Meteor's Jonathan Wilson reveals how to understand and address the issue in his talk 'Peeling the inkjet onion: Nozzle out detection and compensation' at the IJC 2018.
Peeling the Inkjet Onion: Demystifying Nozzle out detection and compensation
1. Peeling the Inkjet Onion
Jonathan Wilson, Meteor Inkjet
• Demystifying Nozzle out
detection and compensation
2. “The only simplicity to be trusted is the simplicity to be
found on the far side of complexity”
Alfred North Whitehead
3. 1. Peel back the layers of the Onion
2. Nozzle Failure
3. Defect Detection
4. Defect Correction
5. Nozzle Compensation
6. Identify how things should work
5. How can a
nozzle fail
without
being
detected
How can a
defect be
detected
without a
nozzle failing
The Onion’s Conundrum:
Conundrum: A question asked for amusement, typically
one with a pun in its answer; a riddle.
6. • We have established a benchmark of the
system.
• Failure has subsequently occurred.
Assumptions!
8. Visual
Detection
through some
optical sensor
or photometer
Electrical
Feedback
through
voltage
variations in
the piezo
Detecting Failure:
Where and how
Pressure
Wave
feedback
through the
fluid delivery
system
9. Visual
Detection
through some
optical sensor
or photometer
Electrical
Feedback
through
voltage
variations in
the piezo
Detecting Failure:
The practical reality
Pressure
Wave
feedback
through the
fluid delivery
system
10. • Vision inspection systems
– Efficient way to identify
defects
– A number of providers in
the market
– Can be retrofitted to
existing devices
– Can also identify tonal
shifts (critical for some
applications)
• Vision inspection systems
– Can be expensive
– You may need to create a
mechanism for adjustment if
you want near real time
correction
– There will be latency
between detection and
either pausing or correction
– Part of the post image
process
Detecting Failure:
The practical reality
11. • Electrical
– Fast detection & reaction time
– Can be incorporated into drive
electronics/print head
capabilities
– Can be monitored from the
machine UI/DFE
– Automated correction conditions
can be applied
– Part of the printing process
– Works with non-imaging
applications
• Electrical
– Detection within large head
arrays could be difficult if there
are extraneous factors
(electrical noise)
– Systems with mixed head
types increase complexity
– Not all head types will allow
this for this method of
detection.
– Environmental factors could
influence detection.
Detecting Failure:
The practical reality
16. Post Imaging
Process
Once defect is
detected rescreening
of the image can be
undertaken, this
screening will then be
applied for the duration
of the production cycle
Imaging Process
through voltage
variations in the piezo,
can has masks
applied within the
drive
electronics/dithering in
the head to use
adjacent nozzle to
compensate
Correcting Failure:
where and how
17. • Post Imaging Process
– Screening for individual colours
(white inks can be more
problematic)
– Ideal for addressing larger failure
modes
– Can take direct input from vision
systems
– Can take direct input from
Electrical systems
Correcting Failure:
where and how
• Post Imaging
Process
– Can take time causing
more media to be
imaged and increasing
waste.
– Can become
expensive if you need
to keep up with the
press
18. • Imaging process
– Only small adjustments
may be achievable
– Environmental factors
could contribute to
‘ghost’ defects
Correcting Failure:
The practical reality
• Imaging Process
– Near real time correction is
achievable
– Density based same plane
compensation, inter-plane
compensation,
– Disabling nozzles that are
problematic
– Grey level compensation
– Detection and failure are
addressed at one level
within the system
20. Defect
Detection
Nozzle
Compensation
Nozzle Failure
Defect
Correction
• It’s not easy!
• Failure and detection are not always
obvious
• Working at real time is expensive and
intensive
• Best approach is to use a combination
of image process and post process
technologies, where needed.
• Setting a benchmark or ‘golden state’
for each machine before shipping is a
great way to ensure you know when
something has changed.
Inkjet is nothing if not complex, we need an NDA to just talk to each other, let alone work collaboratively together on complex issues facing us in the adoption of the technology. One such issue is the our Onion, the nozzle failure on a print head, inside a printing system.
Welcome to my attempt to demystify the complexity that surrounds one of the hottest topics in our industry today. Nozzle failure, the subsequent detection and compensation of the given failure.
I will start by peeling back the layers of the onion, to try and understand what technologies can be used and why, how they interact with each other and what this means for systems builders.
The first layer we see if the most obvious, nozzle failure. When do we know if a nozzle has failed? This sounds obvious, however, certain printing processes can hide failure. Of course, we must ask the question, do we care?
For some printing processes nozzle failure might not be so significant, scanning printers doing multiple passes can easily hide the issue, however, there are printing processes where failure might not be so obvious but the failure might be just as critical, an example could be for printed electrons or 3D printing (powders) where the internal structure is critical.
Defect detection is the second layer of the onion, assuming we can actually detect a failure then we must have a process that can identify it. During this talk I would like to introduce three ways that can potentially be used for defect detection, we will take brief look at each and look at the pro’s and con’s of each.
Given the nature of inkjet and the complexity of print heads and drive electronics it is possible to detect a failure before the defect is detected, feedback from the Piezo via the drive electronics and print head. I will shortly introduce the conundrum that is associated with the onion.
The third layer of the onion is defect correction, here we will find the mechanisms to adjust for failure, we will look at thee different types of adjustment, where the adjustment takes place and what the associated benefits and impacts each can have.
The fourth layer is nozzle compensation, this, in reality, is the output from defect correction, here we have compensated for defects. The key to note here is that this final, but crucial layer of the onion is actually the starting point for benchmarking the system. Why do I state this? To truly understand the failure we must first capture the initial state of a given system, nozzle compensation, or as we like to call it, MetFlat, this is one tool in our tool box to help is define the initial state of a print system, characterize it and adjust for sub-optimal nozzle states even before the device is in the field.
