Note 23:

How Brazilian pronunciation differs from that in Portugal

Because I use the pronunciation in Portugal as the starting point here, it may look as if this (incomplete!) list of pronunciation habits in Brazil gives deviations from the Portuguese style. But in fact, historically speaking, some or many of the features of Brazilian pronunciation are probably older and more original, and it is the pronunciation in Portugal that has changed in recent times. But I don't know enough about this to be more specific.
Most people in Brazil use [s] for the written s at the end of a word or before a consonant. Likewise, final /z/ is [s] in that pronunciation style.
It is my impression that this only, or more often, happens at word end, before a pause. Before another consonsant (not being s), the sound does often sound as [S] or [Z], like in Portugal.
When a final s and an initial s meet (usually between two words, but it may even happen within a word: nascer, descer, and their conjugations) they merge into a single [s] in Brazil, but stay separate (i.e. [Ss]) in Portugal. As a consequence, Brazilian Portuguese is not as "essy" as Iberian Portuguese.

In the states Rio de Janeiro (and according to Standaard pt>nl Dictionary) also Maranhão and Paraná), the same system as in Portugal is used: [S] at the end of a syllable.
In large parts of Brazil (Paraná, São Paulo (?) Rio de Janeiro, Minais Gerais, north-east, but not Caipira and Rio Grande do Sul) the /t/ and /d/ before a high front (/i/) or high central (/1/) vowel (but these are both [i] in Brazil), are palatalised. This means that they become [t'] and [d'], or maybe [tS] and [dZ]. This palatalisation tends to happen also with /l/ before such vowels.
Examples: aquela tarde ^©, gente, triste, grande ^©, um gole ^©.
In Brazil, the high schwa /1/ does not exist. It merges with /i/ when word-final, and with /e/ elsewhere. Cf. triste ^©, in Portugal and triste ^© in Brazil.
Perhaps this difference is due to French influence in Portugal? This page suggests such influence in the 18th century, but doesn't get specific. I’d certainly like to know more.

The (non-nasalised) low schwa /3/ is absent in Brazil too. It merges with low (open) a (/a/). But /3~/ does occur in Brazil too, just listen to this: dança, grande, criança ^©.
I know nothing about the history of this change, but my guess is that /3/ and /a/ once were a single phoneme in Portugal too, and that they got a distinct sound depending on being stressed ([a]) or not ([3]). A closing influence of subsequent nasal consonants then produced a stresses /3/ too, but not always (difference between falamos and falámos), which produced minimal pairs, and hence two phonemes.
In Brazil, the uvular r is often not a flap or a roll, but rather a rasping, fricative sound like a [X], or a glottal sound like a rough voiceless [h]. Examples: ver Irene rir ^© and Irene ri ^©.
In parts of Brazil the single r in VrC en VrF positions (pre-consonantal and final) is uvular too, not lingual as in Portugal. This is the case in Rio de Janeiro, Minas Gerais, Bahia and Ceará, but not in more southern parts of the country.
Examples: mortalha ^©, aquela tarde ^©.
This never happens in Portugal, so it is one of the things that immediately betray a Brazilian accent.
In VrV (intervocalic) and CrV (prevocalic) position the sound is always lingual in Brazil too, just like in Portugal.
Brazilian Portuguese (pt-BR) probably distinguishes written éi and ei (but I have no proof of that), which in Portugal are usually both [3i]. Written ei sounds as [e:] in pt-BR, for example in the word capoeira [kapue:ra], which would be [k3pu3ir3] in Portugal.
Likewise, written final em and en don't become /3~i~/ as is often the case in Portugal, but they are consistently /e~/ regardless of position.
In fast talking, pt-BR tends to shorten consonants and vowels alike, so the length ratio between them stays roughly the same. Stressed vowels can become a lot longer than they normally are. This lengthening is also used as a pause for thought sometimes, like in Spanish and Italian.
Iberian Portuguese (pt-PT) speeds up by first reducing the length of unstressed vowels, making them almost disappear, and it then also shortens stressed vowels. Consonants are less affected. This can make the language very difficult to understand, because in extreme cases it changes into a seemingly endless stream of consonants, nearly without any vowels in between.
On the other hand, as if to compensate for this, the language (in Portugal and Brazil alike) is also characterised by frequent vowel and diphthong encounters, which often results in long stretches of consonantless speech.
The intonation of pt-BR is very different from that of pt-PT. I couldn’t describe any details, but in general pt-PT is much more level than pt-BR, which can have very large pitch changes.
In Caipira (Brazil), a retroflex sound (tongue curled up backwards) is used in VrC en VrF positions (sometimes also VrV).
Examples: vortá (dialectal variant of voltar) ^©, and porém, apartou ^©,
The fricative or approximant allophones of /g/, /d/ and /b/ are not used in Brazil.
For lexical differences (which are off-topic here), see this Small Portuguese-Brazilian Dictionary.

The following is a checklist to trace Brazilian accents geographically. It's probably way too simple, or even inaccurate, but it's as close as I can get:

Item	south	São Paulo	Rio de Janeiro	Rest / north	Portugal
syllable-final s	[s]	[s]	[S]	[s]	[S]
t before front vowel	[t]	[tS]	[tS]	[tS]	[t]
final r (or before other consonant)	[r]	[r]	[R\]	[R\]	[r]