Arrakis 2.0 Sand Table at Great Lakes Distillery

Arrakis 2.0

You're looking at the Arrakis 2.0 sand table, designed, built, and programmed by me, Mark Rehorst (whose name appears on GLD Gin bottles). The table is normally used as a coffee table in my home. It is on temporary loan to GLD.

Arrakis 2.0 at home

The sequence of patterns currently running on the machine can go for about 12 hours. There is no real limit, and it is possible to program a sequence that will run for weeks without ever repeating a pattern. As you may notice, the sand tends to build up around the edges of the table. Once about every 50 hours of operation I push it back toward the center of the table.

In the current sequence, some patterns finish in a few minutes, others take over an hour. After each pattern finishes, there is a 60 second delay to allow you to contemplate the Sisyphean pointlessness of existence, "ooh and aah", take pictures of the pattern, or use the restroom. Then it erases and starts drawing the next one. 

Despite rumors you may have heard to the contrary, the sand is definitely not cocaine. It's actually baking soda, used because its fine grains capture pattern details, it's pure white so the LEDs light it up nicely, it doesn't get sticky when it's humid, bugs don't eat it, it's cheap, and available everywhere. There are about 2 lbs of baking soda on the table.

Almost all the electronics in the table are off-the-shelf stuff, most commonly used in making 3D printers. Many of the mechanical parts in the mechanism are 3D printed or are parts commonly used in 3D printers, such as GT2 belts and drive pulleys. Unlike similar tables you may have seen, this one uses servomotors to achieve high speeds and quiet operation.

Those of you familiar with 3D printing will understand when I say the table uses a CoreXY robotic positioning mechanism to move a magnet under the table. The magnet pulls the steel ball on top of the table through the sand leaving behind beautiful patterns and messages.

The pattern generating software, Sandify, is free, online, and saves patterns as simple gcode files. I wrote a post-processor for the pattern files that assigns two speeds to the motion, fast along the edges of the table, and usually slower for drawing the patterns to preserve details.

The process of generating patterns using Sandify starts with selecting a basic shape and applying different modifiers. As changes, are made, their effects can be seen in a preview window. When I like what I see, I tell Sandify to save the pattern. Then I run a post-processor on the pattern file that assigns two speeds to the motion. Finally, I upload the pattern to the table via WiFi and it's ready to run. Patterns can be strung together in sequences using a macro file. It takes me anywhere from a few minutes to a few hours to generate patterns that I feel are worth keeping.

You may notice that the ball is "smart" in that it always takes the shortest path around the table when it is traveling along the edges of the table. That behavior was added to the pattern generator after I wrote a post processor to delete excess edge motion from the pattern files. The authors of Sandify liked the demo video I sent them so much they added that function into Sandify. They have yet to add dual speed operation, so I continue to use my post-processor for that

In case you're curious about the mechanism, here's a video tour of it:

The table has other uses. It's great for entertaining cats, especially when the ball is traveling at high speeds:

I have also used it to paint pictures by having the ball roll paint on a canvas board:

History

Shortly after I finished designing and building a CoreXY 3D printer (UMMD) at the Milwaukee Makerspace, I got the idea to build my first sand table, "The Spice Must Flow", for the 2018 Milwaukee MakerFaire using the same type CoreXY mechanism and 3D printer controller board. That table was larger,  louder, and slower because the mechanism was driven by stepper motors. 

After the MakerFaire I continued to work on the table, with goals of making it run faster and quieter, so that I could have it in my living room. Many of the parts in the mechanism, including motor mounts, the magnet carriage, and pulley blocks were designed by me and 3D printed on UMMD. 

At first I worked on speeding the table up, and managed to get it to work up to 500 mm/sec. That wasn't fast enough, and it was pretty noisy because it used stepper motors, so I switched to servo motors that are capable of running much faster and quieter. With the servomotors the table can run up to 2000 mm/sec.

Once the speed and motor noise issues were solved, I started working on other noises that the table made, hunting them down and eliminating or reducing them one by one. The result is in front of you:  Arrakis 2.0.

Do you want to build one yourself?

If you have kids building a table like this would be a great way to introduce them to mechanical and electrical engineering and programming. My blog contains enough information about the table that if you are skilled mechanically, and know a little about electronics and programming, you can use it to build your own table. I am always happy to answer questions and make suggestions, including sources of inexpensive parts to use.

