I started my work on the phase detection feature by reading the manual of a Videojet Excel 170i which I bought for testing. I read that it uses a 420us long 10VDC pulse for charging 28 droplets, which later get analyzed for selecting the right phase. This printer is using a piezo frequency of 66kHz which has a period of around 15us. 420 divided by 15 equals 28 which explains the numbers. This printer can select between 4 phase settings. It uses the gutter instead of a dedicated probe for reading the charge of the droplets. Unfortunately, it was not described how the printer analyzes the signal, so this did not help me much.
At least I learned from it that phase detection is an important part of CIJ printing and that it is done by sending a signal to the charge electrode which later gets read from the droplets to get a feedback system.
The remaining question at this point was, what a phase actually is and how the signal could help find out the right phase.
Fortunately, @Paulo Campos could help by telling me about an improved method that uses 16 instead of 4 possible phase settings and a phase detection sensor/probe instead of the gutter.
Thank you Paulo for helping me again :)
Now I had to implement Paulo's method into my DIY printer, which is no 1:1 copy of another printer, but my own design, so there was some work needed to get it going.
Phase Detection Signal Properties
The signal has to be a sequence of sixteen -12V pulses with each pulse starting with a different phase shift to the piezo's sine wave signal.
This is done to get 16 successive droplets each having a different phase shift setting. These droplets get analyzed by the printer to find out which droplet got the most charge from the -12V pulse for selecting the best phase shift setting, in short phase.
By doing so the printer can choose from Phase 1 with 0⁰ phase shift to Phase 16 with 337.5⁰ phase shift to get the best possible charging quality.
Piezo Signal
Starting with the sine wave that drives the piezo:
For my printer, I chose 50kHz for driving the piezo, because this frequency is compatible with the 40kHz piezo rings, that you can find in ultrasonic cleaners and because it has a period of 20us which makes calculations with it easier than using an odd number.
A single Piezo Ring
Before talking about the phase detection signal's waveform it's worth mentioning, that droplet formation happens over the whole sine wave period while on the other hand, the charging puls can be pretty short, but has to happen at exactly the right moment of the droplet formation process.
The phase detection feature finds the right moment by looking after the phase with the highest charge so that future charging pulses can be executed with this phase setting.
If parameters like ink pressure or viscosity change, the right moment for charging also changes which gets detected by the phase detection feature, which then selects another phase, that fits better to the new parameters.
Phase Detection Signal Waveform
Given that the piezo frequency is 50kHz with a period of 20us and the goal is to design a phase detection signal that completes a complete 360⁰ phase shift in 16 pulses I used a period of 21.25us for these pulses.
My assumption was:
20us / 16 = 1.25us
21.25us * 16 = 340us
20us * 17 = 340us
So, 16 pulse periods take as long as 17 sine wave periods which means that every pulse starts 1.25us later in the sine wave than the one before. The phase detection signal and the sine wave have to start at the same time and after 17 sine wave cycles / 16 pulse cycles, both waves are in sync again.
For the pulse, I just used a 50% duty cycle square wave, but I could just change the duty cycle if it turns out that another duty cycle would be better, just the period has to stay the same, so that 360⁰ phase shift takes 16 pulse cycles.
I added a 340us pause after the phase detection signal so that the period of the whole signal is 680us (340us signal, 340us pause) which equals a frequency of around 1.47kHz.
In many CIJ printer manuals and papers about CIJ printing it is described that CIJ printers use a closed loop system to monitor droplet breakup. It works by charging some of the droplets with a test signal, measuring the charge of these droplets and analyzing which one got the best charge from the test signal to choose the right timing for droplet charging.
Until lately, I often read that, but didn't fully understand how exactly the printers do that.
Great thanks to @Paulo Campos for helping me understand how Phase Detection works.
Here is my description of it:
A CIJ printer works by using a piezo for breaking up an ink stream into equal-sized droplets. These droplets get charged to a certain voltage level when they fly through a charge electrode and get deflected out of the stream's flight path according to their charge when they pass by a high voltage deflection plate. According to their deflection angle they hit the surface at a lower or higher position of the vertical line that the printhead can reach from it's position.
For droplet charging the right timing is crucial. The charging pulse needs to take place at a certain moment in the droplet formation process.
That's where phase detection comes into play:
The droplet formation process is based on the piezo's vibration which is based on the piezo's drive signal which is a high frequency sine wave. That means that different stages of drop formation occur at different phase angles of the sine wave. To find out at which phase angle of the sine wave the droplets receive the most charge, a phase detection test is used.
The phase detection test signal is applied to the charge electrode whenever the charge electrode is not used for the charging of droplets that are meant for printing.
The phase testing signal is a burst of 16 pulses that cover 0 to 360 degrees in 22,5-degree steps of the piezo signal's sine wave so that each pulse starts at another phase of the sine wave.
This is done to charge 16 successive droplets with different timings/phase settings which will later be read by a phase detection circuit and the printer controller which then can select the best timing/phase setting to get the best possible droplet charging.
In contrast to print charging, test charging usually uses a low negative voltage instead of a high positive voltage to make sure the ink droplets that are used for testing hit the gutter instead of the printing target and to make it easier for the printer controller to differentiate between both types of charge signals.
A positive voltage leads to negatively charged droplets and a negative voltage leads to positively charged droplets. With that, the printer controller sees the positively charged phase detection droplets as positive spikes and the negatively charged printing droplets as negative spikes and can easily differentiate between them. The duration of the measured spikes is dependent on the time of flight of the ink droplets - It shows how long the ink droplet is present to the sensor.
The signal that gets read from the phase detection sensor usually has the shape of a hedgehog, with the highest charge droplets in the middle and the lowest charge droplets on the outside.
The printer controller then selects the phase setting with the highest charge on its droplet and applies it to the charge pulse when printing.
And that's basically it.
It's worth mentioning, that every changing parameter like ink pressure, viscosity, conductivity, dirt in the nozzle and environmental conditions has more or less effect on breakup or charging, so it's likely that the best phase for charging doesn't stay the same during printing what makes a phase detection function essential for operation.
It's also very useful to have the signal from the charged ink droplets on an oscilloscope screen while adjusting parameters on the printer to see which effect they have on breakup and charging.
To date, I could successfully generate the phase testing signal and just recently I could make some progress on the feedback signal reading.
To prove that the 312V charging voltage can be switched, I used an IRFP460 MOSFET for shorting the charge voltage to ground, which turns the charging off.
The charging voltage is already current-limited by a 100k ohm resistor on its output, so shorting it to ground is not a problem.
Now, I can try to apply some signal to the charge electrode and try to measure something on the phase detector next.
While with a single MOSFET, it's only possible to send a square wave to the charge electrode, I will ultimately need a sine wave that has the same frequency as the signal that is sent to the piezo.
Next, I will try to measure a signal with the square wave and if that doesn't work I will try to build a "sine wave driver circuit" and try again.
Update:
I figured out that the phase testing signal is not a sine wave, but a burst of 16 pulses which cover 0 to 360 degree in 22,5 degree steps of the piezo signal's sine wave...
I noticed something really interesting when I connected a cheap function generator to the charge electrode without connecting the ground (of the charge electrode).
By doing so I get a pretty high AC voltage reading on the charge electrode. The voltage is also not just a misreading, but actually present. I know that because it can influence the deflection of the ink stream when I turn on the high-voltage deflection plate. Turning the high voltage on while the unknown signal is present leads to an oscillating ink stream - the ink stream moves back and forth from the middle position at around 50Hz.
The function generator is powered by a common 5V USB smartphone charger.
While trying out a different USB charger the signal changed and by running the function generator from a power bank or my notebook the signal disappeared completely.
So, because of the 50Hz and because of the absence of the signal when the function generator is battery powered, I guess that the unknown signal is AC noise from mains that could somehow pass the USB charger and function generator to reach the charge electrode and also charge the ink stream.
I have to admit that I'm pretty happy about this noise because it is helpful to prove some things:
- It shows that the charging and deflection work in general with my current setup.
- It is reproducible. I had unwanted deflection before which was unfortunately not reproducible and so I couldn't dig deeper into it.
- It shows that AC charging leads to deflection in both directions, left and right from the middle position. This means it can charge the ink stream positive and negative. Before, when the high voltage cable was too close to the not connected charging cable, deflection happened only towards the high voltage plate which I think means only negative charging of the droplets caused by a positive voltage on the charge electrode. With the AC noise, it happens in both directions, tho.
- It visually confirms that there is no high voltage to ground leak, which had been a huge problem in the past.
While I was thinking about the source of the signal I decided to do a test to find out whether it comes from the printer, the function generator, the charger, or mains.
For that, I cut off the micro USB plug of an old USB cable to access the +5V and GND USB charger connections.
I then tried to prove my assumption by connecting the +5V USB Charger connection to the oscilloscope and the GND pin of the oscilloscope to mains PE.
By doing so, I could successfully prove that the noise came from the USB charger or mains, but not from the printer or function generator.
To be safe I also tried powering the oscilloscope via battery and could still measure the signal.
While I still don't know how exactly the signal gets generated - what I guess also doesn't really matter for the project - I think it's a pretty nice coincidence that I didn't connect the function generator's ground to the printer's ground and so could find this noise signal that was very useful for testing.
I just saw that I had even more luck because the other USB charger that I used for comparison creates much less noise and so it likely would have not caused any visible deflection of the ink stream so that the noise would have been overlooked.
I did the measurement without connecting the GND of the USB charger to anything and left it floating.
When I then connected the charger's GND to PE, the noise disappeared - it also worked when I swapped the charger's output lines.
My uneducated guess would be, that the noise is created by the switching of the USB charger which couples into the charger's output lines as well as in the mains' live and neutral. Because at some point in the building's electrical system, the neutral line is connected to earth the noise is also present on the PE line.
When now the charger's output, which is not connected to the oscilloscope, gets connected to PE where the oscilloscope's ground pin is also connected to, the oscilloscope receives the 5V DC from the charger's output which can provide much more current than the noise signal. With that the noise is no longer detectable.
While using the function generator and connecting the printhead's ground to it's output the noise was likely shorted so that it did no longer traveled across PE, but through the prindhead's ground to function generator's ground connection instead.
I also could be totally wrong with that...
Anyways... next up is creating a signal that can charge the ink stream as the noise does.
Over the last months, I did some testing with different printhead designs, self-mixed inks, and some new parts and features for improving the reliability of the printer, to prevent frustration while testing.
Great thanks again to @Paulo Campos for helping me with that.
There were two main problems I tried to solve with the improvements:
- Clogging of the nozzle
- Electrode output leaking to ground due to ink/dirt forming a conductive path to ground on the isolators.
Both errors prevent the printer from working, so it's best to prevent them or solve them quickly when they appear.
Maybe I will later add a feature to the printer to detect (and maybe even solve) them when they appear while the printer is printing.
I also improved another problem that caused a rise in ink pressure while reloading the timer chamber by adding a water pressure regulator to the ink line.
This rise led to a shifting of the ink stream break-up point and could shift the break-up point out of the charge electrode which would prevent the printer from working.
Adding a water pressure regulator reduced the reloading rise from about 10psi to 1psi which I think should be OK for now.
For getting more info about the ink in the system I added a temperature and a TDS sensor to the printer. It's important for the printer's operation to keep the ink at the right conductivity and viscosity.
With the TDS sensor and timer, it's possible to keep an eye on these values.
In the future, I want to mount the printhead on a moving carriage for labeling and 3D printing. While the printhead moves and the feed lines bend back and forth the ink pressure would float which could affect the operation and printing quality. To prevent this, I added a small spring-loaded puffer/damper to the ink line that should later act like a capacitor for keeping the ink pressure constant.
