Modern Amiga PSUs: deeper than a rabbit hole!
- Thread starter fpmpaolo
- Start date
- Replies 128
- Views 4848
Ground in AC vs. DC Circuits
In AC (alternating current) systems, "ground" refers to a common reference point for electrical power transmission, whether from lightning, power lines, or the earth itself. For example, ground in Italy is functionally the same as ground in Indiana. However, in DC (direct current) circuits, "ground" is a reference point relative to the power source. Unlike AC ground, DC grounds from separate power sources are not inherently connected (i.e., they lack a quantum spin connection).Buck Converter Wiring Explanation
The question "Where do the cables go?" is too broad to answer definitively, but for a Buck converter:- IN+ receives the input voltage (e.g., 12V).
- GND connects to the power supply unit’s (PSU) ground, serving as the 0V reference for positive and negative voltages.
- EN (Enable) acts as a cut-off switch:
- High (positive) or floating = ON
- Grounded (tied to PSU ground) = OFF
- V-Output delivers the converted power.
PSU Voltage Regulation Analysis
If a PSU regulates only one rail (e.g., 5V), the ±12V rails may rely on a secondary circuit for regulation. According to the *8-Bit Dreams* post, the ±12V rails lack regulation, necessitating a Buck converter to prevent overvoltage—though my testing doesn’t support this claim. Two possibilities arise:- The RT-65B poorly regulates ±12V, while the PT-65B maintains them within 10% of expected values.
- Both tests suffer from a low sample size (N), making the results statistically unreliable (i.e., anecdotal). Further testing is needed to confirm whether either outcome is an outlier.
Addendum: Statistical Significance
SAS/STAT software calculates the required sample size (N) to detect an effect of a given magnitude (e.g., a 10% change). For instance, if a 10% difference is expected, but observational variance is 2%, the software determines how many samples are needed to validate the effect confidently.
Last edited:
Interesting thread!
I've replaced the innards of a couple of Amiga PSUs with Mean Well PT-65B PSUs.
Until now, I hadn't realised about minimum load requirements, even though it's written on the PSU's data sheet in plain sight! Here's the PT-65B one:
Mean Well's website also has a page about minimum load requirements:
meanwellpowersupplies.com
Based on the above, an Amiga 600 would meet the minimum load requirement for the +5V line, but not the +12V line. There is no minimum load requirement on the -12V line.
So far, none of my Amigas have died while using these PSUs (even with the 12V line running a little high at 13.7V - could that be down to poor regulation caused by the minimum load not being met?). However, I want to get these PSUs set up with the proper minimum load.
If I aim for adding a 0.3A load on the 12V line, then R=V/I = 12V/0.3A = 40 Ohms. And P=V*I = 12V*0.3A = 4 Watts.
So I'm thinking I'll put a 25 watt (way more than needed) 40 ohm resistor between +12V and GND on the PSU output.
I've replaced the innards of a couple of Amiga PSUs with Mean Well PT-65B PSUs.
Until now, I hadn't realised about minimum load requirements, even though it's written on the PSU's data sheet in plain sight! Here's the PT-65B one:
Mean Well's website also has a page about minimum load requirements:
What is minimum load requirement and where to find this information? - All About MEAN WELL Power Supplies
There are some minimum-load requirements on MEAN WELL's multi-output power supplies. Please read the specification first before connecting to the load. To allow the power supply to work properly, a minimum load for each output is required, or else, the output voltage level will be unstable or...
Based on the above, an Amiga 600 would meet the minimum load requirement for the +5V line, but not the +12V line. There is no minimum load requirement on the -12V line.
So far, none of my Amigas have died while using these PSUs (even with the 12V line running a little high at 13.7V - could that be down to poor regulation caused by the minimum load not being met?). However, I want to get these PSUs set up with the proper minimum load.
If I aim for adding a 0.3A load on the 12V line, then R=V/I = 12V/0.3A = 40 Ohms. And P=V*I = 12V*0.3A = 4 Watts.
