You definitely need to have live demos of each system, using your specific assembly if you want to compare apples to apples.
Each of the AOI OEM's you mentioned have plus amd minus to their unique systems. You will need to evaluate your specific needs, what attributes are super critical to your success and any compromises you are willing to accept.
Are you price sensitive first? Or do the actual technology and end results become primary considerations.
It's been nearly 3 years since I've conducted an in-depth study of AOI so I cannot advise on current offerings. I will say that I was most impressed by MVT at that time. I liked the 3 color LED array they used for a lighting source. Others use gray scale imaging, which in my opinion was limiting and caused too much operator/engineer intervention to make a go/ no go decision on the spot. I understand that MVT has since been purchased by a large organization and some of the desirable flexibility in custom tailoring the application software may not now be available.
What will be your intended use of the AOI? will you conduct post print inspections? pre-reflow inspections for correct part, orientation, alignment, presence/abscence?
Will you only do post reflow inspections for solder quality?
Consider that actual use first. Then find the best machine to serve your needs at the price you can afford.
In the past year I was responsible for developing and leading a Team focused on AOI, Evaluation, Justification, and with a full ROI required, supporting a HMLV manufacturing environment.
If you have in these economic times the additional resources to go off and conduct such a study..I highly recommend you do so. If not I'll share some of the highlights of our evaluation which can possibly help in your decision making. First we targeted a price range based on a budget amount of 140-160K. Several Key Suppliers were identified per our initial quoting activities and supplier response, then invited to face off in an 8-12 week eval., Cyber, V.I. Technologies, and CR Technologies (now Photon) were invited to the show. The measurable were identified and an extremely complex assembly was delivered to each Supplier, assembly had 48,000 opportunities. Multiple BGA's, 0402, 16mil., 0603 resistor networks, and more. Big board 18 X 21. These were the types of assemblies produced on a daily basis.
One of the main AOI deliverables among many was the ability to generate an NPI program that would require minimal first pass adjustments, with little impact to production requirements and scheduling, high capture rate, and low false calls, flexible quick turn support. A machine able to significantly lower the first article inspection times, post reflow, with ability to identify solderabilty and joint integrity as well as absence, presence, polarities, and component identification. Limiting the down time associated with first article inspections, and offer ability to lower DPMO's through continuous process improvements and early defect detection.
I got ready to give you a brief a summary of each Suppliers performance/issues and thought it better to keep those details offline and simply tell you which supplier was chosen and a brief summary of why.
If you'd like to know more of the particulars of the evaluation you can give me a call.
The AOI Team identified CR Technologies as best of the bunch in the price range given. It performed well with the lowest programming times, lowest false call rates, and the ability to identify and confirm solder integrity of visible solder joints. User friendly software, color system. When we were in the final phases of evaluation, head to head, the CR identified an incorrect PBGA based on its optical recognition/identification capabilities, it identified 1 incorrect number out of as many as 30 characters in the components legend where as the competitor passed the assembly. The run was 15 boards, and it was an 8 time user. Imagine the rework and head-aches that would have occurred on that one.
The machine has been installed and in an offline production/NPI operation for 9 months will little or no issues. It helped signicantly reduce total DPMO's and increase first pass yields associated with NPI builds and production level builds. Technical support, and response time has room for some improvement. But, with an uptime of high 98% range, not many calls have been necessary.
I hope this has helped a little, if you'd like to understand more of the evaluation measurable's and how to kick off an effort like this...give me a shout.
PizzaFred, Scott mentioned lots of good ideas. Take one of your boards to each vendor and see what it takes to program the board how long it takes for it to become online. I've found some that are easy to program, but require 40 or more assemblies to be ran before false calls are reduced. If you could tell us a little about what you plans are for the machine (component checking soldering, polarity, etc) and at what location you plan on placing it (prereflow, post relfow, etc) and finally what type of a manufacture you are (low mix, high volume or high mix, low volume) As you are starting to find out, AOIs are like PnP machines, tey are geared towards certain types of companies.
Another issue to look at is the efforts that it takes to become familiar with the machine. Where I work we are also looking into AOI machines, and one of our requirements (based on how simple the vendors say their machines are) is that after no more than 40 hours of training and practicing, the average SMT production engineer must be able to make a good inspection program in a more or less efficient way. What I have seen so far is that programming an AIO system is much too complicated to meet this requirement, and that it actually takes a huge amount of expertise and training to obtain the necessary skill-level.
Daan has made a valid point with regards to expertise and use of a machine.
Yet I recognize that with advanced technology, trades must occur. A hghly sophisticated AOI unit can and will help reduce overall cost of production. Therefore, a need to have a dedicated programmer may not be as expensive as it seems in the long run. Just think of all the time and expense saved by not having a gang of inspectors and rework persons. Surely some of the "savings" could be allocated to support the salary of the dedicated engineer/programmer.
This would also free up the efforts of the manufacturing and process engineers that have enough to do already, let alone spend up to 40 hours to write a program.
An analogy from the medical industry comes to mind- for the daily aches, pains and illnesses we go to a GP - General Practitioner doctor. For more serious issues, there are specialists available.
Must all GP's have the complete knowledge of a specialist to effectively do their jobs?