It is not for me to reveal his intellectual property, but the circuit neatly uses the fewest number of components to get the job done with the voltage available.
In a round-the-houses way, it got me thinking how much it was possible to share components between a low-drop active rectifier and a low-drop-out current regulator to make a low-drop ac-powered LED driver.
And this is what I came up with.
To the right is the classic bipolar active bridge, fed from an ac source and making dc.
The transistors are the 'wrong way around' (emitters and collectors swapped) to remove the problem of the parasitic 7V base-emitter Zeners shorting the bridge when the active diodes are reverse biased.
It was Dave Casey of Zetex who first showed me this circuit. He invented the firm's matrix transistors, which not only have astoundingly low saturation voltage has high Ic/Ib ratios, but also have significant gain when connected in reverse - something which most transistors lack.
FMMT617 and FMMT717 might be suitable, although I suspect there are more modern types available. Click that link (app note AN15), and look at page 6 to read: "Since the reverse hFE of the FMMT717 is very high, peaking at around 200, the transistor operates very efficiently, giving a saturation voltage drop of only 85mV in reverse mode for the....".
One of the only other times reversed transistors are used, as far as I am aware, is in signal choppers, because saturation voltage is lower in reverse connection.
I have had one of these active bridges running for years at 500mArms.
These days, the active mosfet bridge offers lower saturation and no base current losses, although also no reverse blocking so capacitors cannot be fed. The bipolar bridge, afaik, blocks up to 7V out when the base-emitters conduct.
If you are interested in active bridges, Linear Tech released a beauty last year - the LT4320
And to the left is Bob Pease's wonderful LM334-based current regulator which uses the fact that the chip actually regulates to keep 64mV across its bottom two terminals.
It also features in the Nat Semi (now TI) LM344 data sheet as figure 20.
The three stability components are optional. Mr Pease suggests the base-emitter capacitor if stability is a problem, while the data sheet has the series R-C for the same reason.
Interestingly, in its standard connection, the LM334 itself behaves well when connected up-side-down, drawing "only a few dozen microamperes" under reverse voltages of up to 20V allowing it to act as both regulator and rectifier in ac application, it says in the data sheet.
By the way, there is another bit of LM334 musing here.
Can they be combined?
I think so. And to the right is my attempt. It uses the upper pnps of the rectifier as control elements, needing only a couple of diodes to stop the bases leaking into each other - which will increase the voltage at which the bridge gets into low-saturation mode.
BTW, base-emitter resistors are not allowed on the pnps, I think, because they would turn the transistors on 'forwards' when they are supposed to be blocking.
Quiescent current through the LM344 will turn the transistors on a bit, so the load will have to be higher than this times the gain or more current will flow than intended - I was thinking 20mA or 40mA in the led string.
There may be all sorts of mistakes here. Sadly, I have no time to model it, make it, or even find a use for it.
One final note, page one of the LM334 data sheet points out that (up to 20V) the chip does not leak much and can be used as a rectifier. I wonder if that could mean two fewer transistors for only one more 334?
* Mr Kurt, for example, pointed out to me that one should think hard before using rechargeable Li-ion cells in equipment if it is to be used outside in winter, as most Li-ion cells must not be charged below 0°C.