Laser Direct Imaging: A User's Perspective


LDI is a process of imaging printed circuit boards directly without the use of a phototool or mask. The exposure of the photo-sensitive resist is done using a laser beam that is scanned across the panel surface and switched on and off by means of a computer control system. The laser used in this process is often assumed to be in the UV spectrum, as this tends to suit most of the commonly available photoresists. However, systems exist that operate in both the visible and infrared spectrums, working with specially formulated photoresists. Papers have been written on both alternatives over the years, but it is the UV-based systems that interest most of the PCB manufacturers as it offers the opportunity to be a direct plug-in replacement into the "yellow room" environment. Consequently, we concentrate here to our experiences with a UV laser system.




LDI systems first started to find there way into the printed circuit manufacturing arena in the late 1980s (Ref. 1). These early systems were much slower than the conventional contact printing, and it has only been in the last few years with a new generation of faster LDI systems that the process has become a viable threat to contact printing with a phototool. With the availability of higher-powered lasers, giving in the order of 4 watts at the work piece, and with the introduction of a new generation of photoresists requiring exposure levels in the order of 8-10 mJ/sq.cm along with faster computers for data rasterization, it is now possible to get LDI exposure and handling times down below 30 seconds per side for an 18 x 24" panel.


Benefits of Laser Imaging

With imaging times now getting close to that of Contact Printing, the use of LDI now offers the manufacture an alternative. With the current LDI systems, using high speed Dry Film resists, productivity rates of 60 double-sided panels per hour are not unreasonable. It is now that the system can start to be viewed as a production tool and the benefits given serious thought. Much has been said about the lower cost of production with the use of LDI. It is not too difficult to see how we may reduce the cost of manufacture from any one or all of the following:
In addition benefits are gained from potential quality improvements, such as:
As well as the cost saving benefits and the potential yield improvements there are a number of technical advantages and it is these that we will discuss and explore. These benefits include:
Whilst to the board fabricator yield improvements and cost reductions in the manufacturing process are essential in an ever-increasing cost-conscious supply chain, it is the technical benefits that LDI offers that will give him the advantages in the future. This will be either by widening the operating window or allowing more advanced design rules to be used to provide higher density products.


Resolution

On installation, LDI will give all but the most advanced fabricators an immediate improvement in their dry film image resolution. The printing of fine lines and spaces of 50um and below starts to become a reality, without the risk of low yields due to printed and film related defects. With process optimization, it should be possible to produce fine lines in the order of 35um in a 40um resist. With the predicted future LDI system improvements, 25um lines and spaces will become a reality.


Figure 2.


Figure 2 shows a typical result of fine line printing using a standard 5 kW UV light source and silver halide films. It is easy to see the track widening that occurs and the under-developed gaps.


Figure 3.


In Figure 3, however, it shows an example where it has been possible to resolve the lines and spaces but an area of local track widening has caused a print defect. Also visible is the loss of line clarity and the tendency for the track width to vary slightly.


Figure 4.


Compare this with the image shown in Figure 4. Here 50um tracks and gaps are shown having been printed on the LDI system. Here in the vertical, horizontal and diagonal axes, the image resolution appears sharp with straight sidewalls and good development of the gaps.


Registration

Another major benefit that LDI can give us is in improved registration. Improvements are achieved by eliminating the phototool, which has always given rise to issues, especially as the tools move anisotropically with temperature and humidity changes (Ref. 2). This coupled with the tendency for contact printers to run hot, the alignment of film and panel has been an ongoing challenge. With LDI, however, it is possible to use a CCD camera system and target fiducials on the panel to align the print image and panel. It is also possible to use these target positions to calculate any panel or drilling movement, and scale the rasterized print image so that we improve the registration and achieve best fit.


Figure 5.


The LDI systems are able to use a variety of fiducials to allow registration and scaling. Figure 5 shows some that may be used, such as a matrix of mechanically drilled through holes, a matrix of laser ablated microvias or a laser drill alignment fiducial on layer X-1. In addition, the CCD camera system must be flexible and able to recognise target fiducials that may vary in the way they are prepared. For example the target may be electroless copper plated, electroplated copper, coated with direct metalization, brushed, micro-etched or pumice-scrubbed. To add further variability, the copper may have tarnished to some degree. To the CCD camera, all these finishes have a potentially different appearance, and are likely to need a different algorithm to enable them to all be recognized, so it may be necessary to set these variables up in the early stages.