So I'm thinking I'll put a 25 watt (way more than needed) 40 ohm resistor between +12V and GND on the PSU output.
Please post some pics when you do that!
Will do! I've ordered the parts. Hopefully they will arrive before the weekend. And hopefully I'll get some time on the weekend to fit them!
Quick update (turned out not so quick once I got going!). The resistors haven't arrived yet but I've put the wires in place on one PSU already. The PT-65B has two through-holes for GND on the output side. With one already in use, I added a wire through the spare one. On the +12V line, there's only one through-hole, so unfortunately, I had to solder a wire to the underside of the PCB, as you can see in the pics below:
Two spare holes - one on GND (we can use this
), and one on +5V (we can't use this 
):
One wire soldered to the GND through-hole. The other wire soldered to two pads (+12V) on the PCB:
This is what it looks like from above. Not ideal, but the wires are connected:
The resistor has not been delivered yet, but then I had a thought: I've got a very old 12V car tail light bulb in the garage - could I do a test using that? I checked the bulb, and it's rated at 21W. 21W/12V = 1.75A - well within the 2.5A rated current of the +12V line on the PT-65B while leaving more than enough headroom for my Amiga 600's +12V requirements.
I temporarily hooked up the bulb and fired up my Amiga 600. Here's how the 5V / 12V line voltages as measured at the floppy drive power connector differed:
- Before attaching bulb: 5.04V / 13.7V
- With bulb attached: 5.26V / 12.8V
Leaving the bulb attached, I then used the PT-65B's adjuster to bring the 5V measurement down to exactly 5.00V, which brought the 12V value down to 12.26V, as you can see here:
So with a more appropriate load on the 12V line, the PT-65B's output is looking better.
I've ordered 25 watt (though rated at 9 watt without heatsink) 47 Ohm resistors. One of those will draw 0.255A (3.06W), which is a lot less than the 1.75A (21W) than the bulb draws, but still meets the PT-65B's +12V minimum load requirement. It'll be interesting to see what kind of numbers I get with the resistor vs the bulb - I'm expecting the 5V and 12V values to diverge, but we'll see.
I'll update again once the resistors have been delivered.
Two spare holes - one on GND (we can use this
), and one on +5V (we can't use this ):One wire soldered to the GND through-hole. The other wire soldered to two pads (+12V) on the PCB:
This is what it looks like from above. Not ideal, but the wires are connected:
The resistor has not been delivered yet, but then I had a thought: I've got a very old 12V car tail light bulb in the garage - could I do a test using that? I checked the bulb, and it's rated at 21W. 21W/12V = 1.75A - well within the 2.5A rated current of the +12V line on the PT-65B while leaving more than enough headroom for my Amiga 600's +12V requirements.
I temporarily hooked up the bulb and fired up my Amiga 600. Here's how the 5V / 12V line voltages as measured at the floppy drive power connector differed:
- Before attaching bulb: 5.04V / 13.7V
- With bulb attached: 5.26V / 12.8V
Leaving the bulb attached, I then used the PT-65B's adjuster to bring the 5V measurement down to exactly 5.00V, which brought the 12V value down to 12.26V, as you can see here:
So with a more appropriate load on the 12V line, the PT-65B's output is looking better.
I've ordered 25 watt (though rated at 9 watt without heatsink) 47 Ohm resistors. One of those will draw 0.255A (3.06W), which is a lot less than the 1.75A (21W) than the bulb draws, but still meets the PT-65B's +12V minimum load requirement. It'll be interesting to see what kind of numbers I get with the resistor vs the bulb - I'm expecting the 5V and 12V values to diverge, but we'll see.
I'll update again once the resistors have been delivered.
Last edited:
@GentleBen
PT-65B has 3 GND holes on the output.
IIRC pins 3-4-5.