As built, there's about $400 worth of electronics in the table, including the servomotors, but there are much cheaper ways to go. The easiest way to cut costs is to use stepper motors if you don't mind the table speed being limited to about 100 mm/sec, and you can use an inexpensive Arduino based 3D printer controller. Many people build sand tables by modifying glass top tables from Ikea.

If you would like to build a sand table, or other things, but don't have work space or tools, stop by and tour the Milwaukee Makerspace in Bay View or St. Francis. We have all the space, tools, and expertise you'll ever need for almost any project you can think of.

Do you want me to build one for you?

I can do that, but it's not going to be cheap. That said, I'd love to talk to you about it and maybe we can come to some agreement. 

Links to more info:

Milwaukee Makerspace

Duet3D (3D printer controller boards)

The Spice Must Flow sand table- my first

Arrakis sand table

Arrakis 2.0 sand table

UMMD 3D Printer

Sandify sand table pattern generator

A New 3D Printed Lamp

Several years ago I made a 3D printed lamp for my son using guts from a WIFI controlled RGB LED bulb. The lamp had a large 3D printed fractal vase. He recently reported that the electronics had failed and asked if I'd make him another lamp. Sure!

I designed a new vase using a 3rd order Julia set with swept parameters. About 600 fractal images were generated using ChaosPro. The images were imported into ImageJ where they were stacked and converted to a solid and exported as an STL file. I sliced it using Cura's vase mode (did they ever fix vase mode in Prusa Slicer?) and printed it on UMMD. The top and bottom of the vase are the exact same tri-lobe shape, but rotated 60 degrees. At the middle layer the outline of the vase is almost a perfect circle. The vase has a very interesting surface texture that results from the limited resolution of the math used to generate the fractal shapes.

The new vase is 638 mm tall, printed in about 13 hours using 930 g of Keene Village Plastic's Edge Glow Glass PETG filament with 1 mm walls, and 0.25 mm layer thickness.

The Edge Glow Glass filament is a transparent filament with some "bluing" added to make it look like glass under normal lighting conditions. But if you light it with blue light, the material fluoresces a very cool looking pale blue color.

This is what it looks like in daylight. The bluish color that makes it look like glass comes from some dye that's added to the filament. 

In the original lamp, I took apart a normal shaped LED bulb so I could put the LEDs as close to the bottom of the vase as possible. That required making a very odd shaped heat sink to mount the LEDs on. Since I made that first lamp for my son, LED bulbs have become cheaper, much more widely available, and in many different configurations. For the new lamp, I used a WIFI controlled GX-53 RGBW LED bulb and didn't need to take it apart as it is pretty flat and sits very close to the bottom of the vase. I just bought a socket for the bulb and installed it. 

The only way to get "white" light from the old lamp was to turn on all the R,G, and B LEDs and the color temperature that resulted wasn't very nice. You could tweak it a little by adjusting the relative brightness of the LED colors, but it never produced a pleasing white light. The new bulb is an RGBW type that includes warm and cool white LEDs and the control app allows you to set the color temperature anywhere between a warm yellowish light to a cool bluish white. 

I like to set the bulb to a purple color. That turns on red and blue LEDs. The blue light makes the vase fluoresce the beautiful pale blue color and the red light shines through and looks pink through the glowing blue vase. It's a very nice effect that's not fully captured by the camera. The blue light also makes nearby fluorescent objects glow nicely, too.

The new lamp, set to purple light, making its own vase glow blue and the one next to it glow green. There are no glass pebbles in the bottom of the vase in this photo.

Wiring couldn't be simpler. I drilled a hole in the side of the vase near the bottom, fed in an 18 gauge line cord, split the end of the cord and tied it in a knot so it couldn't pull out of the lamp, stripped the ends of the wires, and soldered them to the wires from the lamp socket, then used shrink tubing to cover the solder joints. Finally, I put some hot-melt glue on the bottom of the lamp socket and mounted it in the bottom of the vase. Done!

The vase is large, tall, and relatively light weight so it can be moved or knocked over pretty easily. I added a few lbs of glass pebbles to the bottom of the vase to help keep it stable. I haven't decided if I'm going to epoxy them in or just leave them loose.

Surface texture results from the low resolution math that generates the fractal shapes that get stacked to make the vase. The print came out with a very shiny surface- like glass.

A look into the lamp- the wiring is as simple as can be. The GX-53 socket is hot-melt glued to the bottom of the vase. Wires are soldered and covered with shrink tubing. There's an on/off switch on the power cord, but it's mostly used to put the light bulb into pairing mode. Once paired, you can program schedules and control colors, brightness, and on/off with the phone app.

The nice thing about this lamp, unlike the original, is that if the bulb ever fails the dead one can be easily replaced - assuming they are still available. GX-53 bulbs are commonly used for under-cabinet lighting, so I think they'll be around for a while.

I added some rubber "feet" to the bottom of the vase to make it less likely to slide if bumped. These feet are some soft silicone material that is sticky on one side but not the other. The stickiness comes from the plastic, not an adhesive, and they don't leave any residue if you peel them off. They also don't seem to damage the finish on furniture. If they get dirty and "unsticky", just wash them off and they're sticky again.

Rubber feet on the bottom of the lamp.

Restoring a Pair of Infinity Beta 50 Speakers

I was scanning Craig's List a while ago and found a listing for a pair of Infinity Beta 50 speakers that looked pretty good. I checked specs and reviews and found they were new around 2004 or so, making them 20 years old. They are floor standing towers that have two 8" bass drivers each, so they are capable of producing a satisfying amount of bass down to 30 Hz. 

Review #1

Review #2

Review #3

I dug up a manual on the speakers, some info on the CMMD (ceramic metal matrix diaphragm) drivers, and tech specs that include the crossover schematic. You can access all of that here.

The boxes are MDF covered in a black wood-grain vinyl that has seen better days- they were a bit dinged up especially at the bottom front edges, and one speaker had some "water"  damage where the joint between the side and bottom was splitting open. The seller hooked them up they speakers sounded OK, so I bought them with the intention of fixing up the cabinets. The seller didn't have the grill covers for them. I paid $125 for the pair.

Area affected by "water"- when MDF gets wet it swells up and then stays that way when it dries out. Whose idea was it to use this crap to make speakers?

An inside look at the water damaged area. It looks like a couple corner braces went missing, too. I replaced them with a couple wood blocks.

Damaged top-front edge. Both speakers were about the same. You can see the seam where the front panel was glued to the rest of the cabinet after the vinyl was applied to both.

Damaged bottom front edge- both speakers were about the same.

One bass driver had a dented dust cap, and a tiny dent near the edge of the cone. Not much can be done about that. Fortunately it doesn't affect performance. Grill covers would have been nice to hide this. Maybe I'll print some...

When I got them home I opened them up to check the interior and found a large handful of dog food inside one of them. I can think of a couple possibilities- either a rodent was stashing food inside the speaker, or maybe a little kid was playing with the dog food and dropped it in via the bass port on the back of the speaker. I didn't find any rodent poop, so I'll go with the latter explanation.

Some of the dog food can be seen here. Note- stuffing on only one side of each cabinet. Hmmm.

Recapping the crossovers

The crossovers had the usual cheapo parts- three nonpolarized electrolytic caps and four iron core inductors. I ordered replacements for the caps from Parts Express for a total of about $27 for both speakers. Considering the condition of the speakers, it's probably not worth upgrading the inductors to air core parts or the caps to film type. C4 and L5 that connect at the tweeter are film and air core parts.

It was nice of Infinity to provide such a helpful schematic! The numbers in the boxes on the right side are the sizes of the spade connectors on the ends of the wires (inches, of course!). They used different sizes so it would be impossible to connect the drivers with the wrong polarity.

One of the crossovers before I recapped it. The bass and mid/treble are set up for biamp/biwiring. There are three electrolytic caps (C1, C2, and C3) on the board.

When the parts arrived I opened up the speakers and got to work. The wires use spade type connectors at the drivers with the other end soldered to the crossover board. You have to remove and disconnect both bass drivers, the midrange driver, and the connector plate in order to get the crossover into position where you can work on it. The wires are color coded and the spade lugs on the + and - terminals of the drivers are different sizes, so it would be very hard to make a mistake when putting things back together.

Recapped crossover. Note the 30 uF cap (C2) is made by wiring two 15 uF caps in parallel and the 18.5 uF cap (C3) is made from a 17 uF and a 1.5 uF cap wired in parallel. The original caps were glued to the PCB, and you can't see it in this picture, but before positioning the new caps on the PCB I put a healthy beads of hot melt glue down at each cap location to minimize vibration that could cause microphonics.

