This post is a follow-up to The Learning Circuit episode , Replacing MLCCs with Polymer Capacitors. In that video, I used a Commodore 64 as a target to show voltage rail measurements before and after changing from traditional wet electrolytic capacitors to newer polymer capacitors.
In the video, I commented it looks like 9 volts AC are directly applied to C90. To stay on topic, I decided not to dive into what was going on with the capacitor. (And to be up front, I was only partially sure while recording what was happening.) Now that I have studied the schematic, checked with a simulation, and compared to my measured results, I understand what is up.
Before I explain what is up with the C64's C90 electrolytic capacitor, let me briefly explain the power supply design of the Commodore 64. I'll also correct something I said in the video.
The Commodore 64 and its variants do not take mains AC into the motherboard. At least, not directly. There is an external brick that provides two voltages which are a regulated 5 volts and 9 volts AC. The 9 volts AC is provided by a transformer from mains AC. It is essential to generate this voltage from an AC source. The wall's 50 or 60 Hz provides timing to some of the C64's circuits.
The 5 volt supply is the bulk 5 volt supply for the majority of the logic chips. A subtle point, and one I forgot while recording, is that the C64 has two different 5 volt busses. But let's count up the other busses first.
Combining the external power supplies with the onboard regulators gives the C64 four different power rails: +12, +5can, +5, and +9 unregulated. As I mentioned before, an external supply provides the +5 rail. Then you might wonder, what is up with +5can? Why are there two supplies? I'll get to that.
The 9 Vac from the power brick goes to a rectifier (CR4), a capacitor (C90), and to the two CIA (Complex Interface Adapter) chips. The diagrams on this page ignore the CIA circuit. However, for reference, the signal drives the "Time of Day" input which operates a real-time clock. That leaves two other circuits to consider.
The rectifier connected to the 9 Vac input is rather ordinary. It is a single chip 4-diode construction with some smoothing caps. Its primary function appears to be to supply a full rectified signal into the 7805. That familiar regulator creates the rail called +5can.
So why are there two 5 volt rail in this otherwise simple design? Well, when looking at a larger schematic, it appears this rail only provides power to the VIC-II "RF can." (Pictured above.) My guess, and at this point is only a guess, is the designers knew how critical it was for the analog video signals to have clean voltage. So they opted to provide the sensitive VIC-II with its own 5 volt supply, free of all of the digital logic's switching noise.
Here is what I need to correct. In the video, I changed C102 with a polymer and declared: "now the system will be more stable! The 5 volt rail has less ripple!" While it is true that 5 volt rail has less ripple, the statement is only partially correct. The problem with that statement is that this 5 volt rail does not supply any of the digital logic chips. Later, I want to go back and see if there is a video quality change by using polymers vs. electrolytics since the polymers should make better filtering capacitors.
Okay, that's one mystery solved. But what about this other section with the capacitor C90.
C90 - AC on an electrolytic?
The first point I need to make is that C90 is a polarized electrolytic capacitor. In the schematic, the cathode, or negative side, connects to the 9 Vac and not a ground signal. That configuration puzzled me a little bit, so I measured the voltage across the capacitor. In this case, I put the probe tip on the anode and the ground clip on the cathode.
Even though the capacitor is connected directly to an AC signal, it only sees about a 1.5 volt peak-to-peak swing with a low side voltage of around 7.7 volts. Okay, so that's good, it isn't getting reverse biased. So then what the heck is going on?
My next step was to re-draw the majority of this circuit into a simulator. While I could measure the same points with my oscilloscope, the simulator lets me do something my scope can't. In the simulator, I can remove components to see how the circuit's behavior changes.
While looking at the schematic, I realized a couple of things. This strange C90 is feeding into a 7812, which creates a 12 volt output signal. Then it hit me, C90 must be a voltage doubler. Embarrassingly it took me a while to see the circuit. To me, the circuit is drawn backward. (My fault, not the original's designers.) So I was not reading the components correctly.
Here's a voltage doubler I simulated for reference and then I "reversed" the components to match the C64's service manual. This voltage doubler is made up of two circuits. The first is a clamper provided by the diode I named "Clamp" and a half-wave rectifier. Thinking back to my scope measurements for the video, I showed the output of this doubler as the input to VR1.
In the video, I made a point about the approximate 2 volt peak-to-peak signal. I ignored the DC offset. While it might be hard to see in the screenshot, the waveform's voltage ranges from about 16.9 volts to 15.5 volts. Ah-hah! I completely missed what this measurement showed:
- The measurement shows the 9 Vac effectively doubling. Compare the RMS voltages and you'll clearly see the doubling.
- The diode CR6 is a positive biased clamper, which adds a DC offset to the half-rectified waveform. This circuit is such a clever (and cheap!) way to double the input voltage!
With this biased clamping doubler circuit, the 7812 gets plenty of input voltage to provide a stable 12 volts out. By the way, the primary chip that benefits from this regulator is the Sound Interface Device, or by its more popular name, the SID.
(Please note. There are two versions of the SID, which means there are two versions of this circuit. Later generation SIDs ran on 9 Vdc instead of 12 Vdc, which makes its circuit different.)
Sometimes when making video content on a deadline, things get said (or promised) that can't always happen. In this case, I was confident I would figure out what C90 was doing. However, at the time, I missed a couple of key things I wish I had pointed out. Fortunately, I captured plenty of screenshots and waveform data, so I could come back and analyze the data later.
At some point, it might be worthwhile to dig deep into this circuit for a video. But for now, it's time for me to get back to my bench.
P.S. There are many revisions of the C64 schematic (and PCB.) The component designators in this post are for the most common board, but will not match all of them.