[info] Power supply - kb1gtt

[kb1gtt]

Nice article. Thanks for posting it. I'm absorbing it now.

I'd like to replicate the simulation equation they note which appears it was done in MathCad, I wonder if that can be done in a free tool. I don't have MathCAD and I believe it is a paid for program. I suspect the equation can be plugged into SCILabs or similar, but I'm not sure how.

The example uses an 87uF cap, we have a 330uF cap. I believe that's going to drop our front end impedance significantly, but who knows as the larger cap is likely adding more circular / inductive bits with the electrolytic cap. I also don't understand why they have a near DC impedance of 25 ohms. I suspect they had a 25 ohm resistance up stream, which is not shown in the example schematic. I expect the typical 12V battery to be in the mOhm range, and the wire to be significantly sized such that it doesn't have that kind of resistance. I should probably note on the schematic that I expect a minimum of 18awg for the 12V+ wire (that's about .13ohms for 10 ft) with a GND conductor that's minimum 16awg. I think we have a much lower front end impedance than what is shown in that PDF.

I expect the ECU to have it's own GND wire running to the battery. Frame GND's are not friendly. I should probably add a note to the schematic. Keep in mind you will get many issues if you have around .1 ohm to .01 ohms of resistance on your GND leads. Also remember such low resistances are hard to identify with a normal DMM which only has a chance of measuring 1 to 5 ohms.

[kb1gtt]

I didn't know about TI's WebBench simulation. I ran the simulation, and was surprised they had all but one of the components in the component libraries. The only thing I had to kind of guess at was the inductor, and they appear to have one that was really close to what we have been using, but from a different MFG, so this simulation is likely very good. Also some of the claimed efficiency and such data points match with things I've posted. So that seems like a good real world validation.

This is the PDF copy of the simulation which includes graphs noting an efficiency between 79% at .25A and a peak efficiency at 1.25A of about 85%. As well it notes the average input current is under 1.5A. Unfortunately I don't see an input impedance prediction.
https://sourceforge.net/p/daecu/code/HEAD/tree/Hardware/trunk/simulations/LM2596_5V_power_supply/Design_Report_12V%2C14.5v%2C17V_webench_design_1077162_276_721820786.pdf?format=raw

Here is the thermal simulation, which notes the junction temperature is well below the max of 150C when the ambient is 85C. It also notes how many watts each device is dissipating and shows that those devices are below their max temperatures. All in all it shows that it's meeting the design specs, which is showing good things.
https://sourceforge.net/p/daecu/code/HEAD/tree/Hardware/trunk/simulations/LM2596_5V_power_supply/Thermal_85C_ambient_webench_tsim_1077162_276_277452011.pdf?format=raw

I had it in my head that it was good for 5A, but it's only good for 3A. Good thing the notes printed on the PCB have that correct, and match the simulations. So I'll have to remember it's 3A not 5A. I hope my brain fart didn't miss lead anyone.

[Rhinoman]

The pdf shows an input ripple current of 1.5A at an output current of 1.25A, the 'standard' wire length is 2m which would be around 0.08Ohms. that will give you a ripple current of 120mV, on the 12V wire, the ground wire is carrying lots of other currents and those will either add or subtract to the total potential difference, typically you will see around 0.25 to 1V of total ripple, you can measure that with a scope. The electrolytic will deal with some of the low frequency noise but the switcher will generate a lot of noise up to around 10MHz, if you can't see any then there is an issue with either the equipment or the test method. The simulation also relies on the electrolytic being able to cope with over an amp of ripple current which is a lot for a small cap. You also need to consider susceptibility, there is nothing to stop high frequency noise coming into the ECU through the 12V supply line.

I'm sure that I had a demo version of Mathcad at the time. The source impedance increases with frequency up to |Z| = 50R above something like 10MHz, I will find the details on the LISN.

[kb1gtt]

I think you are referencing the "IC Ipk" graph on page 2 found here https://sourceforge.net/p/daecu/code/HEAD/tree/Hardware/trunk/simulations/LM2596_5V_power_supply/Design_Report_12V%2C14.5v%2C17V_webench_design_1077162_276_721820786.pdf?format=raw I find this simulation results a bit unclear. I believe that chart is conveying the peak switching current, and that the higher frequency stuff is coming from the cap, and not over the 12V wire. I believe it's the max switching current per page 4 operating value's brief description. However I really don't know what the is really trying to convey. When I tested it I tested it with alligator clips plugged into a bench supply. I'm guessing that's at least 16AWG perhaps 14AWG, and it was perhaps 1M long, so 2M total length including the GND wire back to the supply. In that setup seeing 75mVp-p to 225mVp-p might indicate significant 12V ripple current, or it might indicate my bench supply was a bit soft. I believe that front end of this GPS-3303 has significant capacitance, but I have no idea if it's impedance is close to a 12V battery. I should probably put in a series sense resistor and take some measurements, which I could then use to compare against the PDF simulation.

