Before I announce availability of our new C64 smart supply, I'd like to share some findings about the C64 that I've had during development of this product, combined with some common knowledge about shortcomings of the original C64 design. I believe I've also found a new flaw in the C64 circuit, but I'm getting ahead of myself.
Before you design a product, you need to set goals. It shall be a power supply for the C64, so it needs to provide 5V DC and 9V AC. It shall be capable of powering modern and legacy expansions. The Commodore RAM expansions 1700, 1764 and 1750 (*) were probably the most power-hungry devices that were ever produced for the C64. They were delivered with a 2.5-Amps power supply - instead of the 1.5A power supply that was delivered with the original "breadbox" C64. So a power supply with 3 Amps on the 5V rail will do the trick - anything more would be over-powering the connector on the C64 (**), sitting idle and operate outside optimal efficiency. The original (fairly strict for 1982!) ripple level of 50mV or lower for 5V shall be met. The 9V AC part has "always" been solved with a linear transformer, so let's just do that, right?
WRONG.
Power supply design is way more than just slapping together the power sources and call it a day. In fact, I'd call this approach careless, as it's neglecting knowledge that has been out there for many years. Instead, the question you must ask before approaching this design is:
How does a C64 break?
There's of course lots of ways that the C64 can cease to operate, but we're approaching this from the power supply side, and there's two main reasons for defects directly caused by the power supply: Over voltage and power flowing in the wrong direction. Over voltage is easily taken care of: Modern DC-DC regulation has over-voltage protection as a standard property, so with the right component choice, we're safe. The second reason is a bit more complicated, but it is getting more common these days, when careful owners of a C64 are aiming to reduce standby power consumption. The old power supplies from the 1980s are all built with linear transformers that are always connected to the mains grid. This means that they always consume a little bit of power, even if the computer is switched off. This is almost 4W of power that is always drawn by the wedge-type power supply - almost 8 times as much as legally allowed in Europe for new devices.
A natural way to reduce standby power consumption to 0W is to use a multi-socket with a switch - something like this:
This is a really practical solution, as it can switch all components at the same time: The computer, floppy and monitor. Problem solved, right?
Nope, this is an accident waiting to happen. The reason is a design flaw that has been in the C64 from the very beginning: The 7805 and 7812 voltage regulators in the "long board" computers are not designed into the C64 as recommended by their vendors. What's missing is a diode that makes sure that power won't be flowing in the wrong direction through the regulator, which would break the component. Here's an exerpt from a Texas Instruments datasheet, but you'll find this in many other vendor's datasheets as well:
The C64 doesn't have these diodes, as you can see in this exerpt from the original schematics (note that power flows right to left, whereas it's left-to-right in the above image from the datasheet):
This is not a problem if you only switch your computer on/off on it's local switch. For bad things to happen, three (not so rare!) conditions must apply:
- The C64 must be switched on
- power is switched off at the multi-socket
- input caps C19 and/or C88 have aged worse than C57/C102 (***)
In this case, power from the output capacitors C102/C103 can flow back, passing through VR2, destroying it on the way. This alone wouldn't be too bad, as these voltage regulators are still easy to get, and they're also easy to exchange for an experienced hobbyist. However, really bad things happen when you switch on next time, as the 7805 may be defective in a really bad way: It now passes high voltage (over 9V!) nearly-unregulated to the VIC chip, which is the receiving end of the +5V'can' rail. In turn, the VIC chip breaks, and during it's dying gasp, it passes this too-high voltage to the RAM chips, which are directly connected to the VIC on both address and data lines. This is a tedious repair that I've done quite often in my life.
One way to mitigate this risk is to add the diode to your C64, but that's only half of what needs to be done. There is another 7805-type regulator in every original Commodore 64 power supply without that diode, and these are generally considered "unserviceable". So what may be an easy fix for an experienced hobbyist turns into an hours-long endeavour to remove epoxy to finally get to the spot where you can add a diode. There's a better way: If the power supply detects a brown-out condition, it can cut all power to the C64 and protect it that way. This is exactly what the iComp C64 smart supply does.
You think that's enough? Nope, there's more.
Many of you have gotten used to "resetting" the C64 by quickly switching it off and on again. I've seen this lots of times over the years, and it never changed over all the years that I've visited demoparties all over Europe. The habit of switching off and on at speeds that approach karate-style is in many user's muscle memory, and it may be OK a thousand times, but when a CMOS circuit is in the computer (such as a Kernal switcher with a CMOS Eprom or CMOS memory chips), this bears the risk of SCR-latchup for those components. SCR-latchup is the technical term for a semiconductor being BBQ'd beyond repair. Since modern expansions are practically always pure CMOS these days, the risk is a lot higher today than it was in the 1980s. If we just supply "dumb power" like all power supplies do (including all "new" units offered until today), the risk of SCR latchup is the same as with the original power supply. A really good power supply detects that the switch has been turned off, and cuts power for a certain minimum amount of time: I chose one second. So if you turn your C64 off-and-on-again karate-style, power will be cut off for one second, allowing capacitors inside the C64 and it's expansions to discharge to safe levels. When one second has passed, power is switched back on to the C64, and you can continue to work.
There's more, and it gets more technical. Stay with me, as the next topic a finding that might explain thousands of dead CIA chips. I believe this has not been documented before, but Im happy to be proven wrong (I know that the C64 community is very strict when it comes to giving credit to the right person, so I won't claim to be the first to discover this until it's commonly accepted that it's a new finding).
