Very interesting info you found, @Guillaume.
However, it is now possible to disprove this bit:
The fuel pipes above the engine are quite fine bore and exposed to the full blast of the icy airflow, so possibly the most likely place to encounter the problem.
from data here Perhaps surprisingly, the temperature drop along those thin pipes is negligible.
Fuel freezing is also a possibility. Avgas freezes at about -58C but can contain up to 0.1% water in solution, and any more than that collects at the bottom of the container, or at low points in the fuel system. There is a “famous” UK Aztec pilot who has repeatedly stated that he gets a guaranteed double engine failure at -15C (in VMC!). Clearly his aircraft is not airworthy… but why? When asked for more info he never supplied any, but I suspect his fuel system has something wrong with it and water is collecting somewhere, maybe in a dip in a pipe.
I have another bit of info.
If you do use alternate air, the air filter can easily ice over – because there is no airflow to keep clearing it. Basically any crap will just end up sitting against it.
So if you have been flying on alt air (e.g. due to icing conditions) and then close the alt air door, the engine could just stop. I have just got an email from one pilot (TB20) who this happened to. So, whenever closing the alt air door, be ready for a total loss of power!
I should add that I have very often flown in airframe icing conditions, right up to getting ~30mm of ice on the wings (-5C or so) in 5 minutes, but never had any loss of power, and I never used alt air. However, I most definitely was using the prop TKS then. What this means is that either the prop TKS stops any icing in the entire air inlet path (probably true, because if water gets in through the air filter – which it obviously does, from the frozen-alt-door incidents – then TKS fluid will also) or the non use of alternate air prevents the air filter getting blocked by ice (and there is no icing further along the air path anyway, due to the +10K air temp rise due to compression).
Peter wrote:
and there is no icing further along the air path anyway, due to the +10K air temp rise due to compression
I pointed it out before: the MP at the manifold is less than ambient air pressure in a normally aspirated plane so how do you conclude there is compression heating?
Ram air compression. I have some more data now and it shows the rise is IAS related.
There is also a 3-4K rise between the true OAT and the temperature measured about 3cm inside the cowling, but I already knew that from previous data. It takes very little to get significant compression heating if you actually force air into a hole. In comparison, there is much less of that on a wing, because the air just gets out of the way, mostly.
Peter wrote:
With alt air on, the delta-T from OAT to the RSA fuel servo inlet is 30K.I thought about the alternate air some more. If the outside air is at 100% RH, warming it by 30 degrees is going to dry it to O(10%). In turbulent flow, that is probably enough to evaporate any liquid drops before they get to the narrow passages (but I need to check that with some kind of back-of-the-envelope estimate). That must be the idea behind the alternate air.
The servo inlet hole is about 60mm dia and the openings in the four sensing tubes are about 3mm dia.
That rules out ice particles as the source of the blockage.
It will drop the RH for sure, but the H2O molecules won’t magically vanish. The same mass flow of water vapour will still be present, because the stuff has nowhere else to go. All of it is heading for the fuel servo inlet.
There isn’t any “water separation” mechanism that I can see.
But I think you have a point in that air at -35C and 100% RH will have a very low RH at -5C and should therefore have a very low structural icing potential. But what about the fact it is compressed? Doesn’t that negate the RH drop to some degree? I would have thought that a low RH means the water molecules are simply further apart (on average) and if you squeeze it, they become closer together, so the RH must go back up again. Obviously the flow velocity must go up in inverse proportion to the cross-sectional areas – of the order of 20:1 – but I don’t see that as changing the effect.
I know there are people reading EuroGA who are serious experts in this field and hopefully one of them will contribute 
Well, yes, if you compress the air enough, the partial pressure of the water vapor will eventually go over saturation. But, as Achim has pointed out, the manifold pressure at wide-open throttle in a normally aspirated plane is (slightly below) ambient, so the pressure differences aren’t great. So once the liquid water in the air stream has evaporated, it should stay in the vapor phase.
I believe the Lycoming fuel servo has a venturi, so perhaps the air upstream of the venturi in the inlet box is compressed and the air downstream is de-compressed, producing a temp variation between the two areas. Manifold pressure would be below ambient, being downstream of the venturi (NA engine). As the alt air door would be upstream of the venturi it is in the compression heated area.
FWIW I’ve had fuel icing several times and that produces a slow increase in EGT as the mixture leans. I’ve never had filter or impact icing as the filter is cylindrical and mounted at the rear of the engine. Also the throttle body design of the TCM is quite different from the Lycoming.
There is no venturi in the RSA5 servo (I have a spare at home so can get a photo) but, yes, the initial manifestation is apparent mixture leaning.
You see that because at any altitude relevant to this discussion, in a non-turbo aircraft, you are flying “best power” i.e. 130F ROP so initially you see the EGTs rise. This takes a few seconds and then of course they plummet. The first time I just happened to be watching the EDM so I saw the whole thing. The second time I was doing something else…
The other point is that using the alt air loses so much power that one isn’t going to want to be doing that unless absolutely necessary. It’s probably worth 1000-2000ft off the operating ceiling, and one would expect that from the effect which a 20K delta-T would have on the engine efficiency – a 7% loss of power?
I thought that another reason for the power loss with alt air was the loss of ram-air pressure through the tapering filter box duct. My TCM engine shows no noticeable difference in power with alt air, the filter is at the rear of the engine but is in a high pressure area, I think that the loss of ram air is offset by the loss of the filter drag. Alt air leads to behind the rear baffle, an area which is ventilated by NACA ducts so perhaps I don’t have the same temp difference as in the TB20.
I don’t like the idea of a forward facing filter – it leaves you exposed to impact ice blockage and very reliant on a functioning alt air door.
