Attempt to stop the RFI at DAB frequencies by decoupling and filtering the driver PSU have been ineffective because the RFI is being generated by the LED lamps themselves.
This photo shows one of the MR16 LED lamps opened up. The boost regulator boards can clearly be seen siliconed into the connector pins housing. This is the source of the problem. A direct DC test was set up as below
A very simple test, nothing more than one of the lamps croc clipped up to a 12v 7Ahr SLAB, with a DAB receiver a few inches away. What cant be seen here, other than the telescopic antenna positioned above the lamp, is the spectrum analyser.
With the lamp powered, there was a definite set of spurii between 250 - 300MHz indicated on the analyser. DAB transmissions are just a little lower than this, around 220MHz (this will have significance for US radio hams wishing to use the 1.25m band!) There was no obvious peaks anywhere else from 0 to 1000MHz! That makes it a particularly awkward problem, as its affecting just this spot on the dial!
Although the peaks were not measured, only observed, the addition across the lamps terminals of a 100nF ceramic capacitor dramatically reduced their amplitude. The DAB receiver, which was almost entirely wiped out six inches from the lamp, was now usable somewhat closer. Only by bringing the antenna within a few inches of the lamp was the RFI apparent.
This lamp has now been refitted into the holder in the bathroom, with its 100nF capacitor wire-wrapped to its pins with minimum lead length. With this lamp in place, and the DAB radio back in its usual location on the windowsill, powering the lamp circuit results in badly mushed DAB reception, but it doesnt now drop out entirely.
The position of the receiver is almost equidistant between the modified lamp and the next unmodified one of the three in this circuit. I anticipate that adding the 100nF decoupling capacitors to the other two lamps, plus the addition of a few snap on ferrite beads to the positive supply wire close to the lamps, will resolve the RFI issue.
Really though, the 100nF decoupling capacitor and the ferrite chokes should have been included by the manufacturer within each lamp. Another case of saving a few pennies in production, at the expense of increasing the RF noise floor and causing unacceptable RFI.