|
Post by dshadoff on Feb 8, 2019 0:09:01 GMT
I was thinking about building something to attach to the controller port on the PC Engine, and I was wondering:
What is the maximum current that is expected to be drawn from the port ?
That is to say, if you have a 5-way multitap with 5 controllers attached, which all have turbo, you'll have quite a number of chips attached, and the cables will add to the power demands.
...Just wondering if anybody had measured how much this is (or is able to measure it).
Dave
|
|
samiam
Punkic Cyborg
Posts: 100
|
Post by samiam on Feb 8, 2019 5:46:14 GMT
It looks like controllers draw power directly from the system's 5V rail. In that case, I would think the bottleneck for power would be the limits of the 7805 regulator. The specs for the core PC Engine system say that its power consumption is a measly 4W; let's suppose that this is not with more than one controller attached or any extra peripherals on the EXT port. With 9V coming out of the AC adapter*, 4W means that the system is using ~440mA and that the 7805 is dissipating ~1.8W. Per the datasheet, a 7805 with no heat sink is able to dissipate ~2W at room temperature. With a really good heat sink, that jumps to ~7.5W. PC Engine systems employ various solutions for sinking heat from the 7805, none of them looking "really good" but none of them really bad, either. If we suppose that 3W is a safe amount to dissipate from any one of them, we could probably attach something to the controller port that drains around 300mA all by itself. According to this, a SNES mouse requires 50mA while the SNES Multiplayer 5 adapter only allows 17mA per controller. If a typical no-frills controller of the era only needs around 20mA, we might suppose that five PCE controllers and a multitap wouldn't use more than 150mA. Of course, I'd love to see proper measurements. *I've looked at a few game-console "9V" AC adapters on an oscilloscope while under load, and they are almost always averaging more like 10V, and sometimes more like 11V. That changes the calculations with the 7805 considerably.
|
|
|
Post by dshadoff on Feb 8, 2019 6:16:48 GMT
I was looking at making my peripheral operate within the power envelope of existing equipment, so I was hoping for actual measurements. I was also under the impression that each controller would be in the tens of milliamps, but I have no real basis for this belief, other than a vague feeling about how much power things consumed, from back in the day.
Maybe I just need to go buy myself a multimeter made this century, and measure it myself.
On the bright side, if up to 150mA is available, I should probably be fine within that envelope without working to hard on optimizing the efficiency.
Dave
|
|
samiam
Punkic Cyborg
Posts: 100
|
Post by samiam on Feb 8, 2019 9:32:19 GMT
Yeah, that makes perfect sense.
Low-end multimeters these days (in the $30-40 range) are pretty good. You'll get 99% accuracy with virtually every type of measurement.
Another cool thing about now is that you can get decent analogue oscilloscopes from the 90s for very low prices. With one of those, you could stick a weak resistor in series with the 5V line to the controller(s) and observe the voltage drop across it at very specific moments, getting an even clearer sense of what kinds of current tolerances Hudson had in mind.
One thing I noticed when recapping my systems was that earlier PCEs have one 47uf electrolytic cap decoupling the power rail while the Duo has one 330uf electrolytic cap in the same role. That's not because of the CD drive, either; the drive and controller chips all sit behind their own second 7805 regulator. Of course, all systems have lots of little 120nf ceramic decoupling caps as well.
While the 7805 is supposed to be able to keep a power rail's voltage fairly constant even without large decoupling caps, I bet that Duos would be able to withstand larger swings in power demand before having any problems with interference.
|
|
|
Post by SignOfZeta on Feb 9, 2019 3:03:17 GMT
I know that if you make yourself a controller and you f it up you will blow the fuse in the system. For this reason I recommend using a system to develop this that has a fuse in it, like CG1 You could also just pull that fuse out, hook your meter in line with the socket and measure different things by plugging them all into the port one at a time. My assumption is that it can handle double what the controllers normally draw. I don’t have any data on this but the potential for spikes and rush is significant with five pads and a tap so it would need headroom to avoid occasionally blowing fuses. The estimate of about 400mA is correct for a CG. They can run down to like 250mA but the video will dim if the colors get too bright. The. CDROM pulls probably half an amp by itself when the laser moves. Otherwise it’s draw is much lower.
