First, I should point out that, in a 2-house Parliament or legislature, one house could be changed to pure at-large PR, no districts, and the other house could be left unchanged, with its single-member districts elected without PR.
Leaving one house as-is could satisfy the concerns of the person who wants small district representation. Additionally, leaving one house non-PR could ease voters' concerns about change, by making it an incremental change, beginning with just half the legislature becoming PR, at first. This of course is true even if the PR house has districts.
So that simple possibility should be considered, along with the PR district systems.
There are about 3 main types of district systems:
1. Pure at-large election, with the whole country as 1 district.
2. Smaller multimember districts, which could be whole provinces or states, but which could also be smaller.
3. An additioal-member system like the German system.
The German system: In Germany each "Land" (equivalent to a Canadian province or U.S. state) holds its own separate election for MPs to send to the national Parliament. Canada or the U.S. could follow that example, or could have just 1 big national election.
The seats in that Land are half district seats & half at-large seats. The voter casts 2 votes: an at-large vote for a party (or for an independent or an independent list, if they do it as I would), and a district vote for a candidate in the voter's district election.
Say the Greens get 10% of the at-large vote. That means they are entitled to 10% of the seats, system-wide (the "system" being a Land in Germany). Say it's a 100 seat election, then the Greens are entitled to 10 seats. Now say they've won 3 district seats. That means they're awarded 7 at-large seats to fill from their list, to top them up to the 10 seats to which they're entitled.
Of course candidates could be elected to these listes either in primary elections or in an open list election, as I've described elsewhere in this thread.
The German system really seems the best party list sytem system for a country like the U.S., the U.K., or Canada, which is accustomed to single-member districts, because the German system combines single-member districts with systemwide PR.
It's a fallacy when people say that list PR doesn't let you vote for indiduals. The fact is that an independent has a much better chance with list PR than he/she does in the existing system, if the PR system lets people voe vote for an independent. That's because he/she can get elected with just 1/N of the vote in an N-seat election. Also, I've spoken of how voters can control whom they're electing, even in party list elections.
But if voters are still prejudiced against party list PR, then stv would have to be used instead. In its usual implementation stv uses small (3-7 seats) multimember districts. This isn't so good for really small parties, but if voters insist on stv, then it's unavoidable.
Almost unavoidable. Because stv can be the at-large system in an additional member system which could combine systemwide PR with single-member districts. It wouldn't be done like the German system though. Some of the seats in the system would be at large seats (say half of them), and the rest would be district seats. As before, you cast a district vote & an at-large vote, but this time your at-large vote consistes of an stv ranking of at-large candidates. If your district candidate doesn't win, then your "wasted vote" is transferred to the at-large level to follow your stv ranking, as in an ordinary stv election.
Perfect proportionality between district-winning parties & non-district-winning parties isn't automatic, as it is in the German system, but it can be achieved if the percent of the seats that are distict seats is equal to the percent of the total vote expected to be received by district-winning candidates.
But even rough at-large proportionality is better than none, and this percent can be estimated accurately enough to make the proportionality pretty good. It turns out that errors in that estimate don't put the proportionality off much.
This kind of additional member system in which wasted votes are transferred to the at-large level can be called a "transfer system". Several countries use a transfer system (but with party list instead of stv as the at-large system). The German system is a "topping-up" system as opposed to a transfer system. So then, transfer systems aren't new, and stv isn't new, and there's no reason why the 2 couldn't be combined, as I described above, if people insist of single- member districts and stv. You can have both, as I've described.
Though small multimember districts wouldn't be good for really small parties, any of these systems would be much better than what we have now (no PR). To review the alternatives:
1. Whole country or province or state 1 big at-large system.
2. Smaller simple districts
3. Additional member systems
a. German system (a topping-up system)
b. transfer systems
Except for the German system, any of these could use STV or party list PR.
Of these possibilities, the German system; or stv in small simple districts; or a transfer system with stv seem the best possibilities depending on what people want, what is most acceptable to voters.
A transfer system with stv has no precedent, and combines the complexity of stv with that of an additional member system, so it probably wouldn't be the best proposal unless voters insisted on single-member districts & stv (in which case stv transfer would be the only answer), or unless you want stv, and you don't want small parties hindered by small simple districts.
So those are the PR district systems. As I said concerning the issue of how to elect candidates to the lists, the choice of district systems, the choice between the abovementioned 3 best choices among PR systems (German system; stv in small simple districts; transfer system with stv) should be made entirely on the basis of which proposal would have the best publilc acceptance, and therefore the best chance of winning. This of course should be determined by polling before choosing a proposal.