While I had the crossovers out I measured the resistance of the four iron core coils, in case anyone wants to try rebuilding the crossovers with air core coils. Values are as follows:

L1:  0.81 Ohms

L2: 0.45 Ohms

L3: 0.58 Ohms

L4: 0.56 Ohms

If you replace the iron core coils with air core parts, you'll have trouble matching the exact resistances of the original coils. However, if you use coils with lower resistance, I don't think there's going to be any negative effect. I did some digging at Madisound and Parts Express and found most of the parts are cheapest at Parts Express. Among the parts I found, there are film caps that cost $10 and same-value film caps that cost $90. I wouldn't spend $90 for one cap. It can't possibly make that much difference in the performance of the speaker. If I were willing to spend so much that $90 caps wouldn't bother me, I'd probably just build an active crossover and bi- or tri-amp the speakers instead.

Here are the lowest cost air core inductors and film caps I found for the bass and midrange crossovers. The tweeter crossover already has an air core inductor and film cap, so no need to change anything there. Parts listed are for one speaker.

LF crossover, 600 Hz, 3rd order (18 dB/octave):

Jantzen 2.7 mH 18 gauge DCR=0.92 Ohms   $17.05 

Dayton 0.5 mH 18 gauge DCR=0.33 Ohms  $4.40

Solen 68 uF  film cap    $30.98

Total: $52.43

Mid/High crossover 3.3 kHz, 2nd order (12 dB/octave)

Dayton DPMC 30 uF film cap  $7.39

Dayton DPMC 18 uF film cap  $4.59

Jantzen 0.68 mH 15 gauge DCR=0.24 Ohms $17.44

Jantzen 0.75 mH 18 gauge DCR=0.42 Ohms $6.71

Total: $36.13

So the per speaker. the total would be $88.56, not including cost of a new PCB if needed to accommodate these much larger parts. 

You could lower the cost a bit by winding the coils yourself. You'd have to buy the wire and make or buy forms to wind it on, and then have some means to test to make sure you've got the right inductance. There's a calculator here that will tell you the amount of wire required for a specified inductance and form size, and calculates its resistance. Copper magnet wire runs about $20-25 per lb on ebay. For reference, the 2.7 mH inductor uses about 1/2 lb of 18 gauge wire, and more copper when using larger diameter wire to get lower resistance.

If I were going to be restoring speakers and rebuilding/improving crossovers on a regular basis, I'd probably design and build (3D print) a coil winding machine powered by my electric drill and buy wire in bulk on 5-10 lb spools.

Tweeters

When I listened to these speakers before I bought them, all the drivers were producing clean sound. After listening to them at home for a while, I decided that the tweeter output wasn't what I would have expected. I did some research and found out they used ferrofluid, and that it gets gummy over time and affects the performance of the driver. I looked up the part number and emailed Harmon International about the possibility of ordering a pair of them. They emailed back asking for my address and promptly sent me a pair of the tweeters for free! I installed them and listened to the speakers again and now the high frequency output was what I expected it to be.

Cabinet Rework

I pulled all the drivers, crossovers and stuffing out of them, then peeled the vinyl off to assess the damage and make cabinet repairs. I used a razor knife to cut the vinyl into 2-4" wide strips before attempting to peel it off the speaker boxes.  One speaker had some "water" damage, so I chiseled and ground off some of the MDF and used bondo wood filler to replace it, did the same for the damage at the front edges and any other dings, and the seam between the front panel and the rest of the cabinet, then sanded everything smooth.

I attacked the worst one first. This one has the "water" damage at the bottom of the cabinet. It's the one I found dog kibble inside. It's not hard to guess what kind of "water" damaged the speaker.

Close up of the "water" damage. Surprise! MDF, aka cardboard, doesn't like to get wet.

Here it is after removing the worst of the damaged stuff with a chisel and coarse sand paper.

Bondo Wood Filler applied. I filled in the sockets for the grilles because I don't have them and don't plan on making any. I also filled in the seam between the front panel and the rest of the cabinet.

After the first round of sanding. I added a bit more Bondo to fill in the voids and sanded again before  applying the vinyl.

I could have refinished them with more wood-grain vinyl, but there are a lot of other, more interesting options available in the vinyl that's used to wrap cars. I ordered a 5' x 10' roll of 3M 2080 satin flip psychedelic vinyl film from rvinyl for $170. Yes, that's more than I paid for the speakers.