I am working on a PSU like yours and I chose PT instead of RT model, as it has a 50mV ripple instead of the 80mV of RT's.
PT-65B has 3 GND holes on the output.
IIRC pins 3-4-5.
I am working on a PSU like yours and I chose PT instead of RT model, as it has a 50mV ripple instead of the 80mV of RT's.
Yeah, I like the fact that the PT has the ripple within Commodore's original spec (I'm sure I read that somewhere), unlike the RPT models.
Maybe you're thinking of the PT-65D with the three GND pins? The PT-65 spec sheet (https://www.meanwellusa.com/upload/pdf/PT-65/PT-65-spec.pdf) shows that the PT-65D has three, but the others have two.
Yes you are right
I tried out a resistor - 47 ohms, 25 watts. It works but poses another problem to solve.
Without the PSU plugged into my Amiga 600, I measured 0.254A being drawn by the resistor - pretty much in line with my expectation of 0.255A (12V/47Ohms).
With the PSU plugged into the Amiga, I measured 0.27A current on the +12V line. As expected, the voltages diverged compared to my car bulb experiment I mentioned above, but they still look a lot better than when the minimum load requirements for the +12V line weren't being met. Before adjusting the 5V line to exactly 5.00V, I saw 4.94V / 12.60V. After adjustment, I saw 5.00V / 12.74V - not too bad!
But now I've got another problem - HEAT! The resistor gets pretty toasty - after a few minutes of operation, I can touch it, but only for a second. I'd really like a thermal camera to measure it, but I don't have one. It feels like a lot of heat output to put inside the Amiga's plastic PSU case which already contains the PSU board which is trying to get rid of heat itself.
So I've got no pics of a nice solution to this to upload, unfortunately.
I'm not sure what to do now.
Give up?
Reduce the amp draw down to exactly the minimum required, which is 0.2A (12V/0.2A=60Ohms) ?
Use an even larger wattage (i.e. bigger) resistor that can dissipate heat more easily?
Design a metal case for the PSU (which will need to be earthed, obviously) to use as a heatsink?
Think of a way to install a heatsink?
Install a 100mm 12V fan? Would that create noise on +12V and GND? Does it matter?
A combination of some of the above?
Without the PSU plugged into my Amiga 600, I measured 0.254A being drawn by the resistor - pretty much in line with my expectation of 0.255A (12V/47Ohms).
With the PSU plugged into the Amiga, I measured 0.27A current on the +12V line. As expected, the voltages diverged compared to my car bulb experiment I mentioned above, but they still look a lot better than when the minimum load requirements for the +12V line weren't being met. Before adjusting the 5V line to exactly 5.00V, I saw 4.94V / 12.60V. After adjustment, I saw 5.00V / 12.74V - not too bad!
But now I've got another problem - HEAT! The resistor gets pretty toasty - after a few minutes of operation, I can touch it, but only for a second. I'd really like a thermal camera to measure it, but I don't have one. It feels like a lot of heat output to put inside the Amiga's plastic PSU case which already contains the PSU board which is trying to get rid of heat itself.
So I've got no pics of a nice solution to this to upload, unfortunately.
I'm not sure what to do now.
Give up?
Reduce the amp draw down to exactly the minimum required, which is 0.2A (12V/0.2A=60Ohms) ?
Use an even larger wattage (i.e. bigger) resistor that can dissipate heat more easily?
Design a metal case for the PSU (which will need to be earthed, obviously) to use as a heatsink?
Think of a way to install a heatsink?
Install a 100mm 12V fan? Would that create noise on +12V and GND? Does it matter?
A combination of some of the above?
Last edited:
Just a note that RT(129x98mm) will definitely not fit in the newer PSU enclosures (600/1200) due to width.
The PT(127x76mm) has a shot to go into those lovely enclosures, at lower cost as well. I think I'll give that a shot.
I get 74mm internal width on those "ribbed" enclosures, you think the PT has 2mm width to spare, if needed? Angle it a bit?
