Triple output meanwell modifications (fail)

For a recent project I wanted to find a unified solution for a nice triple output 8,16,-16 volt power supply. My first thought was the triple output meanwell power supplies. I can get them in 5, 12, -12 variants (RT-65B), and they have adjustment pots to maybe I can tweak them up a bit. They also come in 5, 15, -15 (RT-65c) so that one I’d only have to change the 15v rails a little and the 5v rail a bunch. I got one of each and proceeded to try to tweak them. I learned a lot but I didn’t get them to do what I wanted. These are specifically the 65W variants because that’s what fit my project’s size and power specs, but there are others that probably use a similar topology.

You can tell this is a triple output and not a quad output variant because of the missing components, I’ll get to that soon. You can see there’s one transformer and one adjustment knob so that’s not very promising for tweaking the voltages individually. That’s why I got the 5/12/-12 so all the voltages would go up roughly the same amount with one adjustment.

What you see on the upper right is the adjustment and overvoltage protection, near the bottom right are optoisolators to give that feedback to the regulator chip in the high voltage section, and that regulator is at the bottom.

I’m going to start with V3, the negative voltage populated on this board. You can see that one transformer tap is half wave rectified with D56 before being smoothed with C63. You might think that the diode and capacitor are backward, but this is a negative linear regulator so that’s the correct configuration. It looks to me that D56A is an optional alternate footprint for that diode so they could use either package depending on availability. We then have the linear regulator, a 7912 which burns off all excess voltage to create a regulated negative 12v source. This means no adjustment for this regulator, but you could swap it out with another one if there’s enough headroom on the input voltage to support a higher voltage output. The output smoothing/tank of C62 is expected and the 2watt resistor in R59 may be to keep a load on this leg of the transformer even without any load on the power supply. I don’t know why there’s another parallel footprint for R81, perhaps if there are variants that need to dissipate more than 2 watts on this leg (this board is used for many designs). D37 is interesting, it’s not populated because if it was then the regulator would be bypassed, but in the wrong direction. Perhaps it’s a protection device if some voltage above the negative 12 volts shows up on the input to V3 then this diode protects the regulator.

The non-populated section is the same as V3, but for a second negative rail on the same power supply. There are quad output meanwells that use this same board and this is how they do it. I also don’t have measurements of that either of the power supplies I got put out on this transformer pin (if any) but I will update the post with information on that here [nothing, this transformer must not have that winding, but the board is common].

V2 is the high voltage output and it is regulated as a secondary function of the low voltage being regulated. What I mean by that is V2 is a series of diodes, smoothing caps, and loading resistors on a couple pins of the switching transformer. There is no 7812 regulator because this leg is rated at too high a power to be able to use one of those and this helps keep costs down. The ratio between this V2 leg and the V1 leg are such that when V2 is outputting about 12v then V1 is outputting about 5v, that’s the ratio of turns between them in the transformer so regulating one regulates the other.

V1 is where all the magic happens. This is the high current low voltage leg. The right section up to the series inductor is about the same as V2, a bunch of filtering and half wave rectification. The filtering after that inductor is about the same as well, but it includes the power indicator LED. The important part about this leg however is the feedback and overvoltage protection. It is my understanding that the SHR1 device is a SOT23 overvoltage protection device that opens the connection in the sense circuitry if the voltage at the output goes too high. That would keep the DC voltage from getting back to the optoisolators we’ll talk about in a bit that regulate the duty cycle of the switching regulation. The trimpot also seems to be part of a voltage divider to allow the 5v output voltage to be varied a little bit to make it exactly 5v. My hack around all this was to remove the voltage divider resistors and fit a pot that can go all the way from 0 to output voltage for feedback (not just tweaking one leg of the voltage divider) and to short across the overvoltage protection IC. In doing so I could drive the voltage up to 6.16v and no higher. It seems that this is the maximum I can get from this transformer even at 100% duty cycle and no load.