The LM2596 datasheet draws the switches as transistors, while the impedance matching PDF's example used the LM2675 which shows MOSFET's for those switches. I would expect the transistors slew rate would limit switching frequency well below the MHz range. I would expect potentially higher frequency content from the LM2675. I haven't seen significant frequency content above 1MHz on the quad.

[kb1gtt]

It appears this Chinese switcher chip is likely a LM2576 marked LM2596. You can see on the a primary switch frequency of about 50kHz as noted in the picture I posted here http://rusefi.com/forum/viewtopic.php?f=4&t=569&start=168 Notice that's 150kHz/div, with the first spike about 1/3 of the first block. I found this claim in comments of the below web page.

http://www.kerrywong.com/2015/07/24/lm2596-dc-dc-converter-module-testing/

I'm tempted to add an inductor location to the power supply input, such that we have a physical place to put such a choke if it's needed. As it stands now I don't have solid indication that it's needed, and I don't know what values to put there if it's needed. Hmmm, how to resolve this concern..... For now I guess I'm reading EMC PDFs about switch mode supplies. I'm currently absorbing this one http://www.ti.com/lit/an/snva489c/snva489c.pdf

[kb1gtt]

From this link, http://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/15/ it looks like the LISN is a 5uH device wired as noted in this schematic.

CISPR 16-1-2 5uH Schematic.png (18.13 KiB) Viewed 21461 times

Here's a picture of my conducted emissions probe. However I have no idea if that's 5uH as noted above or not.

probe.JPG (23.41 KiB) Viewed 21461 times

This web page is the closest thing I have to a datasheet.
http://www.solar-emc.com/MIL-STD-461F.html
I guess I can connect that probe and sniff it even if it's not the correct impedance.

[Rhinoman]

I came across this today, I haven't looked at it in any detail:

http://www.diodes.com/downloads/9551

[kb1gtt]

Nice to see they make these specifically for automotive applications. The one I'm using now is 30V 4.0A and Rds = 50mohm. The one mentioned in that PDF is physically much larger, 40V and Rds = 11mohms. Seems the ones that diodes inc. list for automotive and SOT23 are only in the 1A range. I wonder if an automotive 4A or more SOT23 can be found.

[Rhinoman]

P FETs have higher RDSon, in the past I've used an N FET and bootstrapped it from a switching regulator. You won't get a very big die in a SOT23 and the thermal performance isn't that good which limits the current it can handle.

[kb1gtt]

Looks like this is an option http://www.mouser.com/ds/2/427/sq2301es-534673.pdf it allows 3.9A. The upstream voltage clamp was 20V, I just changed that to 19V.

[Rhinoman]

Looks like this is an option http://www.mouser.com/ds/2/427/sq2301es-534673.pdf it allows 3.9A. The upstream voltage clamp was 20V, I just changed that to 19V.

3.9A at 120mR is 1.83W, Tja is quoted as 166C/W with Tj max 175C.

[kb1gtt]

The gate is being driven by a 7.5V zener, the 125mohm is when driven to 4.5v. I expect that when driven to 7.5V it will have an Rds closer to .075 ohms. See attached picture that shows my eyeball of the graph that indicates the Rds. So that's 0.31V/4A = 0.075ohms instead of 0.125ohms. The design is for 3A as noted on the power supply schematic. So watts would be 0.675W. At hotter temperatures it could have a 1.5 multiplier on the resistance. So at a hotter temperature it could be 1.01W.

Rds.png (36.89 KiB) Viewed 20506 times

The chip has 50C/W to the pads on the chip, the additional thermal resistance is dependent on your PCB layout. The 166C/W assumes reasonably small traces connected to the pads, with a certain distance to your heat sink. My layout has fat traces, with thermal via's, so I expect a more thermally conductive path to the heat sink. Unfortunately I don't know the exact thermal resistance. See "figure 8" and "table 4" on this PDF for how they measured the thermal resistance and got about 200C/W instead of the MFG's datasheet which notes 166C/W.

http://www.njr.com/semicon/PDF/package/ae02220.pdf

I'm expecting ambient to be up to 85C as noted here http://rusefi.com/wiki/index.php?title=Manual:Hardware#General_suggested_environment, the chip can handle 175C, so we have 90C to work with. That would be a need of between 90C/1.01W to 90C/.657W which means we would need to be 90C/W to 133C/W or less. We know we have 50C from the die to the lead, but will the PCB layout be less than 40C to 83C to ambient. I don't really know. I suspect it's reasonably close. As well I suspect we won't be running a continuous 3A at a continuous 85C ambient.

