Putting a PT-65B switchmode power module into an Apple IIe

In post #6, softwarejanitor wrote:

"How are you getting the -5V output?  Adding another regulator like an LM7905?  If so, which other output do you tap to get the power for it the +12V?"

Uncle Bernie answers:

The PT-65B makes -12V and so it's cheap to add a 7905 regulator to provide the regulated -5V rail for the Apple IIe. But exactly this is the "mechanical problem" I ran into and botched on my first attempt. I found a solution, though, but this might not be the preferred solution for all users.

You see, my goal was to really 'hide' the mod that no casual observer could find it out. It would have been very easy to just drill a hole into the front side wall of the power supply metal enclosure for the screw holding the 7905 in place (it should have a nice heat sink, just in case some random slot card is discovered which loads the -5V rail all too badly, not very likely to happen, buy why take chances when it can be done right). I used a TO-220 mounting kit sold by JAMECO to mount the 7905 in an isolated way. It comes with a mica sheet that goes between the 7905 and the metal enclosure (thin coat of thermal paste on both sides) and the screw itself goes through an insulating plastic insert for the mounting hole of the TO-220 package. A very nice solution which is hard to put together from individual components.

The mechanical problem came from my desire to avoid that visible screw on the front side of the power supply metal box. So I tried to mount the 7905 to the bottom of the box. I thought that I could just drill a hole and then chamfer it such that the screw head would be sunk and not stick out, preserving a perfectly flat bottom of the power supply. I did not want to drill a hole into the bottom part of enclosure of the Apple IIe itself, which would have allowed for the screw head to stick out.

But as it turned out, the chamfer, when deep enough to hide the screw head, weakened the sheet metal enough that when tightening the screw, its head just got pulled through the sheet metal, leaving a larger hole, and a loose 7905. Botched ! What a disaster ! Here is that hole (red circle):

PDRM4285.JPG

Despite of this failure I insisted on mounting the 7905 on the bottom of the metal box. I would have hated to cut all the nice wiring and heat shrink tubing open to redo it again, with longer wires, to mount the 7905 on the front side of the metal box. Again, lesson learned: always do all the mechanical work before wiring anything up. Then, the attempt to tighten the screw would have had the same disastrous outcome, leaving this ugly and useless hole, but then I could have opted to mount the 7905 at the front side, with longer wires, and accepting that a small screw head would be visible. This is the way I'd recommend for anyone why wants to put such a PT-65B into an Apple IIe power supply shell. Form follows function. Not vice versa.

The nasty (and not recommended) solution I used to avoid rewiring of the 7905 was to mill a flat recess into the bottom of the box, just deep enough to sink most of the screw head, but not too deep to weaken the sheet metal too much. This was done with a mill bit in a drill press - which is not the right way to do it, because a drill press is not a milling machine (which I don't have). The bearings of drill presses are not designed to take any X Y forces caused by typical milling operations, but for just milling downwards (Z axis) they can take it without getting damaged. But there is some wobble which ain't no good, as the mill bit wants to rattlethe work piece away until fully engaged. It really sucks to have no machine tools like mills and lathes. On the other hand, how often would they be used ... not worth the investment for most hobbyists. The outcome was this:

PDRM4289.JPG

Note how I took a file to the screw head, after the 7905 was installed, to flatten it down flush with the metal surface.

Again, this is not the best way to do it. I recommend to mount the 7905 on the front side of the power supply box. near the point where the cable bundle to the motherboard comes out. As I didn't do this in my case, sorry, I have no photos how that would look. Maybe if you do your power supply conversion the right way, mounting the 7905 on the front side (from the  inside  of the box of course), you can snap a photo and post it in this thread.