Figure 6.


Figure 6 shows some of the range of target fiducials that may be experienced following electroplating with copper and some with direct metalization.


Why Registration and Scaling Errors Exist

As mentioned before, the use of phototools with all their potential inaccuracies makes the alignment of the phototool and the panel a challenge for PCBs with tight tolerance specifications. Below, we have elaborated on the potential causes of errors in the image creation process, and tried to analyze the potential scale of the problem for a board with 500 mm X and Y dimensions.


Temperature and Humidity Stability of the Plotting and Imaging Rooms

In every printed circuit facility, the photoplotting and imaging areas are key areas where temperature and humidity control are considered very important. The typical specification of these rooms is a temperature of 22 degrees C with a variation of I2 degrees C and a relative humidity of 50% with a variation of I5% RH. There are three questions to consider. First, do these areas really stay within their specifications all of the time? If not, what are the lengths and magnitudes of the excursions? Finally, what exactly is the scale of the effect of these variations on the phototools and the base substrates themselves?

With respect to the maintenance of specifications, every time the humidifiers inject humidity, there is an excursion. When one looks at graphs of temperature in a clean room, which has a heavy heat load from printers, there are always excursions from the specifications. Yes, the basic specifications are met most of the time, but certainly not all of the time.



Photographic Film is Anisotropic

Although manufacturers try to eliminate this anisotropy, the fact that the carrier sheet of the gelatine is an extruded sheet means that there are different stresses in the axial and transverse directions. This means that the reaction of the film to temperature and humidity is different in size and timing in each direction. Simply put, for tight tolerance PCBs, using the same scaling in both directions could lead to inaccuracies.


Photographic Film Changes Size with Temperature and Humidity

The specification for temperature sensitivity for photographic film is about 11 ppm per degrees centigrade. Thus, for a variation of 4 degrees C (I2?C) on a 500 mm panel, this gives a total potential variation of 22 microns.

Sensitivity of film to humidity variations is about 12 ppm per % relative humidity. Doing the same type of calculation as before, for a variation of 10% RH, gives a total potential figure of 60 microns.



Base Material Changes Size with Temperature and Humidity

Although it is not commonly looked at since it is assumed that FR-4 only changes size slowly with temperature and humidity, there are circumstances where perhaps it should not be ignored. For instance, the difference in the temperature of the panel at different stages in the manufacturing process will probably be different than that in the cleanroom. Therefore, if the material has reached an equilibrium temperature in any of these steps which affect registration, e.g., imaging and drilling, there will be a shift between the image and the drill pattern.

FR-4 has a temperature sensitivity of about 15 ppm per degree C. Going through the same calculations and assuming only a maximum of 4 degrees C difference in temperatures will give a total potential variation of 30 microns. We have not estimated the effect of humidity as its effect is both smaller and appears to takes longer.



Other Potential Sources of Registration and Scaling Errors

There are other potential sources of variations in size of PCBs, most particularly in bonded multilayer panels. Typically in MLB manufacturing, the straightforward variation is size between one panel and another from the same lot after bonding can be up to 100 ppm. This would translate into a maximum potential error of 50 microns.

Non-linear variations can be considerably larger depending on the materials used to build the MLB and the types of multilayer presses used. Figures discussed in the industry put this at up to 500 ppm.

Drilling will also add some variability to the process. Drilling machines in themselves have an inherent accuracy. However, panels which are at the bottom of a stack might suffer from random drill wander of up to I25 microns. This would give a total estimated potential variation of 50 microns.



What Does this Mean for Production?



If you look at all the potential variations as are shown in the chart below, it looks as though it should be impossible to make a high tolerance circuit board.


Of course, this does not happen for several reasons. First, the differences are never the maximum for every variable. Second, the inaccuracies are usually random in the positive and negative directions and so would rarely, if ever, all add up to the sum of all the values listed above. However, what is does mean is that at the imaging stage of the production process, especially for outer layers of MLBs, each panel in a lot has a variation in size which can be quite significant, and which is not the same in the X and Y directions. Therefore, if each panel of a lot--of a tight tolerance MLB--is imaged with the same scaling factor (even if different in X and Y), there is a high probability that there will be unacceptable hole break-out and levels of rejects.

