For those commenting on 18Farads being rediculous, note that my post and the schematic specifies 0.18Farad, or 180,000uF. 18Farads would have, indeed, been crazy.
Quoted from G-P-E:
Actually, minor correction - Data East calls out 0.018F caps, not 0.18F caps. When you see 0.18F such as on the Secret Service schematics - that's a typo. Remember that they pretty much plagiarized Williams for awhile including power supplies. Bulk capacitance of that value in the 1980s would have been horribly expensive and impractical.
I thought 0.18F was awfully big, but then every schematic I came across (admittedly I was just looking at Secret Service manuals) showed 0.18F, including my original paper copy. For reference, the first image is from the Secret Service manual. Not that I don't doubt it could be a typo. More I'm including it for reference.
Quoted from G-P-E:
The noise you are seeing on the input is directly related to the switching supply.
Looks like the replacement board has included the capacitance for the switch mode power supply. But is there really no bulk capacitance? That seems very strange.
The input wave (yellow) appears to be a considerably higher frequency than the standard 60Hz/120Hz(FWB) hum and is actually the switching frequency of the regulator. You would have needed to change your time base on the scope to a much longer period to see the 120Hz hum. Your trace looks exactly like the "Input ripple voltage plot" posted in the Texas Instruments "Input and Output Capacitor Selection Application Report" for no (or insufficient) external capacitance. Their plot had a saw tooth like yours but had a 200mV p-p which they considered high.
Note that the scope timebase is set to 5ms/div. 120Hz is 8.33ms, or 1.667 divisions, which lines up with the scope capture and also agrees with the measured frequency of channel #2. See the measurements in the low left of the scope capture.
Quoted from G-P-E:
What a few of the board makers use in their design is the application notes for their specific voltage regulators. The app notes specify that the regulator requires x amount of low-impedance input capacitance so that is what the board makers use. What the board maker totally disregards is that the regulator manufacturer is basing the design on a DC power source that has *already been sufficiently filtered* at the input - i.e. has sufficient bulk capacitance. A single, 1000uF cap for both bulk filtering and voltage regulator ramp up is insufficient. All that gobletygoop you see on that yellow scope trace would have been seen at the sound board as well (minus the minor inductance filtering of the cabinet wiring).
When I was selling pinball boards - I used 4x input capacitors. Two paralleled large caps for bulk capacitance and two paralleled smaller, low-ESR caps for the voltage regulator high frequency ramp up. I would have expected to see the same sort of arrangement on the replacement board as well.
Yeah, and I made a mistake too. As noted in my original post the bulk capacitance on the +12V, on the XPin board, is only 220uF. The -12V rail has 1000uF, which does match the original schematic. So the +12V rail is about two orders of magnitude low in capacitance.
Quoted from G-P-E:
What you did is unclear by how its written. You took the 12V unregulated power, inserted a regulator to give you...12V regulated output? Or did you take the 5V output and use the boost function to take it up to 12V?
Yeah, not the clearest from my picture or post. I make a male to female connector to put inline with the power connection to the audio board. The +5V,-12V, and GND lines pass straight through. The +12V line has a buck/boost converter inserted into it.
https://www.amazon.com/gp/product/B07NTXSJHB/
Above is the one I used. It will switch between boost and buck mode automatically to keep the output at the set voltage. Honesty I didn't expect it to work as well as it did. It's a pretty simple and cheap board and I expected a glitch when the input crossed the output set point. However it behaves fine, both on the bench and installed. I'm sure you could make it misbehave under certain circumstances, but in an hour of play last night, I didn't have any problems. Suppose I need to dig up the datasheet on the controller and see what it says about the conditions I'm using it in.
Quoted from G-P-E:
I was going to suggest using a pi filter at the input of the DE board. But DE already has them on their board and pi filters are typically for higher frequencies. Is there really only one pin assigned for the power ground connection? Also looks like DE has only one 1000uF filter on the input, that could have been better for lower freq filtering.
If they're relying on the now nonexistant 18000uF bulk cap for filtering - that one is now gone from the replacement board. Plus the fact that the cap was remotely located on the other board only exaggerated the problem. Better isolation between logic 12V and analog 12V is what DE needed. Would not be an easy task to fix. If I were designing a replacement power board - I would include two 12V bridge rectifier circuits on the board - one for logic power and one for analog power. And include real bulk capacitors and a heavy duty, shielded power connection between boards and all kinds of bells and whistles. Gee, aren't "If I could build these, I would do it better" always great!
There are at least three grounds to the audio board. The main power connector, the data cable to the CPU, and the mounting studs. This is why some people are dealing with ground loop hum. If there really was only a single ground, things would be fine in that regard. As far as what I would do in a replacement design, well that depends. If I was trying to be as absolutely generic as possible, I probably would have put boost converters on the +12V and -12V rails, followed by buck converters to bring them back to exactly +/-12V, and a separate buck converter for the +5V line. You's lose a little efficiency with the double conversion, but it would still be way better than the original linear supply.
For others, I had also posted regarding this on the KLOV forums and G-P-E responded there as well. I thought I'd include part of my response over there, regarding the bulk capacitance, or lack thereof on the power board. Again, more information for those looking in the future.
Since the switching regulator on the XPin board is outputting 5V, the headroom is sufficient that the ripple due to missing bulk capacitance doesn't impact the 5V regulator. It outputs good, clean 5V. Unfortunately the unregulated 12V also directly feeds something that is sensitive to that ripple. The XPin board only requires about 20mA of load on the unregulated 12V to cause about 400mV of ripple. The original Data East board, with 18,000uF of bulk capacitance would have needed about 1.2A of load on the +12V line to get the same 400mV of ripple. Now, given that the Xpin is an efficient buck regulator, it could certainly get away with less bulk capacitance, but not 100 times less. A pair of 2200uf caps would have dropped that ripple down to the 20mV range. For reference, the 0.18F specified on the schematic would have put the original power supply at around 30mV ripple under 1.2A load. I don't have an original Data East power and CPU board, but I can certainly believe 1.2A load as realistic, meaning it would behave just like the XPin board -- or that the XPin board behaves just like the original.
Screen Shot 2020-02-16 at 3.46.42 PM (resized).png