-
Hi everyone, I wanted to share my findings on building another DIY soldering station. The main reason behind the project was that I had a JBC T245 handle laying around for a few years and I just wanted to go through the learning process (as it is the most fun part with any project) instead of just opting for Unisolder. Also, I wanted dual channel.
Hardware:
It's largely the same as quite a few documented projects, consisting of high precision instrumentation amplifier, ADC, and a power switch, controlled by a uC. I chose Arduino Nano because... well it's cheap, simple to program and has more than enough computing power for such a simple project.
The front end consists of AD8293G160. These are cheap, have all the good features of zero-drift amps and also have a set gain ratio - less guesswork when measuring microvolts. The only downside is that the chopping noise feeds through if not filtered and the recommended 700Hz bandwidth is simply not enough to quickly measure the voltages between PWM cycles. After some experimentation I settled on 3 stages of RC filters tuned to 5-10kHz, which provided settling time of <0.5ms and reasonably low output noise.
ADC is the classic ADS1118, again because it's cheap and widely available. ADS1018 would probably be more suitable due to higher speed but I haven't bothered to order some. Total measurement cycle (settling delay+2 measurements) is about 3ms, which translates to ~15% downtime for 50Hz PWM frequency I aimed at. Peak power output for C245 cartridges at 24V is around 160-220W so that 15% loss is perfectly acceptable.
The high side power switch idea was borrowed from another project, however the part used there turned out to be quite expensive and hard to get here. Here's a hack - ST makes load of single and dual channel variations with almost identical characteristics. So I just searched Farnell for "VND5" (D stands for dual) and found a discontinued part on sale for 2 or 3 euros a piece - VND5E008ASP. ~100us switch on/off times, all sorts of protections and doesn't even get warm at full load. Oh and it even measures load current, which allows measuring peak power at runtime and by extension identifying cartridges by their heater resistance.
Power supply for the project was chosen 24V 300W from Aliexpress, probably could get away with less but I tend to oversize supplies if possible.
Aluminum case is a standard extrusion enclosure, also from Ali, can recommend - very nicely done.
PCB, and case openings/engravings were done on a home-made CNC, however my smd soldering skills are still lacking
:
Continued in another post
-
JBC - related
I'm convinced that many developers of such controllers get the C245 pinout wrong (including SparkyBG). My experiments show that the common point between the heater and TC is actually the outer shell. If anyone wants to check - measure resistances between all pins, then voltages at cold and at hot tip.
From a practical standpoint - there are perfectly good JBC stand knockoffs and handles on Aliexpress, the only genuine thing you need is the actual cartridge. It can be difficult to discern between genuine and fake ones but I found out I had two fake ones in my assortment by measuring - more on that later. The cable holder is made from a spring steel element from a windshield wiper.
In principle this station could power any TC/heater combination and for powering T12 and C210 I would only need to adjust detection algorithm in software, as their heaters have significantly different resistances. For other cases one would need an ID mechanism (ala Unisolder) or a menu setting, both of which can be added easily but I simply have no need for them.
Software
For LCD I used u8x8 library. It's text-only but supports larger fonts and with some clever glyph manipulation I was able to draw power bars as well. Still, each element in the display takes 1-3ms, so I had to break up the display routine into a state machine and draw elements in small chunks by calling the routine with each cycle.
Now control is where things get interesting because I settled for a PD algorithm, not PI or PID. I term is there but is only active when error is <5%. Reasoning being that in C245 cartridge heater and TC are so tightly coupled that there is almost no inertia and for any reasonable load only P correction is enough and D is there only to damp overshoot.
This is also where knockoffs rear their head - they have more thermal inertia between heater and TC, therefore inevitably generate some overshoot. I have 2 such cartridges out of 12. All others behave more or less identically, regardless of tip size:
Since I have the ability to measure actual power of each tip, I found that limiting the output power to 150W or so evens out the performance across different tips and only barely affects heat-up times.
Another revelation was when I started monitoring dT (effectively temperature climbing rates), which is a function of heater power and thermal mass of the tip. Turns out most tips average at 100 degC/s, with peaks above 120. Then it occurred to me that I can adjust PID in relation to that as well and implemented an output throttling mechanism to limit the climbing rate (sort of second inverse I term, measuring the derivative). Here's an example with 80 degrees per second target rate:
Again, the slowdown is barely noticeable in practice but probably improves tip life and also serves as a safeguard against stupid mistakes like plugging in a C105 tip.
Serial control and menus were initially planned but I don't think I'll add them because it works perfectly fine as is, there is simply nothing to adjust once tuned.
The controller should be fully dual-channel, I tested everything on dummy resistors but didn't check with a real cartridge. The power window of each PWM cycle is split between channels, favoring the one with less power. If both require more than 50%, then the power is simply split in half. PWM is done in software (16bit timer with two ISRs) because of low frequency and high precision required.
And that's it, please let me know if you have any questions. I can release Eagle files but my source code is a mess and I'd be hesitant to make it public, at least for now.
Oh, and I made a video of it in action:
https://youtu.be/v5FK_rFDtts
-
This is quite a cool project! Thank you for sharing.
-
Great work, and that ST IC does look nice, makes a lot of sense to go with a protected driver. The value/$ for obsolete automotive and telecom stuff can be really high.
http://www.farnell.com/datasheets/1876903.pdf
Limiting the power/rate is a good choice for sure.QuoteOh and it even measures load current, which allows measuring peak power at runtime and by extension identifying cartridges by their heater resistance.
So will a smaller tip have a higher resistance generally? From what I read JBC tips are ~3R. If you are saying they range from 160-220W, you are measuring in the 3 to 3.6 ohm range or so.
I've read some speculation in other threads that the manufacturers may be detecting the tip type but I believe nothing has been confirmed. It could be somewhat simple compensation: 160W tip -> apply x compensation factor, scale as tip power goes up. -
So will a smaller tip have a higher resistance generally? From what I read JBC tips are ~3R. If you are saying they range from 160-220W, you are measuring in the 3 to 3.6 ohm range or so.
Good point, those 160-220W figures are at the station and include at least three sources of error:
I've read some speculation in other threads that the manufacturers may be detecting the tip type but I believe nothing has been confirmed. It could be somewhat simple compensation: 160W tip -> apply x compensation factor, scale as tip power goes up.- They include power wasted in the wire, which I just measured to be about 0.25 ohms in total (0.25/3 ~ 10% loss)
- ST driver's current sensing circuitry isn't very linear and I calibrated it on a known 5 ohm resistance
- Calculations assume 23.5V supply under load as measured on a scope but that can vary a little as well
Going back to the tips, yes, the genuine ones tend to have higher resistance for smaller tips (in the range 2.6-3 ohms) but that might be an accidental correlation in my small sample. The point I was trying to make is that it would be very easy to detect a T12 tip at about 8 ohms and possibly C210 at ~2 ohms. I have no interest in the former and don't have the latter so this remains untested. From what I've read the smallest C105/115 tips are also ~3 ohms, and the only way to detect them without resorting to ID pins would be monitor the climbing rate (perhaps limiting to a short pulse and measuring the temperature rise). -
One thing I don't entirely understand about the C105 tips: they are rated for 12V supply, but, would they not run on 24V with a reduced duty cycle?
Assuming you do use the ID pins for detection and scaled back the power as you have done above. -
One thing I don't entirely understand about the C105 tips: they are rated for 12V supply, but, would they not run on 24V with a reduced duty cycle?
Assuming you do use the ID pins for detection and scaled back the power as you have done above.
Absolutely! In principle you can drive anything as long as the cartridge survives one PWM cycle of minimum power. Also C105 has been done already by Xyphro and seems to work flawlesly:
https://www.eevblog.com/forum/projects/diy-jbc-nano/msg1402927/#msg1402927
Even the pulse-count modulation for 24V AC seems to be OK because those TEC extension modules simple rewire the existing connector from C210 to C105. -
nice project, can you share the codes?
-
Thanks, regarding code as I said, it's an undocumented mess - do you want to replicate the project? If not, I would suggest you ask questions and I will do my best to answer them
-
I am curious about the codes, PID, Power bar and I want to do the project
-
Congrats!!!
-
I am curious about the codes, PID, Power bar and I want to do the project
From personal experience, just reading other people's code isn't very helpful unless it's well structured and documented. I will try to do that and then publish the source.
Regarding your questions, I've described PID in detail already, it's a standard control loop with the addition of second I term which decreases the output by integrating overshoot of the derivative in regard to the target heating rate. The effect is very similar to D term but since it's applied after clamping the output, it's active at any temperature, while D term is only active near setpoint, where P term's influence decreases.
All graphics are done with u8x8 library, which only deals with 8x8 pixel squares. To imitate smooth bars I've defined an array of 8 custom glyphs with increasing portion of the 8x8 square filled. Then to draw it one needs to scale the value to full length of the bar in pixels (in my case 0..56), draw value/8 full squares and one extra glyph at the end with value%8 pixels filled. If you only update the squares that need updating (as opposed to clearing and redrawing the whole bar), you get no flickering.
Btw I've found Excel to be excellent for designing and mapping out a text based LCD
-
Posting full schematic and board files in EAGLE format + code for Arduino IDE, tried to put comments in all the tricky places.
-
Posting full schematic and board files in EAGLE format + code for Arduino IDE, tried to put comments in all the tricky places.
Thanks for the info. I might never do a soldering iron controller but I'm sure I will learn something from this data. -
Hi, has anyone ever tried to compare this or any other diy station to the original jbc? In terms of thermal capacity in first place?
