We brewed again today, with a freshly plugged up false bottom and confidence that our recirculation loop would flow nicely.

We certainly improved the flow of things - and that was very satisfying. We made a Dortmunder and efficiency was up and the constant recirc gave us a beautiful clear nectar into the kettle. It has been a while since we've been able to say that!

But there was still something not quite right. The flow was still slow, almost too slow. It was at this point a few months ago that I began to chase my tail.

First thought back then was that it was probably the heat exchanger coil. There is a fair bit of extra copper in the recirc line and it must be sufficient back pressure on the pump to slow the flow.

Then it must have been a problem with the false bottom. And so imagining that there was air caught in the top of the pickup and getting sucked into the pump, I decided it needed a vent. I figured wrongly that the vent up through the mash tun would ensure that it would fix the problem. Umm, not so. The sucking of liquor out the bottom of the tun also sucked air in through the supposed vent.

So a complete circle - and the vent was plugged ... And today we were back at square one - the flow was less than adequate.

We were going to do a double brew today, but the first one took a long time due to the slow flow rate. So instead of attempting the second, we were determined to find the cause of the problem.

I had my suspicions, so I removed the false bottom from the pickup in the tun and checked the flow rate with an open hose through the bulkhead. And guess what? Just a dribble from the tun!

And here lies the unexpected side effect. Quite some time ago we thought it was a good idea to install polysulphone quick disconnects with bulkhead stops in the three tuns. I think at first they worked well. But over time we have had trouble with them, and in particular the mash tun one. We have gone through four of the male connectors - mostly because they get a bit tight and break apart on release. These connectors have a pin in the centre of the connector that pushes the stopper in the bulkhead open when the two are mated. In our system, the stopper is barely opening now, and restricting the flow from the tun severely.

The main reason I went for these connectors is for the convenience of the stop. It means you can pull the hose off without worrying about liquor leaking everywhere. Before this we had ball valves on the tuns, but they stick out and get easily knocked when cleaning them.

I've been thinking about a stainless replacement of these connectors for a while now. I've not found any with automatic stops on them, and maybe that is for good reason. Anyway, it is not hard to rig up a stop even if it means put a stop plug onto an open tun.

There are a couple of different stainless solutions I've come across so far.  My favourite Melbourne brewshop, Grain and Grape, has these connectors which are similar to the kind you get on domestic garden hoses, except in stainless and a little more robust.

But I like these from beerbelly which are the little brothers to the types used in serious breweries. 

And there are even 'stops' you can buy off the shelf.

With decent software to model and re-model layouts, I'm finding the process of designing boards much less of a chore and more fun than ever.

Being able to nut out a control panel design and then re-work it into a versatile set of strip boards has inspired me to do similar things for the tuns.

A few posts back I had drawn a model of the mash tun under CorelDraw and soon realised that I needed a 3D model. After spending some time to learn SketchUp, the modelling has come a long way.

I wanted to model the mash tun for a couple of reasons. One is to make a dough-in mixing paddle which did not interfere with sensor probes or the liquor return manifold. The other is to make sure all the sensors and housings fit together.

The mash tun boards are essentially remote sensor connections back to the main picaxe controller. I wanted to get this bit right so unplugging tuns for cleaning and maintenance would be simple, and the sensors would be rugged and waterproof.

Through modelling under SketchUp, I was able to determine which dimensions and parts were critical on the sensor board. The other pleasing thing is that a simple design is easy to build on and improve. I had never intended to have the pressure sensors plug directly into the board, but rather connect via a fly cable. Using SketchUp, it was easy to move the sensor around and then figure out how the board layout might be improved for the two to go together.

The other aspect of this I've enjoyed is being able to have both schematic and board 'interact' until the design has been optimised. Some parts are not critical on the schematic, but the way they are drawn can affect a board significantly. Being able to draw a schematic, rough layout a board, look at the board critically, refine the schematic and so on has made this task fun and creative.

We are brewing a Dortmund lager today but the dremel drill press should arrive today or tomorrow. We will then have all we need then to make some boards next weekend. The only reason I can wait that long is I enjoy making beer as well.

I've never done a silk screen print on a board before. Partly this has been a software issue, but also when using the blue press n peel sheets it seems a bit of a luxury because the sheets are quite pricey.

We've played around with glossy photo sheets as suggested by US PCB hobbyists, but never got them to work. Here in Australia it is not easy to simply order a particular brand and code of paper from some overseas mega-mart.