Other than that, the conversion is quite straightforward. The PT-65B  almost  fits into two of the the mounting standoffs for the original PCB, but not perfectly, so a little bit of filing a notch into the PT-65B PCB must be done, to allow the use of these original standoffs:

PDRM4277.JPG

Sorry, I have no photo of the notch itself. It is hidden below the washer visible in the yellow circle. Just be patient and slowly file the notch, as the red outline suggests, with a round file, until the PT-65B  can be put onto these two standoffs. The one on the other side of course uses the screw hole on the PT-65B itself. Two washers are recommended (one below, one above) to avoid cracking the PCB. Be gentle when screwing on the nut (not installed yet in the above photo). The nuts from the original PCB mount may be reused, but I did not find a proper socket fitting to them. This solution has adequate mechanical strength, because the other two standoffs for the PT-65B are tailor made: the required holes are marked on the sheet metal, by using the PT-65B as a template for the holes.

The technique I used is to mount nylon standoffs (machine screws work, too) into the two remaining mounting holes of the PT-65B, and then I put some paint in their tips. Mount the PT-65 onto the existing two standoffs and make the tips touch the sheet metal, where they leave some paint points. These are the locations to drill the holes. After removing the PT-65B out of the danger zone, drill the two holes with a drill press, but drill a little bit larger than the screw diameter, to have some leeway for inaccuracies. Then use a tool with a domed tip and a hammer to drive the sheet metal in from the bottom side, just enough that the flat screw heads of #6-32 flat head machine screws (length  3/4") are flush with the metal surface. This tool can be improvised by using a ball bearing steel ball glued onto a steel tube of slightly smaller diameter than the ball. Nylon standoffs are put into a small machine vise and taken to the drill press to hollow one end out, such that it sits snug on the metal domes / eyes, see this photo:

PDRM4269.JPG

Afterwards, the nylon standoffs are cut to the right length with a saw. The proper height is the same as for the exisiting metal standoffs.

Next step is to install a new AC connector. This is necessary because the original connector is soldered into the PCB of the old power supply:

PDRM4270.JPG

The important thing here is to find a connector which fits into the existing cutout. I can look up the type number of the one I used, but not right now.

Also note the bending trick which was used to be able to drill the screw holes from behind, with the connector heat glued in place. This bending can only be done once. If you prefer to avoid bending the sheet metal, you can drill these holes from the front, but getting them in the right place requires more measurements, maybe a template, and if you screw that up, and the holes are off position, the connector will sit with an ugly slant or could not be installed at all. I prefer to use the mounting holes of the connector as a template / guide for the drill bit.

The PT-65B is mounted using flat washers with star washers on top of the flat washers on the PCB solder side, and flat washers on the component side. The nuts are #6 nuts. You could use the ones left over from the original PCB, but these are smaller and I had no proper socket fitting them.

PDRM4275B.jpg

The above photo shows the situation before the star washers are put on. Then comes the PT-65B PCB, then another set of four flat washers, then the nuts. Star washers will make good electrical contact with the "ground" islands around the mounting holes of the PCB. This is an important electrical safety feature. Use the prescribed washer sequence. Do not overtighten the nuts. Just tighten them with reasonable finger force until snug, and then tighten another 90 degrees turn, not more.

You can see now that the mechanical work of this power supply conversion is the most difficult part, and requires a drill press and some small tools not available in a typical household tool box. The electrical wiring as such is almost child's play. After the original cable has been desoldered from the PCB, a new 6 contact MOLEX connector is installed while adding the two extra wires (from V3 = -12V and COM on the PT-65B) to the 7905.

Here are a few photos showing the wiring:

PDRM4264.JPG

(You would use longer wires to the 7905, so it can be mounted on the front side of the metal box)

Here is the line voltage side wiring:

PDRM4297B.jpg

Note that there are two green / yellow ground wires, one goes from the AC connector to the bolt on the right side, and one from the AC connector to the ground tab on the PT-65B.

And here is the final result:

PDRM4299B.jpg

Don't forget to fasten the ground connection nut, in the above photos it's still loose for a reason (some conductivity tests I wanted to do before powering it up - the PT-65B has its own ground contacts on the PCB, and I wanted to measure the resistance they have to the sheet metal).

For the power requirements of an Apple IIe system, the smaller and cheaper PT-45B would do. It is built on the same PCB as the PT-65B, but uses lower profile components. All electrical connections and the mounting hole placements are the same.

