PWM to Analog converter circuit

Tealc · 15 Sep 2012 at 17:05

Finally got a chance to do some wiring today.

Casterina wanted the power to come from the PWM connector so I wired it up like that.

Here's the larger of the 'doyll' heatsinks on top of the transistor. I drilled a couple of holes in the top to get the soldering iron down and into the lower section so I could melt the solder that holds the sink in place.

Looks quite alright I think. As you can see I applied some additional maintenance to the profile of the sink (gave it a bash with a hammer).

It's a good solid lump of copper.

In comparison with the original doyll profile it's quite a difference in height.

Tried it out with 300mA of load and it gets warm of course but it's totally fine. Also threw 700mA on there and it got hotter but I could still hold it in my hand without too much discomfort.

Left it running in the shed for an hour with the 300mA at low duty cycle and it was just fine as for heat.

So a success I think.

Thanks for the heatsinks guys.

Next one I do I might try mounting a profile reduced Resident sink on there and see if I can shift one on the MM.

Tealc · 15 Sep 2012 at 18:36

I know mate. I've really spoiled it.

I thought about the profile and being able to shove my iron in the gap after I'd mashed it on the back gate step I'm afraid. I just used the 3mm drill that I had in my drill but 5mm or 6 mm would have been better as I couldnt get all the tip onto the lower surface, hence the solder on top to aid with the thermal transfer.

Tealc · 15 Sep 2012 at 19:09

Yes that might work quite well.

Tealc · 18 Sep 2012 at 08:04

So here's Casterina's all wrapped up.

and another one I put together.

Tealc · 25 Nov 2012 at 22:54

firehorse said:
Hi,

I've just come across this thread and it looks just exactly the kind of thing I want.

I'm putting together an Asus P8Z77-I in Lian-Li PC-Q08 case. My achievement is that I managed to squeeze a Noctua NH-C14 into there

Anyway, the case has 2 chassis fans and the heatsink has one cpu fan. The motherboard has 2 pwm fan headers. I'm also thinking of replacing the 7850 graphics card cooler with a Scythe Musashi cooler which has a manual control, but I was thinking it would also be great if I could retain the automatic speed control from the graphics card.

My current build is slowly progressing here
http://www.silentpcreview.com/forums/viewtopic.php?f=20&t=65338

I'm thinking of experimenting with a Coolink 1201 fan which starts at 3V, but all the manual fan controllers only go down to 5V, so what you've done here is great, including the fact I will have less wires hanging outside the case.

I'll be back once I've have something more specific

Great thread!

Thanks very much.

I've made a few circuits for graphics coolers and they seem to work really well. I still find it odd that very few aftermarket graphics coolers actually exploit the PWM function of the graphics card. Even the one I bought, the Thermalright Shaman, doesn't by default. The only difference really is how the control circuit is interfaced with the fan header.

Tealc · 13 Dec 2012 at 21:26

So I decided to knock a couple of circuits together now that we are a bit more settled in our new home.

I decided to modify the circuit a touch to use the additional gain of the transistor to drive that much more in fan load.

I had a total of 0.83A (10W) which can be controlled by a single PWM feed from around 5v right up to 11.9v. The Doyll heatsink coped pretty well with this but was pretty hot after an hour at idle speed. I know it is good enough for over an Ampere in this configuration but would need a proper heatsink, like perhaps the ones that Resident sent over. I decided to ditch the Molex power option as it's not really required and just adds to the cost of the build.

Connected up to a couple of fans.

Before heatsinking.

Ater heatsinking.

And here it is powering some fans.

Tealc · 14 Dec 2012 at 14:03

errata said:
ready made channel aluminium is much easier to get hold of and could form most of an enclosure particuarly if the hot component was moved to the other side of the pcb and was bolted rather thans soldered

I hear what you are saying with the aluminium channel. There are a couple of considerations, one is the thickness of the bolt on the other side of the transistor and the fact that I don't actually have any suitable bolts. I could probably make room for the bolt head if I changed the layout of the circuit a touch and moved a resistor out of the way.

errata said:
Would one of these convernt my an older zalman cpu cooler fan to PWM operation they are currently on fanmates at fixed speeds.

Yes it would. That's what it is designed for, well not specifically yours but you know what I mean.

Tealc · 14 Dec 2012 at 15:42

It's 18mm wide.

I could use thermal adhesive but I'd need to buy some, and the channel too. The contact area on the transistor is really quite small and I'm unsure whether it would be strong enough to hold it in place.

Thinking about it I could mount the resistors underneath the board, or invest in some SMD resistors and that would give me space for a bolt. They are nearly the same price as through hole ones.

Just trying to use what I have to hand rather than having to buy stuff all the time as it's really surprising how quickly the parts cost mount up.

Tealc · 14 Dec 2012 at 17:21

The shrink is 25.4mm round I think. It comes flat though. Don't have much left now.

Tealc · 14 Dec 2012 at 18:02

In my experience, and I have a fair bit with heatshrink they are usualy blown slightly oversized so probably would fly over it.