|
|
|
Post by soop on Feb 9, 2019 20:13:15 GMT
Yeah, that makes perfect sense. Low-end multimeters these days (in the $30-40 range) are pretty good. You'll get 99% accuracy with virtually every type of measurement. Another cool thing about now is that you can get decent analogue oscilloscopes from the 90s for very low prices. With one of those, you could stick a weak resistor in series with the 5V line to the controller(s) and observe the voltage drop across it at very specific moments, getting an even clearer sense of what kinds of current tolerances Hudson had in mind. One thing I noticed when recapping my systems was that earlier PCEs have one 47uf electrolytic cap decoupling the power rail while the Duo has one 330uf electrolytic cap in the same role. That's not because of the CD drive, either; the drive and controller chips all sit behind their own second 7805 regulator. Of course, all systems have lots of little 120nf ceramic decoupling caps as well. While the 7805 is supposed to be able to keep a power rail's voltage fairly constant even without large decoupling caps, I bet that Duos would be able to withstand larger swings in power demand before having any problems with interference. That's interesting. I'd be interested to see the differences between an OG PCE PSU and a Duo version. Can't imagine a duo version would be worse tho.
|
|
|
Post by dshadoff on Feb 10, 2019 0:14:58 GMT
So, I got a bunch of stuff from Amazon today, and I did a few measurements on my Duo.
Voltage supplied by the joystick port = +5.18V Current draw of a single Hori Fighting Commander = 0.8 to 1.1 mA Current draw of multitap + TurboDuo controller + Hori Fighting Commander pad + 2 PC Engine pads (4 pads total + multitap) = 2.8 to 3.5 mA
These measurements are actually a lot smaller than I had expected, even though the Hori seems to draw about double what the official PC Engine controllers do.
Separately, a PC-E Mouse consumes about 6.5mA (again, I expected more).
...Of course, I'm sure the machine can *drive* more power than that. The fuse inside of an original TurboGrafx is supposedly rated at 1.25A, even though the AC adapter is rated at 650mA. I just don't know at the moment how much headroom is available, and I'm not sure how much I want to assume.
On a related note to Sam's about the different capacitors in different editions - back around 2001, I had created an alternative for the Tennokoe Bank, using a 5V-only byte-addressable flash memory (I think it was a 29F040).
I tried it on both an original PC Engine with CDROM attachment, and also a TurboDuo; it worked fine on one, but not the other (I don't remember exactly, but I think it failed on the Duo). It appeared to be related to whether the power source was strong enough the drive the chip's needs during the erase and write cycles (read/code execution was fine). At the time, I didn't put a lot of thought into adding a large capacitor (like a couple of hundred microfarad) to stabilize the power, but I probably should have.
Anyway, all this talk about power is for a separate reason: another item I bought was an Arduino beginner's kit; an Arduino is a 5V-compatible microcontroller with various attachable items. I was thinking about connecting it up to the PC-E, and seeing what mischief I can cause.
Dave
|
|
samiam
Punkic Cyborg
Posts: 100
|
Post by samiam on Feb 10, 2019 1:26:10 GMT
So, I got a bunch of stuff from Amazon today, and I did a few measurements on my Duo. Voltage supplied by the joystick port = +5.18V Eeek. While that is probably within tolerances, one thing I'd encourage anyone recapping their systems to do is replace their 7805s. Not only are those a common failure point in old consoles, but new ones tend to be closer to an even 5V. Wow, that is tiny. I suppose it's possible, though. I'm looking at the datasheet for the 74HC00, a simple set of four AND gates, and the basic supply current is 40uA. Driving high logic at the outputs could also require very little current depending on what it's talking to; it gives one example of 20uA per pin. If you went all the way up to 1.25A, the 7805 would be dissipating 5W all by itself if it only had 9V on its input. Another thing that occurred to me is that if I were designing something to distribute, I'd want to take into account all of the people running Japanese power bricks on North American outlets. That's another 20% in voltage, on top of what I was mentioning before about unregulated "9V" adapters often outputting more like 10V. There is a protection diode in there that should drop the voltage around 0.7V, but still, I'd assume that people playing Japanese PC Engines in North America are putting around 12V on their 7805s. With that in mind, I'd be wary of drawing more than an extra 100mA or so from that power rail. If the rest of the system wants ~400mA and you take another 300mA, and if the 7805 has 12V at its input, you're talking 4.9W dissipated at the 7805, and I wouldn't trust any PCE heatsinks to handle that much. (If anyone wants to do a highly scientific test of putting their finger on the 7805 while the PCE is running, I'd be curious to know how hot it actually gets under normal operation. They say 55 degrees is where it gets too hot to leave your finger indefinitely. Of course, if you have a temperature meter, that's cool, too.) A large capacitor may have helped. Did you have a small-value ceramic capacitor near the chip's power supply pin to smooth high frequency noise? The PCE apparently has a noisy power rail - hence the need for some people with RGB mods to do this. If that mischief includes something that can get save data in and out of the system and is practical for people to build or buy...forget the Ys t-shirt, I'd wear a shirt with your face on it for something like that.