This is the actual feedback side of the supply, so this is where the signals go. Looking at this now it seems like there’s two possible paths to regulate the supply and the optoisolators are both pulling down on the feedback pin so they can both affect the output voltage. The top is the one I was messing with, I could manually turn that optoisolator from all the way off to all the way on with the pot setup I described above. It looks like there’s also a zener/scr based regulator as well that can alternatively drive the regulation circuitry. I need to investigate what happens if I remove R61 to take that out of the circuit entirely and reinstate my mods above to see what voltage I can get out of these supplies. Perhaps this was what was holding me back from seeing more voltage gains, but I suspect the transformer’s windings just don’t let the output get that high with my input voltage. If I find out more I’ll add that information here. [I disconnected that feedback and fried something. I don’t know what and I can’s seem to see where that wisp of smoke came from so I give up on this one]

I wouldn’t normally put in so many blanks but I think the information in here is still valuable as-is and since I’m traveling and can’t test these aspects of the power supply I will have to update this when I get home.

Poly88 replica, entry three (power supplies)

Continuing the series from the circuit board here and the lids here. This will be a shorter entry because I have not completed the mounts for this yet, but it does all work quite well. The part I’m having the most trouble sourcing for the poly88 replica is the power transformer. This is a ‘standard’ if small transformer with a 16v center tapped high current secondary and a 32v center tapped low current secondary. For my use case I need a 120v primary. I say standard because this is what just about all s100 machines would have used and rectified into 8vdc and +/-16vdc for their unregulated rails. I understand that some people run their s100 systems at 5 12 and -12 these days, but I’m not about to modify all my cards to remove regulators and make them vulnerable to damage if inserted into an authentic chassis. I thought about modifying a three output meanwell to achieve these voltages, but I can go over why that didn’t work in another entry. This is a brute force method using off the shelf parts available to me and a similar procedure can probably work for you even if you don’t get exactly the same parts.

For this project I got myself a 19vdc 2 amp power supply, a 60W 3-12v adjustable supply, and a 3A step down positive/negative regulator. First we’ll look at my modifications to the adjustable supply.

The potentiometer was very coarse and it was hard to dial in. It also had a power switch built in to it. I desoldered the original 3k potentiometer and bridged over the switch contacts so the power supply is always on when plugged in (the only way to turn it off was to change the voltage, a bad design for something that you want to set to a certain voltage and leave there). I then measured where the potentiometer was at in the approximate 8v position. I replaces the 3k adjustable resistor with an 820 ohm resistor in series with a 10 turn 500 ohm potentiometer. This severely limits my power supply’s range, but that’s a good thing here. That means that those 10 turns are not going between 3v and 12v, but right around that 8v I need to tune in (the response is not linear).

That worked out just fine

The 19v supply is kind of interesting. When I took it apart (with a hammer and vise) I found clipped off leads on the power in and out terminals. This is not the case this power supply was originally designed for. I don’t know if it’s salvage, or if it’s just a bulk purchase of supplies that were no longer needed for another purpose, but it’s interesting. The 19v is a common voltage for laptoops and battery charging and gives enough headroom for the regulator to do its job.

That’s all it is, a couple trimpots for tuning the output voltages to 16 and negative 16 volts and it works great. This seems like overkill for the amount of regulation needed at this stage. The downstream cards will have their own regulators anyway.

Here it is for scale, there’s plenty of room for a 3d printed mount taking off the original mounting holes for the transformer. With this design you can probably do away with all three large capacitors and all the power diodes in the main board and feed the DC rails directly. In the future I may design a stand alone power supply to accomplish this, but similarly to the transformer with specific windings being hard to find, in the custom switching power supply the specially wound transformer is the hard part to find. That being said I could pick something with a bigger range and regulate down like I’m doing here.

Part four involving the something TBD will be linked here.

(I also managed to score a poly 8813 with the original cards, so expect replicas of those at some point for a complete poly88 system cards and all)

Poly88 replica, entry two (lids)

Part one of this series can be found here, and the original post where I got these computers is all the way back here. Ever since that first post I have had these two poly88 machines and they’ve been missing a very important part, the bright orange lids. Well, I’m armed with a very rudimentary understanding of solidworks and a company that will not only fabricate sheet metal to my specification, but also powder coat it for me. Now is the time to complete these machines.

This is how they arrived, extremely sturdy and it honestly feels super cool to receive something in such a nice crate. It’s not a standard size either, this is custom fit to the parts I ordered so they’re adequately cushioned and I had no complaints about the condition.

They look good, don’t they? well, the first pass was not perfect but I mostly fixed the issues with a hand file and now they fit fine. Let’s walk through my process for building these and where I went wrong.

