Re: Power Supply failure

Not to rain on your fantastic parade or analysis work at all, but FWIW the 3.5.2 update suppressed nearly all the clicking as a result of the PWM changes they made. 😉

Not raining at all. I haven't updated my printer yet, and need to. It's got about seven hours left on a 25 hour print job and then I'll see about doing that. I'll be curious to see how the waveforms change.

Re: Power Supply failure

As am I. I noted that rather than blinking on and off the heatbed LED now stays on but pulses between a low, middle and high brightness, so I'm guessing the change is three different current steps instead of on-off

Re: Power Supply failure

As am I. I noted that rather than blinking on and off the heatbed LED now stays on but pulses between a low, middle and high brightness, so I'm guessing the change is three different current steps instead of on-off

Ah, this explains a lot.

What was happening before was that a very low frequency PWM was being used. Looks like around 8Hz. At the beginning of this, current draw would go from 0A to about 8A. It would stay at 8A for some period of time, maybe 25ms or so (depending on temp) and the drop back to 0A. The reason for the clicking is what this does to the power supply, as specifically the front end boost converter. When bed current (I'm assuming the hot end is off) is 0, then the current draw for the printer is very low and current running through the power supply is even lower. Consider this to be the idle state of the power supply. Now, suddenly, the heatbed current shoots up to 8A. The boost converter PWM doesn't change right away. It won't change until the boost capacitor voltage drops below a certain point. I don't know what that point is, but it is likely relatively high, around 380-390V. The problem is that the boost converter can't start recharging the capacitor until the next zero crossing of the AC waveform. This is what preserves the power factor. AC input current is always in phase with AC input voltage. So, in the worst case, it will be 8.3ms before the boost converter can start charging the capacitor again. Assuming downstream from the boost converter, the efficiency is 90%, then at 8A output voltage, we will be pulling about 1.8A from the boost cap. This means the AC/DC converter is presenting a 222 ohm load. A bit of math says that the boost cap voltage will have dropped to below 325V (from a max of about 400V) before the boost converter can start recharging it. Because of this, the boost converter PWM goes from zero to full blast resulting in a very sharp current spike through the boost inductor. This causes the click you hear.

What Prusa seems to have done is increase the PWM frequency to about 60Hz, or so -- 62.5Hz is what it looks like. By using a much shorter PWM period, the bed PWM current looks much closer to an average current to the power supply. It doesn't see is as a hard off/on swing between 0A and 8A, but an average current of maybe 1A at the lowest setting, followed by 4A at the mid setting, and 8A at full on (complete guess at the levels). This does two things. First, it keeps the front end boost converter PWM active, though at a very low duty cycle when the bed PWM setting is low. Second, the transition from low duty PWM to high duty PWM in the boost converter is now more gradual and so the current swing in the inductor is not so sharp meaning the magnetic flux change is slower and so isn't nearly as likely to result in audible magnetostriction.

Because the bed PWM is no beyond persistence of vision threshold, we see the LED as on at different intensities, rather than blinking.

Re: Power Supply failure

Ah, this explains a lot.

What was happening before was that a very low frequency PWM was being used. Looks like around 8Hz. At the beginning of this, current draw would go from 0A to about 8A. It would stay at 8A for some period of time, maybe 25ms or so (depending on temp) and the drop back to 0A. The reason for the clicking is what this does to the power supply, as specifically the front end boost converter. When bed current (I'm assuming the hot end is off) is 0, then the current draw for the printer is very low and current running through the power supply is even lower. Consider this to be the idle state of the power supply. Now, suddenly, the heatbed current shoots up to 8A. The boost converter PWM doesn't change right away. It won't change until the boost capacitor voltage drops below a certain point. I don't know what that point is, but it is likely relatively high, around 380-390V. The problem is that the boost converter can't start recharging the capacitor until the next zero crossing of the AC waveform. This is what preserves the power factor. AC input current is always in phase with AC input voltage. So, in the worst case, it will be 8.3ms before the boost converter can start charging the capacitor again. Assuming downstream from the boost converter, the efficiency is 90%, then at 8A output voltage, we will be pulling about 1.8A from the boost cap. This means the AC/DC converter is presenting a 222 ohm load. A bit of math says that the boost cap voltage will have dropped to below 325V (from a max of about 400V) before the boost converter can start recharging it. Because of this, the boost converter PWM goes from zero to full blast resulting in a very sharp current spike through the boost inductor. This causes the click you hear.

