The Bud Toaster - (currently: Model 14, version 3)

[Hippie Dickie]

Just some photos to document this new phase ...

eleven of the 33 Bud Toaster cubes, and two stacks of pcbs:

This is version 3, so here are the new printed circuit boards ... solder mask (that's why this iteration is mostly green -- only the device pads are exposed), much wider traces, cleaner layout (10 of the 11 resistors in a line), moved the power plug 1/16" to provide clearance with the heat shield (i should do a rant on how amazingly effective the heat shield is at keeping the cube totally cool to the touch), two buttons (instead of three), nice large pad for attaching the clips that hold the coil. This iteration could be pretty close to right -- except i haven't figured out how to keep the solder mask off the coil clip pads - which have to be manually scraped clean (stupid program) (or, uneducated user -- i need to call tech support):

Here's the populated board, sitting in a cube, where you can see the resistors in a column:

i'm most anxious to complete this unit and see what difference, if any, the 2.5 oz copper makes in the temperature control -- i'm guessing it will be much more erratic until i finally tune the gains in the PID algorithm. i think they're way too large for increased current flow.

 

[Tyg]

Add me to the list of vaporists who would love to reward your ingenuity with a purchase! So much thought and effort went into the creation of your Toaster. Been watching and I'm impressed!

 

[B.]

i really, really, really want to get a CNC router. oh boy, oh boy will that ever be so cool.

i had no what a cnc router is, so i googled it. fascinating. what an amazing tool. thats gonna be soooo fun to play with. i'm jealous.

en.wikipedia.org/wiki/CNC_wood_router

 

[willieR]

Really amazing. I am definitely interested in a smaller unit with very accurate temps. There is a fine line from releasing fun cannabinoids and releasing toxins. The iolite isn't accurate, and as much as I like the MFLB my breathing is hardly accurate for this.

I'll be getting a Cube

 

[willieR]

Sorry to ask idiot questions, but how does the user get the vape out? A small hose or such? Also, how accurately can this be dialed in (how accurate is the temp readout vs actual temp in the crucible)?

Thanks!

 

[Hippie Dickie]

Sorry to ask idiot questions, but how does the user get the vape out? A small hose or such? Also, how accurately can this be dialed in (how accurate is the temp readout vs actual temp in the crucible)?

Thanks!

A glass tube (the "stem"), like the one in this picture, is held against the top of the bud vial and you inhale at the open end:

Larger tubes can be filled and passed to a friend -- a vapor shooter.

Here's the schematic:

The thermometer shows a temperature about 20F higher than the air that is pulled through the bud. Rather than accuracy, i am focused on consistency. Once you set a heater temperature, that is maintained (until changed). It is possible to change in 1F steps.

LOL -- those CNC routers are a bit larger than what i have in mind.

 

[willieR]

Makes perfect sense. Consistency is much more important, and adding / subtracting 20 is simple. Very good. Thanks for taking the time to address this.

As I mentioned there is apparently a fine line between initial vaporisation temps and the slightly higher temps at which toxins start being released. I'm suddenly quite interested in the accuracy of this vaporization level, and you seem to be ahead of that curve, even compared to large plug-in units.

 

[Hippie Dickie]

a morning 2-fer ...

Accuracy of drilling is always a problem -- i.e. the amount of error that can be tolerated needs to be considered.

So when i got these 33 cubes, the first thing i noticed was that the 1/8" holes for the light pipes are 1/16" out of place ... it was pretty obvious, because the hole should be in the middle of the top of the step -- the heat shield is right up against the step, and the light pipe fits exactly between the heat shield and the wood body.

i didn't say anything when the cubes were delivered, and then this morning the solution presented itself ... i just need to add a retaining clip to hold the end of the light pipe in correct position at the other (bottom) end, where it is held exactly over the LEDs:

This makes the drilling tolerance a non-issue, although it will be improved for the subsequent cubes.

Two staples work to prove the concept of using a retaining clip held in place by a #2 sheet metal screw, the head of which is hidden behind the thermometer dial. This also solves the long-standing problem of the light pipes getting loose in their holes and falling out of the cube. i love a solution that solves two problems at once.

 

[DaProfessor]

Nice work going on in this thread. Just providing some positive words from a fellow electronics hobbyist. Lol, unless you happen to be an EE

 

[Hippie Dickie]

EE by degree, but not by work experience -- i'm don't even know what EEs do, except my college suitemate did RF design. i've built a half dozen HeathKits, so i know what it looks like when it's done right -- but that was a while ago -- through-hole technology.

This is fun.

 

[Hippie Dickie]

Status update ...

