[CSDb] - User Forums - Power supply upgrade project

You are not logged in - nap

CSDb User Forums

Forums > CSDb Discussions > Power supply upgrade project

2015-04-09 18:20

Medo
Account closed

Registered: Mar 2015
Posts: 4

Power supply upgrade project

I am working on an upgrade kit to improve the wedge-shaped C64 power supplies (left one in the image). According to what you hear talking to some people (which is a very reliable scientific method), these supplies are somewhat likely to self-destruct under heavy load (e.g. C64+Chamaeleon) and sometimes take the C64 with it. This is both due to cheap construction on Commodore's side, and also because the mains voltage was raised to 230V, which means the regulator runs hotter.

However, all the problematic parts are located on a small PCB which is responsible for producing the 5.3V power from the low-voltage AC transformer output. Inspired by this forum thread on forum64, I started planning my own upgrade which would replace the circuit board with a highly efficient switching regulator and should be more robust, have a more stable output voltage and can supply more current (2A instead of 1.5A), while producing significantly less heat.

Originally, I only intended to build one for myself, but since the cost for some components drops really steeply as you increase the number of units, I'm now looking for more people who might be interested in such an upgrade.

I am also looking for defective wedge-shaped power supplies that you might want to get rid of, because they can probably be fixed with the upgrade kit.

So if you're either interested in getting your own upgrade kit or giving me a defective PSU, please reply.

Here are some photos of the current prototype:

Some of you probably saw this already over on pouët.

2015-04-09 19:54

MagerValp

Registered: Dec 2001
Posts: 1074

Ah yes, the good old footwarmers. Aren't they usually filled with epoxy though?

2015-04-09 20:41

Zer0-X
Account closed

Registered: Aug 2008
Posts: 78

The transformer and the regulator are epoxied, the remaining is free.

Tho getting the lid open from some of those might require quite some violence if the protruding plastic parts are deep in the epoxy.

I've been using a 3 amp dc-dc converter on a few PSUs like this.

2015-04-22 19:23

e5frog

Registered: Nov 2010
Posts: 6

Yes a UBEC is a nice and easy regulator solution to stuff in there.

Where are you located Medo, it may be important regarding shipping 1.4kg PSU:s or buying one of these kits or not.

I assume it's a prototype in the image and you'll have a PCB with soldermask for the kit?

Can't see very well in the image but traces from the transformer looks a bit narrow. What copper thickness do you use for the board?

What regulator are you using? Is it being attached to the board enough cooling if you run it at 2A?

2015-04-22 22:00

Medo
Account closed

Registered: Mar 2015
Posts: 4

I'm in Germany, Koblenz area. However, I'd suggest meeting up at a party if possible instead of shipping things around if possible. I plan to be at Evoke and Outline, and will definitely be at Nordlicht.

Yes, the PCB shown is a hand-etched prototype. I just ordered the final ones yesterday, with proper soldermask and silk screen. They are made with the normal thickness of 1oz copper (35µm).

The traces from the transformer are slightly wider in the final design (100 mil), but also longer. They should create less than 0.05 Ohm combined. I admit I didn't really consider the losses there since the traces are pretty wide already, but doing some back-of-the-envelope calculations they should dissipate less than 0.25W under full load (and probably quite a bit less, I made some pessimistic assumptions). Under the same conditions, the diodes would together dissipate over 1W. So it might be something that could have been improved, but the PCBs are ordered now and I won't lose any sleep over it as it is :)

The regulator is an LMZ12002 "Simple Switcher" from ti (formerly National Semiconductor). It's very efficient, and under worst case load it should dissipate less than 1.4W. The pictures I linked are actually from a test run where Rene let the thing sit to bake for 2 hours first at 1.5A load and then at 2A load, with heat images taken after each test. He has some really nice tools at work which he used for this.

The result looked ok, but not great, mostly because of my puny attempt to solder a thermal pad to a big copper plane with a small soldering iron. I did manage to create a connection, but it wasn't too good in terms of heat transfer, and the connection broke entirely in the end, making the switcher run into its thermal limit.

The final version will be properly reflow soldered, so the connection should be much better. It also features thermal vias now (so the heat is also spread on the bottom), which are kind of hard to do on a hand-etched PCB.

We'll repeat the thermal test with the final version, and that should also give better insight into the temperatures on the copper traces / planes. Why? The hand-etched PCB has a bare copper surface which is very reflective, also for the heat camera, so you can't really see the temperature of the copper. However, everything will be covered with solder mask in the final version.

Here's an image of the final layout:

2015-04-23 12:24

ready.

Registered: Feb 2003
Posts: 441

just to avoid stepping back into the same issue we have with the original PSU when it fails, why not add an overvoltage protection (OVP) to it? Also old linear regulators in most cases would fail slowly to do age of the regulator, so that output voltage would slowly increase from nominal 5V. For switching regulators, if the duty cycle regulator part fails, it could either give 0V or full input voltage immediately. Thus an OVP is almost mandatory on switchers. In fact C128 PSU using a switching regulator has an OVP.

Furthermore, why use 5.25V as output?

2015-04-24 21:50

Medo
Account closed

Registered: Mar 2015
Posts: 4

I considered adding extra OVP, but eventually decided against it since the switcher module seems safe enough for my taste.

The module already has OVP built-in, although the voltage threshold is slightly high (~6V). Still, it's low enough that all the commodore ICs should be safe at least. If I understand it correctly, it's also built in a way that should protect the output even if the main pass transistor fails with a short, since the OVP also switches on the synchronous transistor, so it would effectively short its input to ground in that situation which would blow the primary fuse.

Aside from that, the reliability data on ti's website looks pretty good. They tested over 6000 samples for 1000 hours at 125°C, and recorded no failures.

The 5.25V output voltage is to make the C64 work properly :)
The original circuit has a nominal output voltage of 5.3V. They apparently raised it from 5V after the circuit boards were already produced by adding a resistor divider at the GND pin of the 7805 (you can see where they dremel'd through the copper at one place to make the modification). The reason is that there is a voltage drop of ~0.2V-0.3V in the cable and all the connections between the PSU and the actual ICs on the board, and if the 7805 already had a low output voltage, the resulting voltage might be too low for some ICs.

I decided to set the nominal voltage slightly lower than Commodore because the output voltage tolerance on the upgrade board is a bit tighter than that of the 7805.

2015-05-25 16:48

Medo
Account closed

Registered: Mar 2015
Posts: 4

Progress!

I have the 10 PCBs and most of the components now. The last order of components should arrive next week, and then we'll only have the soldering left to do. It might take a few more weeks until we can get this done, since we're trying to get it done cheaply but with professional reflow equipment, using, er, semi-official channels involving chocolate bribes :).

2015-05-26 10:51

ready.

Registered: Feb 2003
Posts: 441

Looks nice Medo! I think you are doing a good job.
Here's my PSU, just for sharing, given the topic:
https://www.dropbox.com/sh/y5pigtqdlooin6w/AABpfVZRc5Qc98TWRyLT..
it's totally different approach to yours, since I built it from scratch, thus not meant as replacement board. But I like yours a lot. The only problem, is that I have never been able to find any C64 PSU without epoxy inside, so I'd say it's impossible to use your board there.

Refresh

Subscribe to this thread: