Friday, June 13, 2014

CNC'ing SEM parts and 80386 and 80486 photography funding

I thought I'd give the whole crowdfunding thing a try:

If you like the stuff you've seen here on my blog and on, consider helping me recoup some of the costs and keep generating data

In other news...

Over the weekend I spent some time learning CNC stuff better and managed to get my Sherline 2000 CNC mill cranking out SEM parts.  Here's a backplot:

Everything is done in one pass with a 1/8" endmill to leave the part slightly attached on the two sides.  Here is a stock piece:

Milling started:

After a more:

Gasket fit check:

 Comparison with the original:

The original is stainless but thats a bit hard to cut on my system so I did aluminum.  Next I will cut a hole in it and attach a KW25 vacuum port.  This will allow me to attach a vacuum gauge to better understand if the SEM has a healthy vacuum system.

Sunday, June 1, 2014

Rolling with the SEM

Minor updates...

The SEM is off the pallet.  Used a cherry picker with some help from a friend to lift it one side at a time:

Now its on wheels and isn't quite so tall:

The rack on the left will hold the pump and maybe computerized control system.

I also got replacement vacuum hosing:

Its reinforced but still would collapse under vacuum.  Fortunately, I only need it a short length so I can connect stainless vacuum hose

Currently working on learning CNC better so I can make some new vacuum ports for a vacuum gauge.

I also got an EDS detector in the mail:

It needs some TLC but should still be fun, more on that later.

Also picked up some wire bonders at industrial auction:

The base unit seems to work but needs a microscope (check but need to make adapter) and acquire a tip ($30 on eBay) + spool/tensioner.  I don't have the latter which shouldn't be too hard to make but may require a lot of tweaking to get it working properly.

Finally, the microscope camera now has frame lock and has been upgraded to linux-next.  I tried to submit to linux kenrel but don't think it e-mailed out.  In case I get hit by a bus, here is my Linux AmScope MU800 camera driver:  I also white-balanced my camera.  Unfortunately, the driver as written cannot reach the right thresholds.  Instead, I hard coded the RGB values in my local build.  Hopefully will fix these in near future...

Sunday, May 18, 2014

Cleaning the Super IIIA

Just like the toys I had when I was younger, some assembly was required.  Fortunately, everything was pretty obvious where it went and I got it all together.

However, there were two major components not supplied:
  • A method to convert from US power to 100 VAC (Japanese power evidently)
  • Diffusion pump cooling system
  • A roughing pump
 The first was solved by using a Variac to step down wall voltage slightly.  Evidently other Super IIIA users got a giant transformer with theirs.

I rigged up a pump I had laying around to cool the diffusion pump and it seems good to go.  Eventually I might add a peltier cooler but for now I'll just use a big water tank that should stay cool enough.

I have a roughing pump but the plastic oil watchglass was in pretty poor condition, causing the pump to leak oil.  Not only was the gasket torn, but the watchglass was cracked.  I ordered new gaskets and machined a new watchglass:

Although, all the dimensions looked pretty good and it fit snugly, it still leaked!  At this point I just epoxied the watchglass in place and called it a day.   No more leaks!

The Super IIIA has a 1" vacuum hose port and I happened to have a 1" vacuum hose adaptor for my pump.  Unfortunately, I didn't have any 1" vacuum hose to join the two.  There was a small bit of hose that had been cut off from the original pump that was just enough to join the two together:

(not vacuum hose on the left but gives you the idea).  However, its short enough that its fragile and so I ordered some reinforced 1" hose to replace it.

Time to plug it in and see what happens.  Here is the startup panel:

First, I pressed the power button and I think a pilot light came on.  No other signs of life yet.  I then pressed the RP button and some fans whirred on (probably for pressure measurement).  As my roughing pump takes 120V I did not wire it to the 100V RP output and so manually switched it on.

I let the pump run for a bit to evacuate the piping leading up to the chamber.  Although I didn't have a pressure gauge on, the pump eventually got very quiet with no burping to indicate any substantial vacuum leaks.  Then, I pressed the button to evacuate the chamber.  Pump got noisy again as it started to bring down the chamber.  But stayed noisy: it leaked bad :(

Next disassembled easily accessible areas so that I could clean it and grease gaskets.  One interesting bit that it exposed:

Part of the beam steering optics right below the Wehnelt.  The Wehnelt assembly is also interesting (and filthy!):

Everything around the filament is quite dirty.  Lapointe recommended Wenol which seems to have worked well enough.  After cleaning with Wenol, components were sonicated in water and then rinsed with IPA.  I don't think I'm at high enough vacuum where water outgasing will be a problem but I can bake them later worst case.

I imagine these old SEMs are generally not well maintained and without greasing the gaskets, they dry out and develop bad vacuum leaks.  This leads to short filament life and subsequently the filament sputtering all other.  And burned out for that matter:

Hmm...what to do?  Fortunately, there are a number of vendors that sell replacement filaments.  I did some research and settled on ME Taylor, the same people that Lapointe bought his filaments from:

They seem fairly simple and I'm hoping I can rebuild them in the future instead of ordering more.  Depending on how easily I fix the vacuum leaks, the 10 could last me some time or burn through real quick.  So far this is by far the largest purchase I've had to make since I already had the Variac and roughing pump.

After cleaning and greasing, it doesn't leak badly but I still hear some burping,  I'm unclear if this is a safe level to use the DP or, more likely, that the DP won't be able to sufficiently lower the vacuum.  Next on the agenda is to machine a standard vacuum adapter (probably NW25) for one of the unused chamber ports.  This will allow me to attach vacuum gauges to give me a better idea if it actually is leaking badly / in a spot I care.  There is a Pirani gauge built into the SEM but its not very accessible.

It came with the schematics which I was able to scan.  Also, thanks to lapointe2 I also have the user manual.  Both are in fairly crude, disorganized forms as of this writing, but they can be found here along with other info I collect.

Saturday, April 19, 2014

A new toy arrives: introducing the Super IIIA

What weighs 865 lbs, is probably older than me, and has a radiation trefoil on it?  Its our friend the ISI Super IIIA, at last sitting in my garage:

This post is about my initial impressions unpacking it.  In general, it is slightly larger than the noisebridge system (ISI TV Mini-SEM) and has several enhancements.

Here is the console:

The left side mostly concerns with zoom and display.  The right side has some focus controls and pump control.  There are several empty bays on the far left presumably intended for expansion modules / customization.  It came with a cover plate that sort of fits there but with no real way to secure it down.  The upper left CRT is for normal viewing while the right is intended for photography.

It also has a cute little warning sign on top:

Physics is not my strong point but I think it goes something like this: the electron beam strikes the sample and excites the atoms it hits, ionizing some of them . Eventually they drop in energy and emit an x-ray, similar to how lasers work.  However, there is a fair amount of shielding and the column is probably only 10kV.

Moving on to the vacuum system, here is the inlet for the roughing pump:

Above I'm checking connection with a 1" tube.  I plan on connecting a KF25 half nipple so that I can adapt it to my Varian DS-42 rotary vane pump.

This port connects to a vacuum distribution network that includes a diffusion pump:

Presumably the pump has oil but haven't checked.

The diffusion pump external cooling lines had been pinched under the SEM during shipping.  I was able to use a car hydraulic lift to prop up one side of the SEM and un-pinch it.  I will need to add a circulating pump to cool it.  There is also an external filtering assembly that I may or may not use.  The noisebridge system by contrast had a smaller air cooled diffusion pump.

The roughing pump, diffusion pump, and chamber are tied together via an electronically controlled valve:

The buttons should allow the chamber to switching between roughing pump, diffusion pump, and vent.  The noisebridge system was manually controlled.

But what is the vacuum system without the electron beamline?  The beam originates here:

I haven't checked to see if the filament is intact.

And enters various electromagnetic optics that allow it to be properly magnified and focused here:

Both these pieces have unusually high wear.  lapointe2 had a note about filaments burning out.  Talked to a few people and the running theory is that it has a vacuum leak that is causing excessive filament wear, maybe same issue in both.  Hopefully I can grease up the system, monitor vacuum levels, and avoid such problems.

Eventually it snakes its way into the chamber.

Hard to see in above picture, but there are three instrument ports with only one used.  Presumably this allows adding EDS or other accessories.  I plan to use one or two to install better pressure monitoring.  Close up of the interesting bits:

On top the beam exits from the beam steering optics to the stage below (not shown, more on it later).  To the left we can see the detector:

The grid is used to bias the sensor to help with sensitivity or up on this before and need to refresh my memory.  Anyway, the white bit behind it should be the scintillator that emits a small bit of light when the sample emits an electron that strikes it.  This is then passed to the photomultiplier tube:

The signal then feeds into the various electrical boxes that convert it into vertical, horizontal, and intensity signals to feed the CRTs:

Going back to the chamber, here is the sample stage as seen from the outside:

It has mechanical digital read out which is kind of cool.  Unfortunately, its also pretty coarse so I'm concerned it won't be easy to position things at high magnification.

And the inside:

It has numerous electrical feedthroughs.  I haven't come up with a use for them yet but it does leave a lot of possibilities open.

Finally, I ordered some sample mounting supplies from Ted Pella (Ted Pella #16216: Aluminum Specimen Mount, OD32 x 10 mm):

The specimen mount on the stage image previously shown has a smaller one placed on top of it that came with the SEM.  It did not fit and so was packaged with aluminum foil:

As well as conductive tape to hold specimens in place (Ted Pella #16084-6: PELCO Tabs, 6 mm OD):

The specimen holders are quite large...should be able to hold several specimens per holder.

Next steps forward are to:
  • Get it off the pallet so it can be moved around
  • Power on the console
  • Setup diffusion pump cooling
  • Attempt to pump down the chamber
  • Image something!
One oddity is that it requires single phase 100VAC.  I should be able to supply this with a variac without problem.

Longer term goals:
  • Turbo pump?  Depends how well the diffusion pumping works
  • Computer imaging
  • Computerized stage control
But those only matter if I can get it working at all.  Fortunately, someone else rebuilt an identical unit which should also help.

Tuesday, April 8, 2014

Not quite dead yet

Almost reached the one year mark without posting, but haven't been idle.  Since my last post I've:
  • Better at lapping
  • Purchased a Scanning Electron Microscope (SEM)
  • Got a drivers license!
    • Gave me access to industrial auctions and ability to move to bigger place with garage
  • Travel to Russia to play with old ICBMs (one of which has since been launched)
    • (wanted to post a picture with me hugging it and the caption "you can't hug with nuclear arms...but you can hug nuclear arms!")
  • Joined Cupertinker to encourage me to work on a project for Maker's Faire
  • Transferred from the software to the electrical engineering team at work so that I can learn computer engineering / FPGAs
  • And much more...
I'll go through backlog and see what I can post updates on.  FWIW, I have continued to update Silicon Pr0n's wiki and archive areas.

Since my SEM shipped today, seems like a good place to start.  I've been looking for a SEM for a while as I find an increasing number of chips that I need sub-optical imaging on.  I even drove up to northern california for a weekend to check out a Cambridge Instruments Stereoscan 150u.  This is a pretty typical size for these beasts:

This was from Craigslist, but I was also looking at eBay and industrial auctions.
Many were available, but had onerous power requirements (three phase), were too large, or cost too much.   None of these were necessarily a problem but were going to be a big enough hassle that I didn't want to deal with it.

Noisebridge had a ISI TV Mini-SEM that didn't seem to work (although did at one time) and I enquired into getting it with the intention of fixing it up and using it.  Overview:

Ultimately someone else snagged it and its out of the picture.

Eventually I found an ISI Super IIIA on eBay essentially for its scrap price:

Made by the same manufacturer, it looks similar to the noisebridge unit.  It took about 4 months to pack it up and work out shipping but today it was loaded onto a truck and should arrive in a week!

Its the same model SEM that someone else rebuilt.  This model was chosen because it was relatively low cost (about $1200: $312 purchase price + about $900 S/H) and reasonably compact.  It also runs on single phase power (unclear if 110 or 120).  To put its size in perspective, the first SEM's middle piece is larger than the entire setup above.  Granted, it was probably more powerful but this is good enough for now.  I'll post more pictures once it shows up and I start rebuilding it.

If you are in the SF Bay area and are interested in helping fix it up or using it, gimme a ring!

Sunday, April 21, 2013

Heatgun H2SO4 decap

I recently had to get through a number of samples which I wasn't that worried about damaging.  This gave me a chance to try something a little different.

H2SO4 has a few advantages over HNO3 for decapping.  First, I can't as easily get ahold of fuming HNO3 and HNO3 tends to cause a lot of corrosion.  However, H2SO4 works at much higher of a temperature so if nothing else takes longer to decap a sample from the time it takes to heat up the acid on a hot plate.

A heatgun on the other hand can rapidly heat something up by moving hot air over it.  I decided to try decapping a few chips by first cutting down down with a Dremel and then superheating them in test tubes.  You can see a video here:

Be sure to read the comments in the video description.  Basically took about a minute to heat up and then another two minutes to cook off the bulk epoxy.  Total decap time of maybe about 10 minutes a chip including Dremel time and cleaning.

However, this has numerous disadvantages.  First, a few of the dies got stuck in the test tube (presumably related to thermal expansion) and broke corners off when I managed to shake them out of place.  A lot of the dies also had rounded corners but I think that was more because I picked the dies up with my fingers (usually I use tweezers) which tends to break off the tips of the corners.

Dies also tended to come out pretty dirty.  HNO3 is more "water based" so tends to clean off easily.  H2SO4 is thick to begin with and dehydrates the epoxy into tar.  Even after ultrasound (acetone and water) I got some chips coming out looking like this:

The bigger problem though is that some small bits of epoxy didn't dissolve off.  When dies come out of HNO3 or longer soaks in lower temperature H2SO4 they tend to be pretty clean.  However, I think that the epoxy mixtures had denser parts that didn't dissolve as well.  When the chips were run through ultrasound they developed areas like this:

Close up:

Notice how the metal isn't damaged, only the overglass.  This suggests that it happened after it was in the acid or otherwise the metal would have been corroded.  Additionally, the pattern suggests something pulled off.

Chips had corrosion at various points on the die, one of them was quite bad.  I suspect two leading causes.  First, although epoxy dissolves faster at the higher temperature I think that the metal was dissolving more faster than the epoxy.  Second, these were older dies with small holes in the overglass.  Newer dies tend to have higher quality passivisation that may have been less of a big deal as generally they would only eat around the bond pads and not seep in very far.

I didn't take the temperature but I imagine was in the 300+ C range since H2SO4 was boiling off.  I've decapped large BGA chips with 300C H2SO4 before (measured with thermometer) but generally stay away with it because the large amount of fumes and general safety hazard were it to somehow spill.  However,since the fumes could condense as they moved up the test tube (which was still relatively cool) not that much fumes really came out.  Also if the tube somehow broke the heatgun might disperse it everywhere.  I had a teflon sheet behind the tube as a backshield for general control in case that somehow happened.

I tried runs with varying levels of acid.  I leaned towards larger amounts than in the video that kept the solution very fluid at all times.  With the volume in the video the solution became thick enough that it bubbled up and had to be monitored closely so as to not overflow the test tube.

In summary, has the following advantages:
  • Easy to do / readily accessible.  All materials can be found at local hardware sore
  • Quick
  • Higher temperature allows using less acid
But some significant disadvantages as well:
  • Very rough on die including both mechanical and chemical stress
  • Several safety concerns: gives off H2SO4 fumes, hot concentrated H2SO4 will vaporize skin instantly
  • Like other H2SO4 decap, dark solution makes monitoring difficul
  • Eats metal quickly, especially problematic if passivation is not sealed well
  • Dies did not come out very clean.  Could be processed longer but will cause more damage
I'm less likely to do this for older PDIP chips but QFNs or other modern chips that are thin, have higher quality overglass, and more even epoxy might work well.  However, I wouldn't advise this for valuable chips.

Wednesday, March 20, 2013

Automatic lapping: a first cut

I played a bit before with using a lapping machine by "hand".  Here's a picture of the setup:

Basically I press a die against the lapping wheel with my finger and occasionally refill the colloidal silica reservoir (white stuff with stopcock) as it runs low.  See tutorial I wrote a bit back for details.

However, this method tends to give an uneven finish that's okay for looking at a specific part of a chip but not good/reliable enough for full chip images.  This is pretty typical (RSA SecurID):

See the rings?  That's optical interference showing more polishing towards the edges than in the center.  There are some things that can be done to improve this (ex: use HF to eat down more initially, change speed/pressure) but ultimately it lacks repeatability.  For example, depending how tired I am my interpretation of  pressing "soft" and "hard" might change.

This post is about my experiences trying to start with low cost lapping approaches, the problems I encountered, and how those problems were solved as my system and techniques developed.

Nowadays most chips use lapping machines to very precisely polish chip layers during manufacturing.  These lapping machines are quite sizable in order to work on full sized wafers.  This tends to make them both more expensive and larger than I can afford.

But what tabletop lapping machines?  Vendors such as Logitech make very high quality compact units.  Some other brands include Ultratech, South Bay Technology, and MTI Corporation.  Most of the fixtures/jigs that go with them are quite expensive, typically starting at $1000 and can get quite expensive (there is an eBay seller asking $17,750 for a new old stock Logitech fixture).  MTI seems to be the price-performance leader and sells a small relatively low priced jig that I was able to get my hands on, the EQ-PF-3H1W2:

Its really designed for getting an arbitrary flat surface for metallurgical applications, but I figured that I could try to coax it into doing parallel polishing.  In any case, it has three main components:
  • Sample holder (upper right): keeps samples at right angle to lapping plate and guides weights against samples
  • Conditioning ring (bottom): helps smooth out lapping solution and filter out debris
  • Weights (upper left): to control pressure against sample
First problem: how to hold the sample.  It didn't come with any instructions but looking at it and their website products it seems you are supposed to put a cylinder in at least one of the holders and tighten the setscrew until it can barely move.  Weights can then be placed on top to vary pressure as needed.  Its designed for 1" samples but since the teardrop causes the cylinder to wedge against the side I imagine the exact diameter isn't critical

MTI's website seems to suggest two options for holding samples:
  • Press sample with powder into a cylinder
  • Pour epoxy into (silicone?) holders with sample
These are well suited for metallurgical applications as both of these can accept arbitrary shaped samples.  However, the first option is out as dies are fragile and might not survive pressing.  The second option is plausible but I'm a little put off by the setting time.  A third option is to affix the sample to a metal cylinder as done in parallel polishing fixtures.  This turns out to work better but we'll explore other options first since I didn't have suitable materials on hand.

The very first thing I tried was to cast the samples in PLA.  I couldn't get it fluid enough to do any good.  I didn't try hot glue but it might have done better.  Looking for something less viscous, wax is cheap and quite fluid when hot.  The plan is to omit the conditioning ring (raises the jig slightly) and lap directly against the plate.  No control over pressure and such but under "try the simple solution first".  Getting ready:

But what to pour it into?  I didn't have anything 1" diameter and was still concerned about it rocking about.  So...why not pour directly into the lapping jig?  Worst case I melt the wax out.  Bottom surface is a solid Teflon sheet for easy release.  Lets go!  Ready to pour:

And poured:

Hmm...its wandered off center a bit.  Not necessarily a problem but also wax crept under it and is now blocking the chip surface.  Lets run it on the machine anyway:

A few things to note.  First, I don't have a "slurry pump", the device that recirculates solution that spins off the wheel.  Instead I have a collection bottle (not shown) and occasionally swap it with another bottle that I use to feed the reservoir/stopcock above the machine.  Second, I don't have a real "York support", the piece that supports the jig as it rotates on the machine.  There is a small bracket on the side of the machine that was probably intended for the slurry feed (or looks so from the complete unit pictures I see online at least) that I'm supporting the jig with.  The point is its not intended to take a load like I'm doing.  Still, it held up decently well.  I did have to weld a nut onto the threaded rod in order to get things really tight.  Chemistry clamps hold everything else together.  I originally had the dripper on the same support as the jig but the jig got moved around too much and so ended up just suspending it from a ring stand off to the side.

Additionally, the cast iron ring clamp I made the support out of wasn't quite big enough.  Cast iron is easy to break so I figured I'd just snap it and weld it at an angle.  I'm not very experienced with welding, but it seemed that heating it up really weakened it.  Eventually I just ran a bead over it on both sides with a MIG welder to completely reinforce it with steel.

Finally, as the wax cylinders don't freely move up and down I can't use the conditioning ring.  This would have helped to even out the lapping solution and remove particulate.

I don't recall the first round of wax lapping really going anywhere...some combination of the wax being slippery and slightly recessed meant it didn't wear away very quickly.  Not easily defeated, it seemed time to add a weight to try to keep it more level and closer to the surface.  Some reading I later did suggested that putting acetone down may have been an effective way to fend off wax.  Back to the original idea, ready to pour:

Poured and cooled:

That's a nasty bit of thermal expansion/contraction.  As the way cooled it really receded from the walls.  Hopefully the bottom looks better:

Nope!  It still drifted off to the side and looks about the same.

Done playing with wax, I had some West Systems epoxy lying around leftover from some fiberglass work:

I found the hardener in the trash at school.  Unfortunately it is not off the shelf type but I was able to e-mail the company for the MSDS and cross-reference it to get a rough idea of mixture proportion and curing time.  Anyway, IIRC it takes 24 hours or so to cure so its not nearly as expedient as pouring wax.  Some people sell 1-2 hour epoxy that would suite it much better if I got more serious about this method (ex: Tedpella 2 hour epoxy).  So I mixed some hardener and epoxy together and...uh oh.  It put off more heat than I was expecting and setup solid in less than a minute.  Some beaker cleaning later, I this time measured the actual volume needed:

And more quickly poured it out:

Also owing to the large stainless "heatsink" it didn't setup nearly so quickly for better or worse.  The next morning I checked it out:

Although better than the wax casting, it still had a thin layer of epoxy over it.  I may be able to do things to improve this but my feeling says that epoxy casts are better suited for odd shaped objects where the exact polished surface isn't as important.  Also the wells were not setting up at the same rate, likely the hardener viscosity causing it to not get distributed as evenly as I was hoping.  Although I decided not to lap it, I had put a thin wax coating on the side walls and was able to push them out without too much trouble.

Next idea: can I cast the dies into lead fixtures?  Surface tension should prevent it from going under and the pressure should help keep it level.  I just happened to have access to some casting equipment:

Starting with a simple cast:

Gave this:

The die fell out!  Decided to try to just glue it in place and press it against a flat surface:

And it was ready to hit the lapping machine.  Note there is a fundamental difference between this setup and the wax castings: the lead casting/weight is free to slide (when properly tensioned) in the fixture and apply pressure to the die.  The other castings grabbed onto the side of the fixture and did not slide up and down.  This also allows me to use the conditioning ring.

But this ran into problems quickly.  With only one weight in there the jig swings inward to reduce friction (outer diameter spins faster => more force).  I cast some matching forms and got it to spin reasonably easily, short video here.  I can't find any pictures of the results, it must have been either not level enough or too far below the surface to do anything at all (I'd lean towards the latter).

Next I added a nail on top to try to weight it down a little better.  I used a small glob of epoxy to try to really keep it in place.  Ready to pour:

During actual pouring there was a weight on the nail to really hold it in place.  Came out reasonably good, much better than previous casts at least:

However, the machine now spun somewhat unevenly as, the nail aside, the casting was slightly different weight causing a pressure difference.  I was able to smooth it out but after a bit of fiddling with things suddenly sparks and magic smoke violently shot out the machine!  When I took it apart I discovered the motor controller wasn't too healthy:

Pretty toasty but fortunately KB controls still sold them and I was able to get a replacement.  I went with the larger amperage model to try to make it more durable but this might not have been a good idea as the minimum speed is now higher.

Unfortunately, the root cause (other than that the machine didn't have any fuse protection...) was either due to temporary stalling or possibly due to damage in the motor.  Getting the motor open is non-trivial though as there's an arbor pressed onto it.  I didn't have a proper gear-puller for it but I had something close and was able to jurry-rig some bolts to get the job done:

Once apart I could see that the motor had commutator arching:

I took some resistance readings and things didn't seem too bad, maybe it just needed to be cleaned up.  One coil was open however:

If this had been caused by just one stall it could potentially be soldered and the motor salvaged.  Although I probably should have taken more resistance readings at this point, I had just gotten a benchtop lathe and wanted an excuse to use it.  Turned the commutator down:

And then took some more detailed resistance readings but they didn't look good.  Further inspection showed that there were more overheated windings:

Enough bad things at this point that I realized the motor needed to be rewound or scrapped.  As this was very labor intensive, I instead went with getting a replacement motor despite that it cost as much as I paid for the machine.