Besides a couple of low voltages, the CIJ printer also needs high voltages for the charging circuit and deflection plate.
Here, you can see the power supply of an old CIJ printer that I bought for research:
It outputs 312V and 3 to 6 kilovolts. On this printer, the 312V is used for the charging circuit and the piezo driver, and the 3-6kv is used for the deflection plate.
On my printer, I'm currently using a 24V audio amplifier for driving the piezo, because they are available for sale and you don't have to build it by yourself which should make this part of the project easier.
I don't have the charging circuit ready so I can not fully test out if it works, but under the strobe led you can see droplet formation, so it looks promising.
However, if it turns out that the audio amplifier is not enough for driving the piezo I will have to build a proper piezo driver that uses the 312V for it.
Now I started looking for ways to generate 312V and 3-6kV.
312V Power Supply:
Here you can see my 312V power supply. I used a YH11068A boost converter powered by 12V. It turned out that the YH11068A needs some load to provide a stable output voltage and so I added three 100kOhm 5W resistors in parallel to the output. I also added another 100kOhm resistor in series with the output to limit the current for safety.
The YH11068A is protected by a 1A fuse on the input that will blow in case of an unintended current draw.
Finding out the thing with the needed load was a bit tricky, but because you can just buy the YH11068A, generating your 312V is not that complicated.
Adjustable 3-6kV power supply:
Here you can see my adjustable 3-6kV power supply. The output voltage can be adjusted by changing the switching frequency of the MOSFET.
Higher Frequency = Less Voltage
Lower Frequency = More Voltage
Because there was nothing out of the box available, I had to build it by myself.
First, I got myself a high-voltage probe, so that I could measure the high-voltage output without destroying my multimeter. The probe divides the voltage 1:1000 so that you can read like 4.20V for 4200V on the multimeter.
After that, I looked for a circuit that can generate adjustable 3-6kV while being easy to build.
I started with one of these 15kV modules that you can find on Amazon, eBay, and Aliexpress.
This module uses a transformer that gets switched at high frequency by a transistor for generating high voltage. Besides the transformer's primary winding, it also has a feedback winding for controlling the transistor which makes the circuit self-oscillating, so that it does not need any sort of frequency generator to run.
Unfortunately, in contrast to a TV's flyback transformer, it has no internal rectifier diodes, so its output is AC. It also has no way of controlling the output voltage, so it can not be used for the CIJ printer without modification.
Besides that, the transformer that comes with the module is very nice for generating high voltages and because these modules are available almost anywhere, it should be no problem to build the circuit with the same transformer anywhere else.
The modules usually come as DIY kits, so there was no need to first desolder the transformer from the PCB.
Instead, it was only needed to remove the feedback winding from the primary side of the transformer to get it ready for further use.
After removing the feedback winding I tried building a new transformer driver circuit, but this time driven by a function generator and an IRFP260 MOSFET. I also added multiple 1000V diodes in series to form a high-voltage diode that converted the output of the transformer to DC.
Here you can see my first high-voltage power supply testing setup
6060V @ 15500Hz
3000V @ 27400Hz
I tested out driving the MOSFET with PWM at different frequencies and noticed that the voltage at the output changed according to frequency as expected so I could continue with the work on the circuit.
The diodes at the output could not provide real DC, but only the positive AC half wave and so I decided to build a bridge rectifier and add a capacitor to get rectified and smooth DC.
For that, I used four 20kV 5mA diodes to build a bridge rectifier and a 20kV 1nF capacitor for smoothing.
I tested it out and realized that I now could get a higher voltage with the same frequency.
I also tested out drawing some arcs and shorting the circuit, to find out what happens.
Unfortunately, in contrast to the last test circuit, with the capacitor added the circuit could provide enough current to destroy not only the MOSFET but also the function generator :/
Now, I had to find a way to limit the current to a safe level to prevent damage to the electronics in case of a short circuit. It later turned out that short circuits happen from time to time if the high voltage deflection plate gets unintended hit by some droplets.