A while ago I found an explanation about why we had so much trouble with paper. Essentially, the quality of glossy photo paper has improved a lot over the time inkjets have been around. Older clay-based papers have given way to newer water-proof ones. The problem lies in the waterproof coatings because newer papers will not dissolve away from the toner.

The suggested solution is to find cheap and nasty glossy paper - and it is the stuff that gets delivered seemingly daily to our doors - the paper under all that glossy junk mail advertising.

Rather than test it on ultra-clean copper intended for etching, it seemed a good idea to try a silk screen on an old board that we were never going to use.

The print on the paper is barely recognisable due to the glossy 'pre-print'.

After 10 passes through a hot laminating machine (the type used to laminate plastic covers on business cards and the like), the paper had stuck to the board substrate. It was now time to soak it in water to see if the paper would release and leave the toner intact.

And hey presto - the paper felt slimy the moment it hit the water and was easy to rub off, leaving a beautiful print on the board - except for the fact that I forgot to mirror the print

If I can get the toner to adhere that well to copper, I'll be very happy. Getting ever closer now to making some shiny new boards for the HERMAN project .

I'm trying to get this version of HERMAN right - where things fit together and wiring is neat and tidy. Because it is a complex project, I'm using a variety of tools to help fit the jigsaw together.

After testing the 3 pushbutton strip for the control panel under PicaxeVSM, I'm now happy with the schematic. From here I used DipTrace to produce a board outline.

These strips can be place side by side or end on and wired together to form a resistive ladder with an output voltage that changes depending of which button down the ladder is pressed.

I initially intend to have 6 strips side by side, so will use DipTrace to create a six-up print automatically.

To ensure that the connectors and cables would fit physically with the LCD screen, I turned to SketchUp to do some 3D modelling.

The LCD screen is a seiko 40x2 with LED backlight.

The pushbuttons were modelled to figure out height between panel board and front of panel, and the IDC header connectors modelled to give an indication of fit under the panel and with the LCD.

These were then combined to give some neat views of a virtual control panel. Everything seems to fit, but the original artwork done in CorelDraw will need to be modified now.

The whole point of SketchUp is that you can look around and model in 3D space, so the three pictures above give a poor impression of what is really possible. The 3D files are available here if you want to have a good look using SketchUp.

While it took many hours to draw the components shown in the last three photos, most of that was learning how to harness the power of SketchUp. I continue to be impressed, and I'm sure I can draw 3D models quicker in SketchUp and 2D models in CorelDraw. My guess is that the whole panel above would only take a couple of hours to do from scratch.

We ordered a lot of 50 PCB drills late last week, and once we arrange for the dremel press we will be good to go. We are planning on brewing again next weekend, so it will probably be two weeks before we start making boards.

I'm glad I haven't done the usual thing and rushed into building HERMAN circuits. While it is very tempting to replace the old control panel with some nice, neat and well built switches and LEDs, I've just looked again with critical eyes and come up with what seems a much better solution - and one that is useful for others too.

The control panel design currently has 20 switches. The first 16 have definite control functions, but the last 4 are for future expansion. I've been trying to think of expansion without needing to design for every possibility, so was thinking about how a ladder array of buttons might easily be extended.

The other thing on my mind was that although the circuit would work and the front cover would make the panel look pretty, the board design itself was quite messy because it was complex.

So applying the simplicity cycle to the project with these things in mind, I've come up with the following circuit.

Rather than a complete control panel, it is a strip of three buttons and LEDs (that in my case are built into the buttons) that can be easily expanded upon. Each strip has the buttons, LEDs plus a header and a footer . The first button in any array is connected to the header via the link LK1. This connects the array to Vcc or in our case 5V. The buttons in between will produce an output somewhere between 5 and 0 volts, so the last button in the array needs to be connected to ground. This is done by closing LK2.

Although there are three buttons on a strip, the strip does not need all of them populated on a board. If a project demanded a two button decoder, omitting a button means it will never be closed. The picaxe code would still need to account for three steps in the divider chain.

Where more than three buttons are needed, extra strips can be added to build up the array. If, for example, 9 buttons were required, the first strip would have LK1 closed, and then a cable from J2 to J1 on the middle strip. The middle strip would have J2 connected to J1 of the last strip. The last strip would have LK2 closed to complete the ladder array.

Over the weekend I hope to use DipTrace to route out a board design but it ought to be much simpler than the previous version.

I realised that as the complexity of HERMAN was growing I needed to do some layout drawings just to make sure things worked in terms of both wiring and fit.

So I immediately turned to my old friend CorelDraw to do mockups of the mash tun and produced this in short time.

But I then realised that as it got more complex, I needed to do a top view to figure out if the liquor in manifold and the mixing paddle would actually work together.

It was at this point that I remembered why a 3D drawing might actually be useful. A while ago I discovered Google SketchUp and immediately became a fan of it. I know that Garrett has knocked up some nice models of his brew sculpture and various components such as ball valves and thermometers. Having worked partly through the learning curve on both VSM and DT, I was not in the mood for learning how to drive another complex program.

But with some spare time on Sunday night I tackled some of the tutorials. It seemed even easier than I remembered and I wish I'd checked out some of the videos when I first came across SketchUp. It is one seriously capable design program. For a person like me who feels 3D challenged, it is extraordinary.

One of my interests is woodturning. I can see how SketchUp can help with designs for this. No more graph paper for me.

But to my own brew sculpture. The challenge was to re-create the drawing from CorelDraw into a 3D one. After about 3 hours of learning, I got this far.

The basic mash tun was drawn in a few minutes. A few minutes later I had a nice straight edge on the top of the tun - like a chamfer. The rest of the time was spent working through the sequence of tutorials to figure out how to use the follow me tool and create a nice tun shape. I can see the power in this program - its just that while some things are incredibly intuitive, other things are not until you learn their particular way of doing it.

Anyway, there are so many advantages to drawing a brew sculpture in 3D, and the program is really amazing, so the learning curve is worth climbing.

After a bit  more work, the tun was refined a little, but still a long way short of being more useful than the 2D drawings from CorelDraw. I guess that a bit of persistence it will make a difference. I've always struggled to be patient ... must wait ... must learn ... must hold off to get it right this time ...

Back in the workshop on Monday, I planned for a gentle day pottering about looking at things with a critical eye and doing small improvements - like fixing the false bottom.

The open hole in the top sucks air in. The tee was originally plugged at the top. It is a tee rather than an elbow to provide a mechanical point for the bottom of the mixing paddle.

So a soldered plug on that offending end and a bit of silicon just for good measure seemed the trick. If it was being built from scratch the silicon wouldn't be needed, but it is difficult to solder such small pieces and not destroy existing solder joints.

A while ago I had a first attempt at some compression fitting disconnects that could be hand tightened by adding home built wingnuts to the fittings. They worked to a point but were hard on the fingers. I thought about a way to improve a promising idea and came up with this. It is definitely better to use, so time will tell if they are mechanically strong enough and how the fingers fare.

The bolts are 3mm with nylon spacers and a nut on the end (the nut is a simple spacer). The nylon with the rounded bolt heads are a nice combination. The holes in the compression fittings are tapped so the bolts screw in. The bolts are then strengthened with a dab of loctite on the threads.

Apart from some minor improvements and tidying on HERMAN, today yielded one more modification of note. The layout schematic shows that apart from the mash tun false bottom there are a couple of other filters in the brewing machine. One is in the kettle fill line to eliminate any particle matter (grains or husks) from entering the kettle. The other is the more conventional pickup from the kettle into the fermenter. For a long time we have zip-tied a hop sock onto the kettle fill line. Cleaning this has been a bit of a pain, and needed a new tie each time. I had some stainless mesh filter material on hand, so now we have a new filter on the kettle fill path.

Just for good measure, below is a photo of the filter at the bottom of the kettle. It uses the same mesh as above and forms a ring at the bottom circumference of the kettle. There is a slot cut into the bottom of this ring (with a nibbler) that is covered by the stainless mesh. This means that hop residue and trub has a difficult time getting into the fermenter.

The work on SketchUp and CorelDraw is important if we want to make everything fit. We are still waiting on some components that will finalise the layout of bits and pieces and still waiting on some tools that will help with PCB manufacture. Once I get my head around SketchUp and get those bits and then have time to deal with the detail of it all, there will be no stopping us as we build the remaining bits of HERMAN version 6.

Oh ... except that there is still a lot of coding that needs to happen. So many things, and so little time.

It has been quite some time since designing my last major circuit board. I've realised again that I'm much better at having ideas than actually implementing them. I think this is why I stumble over things like the false bottom - with its poorly engineered plug - because I'm already thinking about the next idea rather than fixing the current one properly.

