atxpower supply
I have some arcade cabinets with computers in them, and I'm using a regular PC PSU for them, which sucks because it takes up so much space.
I know there are pico-psus on the market, but I don't really get why I can't just take 12v and 5v from the JAMMA harness and wire that in to an ATX connector.
Is it possible to do?
Yes - a 10 bit ADC can be used to measure the voltage, with suitable front end scaling ! :-). A 10 bit ADC has a nominal range of 2^10 = 1024 steps or 30V/1024 = 29 mV/step.
You could make this 25 mV/step for a 25.6V max reading or 50 mV/step for a 51.2V max reading. You don't HAVE to use all the range. At 50 mV/bit you can probably read to an accuracy of about 0.1V with due care in construction.
You could easily have the meter set to dual ranges so you can measure with more accuracy at low voltages. Say a low range of 0 - 5.12V with 5 mV steps, or 10.24V max with 10 mV steps. As before best achievable actual accuracy is liable to be about double the step size. Or more than 2 ranges if desired.
Don't cut the connector off - find a socket to suit - then you can change supplies if needed.
You'll need to find how to start the supply - join pins 14 & 15 with a link etc. GET THE RIGHT PINS. Pinout and details here. Note their comment on need for a load
Above basic diagram of power on connection from here
You may need a minimum 5V load to get the supply to regulate well. This was true for older supplies and is still true for many newer supplies.
You mention a 30V output, which is not available from a standard ATX supply. This can be achieved by modifying an ATX supply. I won't start to discuss that here. It can be discussed in more detail if of interest. There will be numerous web pages that describe how.
Here are two examples of ATX supplies modified for variable voltage. YMMV - I haven't checked these in any detail. Many more on web.
The following diagram is from Example 1 above.
Achieving higher resolution by measuring relative to an intermediate reference voltage.
It has been suggested that you can achieve even more accuracy (or resolution) by measuring the voltage relative to an offset voltage. eg you could read 30v +- 5 or similar.
This is true, but there is a major "gotcha".
By adding a "pedestal" you are effectively building an ADC with more bits. And, any voltage that you use as a reference point has to be accurate to as many bits as the effective number of bits that you are trying to achieve with the new system.
eg Imagine you decide to measure a range of 0-28 volts. If you step the reference voltage in 1/4 of Vmax steps (0, 7, 14, 21V) you can get 0-7, 7-14, 14-21 and 21-28 V ranges, each with 4 x the resolution you'd get using the same ADC to measure 0-28 Volts. You are effectively adding 2 bits to the ADC resolution.
BUT each step of voltage has to be accurate to at least 10+2 = 12 bits of accuracy (preferably a bit or 2 more). If the voltage reference steps are NOT accurate to say 12 bits then you are measuring 10 bits of resolution relative to a semi random number and the answer will be another semi random number. GIGO applies.
eg 0-7 volts with 10 bits gives 7/1024 = 6.8 mV per bit. The processor can look after the fractional mVs and probably a display that steps by 10 mV steps will be fine. However, if you want to measure from 21 to 28 V with 10 mV resolution then the 21 V reference should be at least 21.00 V. ie accurate to 10 mV and preferably somewhat better. Say about 21.000 +/- 2.5 mV. If the 21V reference is in fact 21.023 V then when you report a Voltage as 26.840V it will really be 0.023V lower (as the reference is 0.023V too high) for a true voltage of about 26.82V. If you don't think that reading 26.84 when it should read 26.82 is important then you didn't need the extra bits that we have been trying to attain and the while exercise is not only a waste of effort but also misleads people into thinking that they have accuracies that are not present.
Accuracies better tha 0.1% or 1 part in 1000 are achieved only with design, planning, and substantial care and attention to detail.
Why does current flow from PSU a to PSU b?
If they're ATX supplies then the output GND is connected to protection earth, so both GNDs are connected together to the power strip's earth (or the house wiring). So, no surprises here.
So after seeing the results of that, I had another idea. I wondered if the PSUs could sink current from a 9v battery. So I connected (-) of the 9v battery, and 12v from the PSU according to the following schematic
This sounds like your switching PSU is leaking a bit of high frequency common mode noise. This is normal and expected for a switching PSU, although the actual current should be very low to meet with EMC/EMI regulations.
To have current flow, you need a closed circuit. In your case, what closes the circuit would be the parasitic capacitance between the 9V battery and the ATX power supply metal case. If I'm right, the LED illumination will vary if you bring the 9V battery closer to the ATX supply. It would also work if you replace the 9V battery with any metallic object (like a kitchen spoon, whatever) since what matters isn't the fact it is a battery, but rather a conductive object large enough to form a capacitor with the PSU case.
Try it ;)
I find this question interesting, because the first part is really basic, while the second is rather high-level ;)
Best Answer
You can, provided you know the amount of current your motherboard requires, and that the power supply in your arcade cabinet can supply that amount of current.
Furthermore, there is more going on there then I think you think. The ATX standard has 3.3V, 5V, 12V, -12V and sometimes -5V connections:
You would also need to fake the
PWR_ON/PWR_OKsignals, which shouldn't be too hard (just a resistor pulling the relevant IO lines into the relevant states), though managing power-up and power-down sequencing could be somewhat involved (ideally, you'd want to switch the JAMMA power supply using the signal from thePS_ONline, but that would also require having a working 5V standby source.