Used Oxford Channel 5 HKL. It is effecient and easy to use.

It depends on the expectations you have. The software capabilities are certainly very rich at OIM of TSL. I only heard that speed and stability is not that good. Bruker is very easy to use, but the data processing is comparatively poor. However, if you can use MTEX (free Matlab toolbox) you have effectively the most powerful tools for data processing and analysis, it is under continuous development and you can directly ask the developer for help and specific tools. Oxfords Aztec is somehow in between. Great detector, quite simple operation, basic post-processing. The weak point is that the post processing is "effectively" still the same as about 5-10 years ago since still Channel5 is used. However, this is also comparable to EDAX TSL.

Bruker offers from the beginning as standard opportunity to save the raw patterns. I can only recommend this to do since it offers many opportunities to evaluate the indexing, phases, orientation solutions etc. I am asking you: Who is not saving the XRD-diffractograms or EDS-spectra and only uses the numbers derived from the raw data? This is nowadays not more state of the art only to save the "estimated" phase solutions and determined orientations. You can have a look at the linked presentation, e.g. what you can still do with the standard EBSD patterns after your EBSD mapping.

Finally, there is no straight answer. You have always to make compromises, but after some years I am still very satisfied with the combination Bruker aquisition and MTEX. If you don't want use MTEX , and you believe "black-box"-software, CrystAlign (Bruker) is not powerful enough to make you happy. If you are experienced enough and you understand all settings and parameters, use OIM of TSL. But you have to make compromises as well regarding the detectors sensitivity and software stability. Use Oxford, if you are a standard user and you don't want very specific data analysis. The detector is very good, and the software offers the basic tools, but not much more. Obviously, for many people this is OK.

Conference Paper EBSD: more reliable and precise

Hi, it would be more beneficial if you can provide more information about the material types you plan to analyze, which kind of analysis and information you want to extract from the raw data. As mentioned, there is no single correct answer.

I have 8+ years of experience with the TSL-OIM and 1.5+ years experience with the Oxford systems. I have no experience with Bruker.

Comparing TSL-OIM and Oxford in terms of:

software capabilities: TSL is much better, although the OIM software has many options, when you get used to it, you start to understand the basics better and extract more information. For Oxford, there is no single software (Tango, Mambo, Salsa), and for each analysis (mapping, discrete orienations or texture analysis) you have go back-forth between those software packages. Moreover, subset selection is cumbersome and the provided noise-reduction algorithms (i.e. wild spikes) can easily "polish" the data.  

stability: this also depends on the SEM, environment conditions etc. But if you compare the stability of software during data acqusition and post-processing both are very good, as long as you have a computer with recommended configuration. For my personal experience, Oxford software tend to crash slightly more, specifically during "texture-analysis" of larger datasets (1-2million points)

Performance: Both systems offer a slower but higher resolution camera, along with a very fast but lower resolution one. The specs are similar on the paper, but i think the Oxford camera is slightly better. Moreover, if you plan to invest in a system containing a EDS detector, Oxford's top of the range detectors have larger detector area (would be beneficial when you make EDS-EBSD combo scans)

For the indexing algorithms both systems have different approaches:

Oxford (mean angular deviation), TSL (confidence index based on triplet voting)

Oxford approach is advantageous for indexing of low symmetry (ex: triclinic) phases; whereas TSL approach is better for deconvolution of patterns. Moreover, TSL contains more options to tune the indexing algorithm. The options also contain a parameter equivalent to angular deviation. 

For short, in terms of hardware Oxford is slightly better, for data-collection and indexing software TSL is much better.

User interferance: Both parties have one "lighter, easier, for dummies type" sofware, good for beginners (TSL--> TEAM, Oxford AzTech). Also both parties have "advanced software" options for advanced users. I use the "advanced" options more frequently. 

If I were you, I would have gathered a few demo samples, and ask the vendors or other users of those particular software+hardware, for a demonstration session. I think all the vendors or representatives, TSL, Oxford, Bruker would gladly schedule a demo-session for you. You may have a better idea, when you see the operation of these systems.

Happy researching !!!