C64 PSU upgrade project - Defective PSUs wanted!

category: general [glöplog]
Thanks for the efforts. I will check my PSU's for broken ones. Thank you for your Info.
added on the 2015-03-26 12:49:24 by Exin Exin
@stfsux: it _is_ a switch-mode power supply, thanks for playing.
added on the 2015-03-26 14:39:04 by trc_wm trc_wm
A trick for hand-soldering LMZ:
Make a BIG exposed via under it's tinned ass -- so that you will solder confortably it to the edges of via. This would make cooling worse, however, and is totally wrong for automated soldering.
added on the 2015-03-26 15:44:29 by lvd lvd
@trc_wm: I dont see any transformer at all. DC-DC converter != switch-mode power supply.
added on the 2015-03-28 22:27:18 by stfsux stfsux
Well let's see, based on my elementary knowledge of power supplies, common sense, and Google image search, one end has to plug in to the wall in order for it to serve it's function. Wall sockets provide AC. In this thread MedO states he is only replacing the main board, and that there is still an "old primary-side fuse is still included" which implies there's another part. In order to go from AC to something safe for the electronics board, there must be something else. Let's see what Google says. Oh look! Pictures of the insides show a large metal block containing a transformer.</snark>
Well, it's not a SMPS in that it doesn't chop the mains voltage in order to use a smaller transformer. But I'd say that using the old transformer is the easiest way to get the 9VAC in any case.

Small question to the crowd - I'm considering whether I should add extra overvoltage protection to the output, just in case the LMZ fails. It would be some extra protection for the C64, but it would also add a bit of extra cost again. What do you say?
added on the 2015-03-29 01:20:09 by MedO MedO
Protection against shit happening to a c64 is priceless :)
I dont known much about that LMZ, how safe is it but it seems to have a lot of protection. Im not sure an extra protection is really useful.
added on the 2015-03-30 19:18:32 by stfsux stfsux
Neither am I, that's why I ask. It would probably be a paranoia measure. The LMZ should be fairly robust - it's made by National Semi (now ti), and has been around for a few years now.

I found the reliability information on their website, but I'm not entirely sure how to interpret it. Even so it looks pretty good - no failures in over 6000 devices tested at 125° for 40 days.

I'm not sure if the built-in overvoltage protection feature would protect well in case the pass transistor fails short and puts the entire input voltage on the output, because apparently doesn't use an extra pass transistor. But even then, the same feature is also supposed to ensure that the synchronous transistor is turned on, so it *should* short the input voltage to ground in that case, hopefully blowing the primary fuse before it fries itself.

That's just theory, but I think I'd feel safe enough for my C64 without added protection after considering it like this :)
added on the 2015-03-31 01:14:47 by MedO MedO
You should probably have a crowbar circuit across the output if you're paranoid. Basically just a thyristor triggered by a zener diode so that the output gets shorted in case of overvoltage.
added on the 2015-03-31 16:50:47 by Tjoppen Tjoppen
The problem with a crowbar in this setting is that the shorted output is supposed to blow a fuse. However, since the output of the LMZ12002 is current limited to 2A, a fuse there might not see an abnormally high current. And since this is a switching regulator, the current draw on the input might actually drop when you reach the current limit, so the primary fuse won't blow either.

I accidentally tried this with my first design, which used a cheap chinese regulator board from eBay that didn't keep its output voltage steady enough. The crowbar triggered, and the PSU sat there for something like 20 minutes before I noticed and unplugged it. In the meantime the Triac had almost unsoldered itself. Impressively enough though, the entire circuit still works :)

Shorting the 5V supply might not be entirely healthy for the C64 either though - in my case, the 9V fuse inside the C64 had blown, though it suffered no further damage. Still, I'm not sure which path those currents had taken.

To be safe, it would be better to immediately break the circuit when an overvoltage is detected. There are ICs which help with this.

However, as I said above I'm leaning towards not adding extra protection now.

By the way, the crowbar I built didn't use a zener diode, but a voltage regulator - the trigger voltage is way more predictable this way. The wikipedia article on the crowbar circuit actually shows this arrangement.
added on the 2015-03-31 18:31:16 by MedO MedO
If you are at Revision and want to get rid of your PSU now, you can find me at the Commodore sign (middle/right of the hall).
added on the 2015-04-03 20:33:21 by MedO MedO
If you are at Revision and want to get rid of your PSU now, you can find me at the Commodore sign (middle/right of the hall).

