All Things Embedded

This has been an extremely busy year for me, despite the GFC. That said, I have not yet ‘reaped’ from my efforts yet.

Around March this year I was approached by a partner company to re package a design I had done nearly 14 years ago for a new product to be launched this year. Back then I was still “cutting my teeth” on micro controller based products and hence the outcome was a conservative but (even if i say so myself!) yet relatively (all these caveats!) reliable product. (I base this on the fact that over this period, 12000 units were sold with less than 100 returned, not necessarily because of bad design but more oft than not, physical abuse).

Now fast forward to the 21st Century…ok, so the old product was simple and reliable and used a particular (now patented) sensor mechanism but, this sensing technology I decided, was not suitable for the new design.

Technology had moved on (though the old design is still selling well in its original form and respective market) and 14 years of experience in the field has taught me new tricks and techniques.

The product I am talking about is a touch sensor switch used to flush toilets (ok, its not a glamorous area to be involved with but hey, it pays the bills!). The original design was and still is, ground breaking in that it allowed touch sensing through a continuous sheet metal plate. The technique utilised was differential sound detection between sensors on the back of the sheet of metal. (For which we got and still maintain a patent for.)

The new product however, was not metal or ferrous based material and hence opened up new sensing techniques (or old ones revisited) to be considered.

About this time (Mar 2009), Microchip reincarnated (along with several other vendors such as Cypress Microsystems) capacitive sensing. Something I had read about and played with from time to time but was always scared off by some horror stories over the years.

So what did Microchip achieve in this new reincarnation?

Not much really, it was just the implementation of some smarter algorithms and in the case of the micro I eventually chose, a programmable current source coupled to a 10 bit ADC.

That was a the relatively easy part. Remember, my original sensor was stated as being a rather conservative design and generally the risk factor was low. It was a good “Brute force design”. I also had the luxury of plenty of space and volume to put all the parts in.

Not so with the new design!.

The new design was being driven by new management open to new ideas and potential sales. I was also in the situation where I needed this product as well as them. So I used the cost budget to the maximise and incorporated every feature I could think off in order to get the design job. Big mistake!

The spec was now not so simple and I was definitely “digging myself a hole”, because unbeknown to me, the timeline to launch was to get shorter in an exponential fashion very soon. (I should have known after so many years in this industry!!)

I thought… “hey! everyone is network enabling their products, so should we”. Then I thought another ‘killer feature’ would be sound for user feedback. (This, as I discovered later really gave me a lot of strife and some lessons to be re-learnt.)

Ok so as soon as the thought process goes that way, we are not talking simple…we are talking ’stacks’…communication stacks. Stacks take memory (my original design had only 512BYTES of OTP ROM!), lots of memory. Then there was the sound aspect, I thought, wow, how about some quality audio, not the usual iPod click but a smooth sound…I was sure heading for trouble again. The specs were mounting up but there was a particular apsect that was not keeping pace….

Throughout the years I have learnt that if you do not get your foundations to a design off to a good solid start, you are headed for BIG trouble. Yes, I am talking about the power supply.

My thought process was stupid in hindsight. This was a miserly power sucking processor, how could there be power supply problems. How wrong I was, but now I am definitely aware of how important a consideration this aspect of any embedded design is right from the word go.

With such tight time lines, you really need the ABC’s of a design secured from the word go and solid enough that you are not having to try some magic in firmware alleviate physical problems later on.

One problem I had was the fact that the power source was from a 24VAC plug pack. Even though I was only going to be drawing in the order or 50-100mA max, with the volume I had available, it was not going ot be an easy ride bringing this down to 3.3V efficiently.

I started with a linear regulator, but remember, I have 24VAC as the source half wave rectified, taking this up to around 36VDC (depending on load) peak!. Then I had to consider the inrush current, which with low ESR components pushed potential peak inrush voltages up to 50V plus!!!. most linear regulators cannot tolerate such high input differentials between input and output.

Even If I could find such a linear solution, the heat dissipation would have made the user interface feel uncomfortable. Not as in hot but instilling some doubt into whether the unit was stressed some how/

This then, basically threw the linear aspect out the window. (Perhaps a mixed discrete + asic could have done the job, but space was a limiting factor.)

The next option was to go for a buck converting solution. The issues here was volume. I did not have the headroom for relatively large inductors. Then by luck I came across a little ‘beauty’ from our friends at Linear Technology. The LTM8020 uModule. These ‘babies’ are fully integrated. Inductor, FET and all !!!. Their only downside is that they have a sort of BGA package making them rather hard (though not impossible) to hand prototype with. (I did manage to do this.)

Linear Technologies uModule

The LTM8020 is rated at up to 36VDC. so something had to be done to ensure this was not exceeded. A low impedance resistor after a couple of dual SOT23 diodes did the trick, bringing the input voltage into a safe area. (Semi’s are generally current tolerant but voltage starts to literally breakdown barriers even for an pft… of a mSec!)

I still had the issue of inrush, though this was a rare event as once the unit is commissioned it is powered for ‘life’ barring the odd power outage. This final factor let me get away with stressing some of the power supply input components to a measured extent, these namely being the input capacitors.

Something one would not even want to consider in this case is the use of tantalum capacitors, which are extremely voltage intolerant in a spectacular way!. Luckily there are ceramic capacitors and these are rated at up to 50V for 1210 packages and accordingly de-rated for higher peak voltages. The use of ceramics saved the day as they were ideal in this application in terms of voltage tolerance, ESR and form factor.