This is the new vinyl.

I wiped the cabinets down with a tack cloth to make sure there was no residue or sanding dust on them and then applied vinyl to the tops and bottoms of the cabinets to get a feel for working with the stuff. It went on easily with minimal effort. I started by peeling the backing from one edge of the vinyl, sticking it to the speaker box, and slowly pulled the backing off the vinyl as I went along. I used a vinyl squeegee to chase out air and apply pressure to the film and followed that with a 2" wide brayer to really press it down. 

The top of one of the speakers after installing the vinyl. I used the flash on the phone to light it up- otherwise it just looks gray.

I left the original black woodgrain vinyl on the backs of the cabinets, but had to wrap the film around the edges at the back and cut the corners at 45 degrees. I used a framing square to get the 45 degree cuts exact.

This is how I got the corner cuts just right.

The back edge of the top of the speaker.

I wrapped a single large piece of vinyl over the sides and front of the speaker. This took a little more effort than the top and bottom, but I had help from my hot girlfriend. Thanks Carol! I started with the speaker box laying on one side, then stuck the film on the other side of the cabinet, then rotated the box onto it's back and covered the front surface, and finally rotated the box again to finish the other side. I used a heat gun in some areas to allow stretching the vinyl over curved parts of the front panel. Once the film was stuck down I used a razor knife to trim the edges and cut the openings for the drivers. I also used the framing square to cut the back corners at 45 degrees as I did on the tops and bottoms of the cabinets.

Here's one of them completely covered with the new vinyl. As you can see, there are many imperfections, but the overall effect is pretty nice. A lot of the rough surface you can see on the front of the speaker was the result of the old vinyl pulling off some of the MDF surface. This is the one that didn't have so much doggy damage.

Finished speakers set up in my bedroom. I'm happy with the results!

More impressive in sunlight:

The speakers weigh about 60 lbs each, so it's very easy to damage the edges if you aren't super careful. I made a CAD model of them because I wanted to print TPU skirts to protect the bottom edge from getting dinged up again by getting bumped by vacuum cleaner, or from moving them around. The skirts will print in two parts for each speaker and go all around the bottom edges of the speakers. I haven't decided if I am going to print and install them...

CAD render of the skirt I designed for the speakers and printed in TPU. It's designed to be stapled to the bottom of the speaker to protect the delicate edges from bumps.

Restoring a 37 Year Old Soundcraftsmen DX4000 Preamp

I recently restored my Soundcraftsmen PM860 amplifier, and was offered a deal on a Soundcraftsmen DX4000 preamp that wasn't quite working. I couldn't find any reviews from the audio press, but plenty were found in some of the audiophile forums, most saying good things. I decided the DX4000 might go well with the PM860, so I bought it.

Front panel of the DX4000, not mine- this one is in a little better shape.

Here's the rear panel of my DX4000. Remember when audio gear had "convenience outlets" on the back?  Those were the days- the days before the marketing people decided audiophiles should buy $1k power cords! Note- no gold plating anywhere!

This preamp has no tone controls, but has three loop I/Os for connecting different signal processors such as equalizers, and two tape loops with switches to dub from one to another. I have never seen a preamp with this much switching before. It also has a headphone amplifier with two 1/4" headphone jacks, one that cuts output to the power amplifier and one that doesn't. One other interesting thing this preamp includes is an output inverter that allows you to connect two stereo power amps in bridged mono mode.

My new preamp had a few problems that were immediately obvious. The top cover had a couple rust spots and bubbling paint. It has a bunch of ganged pushbutton switches that all use the same rectangular button caps, 15 in all. A few of the caps were missing and a few of the remaining ones were cracked. The power-on LED was dead. There was no audio passing through the preamp, except that turning the balance pot made a lot of scratchy noise at the output.

The good news is that I measured the power supply voltages and found the + and -15V regulators working. I also checked all the diodes and found them to be OK. The muting relays were also working properly. There were no burnt parts or exploded caps on the PCB, so I figured worst case I'd need to replace the electrolytic caps (this thing was made in '87) and the opamps (4x RC4136, still readily and cheaply available). There is a MM phono preamp board that has some discrete transistors, but I wasn't too worried about those.

You can access the full size DX4000/4200 schematic diagrams here.

Input switching, one channel shown. See the diagram below to see just how crazy this is.