Shit, I forgot to bring the one I have. :<
By the way, the crowbar I built didn't use a zener diode, but a voltage regulator - the trigger voltage is way more predictable this way

Neat, hadn't thought of that. I just went with how's it's done on this pinball machine I've been repairing.

Idea: let's talk electronics stuff if we run into eachother at Revision
added on the 2015-04-04 08:36:33 by Tjoppen Tjoppen
Unfortunately we failed to run into each other, but I guess we will sooner or later at some other party :)

I updated the design a bit. It uses larger capacitors now, and I decided to just go with the exact ones used in the datasheet example design. Additionally, I added a ferrite bead to the output which should filter out high-frequency switching spikes.

The AC input was moved to the side because of some valuable advice from a friend who looked at the project today. He works at a company which specializes in safety testing and certification of electronics, and he wasn't happy with some aspects of the supply, though all of them go back to the original Commodore design, which probably doesn't comply to today's safety standards. Moving the AC inputs will keep them further away from the primary side of the transformer.

BB Image
added on the 2015-04-08 01:47:50 by MedO MedO
I finally got the PCBs in the mail, and the first part of the components:
BB Image
The remaining components should arrive next week.

As you can see, the layout changed a bit from the last image I posted, in part because the PCB manufacturer I went with does silk screen on both sides for no extra money :)

It might be a few more weeks until we can get at the reflow oven and get this soldered, so I'm itching a bit to try soldering one myself ;)

By the way, the small cubes on the left are the 100µF ceramic capacitors. It's crazy how dense they can make them these days.
added on the 2015-05-25 18:56:05 by MedO MedO
The reflow-soldered PCBs were done just before Nordlicht, but I only now got around to soldering on the remaining things and testing everything :)

BB Image

I soldered one of those into my PSU and did a quick check. 5.15V at the output with no load, a bit low but still within tolerances. All my C64s ran well, including with the Chameleon. Output voltage dropped to 5.07V with the C64 running, so the ICs inside probably get 4.77V considering the voltage drop in the cable - a bit on the low side, but hey, as long as they run all is good.

We'll do some further tests including heat at full load (with IR camera :)), and ripple and efficiency measurements.
added on the 2015-07-25 23:59:07 by MedO MedO

So how's it going? I would love to replace my old PSU with a "safer" one thus avoiding the risk of frying both my c64 & 1541u2. Do you plan to sell it or to make a small batch? Any plans to release Eagle Files&BOM? Any chance of demoscene art in white in the remaining space or the backside? ;-) Does it have some kind of capacitor/fuse ready to die in case something goes wrong?
added on the 2015-10-01 23:16:46 by Crumb Crumb
This has been sitting around for a bit now without much happening, but recently Rene has been running some tests to characterize how the thing behaves under different conditions and it's looking very good so far. I'll write more when we're done with that.

Yes, the Eagle files and BOM will be released. I made 10 of the PCBs, they are all done and ready. I'll probably start selling them once we have the tests done and created some form of documentation to go along with it.

There is no extra protection circuit, though the LMZ12002 supposedly has an overvoltage protection feature that should at least save the valuable MOS chips if something did go wrong. The main reason why I think it is safer than the original is that it is designed for higher maximum power, doesn't run as hot and generally uses components chosen with larger safety margins. However, I have been thinking about maybe making a separate protection solution that you can plug between the C64 and the power cable. We'll see.
added on the 2015-10-09 01:54:00 by MedO MedO
Rene finished some way-too-accurate performance tests on a modified PSU with this board. Not that anyone really cares *that* much about the efficiency of a C64 supply, but yay, numbers!

Using the normal 230V AC power, the efficiency of the 5V output under medium load (1A) is 53%, and it's up to 59% for heavier loads (1.5A to 2A). So, no 80+ certification for us ;). Not spectacular, but a fair bit better than the ~30% you'd get with the original according to my napkin calculations.

In the test, the output voltage was 5.25V under no load, and 5.15V for anything between 0.5A to 2A. Going by min/max values in the LMZ12002 datasheet, the output voltage should be in the range of about 5.05V to 5.35V, so everything is well within tolerances.

Note that the tolerances in the original circuit were larger - I find it hard to calculate hard limits from the L78S05 datasheet, but just from the initial voltage tolerance, typical temperature drift and the effects of Commodore's resistor divider kludge, you could get anything between 5.0V and 5.55V.

Next up: We'll put the thing into a 40°C temperature chamber and run it at maximum load for an hour and see how hot the components get :)
added on the 2015-10-18 04:09:41 by MedO MedO