Freelance Years
Getting a new job after National wasn't easy - nobody wanted design managers - and it took me a while to realise that's the impression my CV gave! So, with time running out my only real choice was to go freelance, or contracting.
Contract engineers in the UK electronics industry are a bit of an oddity, not understood much by the employed guys, and vice versa. My guess is that most of them, like me after a while, distrusted employers and would rather be their own boss. Having done both I think the advantages are illusory.
I got a month's work at Matrics, a small design company in in Tetbury (antique city) in southern Gloucestershire, doing layout at first. Not too challenging but very enjoyable and it did keep the wolf from the door.

After a while they let me have a go at design - my first pure CMOS chip, which was to be the analogue part of a battery charger controller. This used an unusual process (single poly EEPROM) from AWA in Australia, which had the advantage of tunnel oxide capacitors at quite high capacitance per unit area. I did a switched capacitor ADC which worked quite well; the only problem on the chip was the bandgap, which didn't always start up - major lesson here! CMOS bandgaps always need a start-up circuit (not many books tell you that - but it's worth reading stuff by Bob Pease on the subject). The project came to nothing - I think our customer sold snake oil.

Matrics was developing a reputation for being good at designing RFID (Radio Frequency Identification) chips, often known as Tags. A key customer for these was Atmel, a fab based in Colorado with, at that time, no analogue designers. I worked on a number of designs with all kinds of challenges - especially things like low power, multiple supplies and automatic switching between them (that's really unpleasant stuff!)

Common reader to tag communications use 100% amplitude modulation, which is easy for the tag to detect - but of course the net power received by the tag is reduced, so you need a sizeable capacitor to hold up the tag supply. Another scheme  was cooked up which uses 10% modulation  - now the power at the tag is much better but the demodulator gets more difficult. Essentially you need a comparator with adaptive thresholds, which you can derive from leaky peak detectors. I developed a rather neat switched capacitor design which was worthy of a patent IMHO but didn't get beyond the thought stage - no money for that! Anyway it has been used in a number of high volume designs.

We did a lot of mixed analogue-digital designs over the years: interpolators for precision metrology, water filters (a really low power RC oscillator), an Ethernet PHY, and various clever pad designs. Two memorable chips were a bipolar interface for automotive motor control - a very large geometry single layer metal design, and an even larger CMOS telephony chip containing many switched capacitor circuits for things like MF tone decoding. Both worked pretty well.

I spent two spells at Ericsson's design centre in Swindon (now sadly defunct). The buzz when I first went was Bluetooth; I was to join a new team developing the 2nd generation RF chip, the first having been done in Sweden with Sven Mattison, the "father" of Bluetooth, at the helm. I was lucky enough to get two patents on this project, one for a new take on a bipolar bandgap, and the other for what I considered to be a much neater idea, a trimmable oscillator (US patents 6,310,510 and 6,326,859).

Later at Ericsson I got to do some higher frequency stuff with an LNA and gmC filter for a DECT (digital cordless) phone chip. In my last session there I did a very low voltage (sub 1V) bandgap and a dB-linear VCA for wide-band CDMA.

More to come, in time ...