Kahenraz wrote on 2022-06-26, 03:45:
I am not an engineer, but I don't understand why building a small VRM is expensive and risky for parts being discontinued. Can't something be achieved with common jelly bean components that are pretty much guaranteed to exist forever in some form or another?
When I looked into it to design the first version I found basically a few old pentium 2/3 and ss7 vrm controllers that were up to the job. Many of them looked like old stock Specifically the one I was designing around was chosen for the specs of it, dual inductor design, 500khz best of the best. Here is the issue:
The only place I found these chips the last time I looked was some guy selling old ones as a lot on ebay for a bunch of money. (A lot of money) I would have had to buy them and distribute them basically. When they were gone, no guarantee that I could have found another. The tweaker would likely have had to be re designed for another probably old vrm controller.
This way completely avoids all of that.
You don’t even have to use CalamityLime’s device. There are a bunch of random chinese power supplies that are for sale for like 10 bucks. That’s like as much or less money as the controller chip of building one yourself on the interposer, then you have to buy like 30 more parts on top fo that and solder them on.
Where as there are like 100 complete solutions of varying grades of output quality available right now that require no work. With the best ones being lab quality which should give a better singal than anything we could fit on the interposer. And the middle of the road ones being at worst equal to what the likes of kingston and powerleap used.
Yeah, this is not only the more secure way to go, but definitely the cheaper, and the easier way.
If you want it to fit all on the interposer though. A really easy way to do it though is the way 486 interposers did it, using a linear regulator and a few capacitors. But this way makes a lot of heat. Probably actually too much without doing dual regulator.
So the easy solution for on interposer is still more expensive, more work and worse result.
Yeah. I’m 100% convinced a external psu is the best way to go here.
But, if you wish to finish the previous design, you may take it and finish it, or adapt this current design to work with linear regulation.
It’s all gonna be more work and more money though.
Yes, as calamitylime said to the point of the crappy buck converters that I just know people are going to go and order for these, no matter what we tell them to do. Yes! a quality one should definitely be sourced. At the very least you should not try one that someone here hasn’t already tried and proven won’t blow up your chip. Because quite frankly. Some of the really cheap ones just might. There are specifications for noise in the datasheets for these units, and there are requirements for noise in datasheets for the processors sometimes. I would invite a member with extensive knowledge about this aspect to comment as to what to look for in a good vrm. What I know of the subject: more output caps the better, higher switching frequency the better, more inductor phases the better. And running the output wire through a ferrite choke may not be a bad idea.
Nice thing about a external unit is you can add capacitors to it and other smoothing equipment to it pretty easily.
One final and very important word of warning. Many of these converters are capable of output voltages far in excess of what the chip can handle. I’m talking it is possible to plug it into 12v and push 11v straight into the cpu.
DON’T DO THAT! Lol
