randytsuch's audio page

So now on to the full story.  I am using one Astron RS-12A.  I have two, but didn't think the 2nd really made a difference. However, I use the 2nd one when I connect a optical drive to this PC, the 2nd one provides 12VDC for the optical drive.  Both Astron’s are adjusted to a little over 12V, around 12.2VDC.  You need to adjust per this:  Adjust as Astron for 12VDC

One Astron is connected to a 10,000uf, 63V Jensen 4 pole cap.  With a 4 pole cap, there are two connections for input, and two for output.  The input goes to the Aston, and the output goes the the 4 pin ATX connector on the motherboard.  I cut off the 4 pin ATX  connector from an old PC PS to use for this.
Notice that as much as possible, wires are twisted together.















 I also added a 330uf Oscon on the wires right next to the motherboard.  I did this just because I have a stash of oscons, and figured it couldn’t hurt.

The Astron also powers the 12V into the 20 pin ATX cable.  The cable can be either a 20 or 24 pin ATX cable, I am using as a 20 pin cable even though my motherboard supports a 24 pin cable.  The 4 extra pins are to support more current/power, but I don’t use much power so I don’t need the pins.

http://en.wikipedia.org/wiki/ATX

I cut the 12V wire (pin 10), and one ground wire.  Then I connected the motherboard side of the cut wire to another 10,000uf, 63V Jensen 4 pole cap, and that goes to the Astron PS.

The Jensen is under the yellow tape, which I added for safety against shorts.

+5VDC

Now for 5V.  I decided to make a 5VDC linear supply using an unregulated walwart, Triad WDU9-2300.  It outputs 9VDC at 2.3A. 

This walwart is not the optimal solution.  I could make something nicer with a Rcore transformer, a diode bridge and some nice caps, and maybe an inductor to make a pi filter, but that is more work, and would cost more money.  And there are liability concerns.  With this, I just need to add a DC jack to my PC, and then add a linear reg into my PC. 

So I have a DC jack on the PC to feed in the 9VDC, its closer to 10V under the load I have, around 1A.  I feed the 9VDC into a Jensen 4 pole, and then to the regulator.  It is a 2 stage regulator, with a pre-regulator who provides a constant voltage to the final regulator.  It also takes some of the power/heat.

The regulator was one I had built for a previous project, that got transplanted.  It is based on this article

http://www.acoustica.org.uk/t/3pin_reg_notes1.html

I mounted the two 317’s on a good size heat sink so I wouldn’t have to worry about over-heating.  I have a bunch of heat sinks I have collected, which helps on projects like this.


Here is the heat sink with the regulators mounted on it.  I tapped holes directly in the heatsink so I can just screw into the heatsink to hold the regulators down.  The extra hole is where I broke my tap, so I had to tap another hole with a 6-32 screw, instead of the 4-40 I used in the first regulator.

Even though I broke a tap, I recommend picking up a tap set and learning to tap holes.  It's not hard after you learn, just be gentle, when it gets too hard to turn back out the tap and clean out all the little metal chips.  I'm sure there are videos to show how to do it online.


Here you can see how big the heatsink is.  I added some caps to the circuit, but that was after this picture was taken.  Caps are on the output of each reg, and on the adjust pin of the 2nd reg.

The regs are mounted with a blue thermal pad to isolate from the heatsink, and a shoulder washer under the screw, same reason.  If the tab is not ground, you need to do this.  I have a bunch of insulators and shoulder washers I bought a while back.  After mounting, make sure the screw is isolated from the tab with a meter.  One of mine was shorted, and I had to redo with another washer.  Only happened because I was using a #6 screw, which is really too large.  With a 6, you need to apply a fair amount of pressure to get the screw through the shoulder washer, with a 4 it just slides in.

So then I connected the input power to the reg, but did not connect the output to the PC.  I powered it up, and nothing.  Found a bad pot, so I replaced it, and bingo, power.  Adjust pot to get 5VDC, looking good.

Then I found a 7 ohm, 10 watt resistor in my junk draw.  Hooked it up on the output of the reg, and let it run for a while, just to make sure my reg could handle the load.  I highly recommend testing the regulator under a load like this.

Before installing the heatsinks in the PC, I drilled some holes to allow airflow past the heatsinks.  The holes get covered by the heatsinks.











So now I cut the wires to disconnect 5V from the Pico, as well as a few ground wires.  I put in a 10,000 uf Jensen next to the PC motherboard.  The 5V and ground wires get connected to the Jensen, and the other side of the Jensen is connected to the output of the regulator, through a switch.

Connect everything up, turn on all the switches, and bios boots up, yeah.  Then I shut down, and hook up my SSD.  Boot up again, bios boots, things start to happen, and then it resets.  I tried a bunch of times, and it always resets 30-60 seconds into boot.  I am depressed, and take a break.

I reconnect the pico, and it boots fine, so PC is still good.  Reconnect my linear, and start measuring voltages with my meter.  Right before it resets, it appears the voltage may be drooping.  Hard to tell with a meter, but I look at a few resets and see the voltage drop right before the reset.  So I theorize that the 5V demand changes a lot during boot up, and my regulator is too slow, so 5V drops too low and the PC resets.

Now what to do?  I bought some Lifepo4 batteries but had not used them yet.  Some people swear by them for digital.  They are charged to around 3.3 volts.  Put two in series, find some power resistors, and make a 20 ohm load (I think).  Discharge down to 4.5 volts (I didn’t pay attention).  By the next day they were back up to 5.5V, so I discharged some more, till they were around 5.1V.  My reg was set to output 5.1V also.