I made two other improvements.  First, I installed a 2 amp fuse (1/4 horsepower motor => about 2 amps @ 115 VAC).  Second, installed a better support bolted directly to the bottom plate:

Ready to roll again.  Ultimately not a complete loss as I now have a better mast and the motor runs smoother.  Also while I was at it I replaced the polymeric pad and backing plate.

Getting back to work, I was done with the lead castings and wanted to try out using the stainless.  First problem: they aren't cut terribly bad but aren't straight:

See those slight curves on the end?  I think its from their saw slipping around until it bit into the rod.  From some combination of this and the 2 inches that it stuck out, using the miniature 3 jaw chuck on my desktop lathe was problematic as it only had a very small flat area to grip.  Fortunately, I found that the 4 jaw chuck could hold it pretty good at the expense that I had to align the work:

I had to make very shallow cuts (0.001" or so, carbide bit) but each pass was reasonably quick and it worked out in the end:

But we aren't done machining yet.  The previous setup wasn't automatic as I had to keep filling the CMP solution slurry tank manually.  To achieve automatic polishing the system needs to be able to run unattended for longer periods of time.  Professional systems typically use peristaltic pumps.  I sort of had a peristaltic pump but there were two problems:
  • The motor was completely burned out but I could turn it by attaching another motor.  This is not a blocker but is very inconvenient to use
  • I don't have usable tubing
This is my pump head, Cole-Parmer Masterflex 7015:

IIRC I found it in the trash at school along with the burned out motor.  Some research showed that L/S 15 tubing would work and I ordered 50' of new old stock Neoprene tubing:

Surplus off of eBay from Hershey's Chocolate USA!  Anyway, to solve the pump motor problem, I did have another peristaltic pump motor with the same bolt pattern but the shaft length was very different:

Notice how the slot is visible on the right unit but recessed on the left?  Fortunately this was reasonably easy to cut down on a milling machine (sawed off unneeded shaft laying on top to show original size):

I had to go very slowly as there isn't much shaft to hold on.  Integrating everything together:

Liquid pumps from a tank on the bottom through the peristaltic pump and drips out from a clamp held by the mast attached to the machine's base plate.  Liquid drains into a hose coming off of the machine and recirculates.  At some point I'll add a filter but for now I just rely on tank settling to remove bulk particulate.

Getting the fixture ready:

Most lapping guides recommend CA glue or wax for easy sample removal.  I used epoxy as I didn't have any CA glue on hand and didn't want to worry about wax coming off at least during first tests.  Learning from earlier on how important it is to balance the fixture, all three wells had an identical sample.  Put on very small globs of epoxy and then let set under the fixture weight:

I quickly realized that using only one setscrew per well on a cylindrical sample still leaves a lot of play.  The other things I tried didn't have this problem since they were cast into the full teardrop shape.  Concerning but best to just try it out and see what happens.  Lapping:

Here's a short video of it running.  I ran it at reasonably high speed (maybe 180 RPM?) in part because I was short on time.  This was about the limit to stop slurry from splashing off.  This was aggravated by the new motor which is slightly higher, enough to make the splash guard much less effective.

Once the machine stopped I was at a new problem: the samples on the jigs are too tall to fit under my microscope.  Solution: raise the microscope:

Here's the first sample:

Definitely a step in the right direction but not very level.  The polish is also much smoother than anything I got with the finger method before albeit not as level.  Not sure how obvious it is in the above picture, but this chip has very nice standard cells.  Anyway, the other two were much worse:

The left side of the die is unpolished and the right side is gone with a very sharp transition in the center.  The third chip is very similar to the second.

So what are the steps for moving forward?  If this jig is going to be useful for parallel polishing I need to find a way to keep it more level.  However, I also just acquired higher end equipment so future work will most likely focus on that instead.