This is borne out by the experience of PCB manufacturers who have always produced difficult PCBs on small panel sizes in order to minimize the problem of registration and scaling. It is also clearly shown in the section "Using LDI In Manufacture" that results for the LIDCAT project, where the tight tolerances on the annular rings of the microvias would have been impossible to achieve with high yields without the ability to differentially scale for each panel in X and Y.



System Evaluation

Figure 7.


For the evaluation of the LDI system and verification of the CCD camera registration system, a special test panel was constructed in a 1+2+1 build using an FR-4 laminate core with Resin Coated Copper (RCCT) bonded to each side. Two panel formats were produced, 406 x 508 mm and 457 x 610 mm. These were then mechanically drilled and laser ablated as shown in Figure 7. The panel layout was such that the features within the panel could also be used as secondary alignment targets to ensure plenty of opportunity to look at target variation if required.


Once drilled, the panels were then processed through the direct metalization process, and half the panels were held at this stage. The remainder were panel-plated with electrolytic copper. All the panels were then prepared and coated with a suitable dry film photoresist specifically developed for laser imaging.

Next, a series of tests, detailed below, were conducted to check the performance of the CCD camera system and its operation in conjunction with the laser imaging system.
1) CCD Image Processing Repeatability. With the CCD camera in a fixed location, the system was set up to repeatability acquire the target image and calculate the target center. This was designed to prove the accuracy of the CCD systems in finding the target center.

2) CCD Image Processing Repeatability (Production Targets). This test involved the location and analysis of targets produced in normal manufacture, and was designed to look at the system accuracy when dealing with process variation as would be expected in the normal course of manufacture.

3) W/CCD Repeatability (Imaging Panels). This test looked at the system printing accuracy when repeatedly printing the same panel without removing the panel. This test is looking at the accuracy of the system, ability to locate the target centres and print an image in register.

4) W/CCD Reproducibility (Imaging Panels). Here we looked at the repeated imaging of panels but the panel was removed and reloaded between each print. This was designed to measure the system accuracy in a simulated production situation.



System Evaluation Results

  • CCD Image Processing Repeatability < 0.9 um

  • CCD Image Processing Repeatability (Production Targets)
    - PTH 3 x 3 Matrix - <1.8 um (Direct Metalisation)

    - PTH 3 x 3 Matrix - <1.7 um (Plated)

    - uVia 5 x 5 Matrix - <2.3 um (Direct Metalisation)

    - uVia 5 x 5 Matrix - <12.1 um (Plated) *

  • W/CCD Repeatability (Imaging Panels)
    - PTH 3 x 3 Matrix - <6.0 um (Direct Metalisation)
    - PTH 3 x 3 Matrix - <9.0 um (Plated)
    - uVia 5 x 5 Matrix - <6.0 um (Direct Metalisation)
    - uVia 5 x 5 Matrix - <6.0 um (Plated)

  • W/CCD Reproducibility (Imaging Panels)
    - PTH 3 x 3 Matrix - <6.0 um (Direct Metalisation)
    - PTH 3 x 3 Matrix - <6.0 um (Plated)
    - uVia 5 x 5 Matrix - <6.0 um (Direct Metalisation)
    - uVia 5 x 5 Matrix - <12.0 um (Plated)

    * Includes some untypical results where the system failed to find target center.



  • Results Summary

  • W/CCD Reproducibility Accuracy (Capability with Scaling)
    - PTH 3 x 3 Matrix (Direct Metalisation) -Inside 75um annular ring
    - PTH 3 x 3 Matrix (Plated) -Inside 75um annular ring
    - uVia 5 x 5 Matrix (Direct Metalisation) -~ Inside 50um annular ring
    - uVia 5 x 5 Matrix (Plated) -Inside 50um annular ring


  • From the tests conducted in the system evaluation, it is possible to achieve a "no break-out" condition on a production panel with a 75 um annular ring for mechanically drilled plated-through holes, and a no break-out condition with a 50 um annular ring with laser ablated microvias, when using the CCD camera registration and scaling feature. During the trials, some issues arose with the CCD camera system failing to find the true center of some panel-plated microvia targets. This accounted for the non-typical results included in the CCD image processing repeatability (production targets) results. Following the trial, modifications to the algorithms were made to aid the plated target location to resolve this problem.