Interesting idea though.

Will look into it if these two sell.

Tealc · 14 Dec 2012 at 22:57

I wonder if it's really necessary though. I mean it gets warm but provided it's kept to 3-4 fans it doesn't get hot enough to cause an issue.

Ideally you'd have a moulded bit of aluminium or copper with passive cooling thickness vanes which would have a nice hole drilled in it. A small bolt would connect them together with a dash of MX3 or something.

There are of course heatinks already designed for transistor that have this sort of shape going on.

http://www.ebay.co.uk/itm/10-x-TO22...al_Components_Supplies_ET&hash=item415ea4a03f

£1.15 for 10.

That's not too bad actually. Might order some if these two sell.

I had one of these running over an Ampere with one of Resident's heatsinks and it didn't get as hot as ther copper slab. It's the surface area that makes the difference. Plus the grooves help the heat move away, whereas a flat surface doesn't.

Tealc · 23 Dec 2012 at 16:08

Yes that's basically the same circuit but I didn't use R4 to make a voltage divider and instead control the current flowing through B772 at a later point. I find that even a small change in R4 has a profound difference on current running through B772 so wasn't really suitable for a totally adjustable circuit.

You could have changed the 1k resistor on the Phanteks board to something like a 620 Ohm 1/2W resistor and that would have had the same effect.

Tealc · 23 Dec 2012 at 22:55

Ah right. I thought you modified the circuit you linked to onto the Phanteks board by adding an extra resistor to ground before the base of the transistor. Now I see you just swapped out the 1k resistor for something else, which is what I mentioned above.

With your 500 Ohm resistor you are actually exceeding the rating of a 1/4W by the way, not my much (0.288W) but you really need some overhead with these things. Might be better to put two 1k 1/4w resistors in parallel.

Tealc · 10 Jan 2013 at 16:56

Yes in fact I've got two already made up but not heatshrunk.

I posted some pictures of them a few posts ago.

Throw me a message in Trust.

Tealc · 14 Jan 2013 at 17:11

Couple of updates.

1. Dockie your converter is now on it's way to you. Hope you enjoy.

2. Firehorse contacted me asking for 3/4 boards for his build.

A standard one for CPU

One for his Scythe Mushashi dual fanned GPU cooler

And then an interesting one which will control two separate fans on the same PWM but Firehorse wants them to have different PWM to fan speed responses. The only way I could think of doing this without doing loads of experiments was to mount 2 complete circuits on a PCB.

You may have noticed that the board is a little smaller than usual and this is mainly down to not having the fan headers mounted on board and also because I rejigged the design a touch to cram a potentiometer directly on the board, rather than off to the side and actually take up less board than before. Everything works but doesn't give me masses of room for heatsinking, but that's fine as these converters are working on 1, maybe 2 fans at a time and therefore won't need a great deal of heat teansfer.

Tealc · 29 Jan 2013 at 23:19

Thanks for the pictures. It's nice to see these things in use.

Tealc · 9 Feb 2013 at 13:23

And still from time to time people ask me to make these things.

This one is a Version 2 Turbo with Molex connectivity. You can see that it was originally a fan connector powered one. I just modded it this morning. Molex connectors are really expensive compared to some of the other components.

Tealc · 14 Feb 2013 at 00:42

Subliminal Aura said:
Oh man ! this is too cool to believe .... I've been looking for one for ages ?

Where do I sign, pretty please ?

Trust.

doyll said:
Control yourself mate! Don't what you to have a heart attack!

Tealc does great work.
Now that I have 7 phanteks fans I might have to get one of his adapter/controllers myself.

Thanks mate.

I thought you were a Thermalright TY fan man?

7 might be pushing it a little far

I'd probably look to split that load across either 2 boards or at least a board with 2 circuits, something like I made for someone else to control twin GPU coolers.

Tealc · 15 Feb 2013 at 17:17

Was out last night and in work today.

Tealc · 15 Feb 2013 at 19:01

Possible...yes....provided you run several RPM outputs.

So as well as the PWM connector you'd also have two other connectors coming off the converter for plugging into your motherboard. Bit of a spider thing going on.

or, now that I think about it...

Another alternative would be to have the RPM line connected via a multi-position switch (and I do have some) then switch manually between 1, 2 and 3 so you could at least get an idea of each individual fan is running at that particular time, but only ever one at a time.

You could also use a microcontroller and output each fan's output to the motherboard in sequence and a LED indicator next to each fan output to tell you which is reporting.

Far cheaper but along a similar thought process you could use a 4017 CMOS IC triggered by a 555 timer to count each part of the circuit, opening a transistor to act as a switch. Or use a SPST momentary switch to trigger the 4017 CMOS IC to go to it's next count. Again a LED could indicate the fan outputting.

So yes it's possible. Just a bit of a fiddle and you'd have to consider the value of going down this router, besides the fun of designing the circuit of course.

I have thought of adding LED indicators to the board for fun but again is just a thought.