To understand this problem, you have to look at the C64 schematics again - this time at the way that the TOD signal is generated. TOD stands for "time of day", and it's the 50/60Hz signal that is generated from the AC frequency of your local grid. It's fed to both CIA chips. Here's how it's generated inside the C64:
R37 in this schematic is 6.8kOhms, but it's as low as 2.7kOhms in other versions of the C64. It has been added in an effort to reduce the probability of "ringing", which would make the TOD clock unprecise. However, is bears the danger that a negative voltage is applied to the TOD pins. The 9VAC line does have some protection to not go below GND-level through the rectifier CR4 seen in the picture above, but that does not have "ideal" diodes - it's real-world diodes with a forward voltage drop of roughly 0.7V, up to 1V under load. The result is that the TOD signal is pulled significantly below GND level when the C64 is powered from a normal linear-transformer-based power supply, as seen in this scope screenshot:
You can clearly see that the TOD signal at the CIA chips is leaving the safe area for considerable amounts of time. Now let's check the datasheet of the 6526 CIA chip to see it's absolute maximum ratings:
As you can see, the absolute minimum voltage that a CIA tolerates at it's input is -0.3V. This means that when the C64 is operated with it's original power supply or any power supply that is built with a classic ("linear") transformer, it's putting the CIA chips at risk. The good news is: The iComp C64 smart supply makes sure that no negative voltage is ever applied to the CIA chips, not even the semi-allowed -0.3V. Why "semi-allowed"? Read the comment in the datasheet exerpt above: If this chip is exposed just to maximum ratings, it's already at risk. The original Commodore design and any transformer-based power supply are applying more than twice as much, whereas the iComp smart suppy never supplies any negative voltage:
I have discussed this finding with Dave Haynie when he visited the Amiga show in Germany back in October 2025. I said that I believe this could be the reason why CIA chips were generally considered "very delicate" and "easy to break". While everybody back then believed that it was due to static discharge into the user- or joystick ports, it could also be that the chips were always operated outside their specifications. Dave just smiled and said "could be, could be!". Please understand that I'm not claiming to have found *the* reason for dying CIA chips, but I believe that I have found a spot where an improvement can be made, just with the right choice of power supply for your precious Commodore.
There is another important measurement of quality of a power supply: Ripple. That's the term for the steadiness of the 5V DC rail. Commodore has specified this number to be 50mV or better - where "better" means lower in this case. 50mV is only 0.05V, meaning, the 5V rail may only "wobble" by a total of 0.05V over time. Commodore chose this low level because the computer will be a lot more stable with an absolutely steady 5V supply. Further, the later "short board" versions of the C64 don't generate local voltages for VIC and SID chips, so on those models, ripple might show in a wavy picture or noise in sound. You may see other power supplies being advertised with "ripple filters", but hardly with a specific number for the ripple they produce. That should be a red flag for the informed buyer: If you need a filter, you have a problem at the source and you're trying to handle the symptom of a bad design. My approach is to not even require a filter, but just do it right at the source. I'm confident enough to advertise a hard number: The iComp smart supply has less than 40mV ripple on the 5V rail. This ensures best picture- and sound quality while providing top stability of your C64.
Based on these (admittedly rather technical) resons, I believe that a good power supply must be smart. This includes:
- make it safe to power on/off on the grid side
- ensure minimum power-off time to protect CMOS parts from latch-up
- avoid a linear transformer for efficiency reasons
- make sure that no negative voltages are applied
- operating the C64 power connector within safe margins
- generate a pure sine wave on 9VAC with crystal precision
- controlled ramp-up and ramp-down during switching
- low ripple on 5V that meets or exceeds the strict Commodore spec of 50mV
And last not least, the smartest decision you can make about a precious collector's item is insurance. iComp products are covered by a product liability insurance. If we messed up in some way and your equipment gets damaged, you have somebody to turn to who has a lot more financial resources than iComp has.
The iComp C64 smart supply is a two-part solution: A tried-and-true 12V/2A power supply with international adapters (as previously cleared by our product liability insurance for C64 Reloaded boards) and a converter that plugs directly to your C64. This avoids any loss that you may have on a long cable, because there is no cable between the C64 and the power converter. Predictable losses in the connector and input filters of the C64 are accounted for in our circuit. Two status LEDs are at the top, showing white when on, dimming green when off, and blinking white during the guaranteed power-off time. A jumper inside the converter case lets you switch between PAL (50Hz) and NTSC (60Hz) frequency for the 9VAC rail.
A version for floppies 1541-II and 1581 and the C65 computer will also be available:
These have less "smart" features, as these drives and the C65 don't have such complex power structures that make room for errors like with the C64.
The C64 smart supply will first be offered for sale at the Revision 2026 party in Saarbrücken, Germany. Shortly after that, we'll take orders in our web shop.
Jens
(*):commonly referred to as "REU", although this term also covers non-DMA expansions like GeoRAM that take less power.
(**): The C64 power connector is a standard DIN part where each pin is rated 1A max. There are two pins that carry 5V. The original power supply only used one of the pins, stretching the rated current of the connector by 50%. Our PSU is applying that same stretch, but on two pins for double the power.
(***): This is more than likely after up to 44 years: The input caps are exposed to very high ripple, leading to accelerated ageing of the capacitors, whereas ripple on the output of a linear regulator is negligible.