Last edited:
I think the Commodore spec sheets notes 10% +/- on 12V and 5% +/- on 5V voltage spec is acceptable, so that's within spec.
Also, was that a base machine you used or one with some upgrades demanding more draw?
Last edited:
I am expecting it tomorrow. I'll give you the exact dimensions even though I don't own a 74mm PSU enclosure.Just a note that RT(129x98mm) will definitely not fit in the newer PSU enclosures (600/1200) due to width.
The PT(127x76mm) has a shot to go into those lovely enclosures, at lower cost as well. I think I'll give that a shot.
I get 74mm internal width on those "ribbed" enclosures, you think the PT has 2mm width to spare, if needed?
Good idea! I think I'll try the 2 resistors to start with, and I'll make both of them 100W ones - just for more surface area. Is it better to have 2 x 120Ω in parallel or 2 x 30Ω in series? Is there any difference?
That's good to know. Is there somewhere online that has the specs of the various Amiga PSUs?
The numbers I've posted so far have been on a test stock A600 board that I have (sockets for most of chips for testing). To get better real-world numbers (for me, at least) I opened up my main A600 which has an actively-cooled Furia installed, IDE/SD card, 1MB chip ram expansion and RGB2HDMI with a Pi Zero 2. With the PSU adjusted for exactly 5.00V / 12.74V on the test board and with the resistor in place, plugging into my Furia A600 gave me 4.86V / 12.8V. Adjusting the 5V line gave me 5.00V / 13.13V. If I ever get this PSU done, I guess I'd go for a happy medium with something like 4.9V / 12.89V (which I also saw in my testing) on my main Furia A600.
Not sure if this is useful, but I've got an old A600 PSU enclosure that looks like this:
...which isn't one of the "ribbed" ones, I guess? Anyway, the PT-65B will fit into the top part of this PSU case, but the screw posts get in the way on the bottom half:
It's these cases I'm going to try to squeeze the PT-65B into. I've only ever got these from UK. 74mm wide internally, ideally I hope it fits on a bit of an angle to clear the 76mm PT-65B spec width. Fun to get that original epoxy block out of these, ain't it?
Unfortunately I didn't get mine yet.It's these cases I'm going to try to squeeze the PT-65B into. I've only ever got these from UK. 74mm wide internally, ideally I hope it fits on a bit of an angle to clear the 76mm PT-65B spec width. Fun to get that original epoxy block out of these, ain't it?
In theory if you place the PSU as is with angle, one side should be 17mm higher.
But if you can trim 1mm the pcb this goes down to 6mm. Not bad.
Sorry I can't describe it better
Over the weekend, I managed to fit a couple of 100W 28Ω resistors in series (so total 56Ω) inside the PSU case with the PT-65B, and it seems to be working ok - the voltages are within spec, but still not perfect.
Here's some pics showing the resistors fitted:
She may not look like much, but she's got it where it counts.
I used some old ZX Spectrum heatsinks that I had knocking about, along with the perfect amount of thermal paste, just to spread the heat a bit. I knew they'd come in handy one day! They increase the surface area by only a small amount, but better than nothing. I had to cut a few mm off one side of each heatsink to get them to fit.
The numbers are looking better than without the resistors in place. The PSUs minimum load requirement is being met, and the voltages are within spec, though still deviating from idea 5V/12V values by a bit more than I'd really like after all this effort.
The aim was to load the PT-65B's 12V rail with 0.2A (the minimum load requirement for that rail). A 60Ω resistor between +12V and GND would give us that. I went with 56Ω just to make sure that any variances didn't take us below 0.2A. 56Ω should provide a 12V/56Ω=0.214A load. I measured 0.219A load with no Amiga attached to the PSU, which is pretty close to expected. With my A500 attached, I saw 0.229A load.
