New Brewery Build

[newguy]

Followup to this topic.

Got the PCB all soldered today:





All that's left to do is to mask the headers and spray it with conformal coating and it's done. The large silver thing on the back of the PCB is a micro SD card connector as I decided that I want to incorporate a datalogging functionality so I can go back and examine total mashing times, rate of temperature increases, etc etc etc anytime I want.

I'm about 80% done the touch screen layout and once that's done I can concentrate on cutting the code that will run the thing. The large chip, U6, in roughly the middle of the PCB is the "computer" (microcontroller really) that will do everything.

I should also soon have pics of the kettle, mash tun, etc as the fittings will be welded onto them soon (fingers crossed). Right now they're not that interesting - just really big SS pots.

 

[hockadays]

Looks good newguy what size are the pots?

 

[newguy]

Looks good newguy what size are the pots?

2 x 78l (82qt) and 1 x 59l (62qt). The big ones are going to be the mash tun and kettle while the little one is going to be the lauter tun/hot liquor tank. I also have three small "asparagus steamer" pots (tall & slender pots). One is going to become my grant, while two are going to become one very tall HE tank. The heating elements I have are about 30cm long and one pot alone isn't deep enough to fully immerse the element. I realise that I could have opted for a much bigger HE, but I really want a small pot that also has a small thermal mass once it's filled with water.

 

[Zizzle]

Cool.

I never did have much luck with PICs way back when. AVRs instead.

Is the board design your own?

 

[newguy]

Is the board design your own?

Yup. Tedious but I find it fun. Tedious because I refuse to use an autorouter (hate the things).

 

[jwsparkes]

That's some pretty imressive electronics. My limits are single sided, thru mount components. I'm maxed out trying to program PICAXE micro's.

Good job.

 

[newguy]

That's some pretty imressive electronics. My limits are single sided, thru mount components. I'm maxed out trying to program PICAXE micro's.

Thanks...I do it for a living so this isn't such a stretch for me. However, the right tools make all the difference in the world. I use probably the cheapest PIC C compiler available for my own personal projects, but it's also the most user friendly. If you want to start playing with PICs directly, I can't recommend the CCS compiler highly enough (www.ccsinfo.com). At work I have to user a couple of really ugly compilers - and they're a real pain to use.....and one of them is expensive as hell.

 

[Zizzle]

I'm surprised you need that much of a circuit. Got a schematic or diagram up anywhere? Surely this build is worth a blog?

And I only bother with GCC. The proprietary compilers aren't usually worth the hassle. Eventually you hit a bug and having the source code to the compiler is invaluable.

 

[newguy]

Since you asked, here are the schematics. Not shown are the off-board components such as the SSRs and ordinary relays. Also not shown are the sensors. I will use LM35DH temperature sensors and Measurement Specialties Inc FC2231-0000-0050-L load cells (mass sensors).

I haven't ever blogged yet so these posts will have to suffice.

View attachment Schematic_1_of_3.pdf
View attachment Schematic_2_of_3.pdf
View attachment Schematic_3_of_3.pdf

 

[Zizzle]

It's always interesting to see another design.

Just curious about some of your choices... I know it's often easier to use what you know.

But did you consider going with a beefier MCU to avoid the separate ADC, EEPROM and GPIO chips?

Did you consider using DS1820 or similar digital temperature sensors to avoid the analog signal stages?

Where do the load cells go?

Any plans for automated hop additions?

Have you done the board bring-up? Does it all work?

What is the total IO count for doing the actual brew? The brewbot mk2 ended up with 17 digital I/O and 2 analog I/O (not including ethernet, flash chip, USB, LCD, and rs232).

 

[newguy]

Just curious about some of your choices... I know it's often easier to use what you know.

But did you consider going with a beefier MCU to avoid the separate ADC, EEPROM and GPIO chips?

The MCU wasn't actually the primary reason for the design. The primary reason was accuracy in converting the analog signals. I have analog temperature signals from the LM35's @ 10mV/C which means an absolute max of about 750mV @ 75C. In order to maximize the delta in readings, that requires an A/D voltage reference as close to the 750mV max reading. However, PICs are peculiar in that references <3V means that the accuracy of the built-in A/D suffers - a lot. That's why I chose a separate A/D chip with a 1.024V reference for the temp sensors. The mass sensors give an analog output that varies from 0.5V (zero reading) to 4.5V (full scale), so that required, for maximum accuracy, a separate A/D converter with 5V reference. For increased accuracy, I chose a high precision 5V reference in conjunction with the PIC's built-in A/D converter.

The GPIO chip wasn't strictly necessary but I happened to have the '2680 PIC on hand and it's only a 28 pin processor. It also has built-in EEPROM but I make it a point to never use a processor's built-in EEPROM because they wear out much more quickly than a separate memory chip. I would have preferred to use a Ramtron FRAM chip (which never wears out), but I couldn't find stock anywhere when I ordered parts.

Did you consider using DS1820 or similar digital temperature sensors to avoid the analog signal stages?

No - they're not available in metal can packages (thermal conductivity/response speed issue), and, last time I used one, their absolute accuracy and linearity wasn't as good as the LM35.

Where do the load cells go?

I am going to use them to measure the volume of water in my lauter tun, which eliminates a pet peeve of mine. I fill my lauter tun using a measuring cup, 1 litre at a time, at present. I hate this (naturally), so I want a system that I can dial in a target volume and the vessel fills to that level automatically. I need 3 to define a plane so that I don't have to worry about balancing the lauter tun on one or two.

Any plans for automated hop additions?

Nope. I was planning to make it much more 'plug & play' but the cost/complexity of hard piping and the associated valves was prohibitive (to me anyway).