What Prusa seems to have done is increase the PWM frequency to about 60Hz, or so -- 62.5Hz is what it looks like. By using a much shorter PWM period, the bed PWM current looks much closer to an average current to the power supply. It doesn't see is as a hard off/on swing between 0A and 8A, but an average current of maybe 1A at the lowest setting, followed by 4A at the mid setting, and 8A at full on (complete guess at the levels). This does two things. First, it keeps the front end boost converter PWM active, though at a very low duty cycle when the bed PWM setting is low. Second, the transition from low duty PWM to high duty PWM in the boost converter is now more gradual and so the current swing in the inductor is not so sharp meaning the magnetic flux change is slower and so isn't nearly as likely to result in audible magnetostriction.

Because the bed PWM is no beyond persistence of vision threshold, we see the LED as on at different intensities, rather than blinking.

The most excellent explanation I have read in quite some. Easy enough for a non-electrical engineer like me to understand.

Re: Power Supply failure

The most excellent explanation I have read in quite some. Easy enough for a non-electrical engineer like me to understand.

Thanks for the kind words.

As a follow up, I upgraded to the 3.5.2 firmware. I haven't gone through the code changes in depth, but it looks like there are 128 PWM levels for the bed. I haven't gotten to put it on the scope again since the upgrade though. It's in its new enclosure busily printing away. I can definitely see the heat bed LED varying in intesity. It does go all the way off at times (printing PLA in an enclosure with 60C bed), but the transition back to on is initially low intensity/gentle.

I would be very interested to know if anyone running 3.5.2 and the stock Prusa power supply is still getting clicking of any significant amount.

In general, I would consider both the tendency of power supplies to fail when running on 120V, as well as the clicking, to be resolved issues. And it only took 65 pages of posts! 😆

Re: Power Supply failure

I would be very interested to know if anyone running 3.5.2 and the stock Prusa power supply is still getting clicking of any significant amount.

In general, I would consider both the tendency of power supplies to fail when running on 120V, as well as the clicking, to be resolved issues. And it only took 65 pages of posts! 😆

I have the 3.5.3 firmware stock on a new Mk3S Assembled I received last Friday. I do notice faint clicking/ticking from the PSU.

Re: Power Supply failure

I would be very interested to know if anyone running 3.5.2 and the stock Prusa power supply is still getting clicking of any significant amount.

In general, I would consider both the tendency of power supplies to fail when running on 120V, as well as the clicking, to be resolved issues. And it only took 65 pages of posts! 😆

I have the 3.5.3 firmware stock on a new Mk3S Assembled I received last Friday. I do notice faint clicking/ticking from the PSU.

There will always be some amount of electronic noise. I put my Mk3 in an enclosure this weekend and when I stick my head in, I can hear the slight hiss from the Einsy Rambo. When the system is idle, I can hear some amount of noise from the Meanwell supply I use. However, the Prusa supply, when I had it installed clicked like someone tapping a fingernail on my desk. It was loud and obvious. I haven't put the Prusa supply in since I've gone to 3.5.2 on my printer so I can't say exactly how the sound profile has changed.

Only having one printer, and actually wanting to use to print stuff, somewhat limits my investigation time. 😉

RE: Power Supply failure

Does anyone know what the panic board does in the old silver power supply?  There is a single red wire that gets connected to the +24.  Is this just a simple voltage meter, and if the voltage sags, it opens the circuit to the einsy board?  The new power supply has what I think they call as 24vdc OK contact on it, so the prusa doesnt need the little board anymore.  Has anyone ever had this input do anything?  I feel that if the supply dies, the whole thing goes off, and never really browns out, for the panic to activate.  I also think it was bad practice to just land the tiny wire along with the spade terminal, it should be crimped inside the terminal as well.

