Also, what's the best way of finding a collection of sources to form the foundation of a dissertation?

The method I like is to find the big sources for your general topic. For the Spanish Civil War you'd grab Beevor, Preston, Thomas, and Payne. Let's say the Battle of Guadalajara jumps out at you in your first read-through of Beevor. Pay attention to his footnotes on that section. Find them yourself. Now do the same with the other three authors, reading their sections on the battle and mining their footnotes. You should probably have a handful of new sources (or even a bunch of new sources) to find. You should also have a bunch of names of people involved in the battle as well. Do any of them have memoirs or biographies? What about the groups they were members of? Look for newspaper mentions of the battle and the people, too.

Basically, each source has threads that connect them to other sources. Footnotes and bibliographies are the easy targets for you to go after. Other threads can be found by following names and groups. Keep pulling threads until you have the sources you want (or perhaps more importantly for an undergraduate, until you have the sources your professor wants you to have).

"For starters, how 'original' does it have to be?"

Doesn't have to be original at all I don't think. I once had it described to me that undergrads cover existing work, a masters takes you to the border of current knowledge and doctorates are supposed to cover new ground. I personally did my dissertation on a subject that was relatively untrodden. The upside was that it was interesting and stood out from the others, the downside was finding enough material and sources. Always take into account how much material you can get your hands on and which dissertation supervisor you would like before you make your choice. I went for a topic with few sources, but had a great supervisor, so it turned out alright.

The head of my undergraduate department explicitly said to everyone writing theses, "It doesn't have to be totally new, but it has to be new for you." I think that's general good advice for undergrad theses.

I ended up going to her with a list of like 12 different topics (I only remember a couple), and having her help me pick which ones were actually feasible to do in the amount of time I had. We all used primary sources, but they were very different sets of primary sources. One woman basically just used Eusebius's Life of Constantine and one or two other primary sources (plus a lot of secondary literature) to discuss the Christian elements of Constantine's death . One woman wrote about the Book of Job. One woman used ethnographic accounts to try to figure out pre-contact Native Alaskan rituals. My primary sources were encyclopedia articles (I was tracking how rosaries were discussed between the 18th century and now to measure anti-Catholicism as based on existing theories). There were several others. Some of them could have been published with some work, some of them probably weren't that original in a global sense. The one who wrote on Constantine's death won our department's AB thesis award, and she had very little original in her primary sources, but did a great job dealing with a vast secondary literature (since this is a very well trod area) and making an original, synthetic argument (I think at least--it was almost a decade ago; she definitely won the award).

If you insist on looking at new primary sources, are you at a research university? See what's in their special collections. See what's on microfilm and microfiche. One of my colleagues did his PhD dissertation on life in late Ottoman Anatolia and the majority of his primary sources were missionary documents (from the ABCFM). There are extensive documents on everywhere they had missions and these collections are widely available on either microfilm or microfiche (I forget which). There are tons of primary source collections like that. Newspaper reports (your library presumably has access to pro-quest, etc) are another great way to study events. Anne Helen Peterson has done very interesting cultural history, for example, on early Hollywood gossip magazines. Local historical societies and the town library also certainly have primary source documents if you want to write local histories. I worked in my town archives and discovered how my home town finally got its two traffic lights, for example, or how the big families in town left their mark by naming lots of things. Great social history has just been done using local county probate records. I'm not a narrative historian so I don't fetishize finding new documents in quite the same way as my colleagues (the historian who used the ABCFM documents was married to a fellow sociologist, and she said to me, after he was bummed out at not finding anything good after a year of searching in the Ottoman Archives, "I don't get why historians are so obsessed with finding new documents that no historian has ever looked at before. As if there's only one way to understand a document.") There are also extensive letter collections (I did a literary history primarily using Tolkein's letters at one point), as well as translated collections of primary sources (all the early non-Muslim accounts of the initial Arab Conquests are collected in Hoyland's Seeing Islam as Others Saw It).

There are a lot of things. I would suggest coming up with a list of topics that could interest you, rather than looking at what documents you might have access to (but by all means, look what's easily available in your university's library, your town, your county, etc).