watchdog' function.
3 output rails:
: 1- 5v @ 800 mA;
: 1- 3.3 or 2.6v @ 500 mA;
: 1- 3.3 or 2.6v @ 350 mA.
Six 5v "trackers" @ 17 mA for sensors which can be paralleled for more current and are over/under current monitored.
All outputs short-circuit protected.
SPI interface for MCU to monitor and detect errors and configure some functions.
AEC qualified.
TLE6368G2

[AndreyB]

$11.54 as DigiKey

I like it! What I like most is that it's one integrated chip.

TLE6368-G2 is a multi voltage power supply system especially designed for automotive applications using a standard 12V / 24V battery as well as the new 42V powernet. The device is intended to supply 32 bit micro-controller systems which require different supply voltage rails such as 5V, 3.3V and 2.6V. The regulators for external sensors are also provided.

Is there any reason not to love it?

[AndreyB]

PS: $25 for 5 from China?

[abecedarian]

Be careful now. It spec's at least one 60v tantalum capacitors, and the inductors can take up a fair bit of board space.

6368_rough.png (33.51 KiB) Viewed 15248 times

It's in the blob at the left.
*note: I'm still moving stuff and searching for inductors, diodes and capacitors and alternatives, but the chip is shown.

[abecedarian]

Also has 4 'interrupt' lines that can be monitored by MCU to determine faults. So after initial configuration, an MCU can respond to 'interrupts' on dedicated wires, then query the supply to determine what the fault was and then decide how to handle it.

[kb1gtt]

Hey, where's the warnings? Don't you know that this is porn for people like me

What's the boot time? It seem to have some sequencing that's of importance but I don't think I fully follow what that boot sequence is. I believe this is intended to be always powered, then switched on and off. Probably via the 'WAKE" line. However I don't know if that simply put's it in low power mode. It looks like it doesn't fully turn off. I think it will try to keep 1mA of power to the MCU unless you do something specific.

I see it's not reverse polarity protected, so it should have the MOSFET reverse protection thing.

Are the recommended components hard to obtain? Even if expensive I think I would suggest making a one off board, then trying to decrease the costs after that's done. It looks like you are planning for a small PCB, which I think is a good idea, as this may have some undesirable complications, which a small board would iron out.

Hmmm, do we need more than just 3.3V or 5V. I guess that some other chips want 5V. So we might as well use 5V for the digital stuff. Can the STM32 be powered by 5V, or is it just 5V tolerant on the IO lines. I guess we want a 5V and a 3.3V. Are there times when we might want that ability to turn off those 17mA sensor sources?

Take note about section 5.8, "Layout recommendation" I think this means you really want the coils on the back, and chip on the front.

I wonder what the efficiency is going to be. I'll assume worst case when fully loaded. So it would provide (5V*.8A)+(3.3V*.5A)+(3.3V*.35A)=6.8W. I'll assume 80% efficiency, so it would consume around 1.36W for a total draw from the battery of around 8 watts. The chip will need to dissipate around 1.36 watts, and will need to stay under 150C. So lots of via's under the chip and perhaps a heat sink. Even a little heat sink like this one from adafruit http://www.adafruit.com/products/1041#Description is 26C/W, so I would expect just this chip with this heat sink to increase the temperature by around 35C to 50C. If we assume an ambient that goes up to 40C, the junction temperature would be about 40C+35C=75C. Hmmm, I also see the thermal resistance could be 37C/W if you use the PCB as the sink to ambient. So it should be reasonably good in terms of thermal consideration.

All in all, I think it looks good. I'd like to see some better specs on ripple, and long term wondering. However those are probably fairly good, or good enough.

[abecedarian]

I was wondering if this would get Jared's attention.
And I appreciate your comments, J.

Yes, it requires power at all times and is awakened by power applied to the 'wake' pin, presumably by turning the key switch to ON, RUN or similar.
I don't see anything regarding the true start up time, as in how long it takes the regulators to power up, though I would assume it's not that long of a period to cause issue.
Standby, according to the DS, is a small 2.5v regulator capable of 1mA out, but only draws 30uA when the unit is powered down. (Page 12, 2.4 of the datasheet).
I've seen recommendations to populate the area under the chip with thermal vias, but nothing regarding a heat sink.

