I've modified this mechanism more often than any other I have ever developed. This time my thrust was for shortest throughput time, and I managed to reduce this to p960 (hence the p240 to p960 modules).
At lower right, in RED, is the set/reset p960 gun. Started by down left signal (in VIOLET) from BCD converter, this mechanism is basically Paul Rendell's memory cell with an inverting leg. Output is to a divide by 4 mechanism by (I believe) Al Hensel. The glider inserted into the memory cell then will deliver a MWSS left every 960 generations, until a LWSS from the 8 digit nibbler, up in VIOLET, deletes it from the loop. The MWSS is the master trigger for binary to ordinal action and does six things: 
1) In VIOLET, Trigger up to the blocking stream of a p30 LWSS gun, in my Meerschaum Gap 16 reaction, to deliver a burst if 16 LWSSs, one of which will ultimately become the ordinal output of the device;
2, 3, 4) in ORANGE, BLUE and RED, the 1st, 2nd and 3rd stages of the converter. The MWSS resets a latch (Paul Rendell' set/reset type A) and initializes the on/off phasing of the 3 stages;
5) in YELLOW, sends a pulse to initialize stage 4 of the converter, this as described in my last post to the p30 thread regarding a 'double action' p480 gun;
and 6) in VIOLET, triggering a final p240 to p480 correcting mechanism I will digress about below.
Input data, in GREEN as 4 bits @p30, run through pairs of opposed p240 guns in the first three stages, and will provide a second pulse to the fourth stage if the 8 bit data LWSS is present. If a given data bit is present, the up left p240 glider for that stage is deleted, and the opposed down left glider sets the latch (by side entry). This action will change the on/off phasing of the given stage to off/on. The output of the latch in each stage serves as one input of a p30 NOT EOR gun (the provenance of which I am unable to recall). the second input of these NOT EOR reactions comes from, successively:
stage 1) a p60 gun (one on/one off);
stage 2) a p120 gun into a p30 stream in an eater forming reaction (yielding two on/two off); and
stage 3) a nifty 4 on/4 off gun from Dean Hickerson's archives.
The fourth stage will eventually fire a glider in Rendell's Gap8 reaction to provide a proper 8 on/8 off stream dependent on master trigger and data bit 4 (8 thru F).
Each of these stages will pare 1/2 of the LWSS stream away until, at the end of the gauntlet, a single LWSS will be left, representing ordinally a value from 0 thru F (0 thru 9 in the case of this entire mechanism).

Here I'll digress. I first noticed the flaw in flow when, for certain length BCD results, the 8s were coming out of this converter as 0s (or 9s as 1s). I tumbled then onto the p240 fault created by the nibble-wise output of the BCD converter. The double action p480 gun in the 4th stage partially mitigated the problem, but then in the next module, the data encoder, I noted that the same problem existed. The individual segment loops in the encoder are 16 bit (i.e. p480). DOH! I was wearing blinders! I then asked myself, "self, could two wrongs make a right?".
As it turned out, two wrongs did make a right (almost). 
I restored the original p480 gun in the ordinal converter and the encoder module generated the correct set of segment gliders, even though the wrong ordinal was being generated. Give 240 gens here and take them back there. The problem was this:
The worst case scenario for consecutive digits is ...xx09. Displayed as generated by the BCD module, digits are printed rightward starting with the low order digit. Two consecutive ordinals for the same value will be 960 generations (480 cells) separated. But nine - zero are only 690 generations (345 cells). The Ordinal is sent parallel with the segment data burst to the diverting/decoding modules and is used not only to start the decoder, but also to reset a latch in each module that diverts the data burst up for that digit, thus clearing the path for the next burst to travel to the next rightward digit.
So the diverting stream needs to be out of the way before 690 generations transpire. IT ISN'T! It's 60 generations too slow. It is only by virtue of the fact that the digits zero and one are devoid of a center segment that the pathway is clear and functions. This forced my hand in making the center segment the leading marker of the seven segment signals.
Now consider the two wrongs scenario and the value ..xx87:
The seven will be ondinally encoded as F and the eight as zero. They encode properly, but are separated by only 510 generations (255 cells) and the resulting collision with the diverting stream causes the mechanism to GTS.
My solution to this problem was to incorporate a final p240 shifter at the output of the ordinal converter. It uses a p480 gun and the master trigger for the ordinal converter. If the p240 'problem' shift exists, a diverting boat is created (same mechanism as in the BCD converter) and the p240 fault is corrected. This is shown at right in VIOLET for control, YELLOW for p240 corrected ordinal, or straight through in GREEN for uncorrected ordinal via a stabilized inverter.