Having said all that, the work towards final circuit boards for the HERMAN 6 control system is going well. While it has taken extra effort to climb the learning curve of PicaxeVSM and Diptrace, it has been a lot of fun as well. The opening screen of VSM says "from concept to completion", promising the production process will be easier with software modelling. It is true, and it is much more interesting doing it this way. The devil is still in the detail though, and modelling something on-screen remains a long way from working through the detail towards a finished working board.

Because I find detail challenging, I've discovered a real benefit from using this software pair. It is much easier now to design to a basic level of complexity, then go back and re-design the next layer of detail. To begin with, some basic modular bits of the HERMAN project were tested under VSM. The 595 LED driver was one of these modules. Firstly I got a basic driver circuit to work, and then when that was fine, added a second picaxe as a 'sender' to test the extended bus network. The extended bus is a way to add extra modules to the HERMAN system. Some of these have been designed (like the LED driver), while others are just ideas waiting for time to implement them. An example of one of these ideas includes a hop-dropper drive for kettle additions. Knowing that I can design this 'after the fact' means I can focus more narrowly at the moment.

The extended bus has both send and receive serial lines which means that the controlling picaxe can either control devices or receive data from external inputs. I am sure that I2C could have been used for this, but I'm comfortable with serial protocols and how to implement them on the picaxe.

As the parts have slowly come together, the VSM/DT combination has meant it has been easy to add things like test points and connecting points for adding buttons to the control panel.

I've ordered the 40x2 LCD display for the control panel and have been working through interfacing and controlling ideas, thanks to Hippy. When I have all the components on hand, final checking of the physical layout will be possible. I think it would be wise once the schematics seem to be finalised to go back to VSM, simplify the circuits once again and test them with picaxe code, just to make sure all the bits join up as they were intended.

By that time I should have all my PCB making tools gathered together. I'm hoping to get a set of high speed drills like Garrett from ebay, and I need to buy a drill press for our dremel. I will also do a test print on cheap glossy advertising paper to see if it is suited to the toner transfer method of PCB etching. In 3-4 weeks we should be producing our PCBs.

We figured it was time to brew beer again. The plan was to create two side-by-side batches of our favourite North German Pilsner. One batch would be made as normal using only Tettnang hops; the other batch would be identical except it would be hopped exclusively with Spalt.

The inspiration for this move came from a recent trip back home to Tasmania where we sampled Moo Brew Pilsner. It is a German Pilsner and they use Spalt for their hopping. We'd not used it before, but Ross at Craftbrewer stocks it.

A double brew takes most of the day when things go smoothly. If you have been following recent blog postings you may remember we have has some issues of late with flow in the HERMAN recirculating loop. I cursed being distracted last time while brewing that I had not done the testing needed to figure out where the problem was ...

All seemed to be ok at the start. The change to pre-heating the mash liquor in the tun by using the recirculating loop was promising. Not only does this mean we can hit our mash targets spot on, it means we get to test the loop before adding grains in the tun.

Dough-in worked out fine, and the mash hit its target of 63 C as hoped. All good so far. But while liquor flow had been fine pre dough-in, the small trickle once the grains were added meant for a painfully slow brew day. At least the temperature was right. Now we needed to figure out what the problem really was ...

We disconnected the hose at the mash tun outlet to allow the liquor to drain out without assistance. It was surprising to see that flow was even slower than with the pump in the loop. This meant that the pump was actually improving flow by adding suction to the false bottom connection (not a desirable thing). This proved that something was seriously amiss somewhere.

So the next bit was to check the quick disconnect right at the tun itself. The disconnect is designed to close when the plug is pulled out, and the orifice itself is quite small. Using pliers to press the bulkhead stopper in (and open it) produced the same result .

There was no option but to transfer the contents of the mash tun into another vessel and carefully check what was happening with the false bottom and pickup tube. So with some water as back pressure and the false bottom exposed, the problem was glaring right at us ...

Firstly, the water back-pressure blew air and grains back up out of the works. The plug that had been put back in place to stop the 'air vent' was no longer there. This meant that grains were able to get sucked into the take-off tube and block it up. No amount of liquor pressure through the false bottom would free up the works - it would always re-block itself with more grain.

What is so frustrating is that we had thought we'd fixed this problem. Obviously the fix was not robust enough. We temporarily fitted in another rubber plug, but I think a soldered cap will be needed.