Regarding placing inductors on back, the recommendation only suggests placing those as close as possible to their respective pins, and having a ground layer under them. My drawing is only two layers, but the top and bottom layers are grounds... note the blue rectangle outlines the bottom layer plane, and the red polygons (one for power and another for analog / op-amps) outline top layer ground planes / fills. Simple guess is that the inductors are less than 1/2" trace length from the pins, as I have laid things out. I'm still working on it, so expect it to change. BTW this is a board to mount on top of the TI board I've mentioned elsewhere, as evident by the 3 dual-row headers... and you can ignore the MAX9926 near the top-right as that is going to move; it was just something I placed out of the way to gauge how much room I had. Heck, I don't even have injector or ignition drivers placed yet.
Rough render:

6368_rough_render.png (36.68 KiB) Viewed 15242 times

I'd be up to donating a bit to help fund a prototype / test board.

[abecedarian]

Should mention the 3 4.7uF caps at above middle are for the 3 output rails, but I haven't connected the rails yet so they are there but not where they should be- close to the circuit they are powering. They are tantalums, if I remember, but could probably be replaced with aluminum electrolytic types.

[abecedarian]

Are there times when we might want that ability to turn off those 17mA sensor sources?

Those default to 'on' when the unit starts up, and have to be turned off via SPI. Pages 12, 17 (and 19 regarding SPI queries of status) of the PDF.

I would assume that if the engine is not running, those could be turned off if they're not needed. Datasheet suggest to leave them unconnected if they're not needed.

[abecedarian]

It seems there is a demo board for the TLE6368: http://www.digikey.com/product-detail/en/DEMOBOARD%20TLE%206368/DEMOBOARDTLE6368IN-ND/1561198
$154 though.

[abecedarian]

Seems they have a 'new' one, with 5v out at 800mA and selectable 3.3 or 2.6v out @ 700 mA, and third output with selectable 1.5 1.2 1.3 (depending on chip), and with 2 'tracking' regulators giving 105mA and 50mA each.

http://www.infineon.com/dgdl/DCDC-Converter-TLE7368-Data-Sheet-V21-Infineon.pdf?folderId=db3a304314dca3890115039cf70d0bf0&fileId=db3a3043163797a601163918026f01cf

[kb1gtt]

I have mixed emotions about the slightly newer chip. Oddly enough, the reasons I like it are also the reasons whey I don't like it. I like monitor chips as they ensure the system is operational, however I also don't like them as they can be a hidden reason why the system doesn't operate. I'm concerned that the monitor features may be to much for a small effort. I also liked the original chip's use of switching buck pre-regulator followed by ripple rejecting linear's. This chip uses two buck regulators, which may prove to be harder to control the ripple. Hmmm, 5V +/- 2%, I think this is less accurate than the original chip as well.

I had another reply but ran out of time to finish it this morning. More to come tonight.

[kb1gtt]

On a good note, that demo board shows that the chip is on the other side, and the coils don't appear to be on this side. However it doesn't show the actual chip and it's actual layout. It also doesn't specify if it passes EMC, or what kind of ripple it provides. Generally when I see the comment in a datasheet about keeping a GND via or decoupling cap as close as possible, it means you put a via that touches the pin, or you position the decoupling caps before other components. I might suggest considering placing the TLE chip, then the coils, then the other components. Those coils are going to be high magnetic fields and will have critical ground loops to deal with.

About tant vs electro, beware the differences in ESR and inductance. Remember an electro with the foil wrapped in a barrel is very similar to a coil. It can pickup or generate it's own magnetic fields, were you don't generally get that with a tant. Often the effect is increased heat, or increased ripple.

We'll that's what I got for the moment, keep up the good work. Me likes this chip.

[abecedarian]

Been digging for photos of something showing that chip and came across this page which has a photo of both sides of the evaluation board:
http://ehitex.com/evaluation-boards/infineon/218/tle6361/68g-32-bit-system-supply

They don't seem to care too much about inductor placement.