RE: Power Supply failure

The idea is to let the CPU know ahead of time power is going away soon (DC hold-up time). It lets things get "put away" more cleanly - like write where in a gcode file it is to eeprom for later recovery.   So yes, the 24vOK is used, and important if you ever want to restart a print after a power fail.

RE: Power Supply failure

So have many of you replaced the silver with a black PSU?

RE: Power Supply failure

i think the new power supply just has the power fail circuit integrated into the supply instead of being an add on

the 24V is just to provide power to the sensor circuit, the sensing is done on the AC mains to catch the failure before the input filter caps start to discharge

RE: Power Supply failure

Posted by: @charles-h13

So have many of you replaced the silver with a black PSU?

Still using original (in NA, 120V). Mine is connected through APC UPS which is doing a great job filtering all the power grid spikes. I still believe it's part of the issue why many PSU failed in NA. A fluctuation in range of 100-140V is common in some regions.

RE: Power Supply failure

A number of people fit Meanwell LRS-350-24 power supplies, they are quite well engineered and not to pricey, the 350 is 350 watts so are higher powered than the original PRUSA PSU's, but you will need to print a new PSU holder to fit it into as it is thinner and the fan... If anyone tells you it is quiet they are lying to you.

I printed a new cover for it and fitted a 92mm Noctua NF A9 FLX fan into it, now that is truly quiet, I thought it wasn't working at first until I looked.

RE: Power Supply failure

@chocki

What did you do about the panic PCB that was inside the old 3d printed housing?  Also are your parts on thingiverse?

RE: Power Supply failure

there are meanwell mounts on thingiverse that have the mounting for the power fail assy

RE: Power Supply failure

The power panic board is soldered to the back of the power input socket and is refitted into the new case.

Here is an example case, they all pretty much look like this: https://www.prusaprinters.org/prints/5393-mount-for-lrs-350-24-to-mk3

This one looks particularly good and if I had seen it first, it's what I would have printed, as is I printed something similar.

I do plan on remixing this to accommodate the 92mm fan cover but have to finish a few more things first.

You may have to buy some new screws though, some domed head and countersunk screws as needed. In all cases, the spacer is screwed onto the power supply and you must make sure you don't screw more than a few mm into the PSU sides.

RE: Power Supply failure

Posted by: @chocki

A number of people fit Meanwell LRS-350-24 power supplies, they are quite well engineered and not to pricey, the 350 is 350 watts so are higher powered than the original PRUSA PSU's, but you will need to print a new PSU holder to fit it into as it is thinner and the fan... If anyone tells you it is quiet they are lying to you.

I printed a new cover for it and fitted a 92mm Noctua NF A9 FLX fan into it, now that is truly quiet, I thought it wasn't working at first until I looked.

Careful with Noctua fan installation. You can get them very quiet (usually there are 2/3 cables included which you can choose from) but the air flow is also decreased. Some PSU relies on the air flow burst.

RE: Power Supply failure

I don't fit any of the additional cables, but the tacho wire will have to be disconnected as it is not used, I cut it back so that it is in the sleeving which protects the wires, and I fitted the correct connector to plug the fan into the power supply where the existing 12v fan went. It runs less duration than the original fan did and less frequently and the PSU remains cool to the touch even during a 200% speed print.

I think the larger opening where the fan sits actually helps to keep the PSU cooler passively as any heat can just escape past the fan when it is not running. 

RE: Power Supply failure

@allen-m2

Does this fan work in the

Mean Well NES-350-24 24V 350 Watt UL Switching Power Supply 120 Volt

Yet Another Power Supply failure

My power supply always clicked during bed heating. It shipped to me on 03/20/2018, and died 1/27/2020. Getting a Meanwell

NES-350-24