The ideal solution seems to me to keep a rigid frame for optimum power delivery, but to add suspension to the seat - and particularly the rear of the seat - to reduce the road shock transfer into the upper back. The front of the seat is mounted to the frame by a lateral bolt through its side panels to allow the seat to pivot as the seat back moves up and down under the action of the suspension system.
Inspired by Alex Moulton's venerable but effective elastomer suspension (See Moulton Suspension) I initially experimented with my own variant, replacing his specially tuned rubber elastomer with a more readily available tennis ball, acquired by searching in the long grass around my local tennis club. Despite the unsophisticated and very Heath Robinson implementation of this concept it actually worked pretty well. The tennis ball compresses and squashes under the weight of the rider, but still seems to retain enough vertical compliance to absorb bumps and vibration. As a super lightweight form of seat suspension it may still have future merit as it combines both spring and damper functions in a single element. However, my initial and very quick experimental implementation lacked the necessary lateral stability required of a rear seat mounting.
Further development of the tennis ball concept was abruptly brought to a halt by the recovery from the local rubbish dump of a full suspension MTB that someone had thoughtfully thrown away. Despite being very cheap and cheerful it did provide a number of useful components, as well as more frame tubes to add to my ever-growing pile of raw materials. Here the tennis ball is replaced by the spring and damper unit taken from the MTB's rear suspension. (I soon discovered that the "damper unit" - no doubt made in China - contained absolutely no internal damping element whatsoever, but served only to look the part!)
This unit was also effective, and responded better to big bumps, but it was heavy (springs are like that!) and still lacked decent lateral stability to act as a good seat support to prevent the seat from rocking laterally when pedaling.
The next step was to refine the system by combining the spring/damper element with some form of laterally supportive seat stays. My approach was to employ a pair seat stays to provide the support and lateral stability, but articulated so as to allow them to "flex" longitudinally in response to road-induced vertical forces. The amount of flex (shock absorption) is governed by a preloaded spring unit with associated damper. For this I took the spring from the suspended front fork of the cheap MTB and slid this over a hydraulic piston unit used to keep vertical kitchen cupboard doors from falling shut when you open them. This damper body conveniently fitted within the spring, and an aluminium tube was slipped over the unit to maintain the spring under preload (the degree of preload being determined by the effective length of the outer tube which was adjusted by adding a number of washers as shims).
A pair of plastic V-brake arms served as the short links in the articulated seat stays and a lateral bolt acted as pivot and held the three elements - long links, short links, spring/damper unit - together. In action, a road bump applies a vertical force to the rear wheel which would normally be transferred directly into the vertical movement of the seat back via the seat stays. However, the articulated seat stays cause some of this force to be resolved into a horizontal component which acts on the pivot bolt. Since this bolt is restrained from moving horizontally by the long seat stays, it effectively rotates forward and downward around a radius centred at the long stay pivots on the rear axle. This action compresses the spring/damper unit which serves to resist this movement.
The overall action is fairly complex to visualise, but the amount of road shock transmitted to the seat back is determined by the combination of the spring rate and preload together with the relative geometry of the long and short links and spring unit pivot point on the frame. By playing with these parameters, the amount of reaction at the seat can be varied from extremely rigid to excessively floppy; the former offers no suspension value, while the latter provides insufficient support to hold the weight of the rider when pushing back into the seat. The video clip below illustrates the seat suspension action when these parameters are adjusted to give a moderately soft suspension.
The same kind of seat reaction is obtained whether the initiating force arises from a (vertical) bump in the road or a (horizontal) force from the rider pushing back against the seat when pedaling hard. This means that a compromise has to be sought where the suspension is not so supple as to allow the seat to react when pedaling hard, but not so rigid that it adds no absorption of road shocks. With the geometry of the link and spring elements used here, the whole unit acts to provide a falling spring rate. Most - but not all - MTB and other vehicle suspensions are designed to give a rising spring rate. What this means is that the resistance to further movement in the system increases non-linearly as a function of the amount of movement.
A rising spring rate therefore gives an initially supple suspension, but hardens up as more and more compression is involved. This is normally a desirable property of most suspension systems, but not in this case. The reason is simply because of the seat reaction to rider-induced forces from enthusiastic pedaling. What is required is a seat that is initially hard to move (to avoid this rider-induced action) but which responds to greater and more rapid inputs from road-induced forces. The falling rate system achieves this, and sacrifices some loss of small shock absorption as a compromise to avoid rider-induced inputs, while offering useful suspension to more significant road bumps and potholes. (However, very large road bumps will cause the falling rate system to bottom out when the short links' further rotation is halted by contact with the seat side panels. In practice, this is still better than having no suspension at all, and such mega bumps are fairly rare.)
With this design the system is stiffer and less responsive when the interior angle between the long and short seat stay links is very open (approaching 180 degrees in the extreme) and/or when the spring/damper unit is fairly vertical. The system softer and more responsive when the links' interior angle is more closed (approaching 90 degrees in the extreme) and/or when the spring/damper unit is more horizontal. The particular geometry of my system used here was determined by certain factors that I couldn't - or didn't want to - vary (e.g. the point of the spring/damper unit pivot mount on the frame). However, with an hour or so of experimenting I arrived at a suitable compromise to give the kind of suspension response I wanted.
This system was adopted on the bike that I took on a 1500km maiden touring trip across the SW of France (including the very steep and unrelenting climbs of the Massif Central) and it worked very effectively.
