As this question comes up quite frequently I thought it may be a good idea to summarize the advice we normally give to the customers looking to print their artwork onto canvas (though the majority of this information is common to most other media types, as well).
We need to start with the question of what type of artwork is to be printed. While preparing photos is relatively straightforward, artwork containing text, logos or other elements requires a different approach (especially when spot colours are involved).
We can define artwork as two types: 'bitmaps' and 'vectors'. Bitmaps are all photographs taken by digital cameras, scanned images and any artwork which has been saved (or exported) as a bitmap.
Bitmaps are created on a pixel grid and have a set size in pixels, for example: a photo taken by an 8-Megapixel camera will have approx. 3264x2448 pixels. You can imagine a grid made out of pixels where each pixel can have a different colour. The amount of available colours depends on the bit depth - most commonly it will be 8 bits, which allows 256 shades of Red, 256 shades of Green and 256 shades of Blue. This gives a total of 16.8 million available colours.
When it comes to print a bitmap, it's size in pixels determines the quality. Let's say we want to print this bitmap onto canvas. What will be the effective resolution of the print? What is the maximum size that we can print it at? It all depends on the size of the print - If we print it as 32x24" then from simple maths we will have approx. 100 pixels per inch. (3264/32 =102). If we decide to print it at half that size - 16x12", the file will have an effective resolution of 200 pixels per inch. Having a higher resolution doesn't pose a problem, but there is no point to create files with effective resolution higher than 360 ppi. It is usually problematic when we want to print a small file on a big scale - if we double the size from above and go for 64x48", the effective resolution of our sample file will only be 50 ppi, which is too low because the print will appear pixelated.
What are the remedies to that? There isn't really a magic bullet which will add quality to an image (apart from movies where the smallest details are extracted from grainy CCTV footage ;) if the magnification required is very big. With files around 70-100 dpi there is usually sufficient information within the file to increase the size and effective resolution. Canvas is a very forgiving media and depending on the overall quality of the photo, (i.e. a photo from a DSLR will be a much better candidate for increasing the pixel size than a photo from a phone camera) it allows for files with a relatively low original resolution to be upscaled and successfully printed.
Upscaling can be performed using Photoshop's 'image resize' window or using designated plugins - our favourite being Perfect Resize 8 from OnOneSoftware - website link. Generally these tools offer a more automated process. While manual resizing may need to be done in smaller steps and additional sharpening is usually required, the well rehearsed algorithms in these plugins can perform these actions instead of resizing by a lengthy manual process. Another take on a large print's resolution/quality is that the 'appreciation distance' is usually quite substantial and the imperfections are not visible from a distance anyway. From my experience, most customers will examine the print thoroughly, and usually don't appreciate this argument; but most banners, adverts we see everyday are printed way below 100 ppi, some even as low as 30 ppi. (A further consideration may be the workable file size when producing artwork at a really large size).
So to summarize, with bitmap images the quality is determined by the size they were created at with a bit of a scope for upscaling if necessary.
Most common bitmap types:
.jpg - (Joint Photographic Experts Group) - most common format used today, the majority of digital cameras will produce files in this format by default. It supports 8 bit as well as 24 bit colour-spaces. Most important from our perspective is that this is a lossy format, meaning that each time the file is opened and saved again it will suffer further quality loss. Original files saved by the camera are fine to work with for very good results. If any adjustments need to be done to the file those should be performed and the file then saved as .tiff or .psd format.
.tiff. - (Tagged Image File Format) - format used by the majority of photo editing software. Can be lossless or lossy (lossless is what should be used with intention to print). A certain degree of compression is available using the LZW method, which doesn't affect quality in any way. This is the format in which all customer's files are saved once prepared for printing.
.psd format - Photoshop's native format which keeps all data editable, i.e. text, layers, masks etc. In recent years the .tiff format is becoming far more superior than psd, because of better compression methods and greater flexibility (many applications that can not open .psd files will easily read .tiffs).
.gif (Graphic Interchange Format) - very limited amount of colours available (256), designed for simple illustrations. Not really a candidate for a printing file.
.png (portable Network Graphics) - despite lower popularity than jpeg, this format may offer smaller files sizes in certain cases, and the compression used is lossless. The aim of this format is to create good quality files for online applications but again, not really to be used for printing.
.bmp - is a Windows format - it is uncompressed and therefore rarely used. It is becoming more of a thing of the past.
Vector images on the other hand offer much better capabilities. They are created using vector application such as: Adobe Illustrator, Corel Draw, Autocad, Freehand etc. Because the artwork is more or less a combination of mathematical formulas, it can be scaled to any size without any loss of quality. It also gives much better control over colours used. Most companie's logos will be created in this way - it is very easy to manipulate them; printing on a business card as well as massive adverts without a problem. Despite this, the file sizes are very small.
Vector files can come in a handful of formats, of which the most popular is .pdf. Not only it is completely independent of the operating system, but also the hardware used. It can carry a vector graphic object without any loss of editing capabilities. However - if the .pdf has not been created in the right way the artwork may be an almost exact replica of a bitmap file and have a similar size. The easiest way to check whether the artwork within a .pdf file is a vector object is by simply zooming in - vector files magnified even thousands of times will still show nice and crisp edges and print perfectly clear.
To summarize - if the artwork we intend to print exists in a vector format, that is the best way to do it. All documents containing text will look best if used in a properly produced .pdf format file.
Any photographs taken by digital cameras will come in a variant of the RGB format, most commonly sRGB. A slightly larger colour space that we normally use (unless supplied files are in ProPhoto RGB) is Adobe RGB (1998). If your photos are already in one of those colour spaces there is no need to do anything.
CMYK format has been designed with offset presses in mind - it is narrower than RGB and converting a photograph containing vivid colours to it will result in a smaller amount of colours available. All of our printers have a much wider gamut than CMYK - they use up to 10 colours and converting to CMYK basically results in lower quality reproduction.
With regards to companies logos we utilize rips which ensure that specific colours are printed to match to specific pantone colours.