Reality has given me some slack time, so i've done another re-write pass through the code: removed some special cases, simplified the dutycycle calculation (and deleted several pages of math routines), reduced the startup time to reach vape temperature while still managing the controlled ramp up -- now reaching vape temp in about 20 seconds (instead of 40 seconds), greatly reduced overshoot on startup (to about 15F).

There was a temperature stability problem that was different between these four pre-production units i've been using/testing -- gf's cube is fucking rock solid with about +/- 0.5 F, while my go-to cube was oscillating +/- 5.0 F -- so, i'm like WTF is going on??? So that propelled me into this most recent rewrite.

It seems to me the main difference between gf's cube and my cube is her's is more current limited due to a very thin copper strip on the power line (due to early fabrication "issues"), while mine has high purity bus-bar copper. So, my thinking was the PID gains were just too high.

This rewrite fixed that (i think). Most recent test was showing +/- 1F stability. Latest version has been uploaded to my SkyDrive (see post #1 in this thread for link).

i wanted to get this issue resolved before i even try the version 3 pcb which has so much more copper in the traces.

But now i'm finally satisfied and it's time to get on to fabricating these first 20 production units.

 

[Wolface]

This thing looks really cool! Can't wait to here some reviews from other FC members.

 

[Anthony]

If the price is right, I am going to pick this up and sell my MFLB!

 

[Hippie Dickie]

This thing looks really cool! Can't wait to here some reviews from other FC members.

me, too!

If the price is right, I am going to pick this up and sell my MFLB!

Price for the complete setup with free shipping and a 5 year warranty will be $500 ... current plan is to fabricate up to 1,000 units of this model -- depending on reviewer comments, of course.

 

[ty14]

These look pretty badass but 500 dollars for this unit? I know it has a 5 year warranty and all but 500 dollars seems steep.. Not bitching or anything just curious about why the price is so high?

 

[Hippie Dickie]

Price is due to American labor cost (mine) plus R&D expense and intellectual property license fee.

The way i see it is if the Bud Toaster is any good, Chinese slaves will be turning them out by the millions by 2012 and it will be in Great-Wall-Mart for $29.95. my target audience is the early adopter, and people who want a limited edition, high quality piece of techno-art, made in USA. by me.

And everyone is fully informed about the design and what is in my device: only Borosilicate Glass!

See, i use the Bud Toaster every day -- and, as i've said before, The Cube Rules!

 

[budballer]

hey hippie dickie, again nice work on the cube, looks spectacular. I'm curious though where do you get your pcb's printed for so cheap? and what design format do they accept? thank you

 

[Hippie Dickie]

Thank-you for your compliment.

i use one of those PCB foundary websites (which one doesn't matter) that offers a free download CAD layout software program. It took me a while to figure out the program -- and i'm still not sure how to do a polygon area that doesn't get a solder mask overlay. But when the layout is done, i just hit the "buy" button and the layout is validated and uploaded to the pcb foundary. Several days later i have the boards.

There's pretty much a minimum order requirement of about $200, which buys 10 boards. Then additional boards are $2.50 each. Just a bit more for 2.5 oz copper with solder mask.

As a side note, i've been doing a lot of rewrite on the software, and it may be a bit premature to make this pronouncement, but i think i finally grok the PID algorithm and i'm seeing the 1F stability i've been trying to achieve for 2 years. So, as soon as i get the first production model assembled i'll post a video of the Bud Toaster in action.

Yes, i find it hard to believe, but i started this quest for computer temperature control in February of 2009 after playing with a PWM module i got off eBay for several months.

 

[hereatlast]

Quietly watching the progress here, keep up the good work Hippie Dickie

 

[Hippie Dickie]

Quick update ... here is the clip (that replaces the staples) that holds the light pipes securely in place over the LEDs on the pcb:

And, with a brand spanking new soldering iron (with temperature throttle) i was able to finally get decent power connector copper strips soldered to the Deans plug:

i still have some mill work to do to get the heat shield and the thermometer installed. Maybe this weekend will finally see the first production unit in operation.

i don't know if i mentioned this already -- the fully assembled cube weights in at 3 oz.

 

[willieR]

$500? Whew! I have to scratch the cube from my "I want one" list

 

[Hippie Dickie]

willieR -- i understand that $500 sounds expensive. But the way i see it, i would gladly spend $500 to have a Bud Toaster (even this somewhat flakey prototype) ... i have already invested $5000 into making a Bud Toaster, and 10 years of my time ... so in my mind $500 is a steal.

And, a vaporizer with the salient features of the Bud Toaster is unattaineum from any source, anywhere, for any price.

Software update ...