Here I start with a simple sketch of the front face of the completed part. This is where I made my first mistake. I referenced the height correctly, placing the origin of the part at one of the upper corners of the front face of the lower part of the chassis. I did not measure the width correctly though. It wasn’t an issue with the calipers, but how I used them. I got one measurement after clamping down with the calipers reasonably hard and then ran with it. In this case I have two chassis so I have the front and back flanges of both to use for reference. I also have the entire top to bottom section to measure. What I should have done is gently run the calipers down and carefully run them up and down each of the four faces keeping only the highest number I found. The number shown here represents the value after I did that, after I got my parts in. This isn’t a show stopper because it’s a tight fit but it does still fit but I have updated the CAD for the next person.

Here’s another potential blunder. My origin is that blue dot, inside the bend of the sheet metal. I think it should really be at the base of the bend so the lower part of the housing doesn’t jam right in that corner and it ends at the flat point. I thought I did this when making the bend, but it would seem not. This distance is so slight that I’m not too concerned, but it’s a consideration for the next model. This is where I say I’m not a mechanical engineer so I get to learn what matters about this stuff by doing it and seeing what I can get away with (honestly how I usually electrical engineer as well).

The next issue was my measurement of the length of the chassis front to back. This gets critical because I don’t have calipers big enough to do the whole length and I’ll be referencing both the front and back faces. Look at the shape of the lower chassis. It’s a folded metal box and the corners are welded so you may thing that this shape is fully constrained. That sheet metal is fairly soft, however. if you grab the top of the front and back flange you can move them in and out quite a bit. The right thing to do in this instance would be to measure the distance at the bottom to make sure you’re measuring where the material is sure to be the right length and not relying on those corners being square and the lines to all be straight over that distance. I did not do that and was off by enough to make the final fitment annoying. Not enough that I couldn’t fix it by hand, but enough that I regret these mistakes.

The next thing was to cut the slot for the s100 connectors on the sides, I used the card slot mounted in the lower chassis to indicate the center of the hole and went up enough to make it symmetrical. I used the cutouts in the front and back of the lower chassis to indicate where they should go on the upper as well. The only blunder here was again not checking both chassis and both sides to get a better feel for where the measurements are supposed to be. The final product I’m measuring is not perfect so I need to sort of split the difference on all these numbers.

The next thing was to capture the hole positions for the screws. I like having things be parametric where I know how so if I want to use a measurement twice I’ll give it a global variable name and then use that for subsequent uses. That’s what the red sigma is indicating here. I have the holes spaced 4.35mm from the front and back of the machine and if I change one it changes both. Both sets of holes are also constrained to be colinear with constructor lines, but because the upper set had offsets front and back I used a variable for that. Here’s where I tell you not only do I have two different circuit boards as mentioned previously, I have two different chassis. Look at the picture above and you can see the former owner drilled holes so he could mount these chassis together in a more permanent fashion and he could do that because the holes were not in the same place on this one as the other. As such I have two variants of lids for my two variants of chassis, but they only differ in the hole location.

This next part is where things get more interesting. I took a swag at these vents from a reference photo. I also cut some corners on the parametrization which means I had to go back and tweak them manually. The height and width were what looked good relative to the reference photo, the number was also taken from that. The spacing was a choice I made. I decided that they should start right at the edge of the metal flange of the lower chassis since any further toward the edge would be blocked and not act as a vent. The lower position was set by the lower flange and that was mirrored on the upper one for aesthetics. That fully constrains this design but in order to actually place all these I created one and duplicated it across until it landed on the end where I wanted it. The spacing should be a function of the total length, the number of slots, and the distance of the end slots to the end of the total length. I did this with a calculator and just plugged in the result, but that means when I went to update the length I had to update that spacing. For the lower slots I did the same sort of duplication, I copied a subset of the uppers and duplicated them down.

Since my origin was at the face of the lower chassis and I wanted overhangs on the front and back my last two operations were to extend the front and back face by a small amount. The simplest operation by far.

So, the fixes to all those issues? I used a small file to slot the nice round holes and for the most part the slots aren’t visible under the heads of the 6-32 bolts. Overall I am as pleased as I could be with my first bit of CAD in a decade and my first sheet metal CAD ever. PCBWay sponsored both of my beautiful orange lids and of you have a Poly88 without a lid you can order them here. Those are my fixed models which I have not ordered, but are at least better than the ones I initially got. My parts were powder coated in pantone orange 173C and are made of 1.5mm thick mild steel 1018 so if you want the same thing I got, do that. They are a bit heavy, but so is the whole machine. You’re obviously welcome to do it in whatever material you like. The color was a guess, if anyone knows a better color that’s more accurate just tell me and I’ll update this post with that information.