We need to dissipate 12V*0.219A=2.628W of heat. I bought two 100W resistors to try to maximise surface area and lower the max temperature of each resistor - the single 25W/47Ω resistor I tested (see above) would get pretty toasty.
With my main Furia A600, I now see 4.90V/13.01V at the floppy drive power connector -within Commodore's spec of +/-5% on 5V and +/-10% on 12V: http://www.kaiiv.de/servicemanuals/A500_Power_Supply_Specification.pdf . On my unexpanded A600 test board, I see 5.02V/12.89V.
I've had the PSU running for a few hours, and it gets warm, but not hot - the screws on the outside of the case are a good indicator because they are attached directly to the resistors. It feels like there's good air movement through convection.
Amiga PSUs are indeed a rabbit hole! I've learned a lot from all the contributions on this thread!
I can see where Jens from iComp is coming from with his disdain for Mean Well and other PSUs (see his YouTube video) - getting a PSU to meet the Amiga's exact specs is hard! The Mean Well RPT range have out-of-spec ripple and also have minimum load requirements on all three rails = +12V, -12V and +5V, and an Amiga is only going to meet the 5V load. The PT-65B has within-spec ripple, but needs a minimum load on the +12V and +5V rails. The Amiga covers the 5V one, but we have to install resistors to get the +12V minimum load. I'm not sure it's worth all this effort when I can buy a CA-PSU from iComp made by someone who actually knows what they are doing. It's fairly expensive, but it really does meet the specs from what I understand.
Here's some pics showing the resistors fitted:
She may not look like much, but she's got it where it counts.
I used some old ZX Spectrum heatsinks that I had knocking about, along with the perfect amount of thermal paste, just to spread the heat a bit. I knew they'd come in handy one day! They increase the surface area by only a small amount, but better than nothing. I had to cut a few mm off one side of each heatsink to get them to fit.
The numbers are looking better than without the resistors in place. The PSUs minimum load requirement is being met, and the voltages are within spec, though still deviating from idea 5V/12V values by a bit more than I'd really like after all this effort.
The aim was to load the PT-65B's 12V rail with 0.2A (the minimum load requirement for that rail). A 60Ω resistor between +12V and GND would give us that. I went with 56Ω just to make sure that any variances didn't take us below 0.2A. 56Ω should provide a 12V/56Ω=0.214A load. I measured 0.219A load with no Amiga attached to the PSU, which is pretty close to expected. With my A500 attached, I saw 0.229A load.
We need to dissipate 12V*0.219A=2.628W of heat. I bought two 100W resistors to try to maximise surface area and lower the max temperature of each resistor - the single 25W/47Ω resistor I tested (see above) would get pretty toasty.
With my main Furia A600, I now see 4.90V/13.01V at the floppy drive power connector -within Commodore's spec of +/-5% on 5V and +/-10% on 12V: http://www.kaiiv.de/servicemanuals/A500_Power_Supply_Specification.pdf . On my unexpanded A600 test board, I see 5.02V/12.89V.
I've had the PSU running for a few hours, and it gets warm, but not hot - the screws on the outside of the case are a good indicator because they are attached directly to the resistors. It feels like there's good air movement through convection.
Amiga PSUs are indeed a rabbit hole! I've learned a lot from all the contributions on this thread!
I can see where Jens from iComp is coming from with his disdain for Mean Well and other PSUs (see his YouTube video) - getting a PSU to meet the Amiga's exact specs is hard! The Mean Well RPT range have out-of-spec ripple and also have minimum load requirements on all three rails = +12V, -12V and +5V, and an Amiga is only going to meet the 5V load. The PT-65B has within-spec ripple, but needs a minimum load on the +12V and +5V rails. The Amiga covers the 5V one, but we have to install resistors to get the +12V minimum load. I'm not sure it's worth all this effort when I can buy a CA-PSU from iComp made by someone who actually knows what they are doing. It's fairly expensive, but it really does meet the specs from what I understand.
Last edited:
