Bit of fiddling around and I think I've got a carrier PCB that fits with the pinout from that datasheet in my previous post. I've used that trick of putting the PCB between the pins of a connector (I think I've seen that before on a programmer header). I'm reasonably happy with the routeing, and there's space to add an 8mm diameter poly input cap at the top left, close to the input to the module. Because of where the Vout pin on the module is there's no where good to put an output capacitor without making the module 8mm taller, but I think it should be ok just using the output caps on the module itself. Something could be bodged in place if it turns out it's needed. It's approximately 40mm wide and 30mm tall, with all the components on the back. I don't know if it matters that the components are on the back, but that made the routeing a lot neater. It's got a right-angle 16 position rotary switch fitted, so that should be accessible from the top, and using the resistor values from the previous carrier should give 1.3V-3.55V in 0.15 steps. It's easy enough to work out different starting points and step sizes. if needed.

I've put a jumper in for the V3 to Vi/o connection, just in case there's a need to have a different voltage for io. I've routed the sense pin to the sense+ input on the module, and also the power good. I've routed the Disable signal to On/Off, but I need to check up on the polarity of the input to the module (from reading Intel AP579 is it high to turn off the VRM outputs).

What do you think? I'm going to add that input capacitor, check on the on/off polarity, then put in some text.

snufkin wrote on 2022-01-24, 15:50:

Bit of fiddling around and I think I've got a carrier PCB that fits with the pinout from that datasheet in my previous post. I've used that trick of putting the PCB between the pins of a connector (I think I've seen that before on a programmer header). I'm reasonably happy with the routeing, and there's space to add an 8mm diameter poly input cap at the top left, close to the input to the module. Because of where the Vout pin on the module is there's no where good to put an output capacitor without making the module 8mm taller, but I think it should be ok just using the output caps on the module itself. Something could be bodged in place if it turns out it's needed. It's approximately 40mm wide and 30mm tall, with all the components on the back. I don't know if it matters that the components are on the back, but that made the routeing a lot neater. It's got a right-angle 16 position rotary switch fitted, so that should be accessible from the top, and using the resistor values from the previous carrier should give 1.3V-3.55V in 0.15 steps. It's easy enough to work out different starting points and step sizes. if needed.

I've put a jumper in for the V3 to Vi/o connection, just in case there's a need to have a different voltage for io. I've routed the sense pin to the sense+ input on the module, and also the power good. I've routed the Disable signal to On/Off, but I need to check up on the polarity of the input to the module (from reading Intel AP579 is it high to turn off the VRM outputs).

What do you think? I'm going to add that input capacitor, check on the on/off polarity, then put in some text.

I like it!

This one is definitely big enough for a voltage display though!

Could stick a footprint for a little voltage display on the top or back?

Pin 9 on my vrm module is connected to ground. There´s a trace from Pin 13 to Pin 1/2 and half way between there´s a tap leading to pin 9.

In general, there are mainboards with 30-pin VRM-connectors that have two (parallel) jumpers in the empty socket to connect the combined supply voltage from the onboard VR to the CPU´s Vcore pins. And there are mainboards with 4 jumpers (2 x 2 parallel) one pair connecting the supply voltage to Vcore and one pair connecting it to V I/O.

The Murata mdule is available as a "p-type" or "n-type" depending on the polarity is needs to be turned on. (Both types are turned on when the input is open.)

majestyk wrote on 2022-01-24, 16:51:

Pin 9 on my vrm module is connected to ground. There´s a trace from Pin 13 to Pin 1/2 and half way between there´s a tap leading to pin 9.

Thanks, sounds like I've got the right pinout at least. I was wondering about the motherboard routeing to the header. Pin 12A should be sense, which on the motherboard ought to be a fairly thin trace heading toward the CPU socket, which connects directly to Vcore somewhere under the CPU. It's possible (but probably not very likely) that there's a similar trace going to one of the ground pins on the header, so that it could be used for the sense- input on the module.

That pin 9 was a right pain. There must be a reason why it's buried in with the Vcore.

In general, there are mainboards with 30-pin VRM-connectors that have two (parallel) jumpers in the empty socket to connect the combined supply voltage from the onboard VR to the CPU´s Vcore pins. And there are mainboards with 4 jumpers (2 x 2 parallel) one pair connecting the supply voltage to Vcore and one pair connecting it to V I/O.

I've put in a single jumper for connecting V3 to Vi/o, if that's done by soldering in a piece of wire then that should be fine.

The Murata mdule is available as a "p-type" or "n-type" depending on the polarity is needs to be turned on. (Both types are turned on when the input is open.)

Yeah, thought I'd seen something like that when skimming the datasheet at the start. So the Intel app note says that Disable is driven high to turn off the VRM. It doesn't say about signaling to turn it on, so either low or float. From the Murata datasheet I think that means 'N' suffix is needed for that to work properly, where 'off' is an input voltage of 2V to Vin_max (14V). If 'P' suffix is used then it will always be on. Which only matters if the motherboard expects to be able to turn it off (auto-detection of CPU voltage requirements?).

Anyway, even though I've routed the signals if they cause any problems the pins can always be pulled so they're not connected.

