L6902 component choice

[Tfive]

Hi, I'm designing a DIY switching power supply around the L6902. Intended output voltage is <22V, input voltage 24V, output current in constant current mode between 0.6 and 0.8A.

My questions:

• the datasheet shows an inductor value of 22uH, is this value independent of output voltage and current? Currently I chose the part SRU1048-220Y, datasheet: https://www.mouser.de/datasheet/2/54/SRU1048-1391431.pdf
• Is the compensation network also independent of output voltage and current?

For neither of the parts I found any further hints in the datasheet, so I thought I'd better ask...

Thanks very much in advance for your help!

[Peter BENSCH]

Welcome @Tfive, to the community!

You should calculate and simulate the design with the eDesignSuite. However, you must apply at least 3V more to the input than you need at the output, i.e. at least 25V at the input with 22V at the output.

• For 25V input voltage and 0.8A, the eDesignSuite calculates a coil of 47µH, for which you should use a saturation current (not working current!) of at least 1.5A, 2A to be on the safe side, so that its magnetic field does not collapse at the peak currents.
• The compensation network depends on the operating point of the switching regulator, but the eDesignSuite will also take care of this for you.

Hope that helps?

Good luck!
/Peter

[Tfive]

Thanks for your recommendations, very much appreciated. Actually I can turn up the voltage on the input supply up to 28V if needed. So that should not be a problem. I will have a look at eDesignSuite tomorrow.

[Tfive]

Kudos to the engineers who designed this tool! First of all I love that it's web based, no need to install anything. Secondly it seems to just work. I work as a web dev myself, so I pretty much know what I'm talking about ;)

I was able fine tune the circuit as now the current and voltage demand are known more precisely, i.e. 21.5V @ 600mA.

I have one last question:
What voltage rating do I need for the compensation network? The design tool calculated some not-so-common values for the capacitors and with 50V rating these get rather expensive. Will I get away with 10V caps? Normally I do want to overrate voltage rating with at least 50%. as these are quite likely to come out as X7R or X8R types their capacity will vary with voltage, so how tolerant is the network to changes in capacitance? 

From my gut feeling I would suspect this is a rather low voltage node here, so probably no concern?!

[Peter BENSCH]

Thanks for the kudos, very much appreciated!

The internal supply voltage is 3.8V, so that the output of the VFB comparator can only assume a maximum of this value. With 10V capacitors you will be fine, 6.3V should also be sufficient.

You can check the influence of the component tolerances yourself with the eDesignSuite by changing the values of the R and C of the network and entering the tolerance limits. If you shift the gain margin to values that are too small (<45°, 60° would be excellent), you would have to adjust the values or tolerances.

Good luck!
/Peter

[Tfive]

OK, thanks again for your answer! As it seems I cannot adjust individual values of the compensation network. I just told the tool that I want more phase margin (it had chosen 45 degrees, I told it I want 60). Not only did it spit out much more standard values but I would assume that tolerance requirements on these are much more relaxed now. It came up with 68nF, 3kOhms and 27pF. All of the caps I can now get in 50V/5% and the resistor in 1%. Guess that is probably way overkill, but hey, it's a one-off design.

The design tool now recommends 47uH, I chose the Bourns SRN6045HA-470 with a saturation current of 1.9A, which should also be plenty. Man, this is fun - first time to lay my hands on switching regs.

I designed and built many of designs in the very high end audio field, used switching power supply modules before, but followed by lots and lots of passive filtering and sophisticated linear regulation to achieve SNR > 120dB even in small-ish power amps. But that's not the goal here, just a small LED arrangement for my seedling grow box - and to keep the temps down efficiency is key here :)

If you have no objections to my component choices I shall mark this as resolved...

[Peter BENSCH]

Glad you're happy, even better if you build it up and it works.

The values sound plausible. The coil is indeed overpowered, above all you certainly don't need expensive AEC-Q200 automotive certification, but such a shielded or semi-shielded design is still advantageous.

You should just be a little careful with the design. Although the currents are quite small, please:

• place the feedback line in a wide arc around the coil, never directly underneath it
• with such asynchronous buck regulators (so-called because of the external Schottky diode), very steep edges occur at node pin 1, cathode D and coil, which is why this node should be designed with the smallest possible spatial expansion and, if possible, as a circle or blob, so that you produce as little RF interference as possible. Almost the same applies to GND VIN, pin 7 and GND VOUT.

For larger currents, I would also make sure that the main current paths from VIN via pin 8 and 1 to VOUT and back via GND are realised as a straight conductor run and preferably without vias. Ideally, you should only work with one layer for switching regulators.

Have fun with the supply of your LED arrangement!

Regards
/Peter

[Tfive]

Thanks for pointing all this out! actually I will have to use rather "large" 1206 parts for easy hand soldering, so it will not be as tightly packed as maybe some industry designs. But I learned through a lot of reading and watching yt videos that "thou shalt keep thy current loops small" ;)

I have ground planes on both sides and only the two D+/D- lines from the USB connector to the GPIO chip need to run on the bottom layer, all the rest I managed to keep on the top layer.

Do you think this is packed tightly enough, not violating any design rules?

Edit: Note - the VR is on the back side an THT for easier access and better hand soldering. The output cap is also THT for easier soldering and is a panasonic solid polymer.

[Peter BENSCH]

You have a mistake in your schematics: cathode D2 should be connected to pin 1 not pin 3.

I would also not "wrap" the track from pin 4 (COMP) around C4, but simply mirror C4, C3 and R5.

[Tfive]

Oh, thanks for spotting that! I corrected that and now as I'm laying out components again I ask myself, who should be closer to the chip - D2 or L1? I guess D2 as it has a connection to ground and thus making the current loop smaller?!

I mirrored C3, C4 and R5, obvious, doh. But I will take my time and revisit this layout for the next several days doing small incremental placement improvements, as I always do before submitting a board to the manufacturer...

OK, this looks already pretty neat and tidy to me now, Kudos to your input!

PS: if not obvious, I place all my ground vias under component pads. Always did so because theoretically electrically the best solution and in practice I never had any problems with them sucking away too much solder.