So much for today's internet session. I have much more photos and information, but now you have the basic ideas how it was done. If you like it, and want to do it yourself, I can provide more info. Just encourage me with comments (and questions you might have).

- Uncle Bernie

In post #2, CheshireNoir wrote:

"I'd like to see your take on this so put me as a +1 to showing uis the PT-65B conversion. "

In post #15, CheshireNoir wrote:

"My "best" trick, for instance, was using some acrylic as a layer between the Meanwell and the case. Insulative and allows you to mount the meanwell without having to drill holes.

Uncle Bernie replies:

Had I known about your website with your conversion, it would have saved me some time, and probably I would have avoided the pitfall with the hole in the bottom of the box. I think the internet has solutions for most problems, but the trick is to  find  them. And when you don't know they exist, you don't know how too look / search for them. So things get reinvented / work gets duplicated all the time. Still, much a better situation than back in the day, before internet. Public libraries were all you had.

I've looked at your website, and your conversion is nicely done. One comment, though, and this is about the little PCB with the capacitor on the 7905. IMHO, these added components are superflous. I am well aware that the datasheets of those regulators recommend adding bypass capacitors both on the input and output of these regulators, but as long as the wire to the PT-65B is not too long (which it won't be in this case), the output filter capacitor on the PT-65B will serve well as the input bypass capacitor of the 7905.  The output bypass capacitor is said to 'improve load step response' but as long as there is a small bypass capacitor on the Apple IIe motherboard, and the power cable has the same length as it has, no such capacitor is needed on the 7905. I think these recommendations in the datasheet were motivated by issues they have seen in the field where power supplies with the regulators were farther away from the load, and once the wires get long enough to add some inductance then load step response may indeed suffer. The input bypass capacitor may only be needed if the distance from the rectifier / filter capacitor to the regulator is longer than 1 ft. At least this is the length when we started to see first signs of instability / ringing when evaluating such regulators. Note that so-called "LDO" regulators react much less benign to such long wires. It's easy to destabilize LDOs or to turn them into oscillators. So better avoid LDOs unless headroom too low. The 78xx and 79xx regulator families are much more robust against such adverse conditions (long wires and poor or nonexisting bypass capacitors).

But it's up to the builders if they want to include these extra bypass capacitors or not. What I would NOT do in any case is to use a solid tantalum (despite, again, the ancient datasheets spell that out). Solid tantalums, especially when aged, are a fire hazard. So better, avoid them.

Your solution with the acrylic as an interposer to mount the PT-65B on the original standoffs avoids a lot of trouble with sheet metal work and is easier to do. But it's an insulator so the ground islands around the PT-65B screw holes make no contact with the sheet metal box. Now, you are down to one ground connection, and this the the one to the ground tab on the PT-65B. If this connection falls off / becomes disconnected, the whole power supply is electrically unsafe and also may radiate excess EMI. To save your concept, you could use sheet metal instead of the acrylic, or add at least one additional ground connection at one of the screws contacting the PT-65B PCBs. They make round contact thingies for screws, same sort you can see at the grounding bolt in my photos of the line voltage side, with a solder tab for a wire. You could use two of those and two star washers to connect at least one of the ground islands of the PT-65B to a screw / bolt which contacts the metal box. This will keep the PT-65B grounded even if the ground tab gets disconnected. The importance of grounding the switchmode power supply properly cannot be underestimated. These grounding islands are there for a reason.

- Uncle Bernie

In post #24, softwarejanitor wrote:

So my strategy here is just going to be to cut a piece of plastic to the size of the original AE power supply board and mount it into place with screws like the original board was.  The PT-65B board will just be screwed down to that.  Then it is pretty much just making the hookups to the power connector/switch and the output side with the added LM7905 like Bernie's example.

Uncle Bernie comments / warns:

When using an insulator plastic sheet as an interposer between the PT-65B and the metal shell (which requires less metalworking), do not forget to use at least one grounding lug at one of the two "safety gound / earth" PCB foil islands of the PT-65B, and connect that to the metal enclosure, which, if eloxated or painted, may need some grinding to the bare metal to make a good contact. Verify that this ground connection is low ohmic from the ground tab of the PT-65B to the metal enclosure. The reason for this was given in post #16. You want the PT-65 safely grounded even if the wire to its ground tab would fall off.