Here are some specs from the Intel 430Hx documentation:

Ahahahahhahaha. Thanks for posting the picture of the header. It struck me odd to have a thick trace going to that 3rd pin, which I had as being another Vio pin on that side. Plus that there's a trace routed to pin 11, which I had as being UPVRM#. Then realised that it would make sense on the other side, where those pins are +12V and PWRGD. So I went back and checked on that RCB002 and it looks like for their top view, they have A and B back to front. Now to rejig some traces...

Honestly, what's wrong with header numbering. It goes 1,2 then 3,4 then 5,6 then 7,8 then 9,10 ... then 29,30. Odds and Evens. Some then confuse things by putting numbers in the wrong order. Then this manages to introduce letters.

Thanks also for those snippets. So it sounds like there's only the Vcore sense, and the pin 9 ground is only there to annoy people trying to route it out through the Vcore. Also sounds like it's up to the motherboard to sort out the power good signal, so should be ok with the module holding it low until the output is ok.

You´re right, I just double checked, this is the true Pinout:

Yeah. I spent a few minutes originally checking I had them the 'right' way around. Then once I saw your photo a few minutes not being sure if they were right or not. Then more than a few minutes trying to figure out where I went wrong, before finally convincing myself that their diagram was wrong.

Still no harm done. I mean, the only consequence of getting that wrong is putting 12V on to Vio, and that can't possibly cause any problems.

Think I've got it rerouted. Now with less magic smoke releasing goodness. On the plus side, it makes the Vcore routeing a tiny bit nicer.

Ok, I think this is it, including gerbers. I've got vias scattered over the ground copper pours, and the Vcore is tucked down at the bottom, so routeing out for a display would be a pain. Could always glue something like this (forgive the ebay link) https://www.ebay.co.uk/itm/392790475976 to the back of it, solder the red+black to the back of the input capacitor and yellow to Vcore on one of the connector pins.

There may be mistakes I haven't caught, so a little bit of testing before connecting it to anything old would be sensible. Say if anything's amiss.

[forgot to say, I moved the resistors to the back to make it easier to change them if different outputs are wanted]

[edit: I've just made a small change and connected sense- to ground. I found a datasheet for a different Murata converter, and that included a bit of a description about the sense pins. Apparently they are internally connect through resistors to Vout and Gnd, so it will work without them connected. But it is better to connect them. So I've got Sense+ on the sense connection from the motherboard, and Sense- shorted to Gnd.]

Ah, having said that was it, I realised there was room on the back for a couple of 7343 SMD poly caps for input and output caps, which I think can be got as 100uF/10V at 2mm thick. Probably optional, but could help with ripple.

I should also have said that I don't actually have anything to try these with (and shouldn't really be spending time on this, but hey..). But if anyone does build one please say if it works.

This is a good idea, filtering for (even) better ripple can be done much more efficient on this VRM design due to the multiple ground connections that the (dedicated) 30-pin connector provides.
The small TO220 module highly depends on the quality and distance of the nearest ground connection of the mainboard.

I have dozens of mainboards with a 30-pin connector and I own just 3 linear VRM modules (that are restricted to 2.8 / 2.9 V) so I´ll definitively produce a small series of this design that makes it possible to experiment with all the K6-2+, K6-III+ and Tillamook CPUs.

Where do you guys have your PCBs made btw. when you need maybe 50 or 100 of a small form factor like this?

I don't have a lot of experience ordering PCBs. For some small work stuff I've used a company in the UK ( https://www.pcbtrain.co.uk/ ). For a small personal project a couple of years ago I gave PCBWay in China a go and had no major problems with them, other than a bit of weirdness around the pricing because I had a single gerber containing separate circuits, which they counted as two designs. The design was 2 sided with some tiny features(*) and the copper and resist all lined up ok. Some of the tiny silkscreen markings weren't very sharp, but I was at the limit of their standard specs. I did the soldering myself so I've no idea what their assembly or paste stencils are like. I can't remember if the vias were tented or not, but for this regulator they should be tented to make sure there's no exposed copper under the Murata module. I was also only getting 10 boards made, so I don't know how their pricing scales. Delivery was on time.

(*) https://uk.farnell.com/onsemi/ncp140amxc330tc … fn-4/dp/2775460 150mA, 3.3V out LDO, in a 0.8mm square package. With 5 pads. Barely visible amounts of magic smoke to release. I also used some 0201 components and had to attach a magnifying glass to my bench magnifier. I was using 0.15mm track/gap spacing and (I think) 0.3mm drill holes for vias.

majestyk wrote on 2022-01-20, 11:35:

This will be a perfect concept for the final layout for all common voltages! Today I started with a first test. Just set the out […]
Show full quote

This will be a perfect concept for the final layout for all common voltages!
Today I started with a first test. Just set the output voltage to 1,9V with 2 x 2K7 and connected 3 wires of the tiny thing to the mainboard. It´s so light-weight, it doesn´t even need any additional fixture.

The attachment murata_ga586hx.JPG is no longer available

The attachment murata_ga586hx2.JPG is no longer available

It´s working perfectly so far, I can´t feel the slightest heat dissipation. It´s go time for Heathunk...

Wow! That really looks like a very interesting modification. Could you please make something like that for me and send it to me with a tutorial? I have a ga586hx revision 1.55 and don't dare to do the old school mod where a larger heat sink is needed. At the moment I have an AMD K6-233 in there but would like to use the K6-2-450+. That would only run at 400 but that's OK because then I would have an additional level 3 cache.