[kb1gtt]

Hmmm, looks very similar to your layout. I say build it and we'll see what happens. Also looking over the circuit a bit closer, those coils are more like RF chokes than a device that switching a heavy load. So the variations in magnetic fields are likely to be less than I was thinking. All in all, if it's wrong, it will probably only be a mild RF issue, or mild ripple issue. So I'm thinking it's not really all that critical. I wonder why they caution about the placement in the datasheet.

How many parts can be made with Seeedstudios OPL? They are low cost assembly.

[abecedarian]

Well, I twerked... um... I mean tweaked the layout a little:

6368_rough_feb152014.png (34.13 KiB) Viewed 15379 times

I see it as viable based on visual comparison to the photos on the site I posted above... may rotate the inductor at the bottom left, 180*, and move the diode below it so as to accommodate that.
Also, Infineon didn't tie the bottom 'pad' to the grounds so that's mucking with vias, but they say:
"bottom heat slug and GND corner pins are connected", just not in the library part.
This results in design check issues when trying to place thermal vias under the chip.

Brought SPI out to the most convenient port ([MIBSPI2]; vertical traces near center).
Routed sensor signals from the connector to the ADC buffers and changed resistors around the op-amps from 0805 to 0603 since they won't carry significant current.
Routed LDO_0 (5v) and LDO_1 (3.3v) over to where they're needed; LDO_2 is in a quandary- don't need it, but....
Blue 'bottom plane' trace is 12v, meant to go to injector drivers... somewhere. Ignition drivers will likely be simple transistors to switch LS2 ignition coils at CMOS logic level and little current.
I might slide the power up and put the injector drivers down in the lower left corner. Either way, I'm not too happy since signals will cross: either outputs from drivers will cross SPI lines, or the logic-level signals will cross power. I think the latter is a better idea, but....


As for building it, unless 47uF/60V tantalums come down in price significantly ($65 at Digikey), it'll be a while before I build anything. So I might be going back to the TI supplies I posted elsewhere. They may have taken up a bit more board space but the BOM's for 5v and 3.3v supplies were under $10 each.

[kb1gtt]

A 50V tant is far more tolerable. In qty 1, it's $15.80 http://www.digikey.com/product-detail/en/597D476X9050Z2T/718-1408-1-ND/1820165 it's still blah, but getting more better.

I see this 47uF has the 47uH in between it and the battery. I seem to recall a surge from the battery/charging system is fairly fast. So I don't think this cap would see the 100V spike that you could get on the battery/charging system. Perhaps that should be simulated to make a prediction about what might be seen at this cap. I wonder if it needs to be such a high voltage rating. Perhaps that's because of the switching mechanism and perhaps the voltage could spike under certain conditions. I also wonder if we could get a cost reduction by using qty 2 caps at say 25uF each. Hmmm, tants are very volatile if you exceed the voltage. Perhaps it's good o add a TVS to clamp the voltage. http://www.digikey.com/product-detail/en/TA6F43AHM3%2F6A/TA6F43AHM3%2F6AGICT-ND/3913288

[abecedarian]

I have to leave the sim up to those with the tools. The basic schematic follows what's in the datasheet.

[abecedarian]

Here's another option.

Unfortunately not automotive certified, but rated -40 to + 85C operating range.
7-36v supply.
Single (either) 5v or 3.3v @ 1.5A output.
Output continuous short-circuit protected.
Direct drop in replacement to TO-220 package 78xx regulators.
$4.30 USD each.

[kb1gtt]

I have used a very similar one of those. The ripple was blah, also it was very picky about having a solid supply. If the supply was a bit weak, or if you simply used some thin wire, the inrush would cause a voltage drop which would cause it to fail to start the switching circuits. These folks have an interesting option, http://www.elektor-labs.com/ sorry my ISP is garbage so I can't find an exact link. I know thy have a T0220 drop in replacement switch mode regulator that's much higher quality than most.

[kb1gtt]

I think I found what I was looking for. See this link for more info about the drop in replacement. http://www.elektor-labs.com/project/switched-7805-replacement-tht.12556.html

[kb1gtt]

Ah here we go, this is the one I remembered seeing. http://www.elektor-labs.com/project/switched-7805-replacement-1202121.11952.html