When is a defect not a nozzle failure? When can a nozzle fail without being detected?
This is the Onion’s conundrum, and as complex as the layers of the onion are, the answers are equally complex.
A nozzle can fail without being detected if the output being produced is non critical (flood coating), part of a scanning system, where the failure can be hidden in multiple passes, and more.
A defect can be detected due to interactions of the fluid on the substrate, Pooling, Chaining and Mottling are common in UV applications where the fluid can interact with the surface tension of the substrate and cause optical defects to appear.
Why am I stating the obvious? It should be a key part of any build process that a ‘golden’ state is defined, it will help track changes within the system and be the initiator for the subsequent process steps.
There are now a number of tools available for system builders to ‘profile’ the engines they build, tools like MetFlat from Meteor will create a profile of the printing system.
When failure has occurred there are a number of ways it can be detected, each one has it’s merits and its drawbacks. Only two are really applicable in real world systems today.
Visual: The use of optical sensors is not new, they are effective at spotting small non-uniform errors even at high speeds.
Nozzle status feedback from the head can be obtained via the electrical interface to the piezo drivers. This is achievable by detecting small electrical signals that are generated by the piezo after a drop is ejected from the nozzle.
Pressure wave feedback through the fluid system is possible but it’s not currently practical for real world applications. In this scenario you would need to know your flow rate up front and then have flow meters in the fluid delivery system, for applications where you are jetting 100% this becomes possible but anything else the changes in flow rate are too small to measure successfully.
For the purpose of the talk today, we will not go into the pressure wave approach, we will focus on what is commercially viable and relatively easy to implement.
Pressure wave feedback through the fluid system is possible but it’s not currently practical for real world applications. In this scenario you would need to know your flow rate up front and then have flow meters in the fluid delivery system, for applications where you are jetting 100% this becomes possible but anything else the changes in flow rate are too small to measure successfully.
For systems with a product detect signal, in other words, there is a gap between printing, this process can be very effective and can increase productivity if there is a correction routine in place.
For systems with a product detect signal, in other words, there is a gap between printing, this process can be very effective and can increase productivity if there is a correction routine in place.
The idea of using the same drop ejection piezoelectric to sense the status of the nozzle is over 35 years old. The first patent was filed in 1985.
Significant testing is being done on ow this can be implemented for real-world use, it is not easy, there are a number of variables that can inhibit the results.
Much more work still needs to be done in this area, it is something we, at Meteor are invested in.
Knowing there is a failure, understanding where the detection has taken place then gives up the prescribed response. The response is dictated by the location of the detection
This diagram has 4 axis, the location of observed failure at the top is where we establish the located of the detected failure (keep in mind we have already established that a we can see what appears to be a defect without having any nozzle failures).
On the left hand side we have detection that is post imaging process, this means the image is already on the substrate and failure has to be detected outside of the process, this is where line scan camera’s and photometers are additional processes to the system.
The second, lower half of the left side is where correction is taking place, again, this is outside of the imaging process, this could involve rescreening and further colour management of the image to correct for the detected defect. Everything on the left has side has a slower response time than those on the right, the response time could be microseconds up to many seconds, in a high speed single pass system this could lead to waste of materials that need to be re printed.
On the right side (the imaging process side) we are much closer to the detection and correction, feedback from the Piezo that exhibits a potential failure can be addressed in head/drive electronics at near real time speeds.
The practical reality is that at the bottom half of the axis both sides will, at one point need to work together, depending on how great the failure is, adjustments at the head/electronics will be a smaller scale that those that can be achieved by passing back through a rip, being rescreened and colour managed.
Defect correction is a mechanism that can be used at higher level (I will explain what I mean), if we assume that nozzle compensation is a process that can take place at an electronics level (fast response, possibly limited adjustments) the defect correction is a broader correction routine that works at a screening level, there is a degree of latency with this approach, from detection to correction output will be taking place, the broader approach can deal with larger errors than can potentially be addressed at the lower, nozzle compensation level.
The key is to have both detection and compensation running on the same system (there is a cost involved in this approach, you need either a line scanner or optical sensor and or a spectrophotometer, depending upon what it is you are looking to capture and adjust.
When failure has occurred there are a number of ways it can be detected, each one has it’s merits and its drawbacks.
Visual: The use of optical sensors is not new, they are effective at spotting small non-uniform errors even at high speeds.
Electrical feedback from the head via he electrical interface is achievable by detecting variations in the Piezo, as highlighted earlier, it is possible to detect and react to the failure.
Pressure wave feedback through the fluid system is possible but it’s not currently practical for real world applications. In this scenario you would need to know your flow rate up front and then have flow meters in the fluid delivery system, for applications where you are jetting 100% this becomes possible but anything else the changes in flow rate are too small to measure successfully.
When failure has occurred there are a number of ways it can be detected, each one has it’s merits and its drawbacks.
Visual: The use of optical sensors is not new, they are effective at spotting small non-uniform errors even at high speeds.
Electrical feedback from the head via he electrical interface is achievable by detecting variations in the Piezo (Fernando)
Pressure wave feedback through the fluid system is possible but it’s not currently practical for real world applications. In this scenario you would need to know your flow rate up front and then have flow meters in the fluid delivery system, for applications where you are jetting 100% this becomes possible but anything else the changes in flow rate are too small to measure successfully.
When failure has occurred there are a number of ways it can be detected, each one has it’s merits and its drawbacks.
The onion is complicated, there are no clear boundaries between the layers as we have already established, failure and detection are not always obvious or as they appear.