The IDC connectors mentioned by macintosh_nik can be used, but avoid the white ones, the slots in their contact springs are too narrow for the wires found in such power supplies. Strip isolation, twist and pre-tin the wires and then solder them into these IDC connectors (from the top). You do not need to drive the wires deeper into the contact springs because once the enclosure is closed again, none of these wires within will be wiggled anymore. Both the +5V and 0V/COM rails need two wires / contact springs, for both pins the PT-65B offers for each of them. The reason is that one of the IDC contact springs alone can't support the full rated output current of the PT-65B.

These IDC connectors  are a good replacement for the MOLEX connector system I've used for my mod - - - these need a heinously expensive crimp tool and the crimp contacts which have to be bought on a reel of 1000 pcs.

While you making your order to Mouser, you may also add this connector for the PT-65B grounding tab (it's a crimp type, but solder the wire into it unless you have the expensive special crimp tool, and not the cheap Chinese made toy pretending to be the tool):

Faston connector for 5.21 x 0.72 mm tab.

TE Connectivity part #640911-1.

Mouser part #571-6409111

and if needed, the line cable connector, too. Oh, and Mouser would also have sold you a PT-65B (or PT-45B) as an official distributor of "Mean Well" products. No need to buy from dubious sellers on Amazon or Alibaba. One of my beta testers for the +8V mod got defrauded by a Chinese scammer who sold him a badly replaired, old PT-45B they obviously had fished out of the electronic junk recycling stream, and for the same price Mouser rings up. All he "saved" was the postage. And the time spent for  hunting down the fault would have cost  thousands  of US$ if it wasn't a hobby.

- Uncle Bernie

In post #28, softwarejanitor asked:

"Do those include the metal parts or do those need to be ordered separately like some connectors ?" 

Uncle Bernie answers:

Yes. These IDC connectors come with the contact springs installed. But you can take individual contact springs out: depress the latching tab with a pointed tool, and then pull the spring out from the top using tweezers (or needle nose pliers). If you do that on positions #2 and #4 of a 6-pin connector, and then cut the plastic shell at the now empty position #4, you have turned a 6-pin IDC into a 3-pin IDC with the middle contact missing (for air gap) and this is the right one for the line voltage input of the PT-65B. This is why I recommend to buy 10 pcs of the 6-pin connectors, once you have Apple-1 around, or want to put PT-65B into an Apple II, you have all you need in stock. These are very cheap and very useful connectors.

- Uncle Bernie

In post #33, Wayne asked:

"Some power supplies come on the instant you flip the switch while others come on a second or two after. Which one is the PT-65B ? "

Uncle Bernie answers:

Under typical load conditions, a PT-65B (or PT-45B) takes somewhere from 1-2 seconds to bring all the output voltages up.

For most applications, this is a non-issue, and the original Apple II power supply also took some time to "start up".

In both the PT-65B and the PT-45B, the +12V rail is developed by having one output winding of the pulse transformer "riding" on the +5V rail (with a rectifier diode, of course). So at least during power up, the +12V rail never should have a voltage lower than the +5V rail. When the power is turned off, however, the situation depends on the loads seen by both rails. But according to my measurements, it's never more than one diode drop difference. But the +12V (V2) rail indeed may have one diode drop lower voltage than the +5V (V1) rail during power down.

Back in the day, most ICs were robust enough not to get damaged with sluggish or random power up or power down sequences. Very trivial solutions - such as diodes between +5V and +12V rails) could be adopted to prevent that the +12V rail ever was lower than the +5V rail by more than one diode forward voltage drop. Same trick can be applied to negative power supply rails. If Schottky diodes are used, with a forward voltage less than a PN junction, then nothing bad can ever happen to the ICs (normally).