Tracing signal path from "digital" input on the upper left through output (follow the green line from the upper left to the lower right), the unbuffered input signal passes through 13 sets of switch contacts! That's probably why you don't see this sort of thing done much. The signal passes through 5 switch contacts (blue line) just to get to the tape outputs.

The DX4000 schematic diagram. The phono preamp is at the top, power supply lower left, and line/headphone amp lower right. The DX4000 phono preamp does not include the cartridge matching switches or the op-amp buffer stage.

Power supply schematic with power off. The yellow switches short the power-on LED and the 47 uF cap (red). When power is switched on, the yellow switches open and the LED turns on, and the 47 uF cap (red) charges slowly through the relay coils (about 1k Ohms) and the 2.2M resistor. Once the voltage on that cap gets high enough -it takes about 4 seconds- the transistors (green) switch on, shorting the 2.2M resistor, allowing more current through the relay coils which switches them, connecting the preamp signal to the power amp. This delay prevents turn-on transients from causing the speakers to thump if the power amp is turned on before the preamp. Note: The 17V connections actually sit at 22.8V.

I did some additional testing and found that some audio went through the preamp when I wiggled the input selector buttons on the front panel. I examined the switches closely and found that the solder joints to the PCB were cracked. The single-sided PCB has oversized, unplated holes, and small area pads for the switch pins so you have to really flood the connections with a lot of solder to ensure that it bridges the gaps. Once I resoldered the pins the input switch worked fine. I resoldered all the other switches on the PCB- quite a job, given the number of switches.

Why was the LED dead? Hmmm. The power-on LED is powered via the +/-17V rails (measured +/-22.8V) through the 10k resistor that drops the voltage and limits current through the LED. If the LED has 1.5V across it (typical for red LEDs), the resistor is dropping 44V, which means there should be about 4.4 mA going through the LED. That shouldn't kill the LED. Maybe just an early failure. It happens...

Yikes! Wirewrap connections were used at the I/O jacks and on the PCB. Most of those wires were stuffed under the PCB- I pulled them out so I could inspect the underside of the PCB. Like the PM860, there is no silk-screen layer indicating part numbers or values on the PCB. Soundcraftsmen didn't believe in keeping connections short or using shielded cable! Was wire wrap really cheaper than soldering?

Side note: to me, this preamp looks like the kind of electronics projects I did when I was in high school. My web searches indicate that the DX4000 cost $499 when it was new in 1988. Adcom's GFP565 preamp from around the same era sold for $800 new. I realize that's a significant price difference, but compare the photo above to the photo below. Which looks more serviceable? Which looks less likely to require service? Which looks like it was designed and assembled by professionals? There really is no comparison. This is similar to the difference between the Soundcraftsmen PM860 and Krell KAV-300i amplifiers I recently recapped. Sometimes it is worth the extra money that some items cost, even if the specs are essentially the same, and even if you can't hear a difference between the items being compared.

Adcom GFP565 preamp, sold at the same time as the DX4000. Today you can buy the Soundcraftsmen DX4000 on ebay for $150-470 depending on condition. You can get the Adcom GFP565 for $250-500. I know which I would rather have, just based on the build quality.

Underside of the PCB, not much to see here, except for the dozens of switch contacts that had to be resoldered.

Parts circled in green are electrolytic caps. Opamp ICs are circled in red. The two blue caps near the center are nonpolar electrolytic coupling caps. The two gray things in the upper right corner are muting relays. The 470uF bypass caps for the opamps are located near the power supply on the left, far from the opamps that are all located to the right. Hmmm.

Recapped phono preamp board, electrolytic caps circled in red. I replaced the two orange, 0.39uF, electrolytic input coupling caps with film caps (white caps in blue circles). The two red caps circled in blue are film caps that replaced four electrolytics wired as nonpolar parts.

This is the active circuit schematic for the DX4000. The phono preamp does not have the cartridge matching switches or the phono gain stage shown. Electrolytic caps are marked in yellow. Only one channel is shown. There are some differences between the schematic of the phono preamp section (upper left) and the parts on my PCBs. 

Modification

For some reason, the designer chose to power only the phono preamp board from the regulated +/- 15V rails, and the op-amps from unregulated +/- 17V (schematic designation). These op-amps, like most, are specced at +/-15V operation (data sheet here), with absolute maximum of +/- 18V.  The actual voltage on the "17V" rails is 22.8V. There are 220 Ohm dropping resistors between the op-amp power connections and the 17V rails that will drop that voltage a bit. I measured +/-17.4V at the opamps which (in my opinion) is too close to the 18V spec limit. 