I changed out the 5V switch to a double pole type.  On the “output” of the switch, I shorted the poles and that is connected to the Jensen and then the 5V into the motherboard.  On the input of the switch, one pole goes to the 5V regulator output, and one pole goes to the battery.

Turn everything on, and it boots up fine.  The battery ensures that the 5V does not drop too low.  Also, since the battery gets connected to the 5V regulator when the switch is on, it will keep the battery charged to 5.1V, the output of the regulator.

I am going to play with different 5V regulators, but am happy this works.  So I think if I had a a better regulator I wouldn't need the batteries.  But the battery solution was pretty easy to add, it it works fine with the batteries and my regulator.

So far, I have connected linears for the +12V and the +5V lines on the  ATX connector.  There is another signal “+5 V standby” or +5VSB on pin 9, it’s a purple wire.  I cut this, and connected it to the +5 VDC output from my linear reg.  Remember, +5 to the ATX on the red wires goes through a switch.  Since I connected +5VSB to the output of the linear reg, it is not switched and is always on.

The last thing I did was +5 was to modify a couple SSD power connectors.  I connected these to the  switched +5, so the SSD’s get power when the switch is on.  I’ll upgrade these later too, but for now it works.

Here is the 5V reg and it's heatsink inside the chassis.  The Jensen Cap on the far right is connected to the 4 pin ATX.  The little board  with the blue LED next to that is a fan filter, goes inline between the MB fan connector and the Scythe fan.  Next to that is the Jensen which is on the input side of the regulator.  And finally the heatsink with the regulator mounted on top.  You can also see the two lifepo4 batteries at the top left of the heatsink.

BTW, regulators get to about 100F, around 20F over room temp, so not bad at all.  So this heatsink is way overkill, but that works for me.




Dec, 2015 EDIT:
I've made some other changes that I wanted to add to my blog.

Here is a newer pic of my 5V circuit.


The significant changes is I changed to use much larger lifepo4 batteries.
I've been using these batteries
26650 clone lifepo4s
To get 5V out of lifepo4s, connect two up, with the positive of one connected to the negative of the 2nd.  Now you'll see about 6.6V across the batteries.  Then I put a small value power resistor to drain the batteries.  It took a while, and when you remove the load, the voltage will creep back up, so you may need to do it a few times, and go lower than 5V.  Once its a little lower than 5V, connect per the above and it should be fine.

The other significant change is that the switch in the figure above has been replaced by a relay.  The relay is connected to 12V for turn on.

This same 5V supply is used to power my two SSD's.  One SSD holds the OS (windows Server 2012 R2) and the other holds all my music.  I plan to try making a separate 5V for the SSDs.


3.3VDC

Now it was time for the +3.3 V linear supply.  I bought one of these, Stancor STA-5760 6V@2A unregulated walwart.  The Stancor has + on the outside of the jack, which is the opposite polarity that I wanted.  So, I cut the cable in the middle, and then I separated the wires all the way back to the jack and walwart, and twisted the wires.  I spliced them back together, but swapped the wires so + is on the inside pin.

I built a regulator with a LT1764 and a LT1587.  The LT1764 drops the voltage down to around 5.2VDC, to feed the 1587 a constant voltage.  The 1587 generates 3.3VDC.

I built up the LT1587 circuit first, to make sure it worked right.  I found a 5VDC walwart, and used that for the input voltage to it.  I built using fixed resistors, and my first attempt was too high, 2^nd was too low , and 3^rd was just right J.  I used a 110 ohm (really two 220 ohms in parallel) and  a 177 ohm, should have been 3.16V, but it was right a 3.3VDC.  So if you want to build with fixed resistors you need to have spare resistors around to tweak the value, and wait until you get it right before soldering for real.

With the that working, I added a LT1765 in front of that.  At first, my voltage was too high.  Then I figured out it wasn’t working because I didn’t add an output cap.  I added 100uf solid poly caps (oscon type) on the outputs of both regs, and everything worked fine.  I used a 1k and a 3.3k resistor here, get around 5.2VDC.

I also added two more caps, 10uf caps between adjust and ground.  Also added a protection diode from input to output on each reg, and a LED to indicate if 3.3V is on.

I tested with no load, and then with a 7 ohm power resistor.  Worked fine under load, so I’m ready for installation.  In the PC, I have a DC power jack connected to a 10,000 uf Jensen, feeding my voltage reg.  The voltage reg output goes to a switch.  Then it feeds another 10,000uf Jensen, which is right by the motherboard ATX connector.  The ATX 3.3V wires connect to the motherboard.


NOTE on diagram

This circuit diagram has a mistake at the switch.  The two poles at the output should be shorted together.  So when the switch is closed, both the battery and the 3.3V regulator output are connected to the Jensen 4 pole and then the PC 3.3V input.

Connect everything, turn it on, and nothing.  Let me try again, success J.  

Here is the MB PS ATX connector after removing the Pico.  I have a three Jensens here, on 12V, 5V and 3.3V.











EDIT Dec 2015
Like for the 5V section above, I'm using these batteries.  3.3 is easy, just takes one at normal charge.
26650 clone lifepo4s

And the switch has been replaced by a relay.  The relay is connected to 12V for turn on.



So now I have a picoless, linear powered PC.