Have you done the board bring-up? Does it all work?

Not completely sure what you mean? I have loaded a simple test program into the PIC and everything works, if that's what you mean.

What is the total IO count for doing the actual brew? The brewbot mk2 ended up with 17 digital I/O and 2 analog I/O (not including ethernet, flash chip, USB, LCD, and rs232).

I have 4 relay outputs (2 pumps, 2 water inlet valves), 5 SSR outputs for the 5 heating elements, 4 float switch inputs, 3 temp sensor inputs, 3 mass sensor inputs, .....and that's it. There's an additional 2 lines (serial in/out) for the touch screen. Since the SD card interface is SPI and so are the GPIO, EEPROM, and 'sidesaddle" A/D, it only requires on additional line (the /CS), so it was good value for IO pin count.

What I haven't mentioned is that I have drivers written for everything already - I haven't used anything on this design that I haven't used elsewhere previously, so the 'code burden' is very low. I can cut & paste from projects I've done previously, and that will save time.

 

[Zizzle]

I was curious since I seemed to have more I/O but a much simpler circuit in the brewbot.

We seem to be designed for slightly different problems.

Design philosophy may have something to with my stronger SW background, and I'm guessing you having a stronger HW background?

I guess there is no "wrong" way here, but my rationalisation goes along these lines:

a) hardware bugs are much harder/expensive to correct than SW, so I tried to keep the HW as simple as possible

B) DS1820 absolute accuracy is not fantastic, but as long as they are reasonably consistent run to run, once the system is dialed in, absolute accuracy isn't vital. We can correct in SW to a certain extent by calibrating the sensor. The complexity they avoid is important. What sort of accuracy were you aiming for?

c) the large thermal mass of the water means that temperatures change relatively slowly. e.g. seconds per degree. The conductivity of the plastic DS1820 case isn't the limiting factor. I actually see to have more trouble with stratification of the water in the vessel. Interestingly when I was using an aluminium brew pot/HLT I didn't have this problem. SS just doesn't conduct heat that well.

I'll be interested to see how your load cells work out.

For filling of fresh water, in the brewbot I use 2 conductive probes. I measure the time it takes to fill between those two probes and can then extrapolate to the amount of time I need to fill for to get the desired volume. Not super accurate, but probably better than us hand measuring water.

I would probably go with one of these if I were doing it again. Actually I have one here and might put it inline anyway:

http://www.seeedstudio.com/depot/g12-water...nsor-p-635.html

Keep us update with the build progress. I'm enjoying seeing the technical rigor and attention to accuracy.

 

[WhiteWolf]

I really have no idea what you guys are talking about, but I'm excited!

 

[matho]

I'm using loadcells to measure the volume in my HLT it has worked very well so far, its got rid of all my volume discrepancy.

Here is what im doing, its like a pre-schooler with crayons compared to what you guys are doing.

love your work newguy and also yours zizzle

cheers matho

 

[newguy]

My design philosophy seems to follow the old engineer joke of "calculate to 12 decimal points, measure with a micrometer, mark with chalk, and cut with an axe." I love accuracy even though I know that, practically, there's no difference in real life between a 0.2% and a 0.6% sensor over the temp ranges most commonly used whilst mashing. My take is that the difference in accuracy is only a few $ and for this one-off system, it's a small portion of the total budget so I go for the best specs I can get. If I was building 10,000 it would be a different story.

You're right, my primary background is HW so I guess that helped to shape what I've created. My primary goal is always to maximize accuracy, and that is the biggest influence on what I've created. Regarding HW bugs, there really isn't any chance of that. The circuitry in this thing is, as with the code, reused from other things. The only complicated parts are the analog signal paths, but they're really only low pass filters with an absolute gain of 1 (as close to 1 as I could get). Lowpass filters to get rid of the 60Hz mains hum, and a gain of 1 to make the SW more straightforward. Same thing when it comes to the 1.024V reference - that just ensures an "even" mV/count from the A/D converter.

Temperature rate of change is also something I know will be an issue with SS as the thermal conductivity of SS is the shits compared to aluminum or copper. My present system uses home made copper thermowells with LM35 sensors (in the tin can package) but for this system I really wanted to go the all SS route for cleaning and durability. Practically, from a temperature ramp point of view, all this really means is that my control loop PID values will have to be tweaked to take this time lag into account so that I don't have excessive temperature over- or under-shoots when ramping. It just means a bit of extra tweaking to the final PID values, but that's it.

Your two conductive probe idea is a really good one. I didn't consider it when I put this together, but I can think of one special brewing case that I've done (and will do again) that would render it ineffective. When I brew really high gravity beers, I buy a fresh wort kit and use that as part of my strike water in my mash. I do this to reduce the amount of malt I need and also because when you're shooting for 1.100+ OG, that requires a huge mash tun and also starts to tempt the stuck runoff issue. With two fixed probes, you're kind of screwed if one or both are already covered with the FWK contents before you start filling with water. If I want to use a FWK as part of my strike 'water' with this system, I just need to zero the scale with the kit in the lauter tun (like the TARE function on a digital scale) before adding the water. That, and I don't have to worry about the wife or kids flushing a toilet(s) while my system is filling and screwing up the two sensor flow rate deal.

Finally, regarding the flow meter you linked to, DAMMIT!!!!!!!!!!! I searched and searched and searched for cheap, accurate flow meters, literally for years, and I never once came across the one you linked to. I honestly do wish I knew about it before I went the mass sensor route....it would have cost about 1/15th what the 3 load cell solution has. <_<

 

[pimpsqueak]

I really have no idea what you guys are talking about, but I'm excited!

+1