    Using LDI in Manufacture

    Following the system evaluation, the LDI system was then tested on a simulated production run. Here a test panel for a fifth framework RTD project, IST-1999-10852, LIDCAT (www.lidcat.com) was selected. The test panel was designed as a high density microvia design on a panel 405 x 508 mm of a 1+4+1 construction. The individual test structures are manufactured 4 up in a subpanel, with 6 subpanels per manufacturing panel, giving a total of 24 test structure panels per manufacturing panels. The structures within the test panel are as follows:




    The design rules used in this board were as follows:


    A trial batch of these boards was then run through production, both inner and outer layers being printed on the LDI system. Inner layer yields at AOI were over 99%, but the design rules for these layers were not challenging. Batch 1 yields at electrical test, reflecting the full manufacturing process, was 28%. Following examination of the defects, fine tuning of the laser ablation of the high aspect ratio holes, plating and inner layer scaling was conducted before the release of Batch 2. The inner layer yield of Batch 2 again was high, over 99%. At electrical test, the full process yield was 96%.

    Inspection of the finished panels after LDI showed excellent registration. Even where only 25um annular ring existed, it was possible to achieve a "no break out" condition. This can be clearly seen in Figure 8. The good registration and the image quality of the outer layer print were major factors in the high yield of Batch 2. The yield figures do give an indication of the primary image quality, but do not indicate the challenges that were faced in achieving a good soldermask registration. Here, due to the tight registration and the fact the primary image was scaled, it is fair to say that the soldermask print was difficult at best. If we are to get the best from our LDI, we will need laser imageable soldermasks as well as primary image dry films.



    Implications for HDI MLB Multilayer Board Primary and Soldermask Imaging

    The results show that as the tolerances on annular rings and positioning accuracy of HDI panels become ever tighter to allow for greater wiring density and, as the number of HDI layers with their higher inherent movement increase, the necessity to allow for this panel movement with some form of registration and scaling becomes imperative. LDI should be the leading candidate for this application based on the results shown in this article.

    We would also like to touch on the potential need for LDI of soldermasks. There are two driving forces. The first is the same as for LDI of primary images, and if the primary image is created by LDI, it might be essential to use LDI for the soldermask image in order to maintain registration accuracy. This was the conclusion from the LIDCAT test boards discussed above as well. The other driving force is the demand from electronic OEMs for smaller tolerances for soldermask positioning around features on the PCB.

    The convergence of these two forces has led several soldermask manufacturers including Taiyo, Coates and Vantico to develop soldermasks with photo sensitivities in the region of 30-50 mJ/sq.cm. Although these will take about one minute per side to image with today's LDI systems, they can confer great advantages for the user. Also, the alternative of imaging several films for soldermask imaging and trying to optimize the registration on each panel take a lot of time and effort, while the net time taken may not be less.



    Conclusion

    We have seen that LDI has now started to reach maturity as a process for the manufacture of PCBs. It is now possible to get productivity rates in the same order as that of contact printing. With potential cost savings and yield improvements over contact printing, LDI has shown it is now a viable alternative. Whilst these benefits are important to the PCB fabricator, the potential benefits in resolution and registration offer most to the advancement in printed circuit technology.

    In the system evaluation, we were able to predict that we would be able to achieve a "no break-out" condition on laser drilled microvias when we had a 50um annular ring, but in production trials we were able to manufacture product with an annular ring down to 25um in localized areas. With this level of improvement in registration, it will not be far into the future that we will be able to start to change design rules and reduce pad sizes.

    A reduction in the pad size or annular ring coupled with the use of LDI could then offer the PCB fabricator a number of benefits:

    For the end-user and designer, there are other potential benefits.

    Better registration could give rise to smaller pads, finer tracks and spaces, and higher density products.

    Clearly, LDI will give us better control of our process with a much wider operating window. To get full benefit of the process however we must wait for the fully commercialized LDI soldermasks to become available.



    References
    1. "Laser Direct Imaging: A Cost-Effective Choice," Dave Stone, Barco Belgium. Proceedings EIPC Summer Conference 2001.


    2. "Laser Direct Imaging: A Larger Volume-Users View," Steve Jones, Viasystems (Tyneside, U.K.). Proceedings EPC Show Maastricht 2000.