I saw a similar thing in the PM860 amp where the main power supply filter caps were rated for 75V and there was about 72V on the rails. That's not a lot of margin. Let's say the line voltage was a little higher than normal, or there was some momentary surge on the power line. Where are those voltages going to go? What's going to happen to those caps and op-amps? Why on earth would a sensible engineer do this?

I considered connecting the op-amps to the +/- 15V regulated rails, but thought there could be some problem with putting the opamps on the same 15V rails with the phono preamp, so I decided to add a second dual 15V regulator specifically to power the op-amps. 

I installed a 15V regulator module that uses LM317T and LM337T regulator chips with a few external parts to provide regulated +/- 15V from input voltages over +/-18V or so. There will be plenty of headroom to maintain regulation because the module is powered by the +/- 22.8V that is present on the 17V rails. Each op-amp IC uses 6 mA at idle, and there are 4 of them, so 24 mA nominal load for the regulators (which squares with the measured voltage drops across the 220 Ohm resistors). That's more than enough to meet the regulator's minimal output current requirement of 10 mA.

The modification is simple. Take out the two 220 Ohm dropping resistors that sit between the 17V (actually 22.8V) rails and the op-amps and replace them with the new 15V regulators. The regulator ground connects to the preamp ground at the ground wirewrap stake. I used a drop of hot-melt glue to hold the regulator module down on the preamp PCB. 

The two 220 Ohm resistors (red) get removed, and the 15V regulator module replaces them. 

New power supply filter caps (big ones circled in red on the left)- 4x 2200 uF @ 35V, and new +/- 15V regulator module (green circle) to power the opamps. The regulator board is held in place with a drop of hot-melt glue. The regulators simply replaced the 220 Ohm dropping resistors. The white wire from the regulator board is the ground connection for the regulators and connects to the preamp ground wirewrap stake. The regulator chips on the new module don't need heatsinks as they are minimally loaded by the op-amps, even when driving headphones. Note- the original regulators are in place and operational- they supply +/-15V to the phono preamp board.

New electrolytic caps and the +/-15V regulator board that was added to power the op-amps. The regulators replace the 220 Ohm dropping resistors that were used to drop the 22.8V down to 17.4V.

The RC4136 quad op-amps used have reasonably good specs, but the pinout isn't typical of most quad op-amps. Some audiophiles would prefer to use better, lower noise, wider bandwidth parts. You can buy little plug-in adapter boards that allow you to use more modern, higher spec op-amps, but that would add another $100 to the cost of restoring this preamp. You'd be changing the op-amps, but there's still the lack of bypass caps, the funky wiring, and all those switch contacts to go through. I doubt changing the op-amps is going to result in improved sound quality when you're starting from such a marginal design.

What About the Buttons?

I searched page after page of switch cap listings at Digikey and Mouser and could not find a same-size replacement for the missing and cracked switch caps. I decided to 3D print them.

I measured one of the un-cracked buttons and the posts on the switches and came up with this design in about 30 seconds:

Back side of the 3D printable button cap. The cutout in the center fits tightly over the switch post. I printed these using TPU filament so it would flex a bit and grip the switch post tightly. They probably won't work if you print with a hard filament like PLA, ABS, or PETG.

The button is slightly tapered like the originals. The original caps had concave tops but I went with a flat surface for the sake of print quality. I printed test buttons with the fronts and backs on the printer's bed. In the end I went with the front-up prints to get a smooth surface on the visible and touchable part of the button.

One of the printed button caps. I used TPU filament and it grips the post on the switch tightly and will never crack like the original button caps. This and the the other 14 were printed in 0.15 mm layers. I printed a set in green and another in orange to see which I preferred.

This is what it looks like with all the printed button caps installed:

I went with a red LED so it would be clearly visible among the green button caps. I added orange caps to the input selector switches.

If you need to print buttons like this you can DL the fusion360 file here or just grab the STL file here. TPU tends to be hairy and blobby stuff, so plan on spending a few minutes cleaning them up with a wire clipper after printing. 

Does it Work?

After about an hour of testing all the I/O paths, I can report that everything is working fine. There's no noise from either the volume or balance pots. Music sounds clear and undistorted.