But with "modern" ICs with various "core" and "IO" voltages (or some special functions, like LCD drivers etc.) a strict power up / power down sequence may be mandatory, otherwise these ICs would get damaged. For this purpose, the semiconductor industry offers various families of "power sequencer" or "power system management"  ICs. Before retirement, I worked as an IC designer for one of the leading semiconductor companies, and believe me this topic can get very complex and mind boggling. Imagine what happens when somebody jerks the line cord of such a system out of the outlet. How can you detect that event in due time such that the (safe) shutdown sequence still can be completed ?

Or worse, what if one of the many dozen lower level switchmode power supplies / supply voltage converters on a $10000 PCB blows up and that output voltage is lost (almost) instantly ? There are IC which can draw 100 Amps at 0.8V core voltage. What to do to avoid damage to the sensitive multi supply voltage ICs on that board ? --- This is why the effort and expense to design and develop a proper power supply concept for such system can become very high, often higher than planned.

For me (now being a mere hobbyist) yet another reason to avoid modern FPGAs having multiple "core" and "IO" voltages like the plague.

- Uncle Bernie

In post #39, softwarejanitor  wrote:

" Now that I'm looking at putting everything together I had a weird thought.  I'm almost tempted to drill and tap a screw hole into the end of that big heat sink on the PT-65B and screw the LM7905 down to that.  Maybe even just drill the hole and use a self tapping metal screw.

Am I crazy ? "

Uncle Bernie answers:

No, doing that is not crazy, but it is dumb, because you would create a short circuit between the V3 output rail  (-12V) and COM/0V.

In other words, it won't work. Most likely, the 7912 regulator on the PT-65B would protect itself from overload and limit the current, so nothing will blow up immediately, but nowadays, who knows who made the 7912 and on which process technology. Back in the day, when USA based wafer fabs owned by reputable semiconductor companies made them, based on the original American design of the late 1960s / early 1970s, these regulators were almost indestructible (I know tricks how to destroy them, though, like the infamous "file test" using long lab cables for inductance).

But nowadays, fly by night no-name Chinese wafer fabs making them at 5 cents a piece, possibly based on stolen / reverse engineered mask layouts, who knows. I encountered one Chinese made 7912 which had such thin leadframe that it delaminated and cracked the die (aka "chip").

The metal tab on 79xx TO-220 regulators is not connected to 0V/GND (like on the 78xx) but it is connected to the input voltage, which is the most negative voltage a 79xx  sees on any of its three "pins". The deeper reason is that it's a junction isolated bipolar process, and the p-ISO structures isolating the NPN collector tubs (N doped), actually are PN junction diodes, which always must be reverse biased, otherwise, everything in the IC gets connected via these diodes opening up, which is bad ! These p-ISO structures are ohmically connected to the the p-substrate (the "die" = chip) and this is bonded to the TO-220 carrier, aka leadframe, with an ohmic contact. So there is absolutely no other way to build these negative regulators cheaply. You could use film attach to isolate the die from the lead frame  but that costs more money, and these IC are penny items. There exist versions of these regulators with a fully plastic encapsulated body and mounting tab. But these are much more expensive and harder to find.

A solution you could try is to take out the screw of the the 7912 regulator on the PT-65B itself, and then use it to attach the 7905 regulator right there, such that the tabs mate. Installation on the opposite side of the heat sink is not possible (at least on my specimen) because they have a bulge there in the metal which contains the thread for the screw. Since you now have connected the 7905 to the raw negative input voltage via the tabs, all you need is to add a small ground return wire and the fat wire for the -5V would directly go to the output of the 7905. Be aware that the ground return wire is essential. If it ever breaks off, the output of the regulator will go to the input voltage, and now you've got unregulated -20V (or more) instead of -5V regulated.

Still, why make such an ugly kludge. You can buy isolated mounting kits for TO-220 like the one I did use.

- Uncle Bernie

In post #41, softwarejanitor wrote:

"And that heat sink does not attach to anything else".

Uncle Bernie cautions:

That heat sink is still connected to the TO-220 tab which is connected to the input pin of the 7912.

(Beware of "invisible" electrical connections --- this also applies to the heat sink of the primary side swicthing MOSFET, which is "hot" = electrical shock hazard).

-  Uncle Bernie