-
Yes, you will find quite a few comparisons on YouTube. However there is nothing to test really, thermal capacity is determined by the tip, not the station.
What you're left with is power output and temperature regulation algorithm differences, which are pretty insignificant due to insanely good technology built into JBC cartridges -
Hi. Thanks a lot for sharing this very cool design! Highly appreciated.
As I have no CNC mill to create the PCB I am about preparing the .brd for boardhouse production (just adding some values to the silkscreen layer to facilitate the population of the parts.
If you want I will glad to publish the modified board file (again, no modifications on parts or locations).
I have some questions on some part values: what is the value of R7 ? and what did you take for L1 ? I did not find a reference in the notes in the datasheet about C20 and C21
Best regards from Switzerland and thanks again for sharing
(I will be glad to send you one of the production PCB's when I have them). I will not fabricate 5 or 10 soldering stations
-
I in PID is there to ensure that you actually arrive at the precise setpoint when error is small but still non zero and P does not affect the output enough to move, by making the small error more significant the longer it persists. You are correct disabling I when error is large, it prevents I windup
-
I have some questions on some part values: what is the value of R7 ? and what did you take for L1 ? I did not find a reference in the notes in the datasheet about C20 and C21
Hi, glad you like the project
R7 is a 0 ohm jumper, to join analog and digital grounds at a single point. You can use a piece of wire instead.
L1 can be seen in one of the pictures, I used a random SMD inductor from those cheap DC/DC converters from eBay, 22uH. Probably unnecessary but certainly doesn't hurt. -
A word of caution - there are several critical constants baked into the code that you might need to calibrate:Code: [Select]
// Specific to the supply at hand and VND device used
#define HEATER_VOLTAGE 23.5
// DS specifies K at at 5000 but with large uncertainty. Has to be calibrated with a known load.
#define CSENSE_COEF 4.8
// adc reads 3705 when disconnected. Depends on uC supply, voltage reference for the inst. amplifiers, gain and PGA range.
#define DC_THRESHOLD 3700
Second, I don't think I mentioned but I used a 4-pin step down modules (attached), these seem to be very reliable and efficient.
IC1 doesn't have to be the model specified, any voltage reference will do, even a zener. Just keep the reference voltage in 0.8V..1.5V range. The higher you go, the narrower temperature range you will be able to measure: (5-vref)/160/Ttc. For C245/C210 (~20uV/C) this is unlikely to be a problem but K type thermocouple (T12 is ~50uV/C) will have a hard temperature limit of ~450C with a 1.1V reference.
Regarding C20/C21 - these form a first order RC filter at the input to decrease measurement noise. I found experimentally that a good compromise between ADC rise times and noise suppression is a cut off frequency of several kHz, therefore all resistors and capacitors around instrumentation amplifiers are tuned roughly to that frequency. DON'T decrease the values for R8-R11, you will blow internal protection diodes of the amplifier.
Also, don't connect an iron straight away. While the code will detect a missing heating element, it doesn't know if a thermocouple is present. So if you accidentally short TC input or connect it to wrong pins, you will get a white-hot iron tip. Also use telemetry by serial (default baud 0.5M), at least for initial debugging, two options are available:Code: [Select]//dumpStats();
dumpTelemetry();
Which iron tips are you planning to use?
-
Hi
Thanks for the infos. Much appreciated. you can also mail me anytime to cgraf[at]rocketmail.com
The boards are on the way from HongKong. I think there was a small error on the pcb (a crossing of tracks), see attachment "pic1.png"
I also attach the board I sent for production (I did no changes to the schematics).
Where did you use the step down ? After the power supply ?
I did not decide yet on the tips ? an idea is the T245 handle with C115 cartridge.
Best regards
Christoph
-
These regulators are denoted in the schematic as VREG5V and VREG_FAN. In my case both are set to 5V, so the 12V fan is spinning quietly.
Regarding the crossing, I did everything on a single sided PCB so crossing wires on the other side was no problem. If you ordered them as is, you will have to scrape one track off and make a wire jumper.
Regarding C115, I'm not sure they fit in T245 handle at all. And even if they do, it's up to you to figure out the pinout and not mess up the wiring
-
Hi
The error with the crossing was corrected on the board (see my .brd file). About the tips you are right, I will use the same as you...
regards
christoph -
Boards arrived. See attached files. very good quality and a matte finish.... a pity this goes into an enclosure
Anyzis if you send me your postal addresss I will be glad to send you one of the PCB's
Best regards
Christoph]
-
Great project!
Just a little esthetic suggestion: Change out the silver cap head screws on the front. They really stick out from an otherwise clean and nice case. Maybe change them to some black countersunk screws. Or some black buttonheads if the case is not thick enough for countersunk screws.