Perhaps the 3A should be derated a bit for thermal purposes. Or perhaps I can find a chip with a lower Rds, or perhaps I can put in a larger chip. For now we don't have any applications that come close to the 3A limit, so the real word applications shouldn't have any problems. Thermal issues are hard to predict until you build it and test it while looking at it with a thermal camera.

[Rhinoman]

I found another interesting app note regarding high frequency switching noise here:

http://www.linear.com/docs/11877

[AndreyB]

It has been some time, where have we left it? Should I populate current version 0.4 for a validation?

[Rhinoman]

I haven't had time to look at this in detail, I will be producing a different board for my Vitara because I need a peak-hold injector driver. Currently I'm working some distance from home and I spend 3 1/2 to 4 hours a day commuting.

[kb1gtt]

I know it's a bit OT for the power supply thread, but for lowz you may find this interesting http://puma.freeems.org/ I was one of the people working on that and it worked reasonably well at least from the very few data points we had to work with the lowz worked well.

I've been busy busy busy. Perhaps at the holidays wind down I can find some time to jump back into this. I still have the spectrum analyzer almost ready to run, and I have the frankenso sitting here almost ready to connect and test. But I haven't found the time to actually run it.

[AndreyB]

Just placed an oshpark order for rev 0.4 board.

[AndreyB]

It does produce 5 volts - now it probably needs some deeper testing.

[kb1gtt]

Very nice, very nice.

[AndreyB]

I remember you ran some tests of the power supply but I cannot find these scope traces, do you remember where were these?

[Abricos]

this device for sale ???
or is still under development ???
and what is its value or Price ???

[kb1gtt]

Hmmm, I agree I should have some scope shots, which show the ripple and such. But I can't find them either.

[AndreyB]

On the previous power supply design (the one users on Frankenso 0.4) the two through-hole diodes were 1N4747, were they? Why do we not mention 1N4747 anywhere?

On the newer design we now have two different through-hole diodes, do these have an 1Nxxxx code?

[kb1gtt]

The power supply on Frankenso R0.4 calls for fairchild,BZX79C20 which looks to be about the same as the 1N4747. I forget if the 1N4747 was used at some point. A quick look seems to show it's the same as the BZX79C20. I may recall that was how it was listed at sky-macau, but it's not listed that way right now.

In Frankenso R0.5 D1002 and D1003 both have 1NXXXX designations. D1001 is BZT52C7V5-7-F which could probably have a 1NXXXX designation. Something like 1N5343BRLG would probably be good enough.

[AndreyB]

In Frankenso R0.5 D1002 and D1003 both have 1NXXXX designations. D1001 is BZT52C7V5-7-F which could probably have a 1NXXXX designation. Something like 1N5343BRLG would probably be good enough.

What is the difference between 1N5343BRLG and 1N5343? eBay likes 1N5343 much more.

Please confirm 1N5343 is the one we want and please give me the other ones when you have time, I will order some from eBay.

Now I see 1N5244B and 1N5359B as part of the BOM

[kb1gtt]

It looks like the base number is 1N5343B, then the RL means it was produced on a real / tape of 4000, instead of a box of 1000. Then the G seems to just hang out. I don't see any real differences in the diodes. I think they difference is simply packaging in how they are stored.

[kb1gtt]

So I finally got around to testing this board. It looks like something wrong with the Q2 MOSFET. It's simply acting like a wire, and not clamping the voltage.

Ooops, I see the tab of the Q2 MOSFET should be drain, and it was the source. Oops. It worked for @russian, as the diode was conducting. I cut the trace, lifted the MOSFET leg and added some wires to re-wire it. Then it worked as expected.

My bench top power supply goes up to 65V, and it works fine for that range. Q2 keep the voltage at about 16.5V and didn't kick in until about 14V. Also some good news the axial inductors worked just fine. i was a bit concerned about the magnetic fields, but it doesn't seem to matter that those fields are vertical.

So I'll be doing R0.5.

[AndreyB]

ordered rev 0.1 of https://svn.code.sf.net/p/rusefi/code/trunk/hardware/frankenso-PWR-Supply/

[kb1gtt]

This board is cut from the Frankenso R0.5, it's just the power supply and it's on a 2 layer board.

[AndreyB]

How about we replace IXTA6N50D2 with IRF640S / RF640SPBF? Is it the same package / size / pinout? $4 cheaper.


[video][/video]