The third part of the PID algorithm is based on the derivative of the coil temperature (over time). i just realized yesterday that in my latest rewrite of the code i confused coil temperature change with change in temperature error, and the calculation inflated the derivative term. This introduced a slight temperature wobble. The fix is a tremendous improvement - the D term is critical in this application.

i think the most important lesson i've learned while developing this integer implementation of PID is how the physical plant (i.e. the coil wrapped around the glass) is so fuckingly annoyingly non-linear. This is where my naive intuitions about temperature control got smacked upside the head by the 2x4 of reality.

"It" seems to be all about averages. The PIC sees a wildly varying sequence of temperatures and yet tries to maintain an average temperature that is the SETPOINT. As a result, control must be very subtle because slightly too large changes make the plant unstable.

i was amazed to see how the control system could devolve into bang-bang control -- that is, full-on / full-off. The best temperature stability with this system was +/- 10F -- this is what was happening until i introduced the non-zero bias value.

The code adjusts the bias value based on a 5 second running average -- the idea is to compensate for the battery pack losing power.

i put in some stop-limits in case the algorithm couldnt respond adequately -- temporarily kill power if the temperature gets too high above SETPOINT or temporarily boost power if the temperature drops too low. This is okay in theory, but i discovered my limits were way too tight (the temperature readings are REALLY erratic) and the system resorted to bang-bang. Once i broadened the range the algorithm started working beautifully.

just some observations while im thinking about it

 

[Hippie Dickie]

Frustrating weekend, but with perhaps a good outcome ...

i thought i was going to finish assembling production unit #1 and be able to report first vapor yesterday, but the damn unit just wouldn't start up. i finally traced the problem to the PIC -- evidently i fried it while doing some soldering. So i replaced it and at least the user input started working -- both buttons were acknowledged. Also replaced the MAX and the MOSFET. Still no joy. No power up, and a very weird on/off pattern with the green LED -- something that is NOT programmed.

it is insanely stoopid to be having problems like this ... the pcb is just too damn simple to be even capable of causing 12 hours of hassle. At least the only tools i need for hardware debugging are a continuity tester and a volt/ohm meter. Plus writing debugging data into the eePROM of the PIC which i can upload to my tabletPC after a test. i could see the PIC was reading and saving temperature data, and trying to turn on the MOSFET to full power. But, the red LED was not coming on. So i replaced it. Same. Then, on a whim, i resoldered the MOSFET. Now it seems to be working. Open air test results are confusing -- i need to make the heat shield and install the heater in the cube before i can believe what i'm seeing means anything.

These kinds of problems will disappear when the boards are stuffed and soldered by robot -- i hope. Practice makes better -- the next couple of assemblies should be better now that i know what to pay attention to.

i did notice from the eePROM data that the MAX is pretty stable just reading ambient temperature - readings bounced around 4 units, where 1 unit is 0.25C. Same as the datasheet spec.

 

[Scripto23]

Hippie, I just spent the better part of the day reading ALL 36 pages of this thread and want to say you're doing a great job and your vape looks awesome! I have a fair amount of experience in electronics and have actually been building my own vape over the past few months, but anyway...
As for your current problems, have you thought about getting an oscilloscope? For less than $200 you can get a really good one and I feel it would be a worthy investment, especially considering the task you've undertaken.

 

[Hippie Dickie]

Thank-you for the oscilloscope suggestion, i haven't used one in about 30 years but i don't see how one will help - a VOM & a circuit buzzer is really all i need. However a USB connector so i could pull eePROM diagnostic data would speed things up a bit.

Status Update
okay, so, although i got the v3 vape to work, sort of, i decided to set it aside and make another one to see if the intermittent problem was the new pcb layout or the fab. This time i was especially careful. i attached the MOSFET first before any other components so i could apply massive amounts of heat and get the source tab properly soldered to the pcb. this appeared to do the trick and the new pcb / vape is assembled, functional, and working well.

Plus i worked out a few problems that remain. i had a totally inadequate method of fixing the K-type thermocouple probe to the heater coil. Now i use a couple of nested polyimide tubes (700F working temperature) to hold the probe securely against a loop of the coil.

And, i modified the coil clips yet again, to make the coil easier to form and fit.

i'll post pictures when i receive some new parts.

So, what started out as a somewhat bummer week has resolved itself into a very nice conclusion.

The new pcb with thicker and wider copper traces does perform better -- about 3F /sec faster warm up -- now hitting 18F per second during initial warm up.

And, now that i know the new pcb is valid i will try to assemble 5 more tomorrow. Meanwhile, performance and reliability testing is underway on the first production unit!.