Anyway, with the grain back in the mash tun, all was tested and finally the flow around the loop was normal.

We'd lost so much time by now that we did not wait for the grains to heat to mash out. Our efficiency suffered because of it, but at least we knew we could run the second batch through with confidence.

The second batch of Pilsner mashed beautifully. We made 85 litres of beer yesterday and I've been reminded again that simple fixes often create more trouble. When figuring out solutions, industrial strength lenses are needed so that things are fixed once and for all.

It was great to see the auto-router in action and gain confidence that the process is working ok. Taking a step back, it seemed important to keep a record of the steps required to create a successful PCB from concept to final.

1. Work out the concept - document aims; research methods to achieve them; read tech sheets; ask questions
2. Draw a circuit and test it - using picaxeVSM I can test circuits and picaxe code together to test the concept. Prior to being able to do this onscreen, I would draw the circuit and get the breadboard out to test things. Both methods can be complimentary. VSM can be useful for smaller building blocks while a complex piece is best tested on breadboard. VSM can actually interact with external circuits so both can be used at once.
3. Do a design check on the schematic - ensure there are no errors, and in particular for VSM ensure that all components for PCB export are checked, and virtual instruments are not going to be exported. Check that the circuit has power and ground connections (VSM will run without them).
4. Source all the components - I usually forget to do this at this point. It is important because a circuit cannot be finalised until critical components are on hand. It is important to have these components before designing a PCB or the design will often not match expected components.
5. Create packages and pads to match components on hand - in particular I need to create larger than normal pads for using the toner-transfer method of making PCBS.
6. Create a rats-nest/airwire view of the PCB - this gives a visual that helps to re-design the schematic. Some wire connections will not be critical and by changing pins on connectors, for example, will make for a simpler layout.
7. Re-work the schematic with the rats-nest in mind. Re-work any other schematics that may be affected by changes as well.
8. Create PCB layout - manually or auto route the layout. Perform design checks to ensure there are no errors. Ensure traces are of sufficient size for manufacture. Ensure pads are sufficient size. Add fill patterns if desired to lessen etching. Add text to identify circuit. Add component pattern artwork if required.
9. And then do it. There is a blog entry about using the toner transfer method here. There is also information on doing double-sided boards here.

I'm currently using the DipTrace free edition. I've currently stated that I am a real fan of this program - for reasons of usability and an auto-router that saves me much pain. The fact that it has a free edition that still packs a punch is icing on the cake.

Except that catching a free wave doesn't always give the best result. The control panel schematic was over the 250-pin limit when I tried to import the netlist into DT.

OK, I'm not complaining. But I did have a look at different packages available and got thoroughly confused. If there is a free 250 pin package, why is the next available option a 250-pin starter package for $75? And then to get greater functionality than I already have, the Lite package gives 500 pins for $145!

I decided to enjoy the current free gift of 250 pins and look for a workaround.

It took a while to figure out that with my compound switch with LED, I had an extra 72 pins I didn't need. For each component I had drawn in all 6 pins, which meant 12 pins for each button where a LED was needed. By redrawing the switches to four pins and the LEDs to two pins, the 72 pin reduction nearly got me down to the 250 pin limit.

The easiest thing to trim on the original schematic was the second extension bus connector. This was a 'just in case' part for expansion purposes, but will be available elsewhere in any case.

At last, a trouble free netlist import.

Ratsnest after the components have been arranged on the board.

The last time I built one of these things was the prototype panel that currently drives HERMANs control. The schematic was a little less complex because it didn't have the inbuilt LEDs. The PCB was built using ExpressPCB. The naming is a little ironical. The only express part of that program is internet ordering for boards. It was a function I never used anyway and after discovering DipTrace there was no reason to keep using it. Under ExpressPCB, all the traces need to be laid down manually. Even worse than this, there are no netlists to minimise errors. A ratsnest layout like the one above simply does not exist.

I was eager to see how DT fared with the auto-router on this moderately complex board. It would have taken me weeks to get it prepared manually, catching a few spare minutes here and there, and driven me mad in the process. I feel like saying "I'm getting too old for this kind of thing."

Anyway, the auto-router did its work in about 20 minutes, with a result likely better than I would have done manually. Fixing up the compound component pins has also made things less confusing for the machine.

The board has quite a lot of jumpers but that is ok. The auto-router left one further air-wire to trace. Later in the week I will do some fine tweaking on the design so that hopefully we can make a board next weekend.