Recently, I noticed in the Elegoo Saturn Facebook groups that many users report a broken Saturn. The symptoms are that the UV backlight is always on. Elegoo has been really helpful and it seems that when a user reports the problem, they send a replacement mainboard.
I was wondering what causes the problems, so I reached out to a few of them if they would be willing to send me the broken motherboard. One of them did (thank you!) and I started to investigate what’s wrong. TLDR: The fix is easy and costs you nothing if you know how to solder.
What’s wrong?
The LEDs are driven via a boost converter that outputs a constant current. This driver is located on a separate board. There are actually three of them – one for each array of LEDs. The turning on and off of the LEDs is not done via pulling an enable pin of the boost convert driver IC but there is a dedicated MOSFET on the mainboard, that switches the power for these boost converters. You can find them on the photo below (it is the left-most transistor). I also attach a schematics of the connection:
The right-most transistor controls the LEDs
The schematics of the switch circuit
After measuring the transistor the problem is clear – the transistor is dead. It is shot-throught and it is always conductive. I have also confirmed this after removing the transistor.
How to fix it? Replace the transistor. Luckily, the manufacturer gave us a spare transitor. The middle transistor is not used for anything, therefore, we can desolder it and solder it instead of the broken transistor. Easy-peasy.
After swapping the transistors, the board works again!
Why does it happen?
The fix is easy, but it would be nice to find the cause. The first suspect is that the boost converter is an induction load that can create high-voltage spikes across the inputs that can easily kill the MOSFET. However, there is diode protection for this. Also, the MOSFET seems to be sufficiently rated according to the datasheet.
On top of that, there is a big capacitor on the input of the boost converter, that should act as a snubber for such pulses. I also took my Saturn and examined the input of the boost converters with osciloscope. There seems to be nothing that could kill the transistor.
However, the problem appeared only recently; therefore it affects only new Saturns (mine is from the first preorder batch). So there could be two things:
- Elegoo changed their boost converters design for the newer Saturns. Therefore, there could be a flaw in the design that kills the transistors. However, as I don’t have this design available, I cannot tell.
- Or there is simply a bad batch of the MOSFETs.
If you are an owner of a newer Saturn that has all three boost-converters on a single board and you are not afraid to disassemble your Saturn, please, get in touch with me and send me detailed pictures of the new boost converters. Maybe we can find the problem or rule them out as the root of the problem. If you are unsure how to disassemble your Saturn, I made a guide recently for my Silencer kits.
Recent news: My open letter to the 3D-printing community
I love the 3D-printing community, but I think there is room for improvement. Let's get better in 2023! Read the full letter.
Support my work!
If you like my work (these blog posts, my software and CAD models) and you would like to see more posts on various topics coming, consider supporting me in various ways:
- You can become my sponsor on Github.
- If you prefer, you can also become my Patreon.
- You can buy me a coffee on Ko-fi,
- or you can buy something from my Tindie store (also see below),
- Or you can just share my work!
If you are interested in knowing what I am up to and recent sneak-peaks, consider following me on social media (Twitter, Instagram, Facebook).
My store offers
I launched new tank cleaning kits for Elegoo Saturn, Saturn S, Mars 1, and Mars 3. You can find them in my store.
How big is the capacitor on the input side of the buck? maybe the fet can’t handle the large inrush current when the cap charges. Perhaps adding an NTC inline might help but it would most likely drive up the exposure time a little bit. The ideal solution would be to modify the buck to work via an enable input in the conventional way.
It is a 470u electrolytic (with relatively large ESR). The MOSFET is rated at 40A so in my opinion, the inrush current should not kill the MOSFET (if it complies with the specification – which I am so sure about). Interesting is that it does not die on older Saturns.
Love your work. But my Saturn failed a few weeks ago with this issue and was from the first Pre-order batch. Several others on the groups are also experiencing the issue with Pre-order machines. Whatever’s causing the problem, it doesn’t seem to be limited to only “new” Saturn’s.
Thanks for letting me know that also pre-order Saturns fails. Therefore, I will try to dig deeper with my Saturn.
Although the datasheet shows 40A, 60V capability for the MOSFET that is only true for a large Vgs voltage.
At turn on the large surge of current that charges the 470uF capacitor will mean the MOSFET is operating in its linear region – it will have significant volts across it whilst passing the charging current. That equates to lots of watts in dissipation each time the lamp turns on.
At the low signal voltage out of the MCU the device will be marginal to turn on fully when the lamp is on. So, the dissipation will be quite high here also. (This is probably what kills it over time.)
Finally, the series 10k resistor (R12) seems a very large value also. The gate-source capacitance in combination with the 10k resistor will mean the MOSFET turns on quite slowly – again resulting in large currents with large voltages across the part causing significant dissipation. (Maybe they did this deliberately to reduce the inrush current that charges the 470uF cap.)
I think the series resistor should be reduced to improve turn on speed and a part with a lower Vgs threshold used.
I have added some pictures of measurements on my own Saturn to the FB group you linked this blog to. I measured a peak inrush current when the lamp turns on of 77 amps. The 24V PSU drops to ~19V. There are 3x 220uF capacitors on the driver board, one per channel, and these need to charge up when this MOSFET turns on – they are the cause of the spike. Its the power dissipation during this phase that is killing the part.
Adding a small, stick-on heatsink would be a good idea to help with the thermal dissipation. Appears to be plenty of space around and in front of the device from your pictures.
Thank you for letting us know! I just find a guy with the new design of the LED drivers (3 drivers on 1 PCB) who confirms your measurements. Personally, I do not measure such significant current peaks on the old LED driver (only about 50A). So this is probably also the reason, why newer Saturns die more often compared to the old ones.
Power supply designer here. My friend just left me one of these to debug. I will look at it soon, but a few comments before I do.
10k resistor as a series resistor?!?! Are you sure it is the series resistor and not the pull down resistor (for when the driver is not fully on). A switcher would not work with a 10k turn on resistor. Full stop.
I have a feeling the layout is the issue with these failing. That is not how a boost is normally laid out. I will repair and take some captures of the switching waveforms and gate drive signal and determine how much margin they have in their design
The PCB layout in general looks horrible.
If anyone has the schematic for the boost circuit, sharing it would be very much appreciated.
I made the schematic a long time ago, but I am pretty sure that the series resistor was 10k. Why wouldn’t it work – you just get a really slow rise time, but the MOSFET will open up eventually. Definitely not a wise choice (if you want to prevent ringing, something between 1R and 1k would be more suitable), but hey, it works.
PS: Just to be sure: the boost converter is a separate, stand-alone unit. They drive it by connecting it to the ground by this failing transistor. What I don’t understand is why they didn’t use the enable pin of the boost converters to the switching. They could have saved a transistor.
With a 10k resistor your turn on time will be longer than the MOSFET’s ON time. Additionally, most MOSFET drivers have a parasitic pull down resistance of around 10k even when on, which will fight a 10k turn on resistance. Normally the series resistor is in the order of 1 to 10 Ohms.
Just a thought, instead of removing it and then moving across the other one, it looks like it is part of a similar circuit? Could you cut the offending ones legs off and then jumper wire the other one if its a redundant circuit?
Board on my friends Saturn does not look like the board pictured above at all. And all components check out fine. Backlight still always on. Do you know when the PCB changed?
We don’t know exactly but somewhere after Saturn was sold normally via Amazon. Could you share pictures of the board?
Could a preemptive heatsink avoid the burnout of the transistor? I have some heatsinks lying around, is there any space for a heatsink anyway?
There is plenty of room for a heatsink. It could help, but I think the effect would be negligible. The heat is generated in short periods of time with long off periods. So what plays against you is the thermal resistance of the transistor package. The heat just does not make it in time to the surface. Also, you should add a heatsink to the back pad of footprint (there is the best thermal conductivity) but you cann’t, as there are no thermal vias to the other side of the PCB. In my opinion too much effort for no effect. If you want to avoid failure, I would start with changing the MOSFET.
Hello Sir.
I tried to remove the middle mofset WSF40N06 and cracked the body….lol…The bad mofset WSF4060 was removed perfect….which is useless. I ordered from Amazon the WSF4060 and they only sell 5 of them at a time…only $7…
My question is can I use 2 of the 5 and replace both the mofsets with the WSF4060 or should I get a WSF40N06 for the middle one to replace, or is it necessary to replace the middle one anyway. As I read this…
The WSF40N06 is the highest performance trench N-Ch MOSFET with extreme high cell density ,
which provide excellent RDSON and gate charge for most of the synchronous buck converter
applications . Any replay will be appreciated. Elegoo Support is sending me a new motherboard anyway and all firmware as what I had.
The transistors are independent. So they don’t have to be the same. I personally use IRLR8743TRPBF as a replacement, that should last for eternity.
The WSF40N06 MOSFETs are proving difficult to find here in the US. Are you saying that the IRLR8743TRPBF is a direct replacement for the WSF40N06? I hope so because I can get the IRLR8743TRPBF today locally. Thanks!
I’m also curious about the IRLR8743TRPBF mentioned. Can someone confirm if it works? And if possible, what about it compared to the WSF4060 in the application will give the impression that it “should last for eternity”?
TIA!
For anyone interested, I used IRLB8743PBF (it was available overnight). With a bit of modding of the heatsink tab, I managed to get it to work! That being said, scroll to the bottom for my comments on contacting elegoo support for a replacement. The replacement is shipping from china, so it’ll be a bit, but in the meantime, I’m back up and running again!
Got the new replacement motherboard in that is promised to never have this issue again. It uses: PTD50N06 an N-channel MOSFET made by PUOLOP
Got a defective motherboard overhere and checked for the resistors values (steering the N-MOSFET). R12 and R49 are both 10K.
My main question for Elegoo is, why not develop an UV-board which runs at 24V ? It’s not hard to design this and it would eventually save a buck-boost converter (and a defective MOSFET).
At the very moment I develop a device for my Elegoo Saturn, which requires hardwiring the switch-signal and 24V from Elegoo motherboard, while the switching will happen between Elegoos buck-boost converter and UV-LED board : More reliable and faster switching. (I’ll likely update here later about this new board when ready)
Forgot to mention : This board I develop enables the use of current Elegoos buck-boost converter, but also a different and more reliable and higher power buck-boost converter (150 Watts). Such a buck-boost converter is for my near future plan : to create an Elegoo Saturn “on steroids” (requires also development of new UV-LED board with higher rated power UV-LEDs).
In the past month I did a similar thing for an AnyCubic Photon. I swapped the 20W UV LEDs into a single 30W (12V) UV-LED, to speed up 3D-printing a bit. Upcomming months I plan to replace this 30W UV-LED for a 50W-version. In such case a fair comparision can be made between printing-speed between RGB- and mono-screen : Anycubic Photon versus Anycubic Mono.
They need a converter no matter the voltage they use – they have to provide a constant current source. The LEDs resistance changes drastically with temperature (the higher temperature, the lower resistance, thus more current, thus more heat thus…). This is why they have a converter – it does not matter if it is a buck, boost or buck-boost.
Thank you for the information and the pictures.
Just changed the MOSFET in my Epax E10 printer (same chitu system, got the “blue screen”) . I didn’t swap the middle one but ordered some new WSF40N06 and it works perfectly. Thanks again.
My E10 just done the same thing is yours still working after changing the mosfet out. If so can you show which one burnt out? I can’t seem to find it on mine.
When I look at the data sheets, simply swapping the Mosfets crosswise may be problematic. The 4060 has an RDS On of 7.2mOhm and the 40N06 has 20mOhm. The 40N06 has to cope with a higher loss. As already written, it would be better to use a stronger Mosfet with a comparable RDS-On resistance. 🙂
Update :
In the past months I developed an upgrade kit (UV Blaster 30 KIT) for the Saturn, with more powerful UV-Boards, operating at 24V DC. It eliminates the use of the inverterboard, while it drastically shortens the time for a print-job. I plan to get them available at Tindie (and maybe eBay). However, there are several design-flaws in the Saturn causing problems, resulting in the “UV-LEDs always on”. Even a more recent board (V1.0.5) contain these flaws.
The 3 main flaws are :
1) Shared powerlines and caps
2) Powerregulators and caps
3) Grounding-problem.
Let me explain.
1) Right after powerswitch, a powerlane goes to a 150KHz powerconverter (LM2596). Also, a powerlane goes to the fan/LED-sockets and to the steppermotordriver.
Also, there are 3 pcs 470uF/35V caps. One near the LM2596 (C13), one in the middle of the powerlane to the sockets and steppermotor (C50) and one between sockets and steppermotordriver (C9). I assume, C13 is used as a buffer for the LM2596. C2 ment for steppermotordriver and C50 as powerbuffer at the sockets. However, because they share their powerlanes, an unbalanced capacity of 1410 uF (3x 470uF) is created.
The LM2596-datasheet suggests a 680 uF cap at input. A different value might cause imbalance. I can’t see what’s under the heatsink of the stepperdriver, but I assume it’s one from Trinamic. If checking the TMC2100 datasheet, a 100uF is advised. Even Allegro stepperdrivers suggests a 100 uF cap for the A498x-series.
The remaining capacity is probably planned as buffer for the sockets, like fans and UV-LEDs (via inverterboard). Adding up all required capacity, the total used value might be correct, but it doesn’t look like the “best” design. Better use 3 separate powerlines and “isolate” each by using a rectifying diode in each lane or even a Schottky diode.
2) Datasheet of LM2596 mentions at output / feedback-line a value of 220 uF. If you check your boards, in the 5V powerlane are a few tiny brown ceramics and 2 pcs 680uF/6.3V (C52 and C121) in parallel. This totals it to a whopping 1360 uF. It’s far from the suggested value. Besides, they’re 6.3V rated, which is just below the edge of safety.
The onboard CJT 1117B-3.3 is 3.3V Voltage regulator. When checking the datasheet, at the output a 22uF capacitor is suggested while a 680uF cap (C59) is used. I don’t have to explain anything further, I guess.
3) I noticed some odd behaviour, which took several days to figure out what caused this.
I replaced the MOSFET, tested succesfully the UV-boards, while not connecting the FFP cable from LCD panel to motherboard. As soon as the FFP-cable got connected, closed the machine (incl. screws), the “UV LEDs always on”-problem returned. Using a multimeter and oscilliscope to check everything, but without results. I grounded the motherboard (via screws) to casing. I used transistortester to check MOSFET for damge. I even replaced several times the MOSFET although it tested to be OK. Still nothing helped.
Just by accident, I figured out, whenever inserting even a single screw to fixate the red plate to the casing, the problem occurs. And when none is inserted, everything works fine. This happens only if the FFP-cable is connected to motherboard.
I can go in depth, but bottomline is :
The PCB is fixated via a metal bracket and 4 screws to the red plate. These 4 screw connect GND of LCD panel to the red plate. The red-plate is electronically connected to the casing via another set of screws…. I removed the 4 screws and used small pads of double sided tape to mount the PCB isolated to the metal bracket. BINGO, it works fine !
My impression : All 3 mentioned flaws together might result -in time- to the “UV LEDs always on” problem for a machine. There’s likely a power-imbalance, because of “non-optimal board design” and remarkable values for caps which were not advised by manufacturers. Besides, there’s a grounding-problem.
I’ve solved the flaws by removing / replacing some caps, cutting a part of trace to the LM2596 while using a Shottky-diode to bridge it to get stable powerregulation from 24V to 5V to 3.3 to even 1.8V DC. Also I replaced Schottkly-diode Q5 by a heavier version. Note : Q5 had to do about the upgrade-kit I use nowadays for my Saturn to shorter curing-times (read : faster printing).
And yes, the inverterboard also causes a problem, generating spikes. The side-effects of several imbalance at power-generation, spikes might not be well filtered. All togehter might cause the problems we face with this nice machine.
I hope you (and Jan Mrázek) can appreciate the story / explanation.
Just my 2 cents
Hi guys and thanks for all the feedback, my machine just got the “always on” “feature”.
Reading all your comments I will try first replacing the broken MOSFET with the non needed one. I ordered 5PCS WSF40N06A MOSFET 60V 50A TO-252 to replace them. I hope these work when I get them and replace them. I’m ok with soldering and have a heat gun. But I know nothing about values and how to. Always thankful for all you experts telling us what to do and how to fix things.
I’ve also had this same issue just recently on my Saturn OG model – the troubleshooting on this blog was spot on and now I have a MOSFET to replace; other than the original model, is anyone knowledgeable to promote a list of alternatives? I’m in Europe and I have tried getting the models listed in the comments as well, but no luck;
Thank you in advance for all the help!
Try TME.eu or mouser.com. They should have alternatives mentioned in the comments.
Hi Jan, thank you for replying – I found only this one to be in stock – for 60V 50A to-252 -> FDD10AN06A0 from ONSEMI – would this one be ok to use as replacement?
https://www.tme.eu/pt/details/fdd10an06a0/transistores-com-canal-n-smd/onsemi/
Again thank you for your patience and help, if ok I would jump in to order a few and prepare for the repair job!
That transistor is not suitable as it won’t turn on a 3.3 V signal. But for instance, IRFR8314, (https://www.tme.eu/cz/details/irfr8314trpbf/tranzistory-s-kanalem-n-smd/infineon-technologies/) should do just fine.
Hi Jan, thank you once again for taking the time, I would have definitely failed by ordering the wrong one!
Hopefully, my next message here will be to confirm success with the repair!
Hi Jan,
Leaving a follow-up on my situation: the burnt MOSFET seemed to be fused with the pad, to the point that when I was trying to remove it using a heat gun, I was already able to remove and re-apply the ones next to it, but that one just wouldn’t snap off – I ended up giving up and ordered a new board…
On the new board, the MOSFETs read as follows, from left to right, the 2 contacts pointing down:
PT D50N06 LD5DJ | PT D50N06 LD5DJ | PT D4080 MD1DJ
(are these basically the same, or better, or worse than the originals?)
After installing the new board, all good again – the UVs turn off as expected, after each layer run and also at the end of printing.
As for the stuck MOSFET on the original board, I decided to go all the way on it to see at what point it would come off – and by then, the actual strip was coming off and the chip was still stuck to it, I had to peel it off forcefully even while using the heat gun… so I’m unsure what I could have done differently (not very experienced with heat guns to remove/re-attach components) but if anyone knows of any good “trick” to apply in a situation like this, please let me know; I still have the good MOSFETs I ordered as backup, in case I need to do the procedure again.
In any case, thank you again for the work you did and let’s hope this new board – and any other models currently shipping out – won’t suffer the same fate any time soon…!
Thank you! This was going to be a disaster for me until I found this page.
I have an old Saturn – second batch pre-order. Driver board has May 2020 printed on it – so not sure if that means it’s old one or new one.
I had some mosfet on hand that I hoped might be ok: NTD3055L104 https://www.farnell.com/datasheets/2354279.pdf The current at 3.5V gate is only specced at around 8 amps which is a lot less than the original WSF40N06 or your IRLR8743PbF – apart from that though specs seemed ok, so I risked it – and it just finished a 6 hour print.
Fine so far – and if it fails I know what to do. Thanks again!
Hi, i have photon s with old v1.4.0 motherboard and come across same problem.. checked whole thing and everything is looking ok to my eye, just wanted to ask if u know where should i check on my board if its same problem? Thanks John
Just wanted to mention, I’m in the middle of doing the MOSFET replacement. MOSFET’s should be here this afternoon. My June ’21 Saturn 1 just started doing this. But before I started working on the board, I emailed elegoo tech support. Now that I’m to the step where I’ve cleaned the pad and can solder in a new MOSFET, I noticed I got a reply from 8 hours ago mentioning they have a new motherboard that prevents this from happening and would send it to me free of charge. They requested a video of it exhibiting this behavior, but unfortunataly at this point, without that MOSFET installed, I’m too far into the process for that. I was able to power it up to give them firmware information. But to anyone still dealing with this issue that reads this in the future…check with Elegoo first!
Cody, how long did it take them to reply and which email address did you use? I have just sent a message to 3dp@elegoo.com. I had my printer in a polystyrene box in the garage to keep it around 20 degree when printing and found everything at 48degC, the print finished and backlight still on.
Hello, i can not find a mosfet D4080 LI3DJ . Is there an alternative i can use instead? It is a Mars 3 Pro Motherboard.
Thanks in advance.