|
|
|
Post by dshadoff on Feb 10, 2019 2:31:59 GMT
Wow, that is tiny. I suppose it's possible, though. I'm looking at the datasheet for the 74HC00, a simple set of four AND gates, and the basic supply current is 40uA. Driving high logic at the outputs could also require very little current depending on what it's talking to; it gives one example of 20uA per pin. Right... my expectations were more like 10mA, based on what I ha known about "Low Power Schottky" (74LSxxx series) chips. I have an old (1984) databook about the "new" 74HCxxx series, and they mention that these use substantially less power than the LS series. I played a lot with electronics as a kid, but I suppose I stopped around this transition, so I wasn't aware about the magnitude of the power change. I thought that the North American mains were closer to 110V... but then I remembered that I just bought a nifty multimeter and hadn't put it away yet. Wall voltage: 120V AC PC Engine adapter (Japanese), rated at 9V: with no load, I read 15.5V TurboDuo (American), rated at 10V: with no load, I read 13.6V (This probably drops when loaded, but I wonder whether component degradation would lead to this too ?) These readings might also be higher than realistic, as they are still DC with a waveform (just rectified AC), so the "RMS" figure may be more like the stated value. Still that means the 9V input is still closer to 11V. (Even though the TG-16 Repair manual says to expect 10.5V) So, I guess I'll go looking for a new power supply. My PC Engine gets warm on the AC adapter side, so the closer to 9V that it actually is, the better. I didn't mod it, so whatever was there to begin with. And the machine was only 10 years old at the time, so less likely to be degraded parts. Today, I would position the (new) capacitor as close to the Flash chip's power leads as possible. Ha ha... yes, something like that. New Develo, or perhaps MB128 replacement, or even a "PC mouse to PCE mouse" converter. Lots of potential on that port. Dave
|
|
samiam
Punkic Cyborg
Posts: 100
|
Post by samiam on Feb 10, 2019 3:26:06 GMT
I thought that the North American mains were closer to 110V... but then I remembered that I just bought a nifty multimeter and hadn't put it away yet. Wall voltage: 120V AC PC Engine adapter (Japanese), rated at 9V: with no load, I read 15.5V TurboDuo (American), rated at 10V: with no load, I read 13.6V (This probably drops when loaded, but I wonder whether component degradation would lead to this too ?) These readings might also be higher than realistic, as they are still DC with a waveform (just rectified AC), so the "RMS" figure may be more like the stated value. The no-load readings will always be much higher than loaded ones with these unregulated wall-warts. Also, the amount that it drops depends on the load itself. Ironically, with an original adapter, the more juice you pull through the PCE's controller port, the lower the voltage at the 7805's input will be, further complicating the wattage calculation. There are only three things inside of those adapters: a transformer, a four-diode bridge, and a big smoothing capacitor. To my knowledge, low-voltage transformers don't really degrade (and high-voltage ones that do only wear down their insulation and short), and the diodes are more of an all-or-nothing sort of deal. It's the usual culprit - the capacitor - that you have to worry about. I've opened Supergrafx, Super-CD, Genesis and SFC adapters, and they all used completely unremarkable 16V electrolytic capacitors that were anywhere from 2200uf to 10,000uf. They weren't high-grade or even high-temp, which is alarming given how warm these adapters can get. Frankly, I'd expect every OEM adapter out there for any older console to have a smoothing capacitor that is badly degraded. The result of this should be a larger, deeper 120Hz ripple on the output. With a brand-new capacitor in it, the SFC adapter's output under load (the console) on an oscilloscope showed a roughly 10.5V peak and 300mVp-p ripple, IIRC. I didn't look at the others. This is with 100V AC from the wall in Japan. According to the datasheet, 7805s need at least 7V at the input to maintain 5V at the output unless the load is very small. 7.5V power supplies are common, but that protection diode is going to put you in danger of going below 7V. The next step up among common supplies is 9V, which is fine. I've picked up a few 9V SMPS from used-junk shops to use with my consoles. On the other hand, if you can get a nice used one, I recommend a PS2 slim adapter. It's 8.5V, rated for 5.65A, and I've personally verified that it outputs little in the way of high frequency noise. 7805s get worse and worse at ripple rejection as you climb in frequency, and some off-the-shelf SMPS output terrible noise. That's definitely one reason not to go with a cheap no-name brand in this case. Exciting stuff. Have fun, and let us know what you come up with!
|
|
|
Post by soop on Feb 11, 2019 13:35:14 GMT
Well, it's not quite the same, but I've had a Famicom open and touched the heatsink on the voltage regulator and it's VERY hot. More than uncomfortable, you wouldn't be able to touch it very long (like a second) without burning your finger. And that 7805 is actually bolted to the RF shield to provide an extra sink. It was so hot, that rather than remove the RF shield completely (unneeded for an AV mod), I purposely bought some tin snips and left a large section of it attached.
|
|