Part three involving the new power supply will be linked here.

Poly88 replica, entry one (PCB)

Since my post several years ago getting a pile of scrap to run as a functional computer I have gotten several requests for people who want to buy my Poly88 machines. I’ve resisted these offers because I know that if I let them go I will probably never be able to get another. Since I have these machines, I’ve decided to do my best to not just document them but to recreate them in CAD so anyone who wants one can just make it. They’re not that big and with manufacturing these days being so easy I see no reason to pine over an old sheet metal box, just make a new one. For this project I have also been sponsored by PCBway so all the parts will be available for purchase through their website as I design them. Part one here will cover the first part of the journey: the circuit boards.

I have two machines and naturally that means I have two different revision motherboards. They’re not very different, but I did bother to do both variants in CAD so you can pick your variant. The PCBWay project has the rev 2.1 board uploaded because that’s the one I ordered, but my github has both. You can find the schematics either on my github or in the manuals, but the only really noteworthy thing is that there’s an optional section for generating a signal on pin 55 for mains frequency based clocks. One of my original circuit boards had that section populated, the other did not, I believe that had to do with these chassis being able to connect for extra capacity. I’m going to detail some of the quirks I found while designing the board.

That red jumper wire had to be added to the rev e board because there was a trace missing for connecting one of the 8v legs from the transformer. That is mostly why I had the rev 2.1 boards manufactured, because they seem to have fixes in place.

This is the above section captured by me in KiCAD (shown from the top view, so mirrored)

This is the same section on the rev 2.1 PCB. You can see the connection is made, but there’s also some alternate connector for the 8vac pins. This is odd to me because you can’t just use those three pins since you need the +/- 16v for the other two rails on the s100 connector. I may have a hint for that however.

Notice the diode I labeled D8 and marked as Do Not Populate. The truth is you can populate it and I don’t think it will do anything harmful because this is how I have it on the schematic:

This diode sends voltage from the 8v rail to the 16v rail. That means that in a fully built system no current will flow as 16v is greater than 8v. If you choose to build the system with only an 8v rail however, this puts that on the 16v rail as well. There may be some use cases where you have an s100 system and you don’t need the full 16v on that rail and don’t need the negative 16v rail at all. I think that’s what the extra unpopulated 3 pin connector is for and this diode. I have not used them in my build but I left them in because I kinda like the reconfigurability of this board.

I have a note up there about the RTC circuit only being on one backplane as mentioned above, but I also note that 1, 2, 51, and 52 do not connect between backplanes. That is implemented on the male side of the connector. That means that the backplane, when acting as an s100 card, passes the signals back and forth but not the power rails. The ground is shared however. This means that you can expand these chassis up to however many slots you want, but each chassis’ cards are powered by the transformer in that chassis. I also said it’s implemented on the male side. Let’s take a look at that.

You can clearly see the pins making no connection with the rest of the board. Why is it important that it’s on the male side? because this means you can plug in an s100 card outside of the chassis on the female side and have a total of 6 cards connected to a single poly88 chassis at once. Why is this important? well, I may want to use the chassis with the lids on with an exidy sorcerer s100 chassis adapter card. Eventually. I have so many projects…

In an earlier picture you saw brown and white twisted wires going off the board, I made a note that this is used for the fan upgrade to the poly88 (not all of them had this feature, it was extra). Those wires go off to a 24vdc fan. That fan has three wires and a 5.6uF 250v cap on it so there’s something they’ve done to make it happy running off the 32v from the linear power supply. I may get into that more later as I replicate more of these computers.

That’s really all there is to these circuit boards, Your biggest cost will be the s100 connectors and if you want them to be exactly right your biggest headache may be the 0.093″ 5 circuit board mount polarized connector pictured above for the power input. On the PCBWay project I have linked to a set of wire to wire connectors that you can use with the board side being a pigtail instead of PCB mount, but you are of course free to use whatever pigtail you like. In the future I will probably be showing alternatives (I’ll certainly have alternatives to the power supply transformer, which probably means you can skip all the big caps and diodes too as my supplies will be regulated DC.

Part two involving the lids will be linked here.