So, I got myself some 20mOhm 5W high voltage resistors to limit the current to a safe level for the electronics and also for the user.
I added the resistor between the negative high voltage output and GND. Now it's possible to short the high voltage to ground without causing damage to the electronics.
While waiting for a new function generator and a high-voltage cable to arrive, I tested out the nice XY-LPWM generator with integrated LCD, that I wanted to use for adjusting the switching frequency and with that the voltage of the high-voltage deflection plate.
The cable arrived and I could finally fix all parts with hot glue onto a hard paper that I mounted into an electronics box which I mounted on the printer frame.
Finally, I added a switch and a 1A fuse to the input of the high-voltage power supply and connected the high-voltage cable to the deflection plate.
With that, the work on the 3-6kV power supply was done.
Here you can see both power supplies running
And here you can see a test of the 3-6kV power supply.
With that, I finally have all needed voltages ready to drive my CIJ printer prototype.
The CIJ Printer needs a couple of different voltages for operation and here are the power supplies/converters that I use for generating these voltages:
- 24V Power Supply
The 24V 10A power supply is the main power supply of the printer. It generates 24V from mains for all other power supplies/converters. It also powers the solenoid valves and the audio amplifier/piezo driver.
- 12V Converter and 5V Converter
An XL4016 step-down converter powered by 24V is used for generating 12V for the plus/minus 12V converter and the high-voltage power supplies (more about them in the next build log).
Another XL4016 step-down converter powered by 24V is used for generating 5V for the Arduino and other logic-level electronics.
- Plus/Minus 12V Converter
A DD1912PA plus/minus 12V converter powered by 12V is used for generating plus/minus 12V for the charging circuit and phase detector in the future. This part of the project needs further research and at the moment it is not in use.
These are the low-voltage power supplies that I'm using at the moment.
In the next build log, I will give you a closer look at the high-voltage power supplies.
To prevent damaging my notebook in case something goes wrong and especially because of the high voltage circuit running at 3-6kV, I was looking for a way to control the CIJ printer prototype without an electrical connection to my notebook.
For doing so, I tried out sending the signal of the serial connection via lasers over optical cables (Toslink cables) and it worked just as well as a wire-based serial connection.
I built the connection like:
PC - USB Connection - Arduino Nano - Optical Connection - CIJ Printer's Arduino Mega
I connected my PC to an Arduino Nano via USB. The Arduino Nano uses hardware serial to communicate with the PC and uses software serial to communicate with the CIJ Printer's Arduino Mega. The software serial TX pin is connected to a laser and the software serial RX pin is connected to a laser receiver. The Arduino Nano just passes the serial connection from the PC to the CIJ Printer's Arduino which also uses software serial, a laser, and a receiver, so that the connection works as if the PC is directly connected to the CIJ Printer's Arduino.
I also thought about using a wireless connection, but I came to the conclusion that an optical cable connection would be more reliable and easier to build - so no wireless connection.
Here are the kind of lasers and receiver modules that I used for it:
I placed the laser and receiver in a 3D-printed part and fixed everything in place with hot glue. This is the part on the CIJ printer side:
Over the last months, I added a lot of new stuff to the CIJ printer project, but I could not find the right moment to report about it.
Because of that I will rather show you the current state of the project and add more build logs about the new features after that.
Here are some pictures of the current CIJ printer prototype:
Here you can see the latest printhead prototype:
I added a cable guide, a new 24V audio amplifier for driving the piezo and a laser sender / receiver to isolate the PC from the printer prototype:
Here you can see the new power supplys for all needed voltages (24V, 12V, 5V, ±12V, 312V and adjustable 3000V to 6000V):
I finally took the vacuum pump from my desktop to the printer frame and secured it in place with 4 springs to prevent the vibration from entering the frame and causing a lot of noise:More build logs with more details about the added parts coming soon...