Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

[Per Hansson]

Hi, these two caps came from a unit that otherwise is full of Nichicon caps:
Nichicon GQ, GU & PJ manufactured late 2001.
Got any idea what model they are? (See picture).

The 33uF cap has an ESR of 40 so it's in dire need of replacement
I could get it down to ca 5 by heating it with my iron, then matching the other cap...

[Behemot]

Re: Help to identify surface mount capacitor

No idea, where is it from?

[Per Hansson]

Re: Help to identify surface mount capacitor

They are from a Mitsubishi Melservo MR-J2S-100A servo amplifier.

[Behemot]

Re: Help to identify surface mount capacitor

Not familiar with that, but most of the SMD caps are GP. Maybe you can use some long-life series like NCC LE, those are in between standard and low-ESR caps, ESR-wise.

[mariushm]

Re: Help to identify surface mount capacitor

I'd say they're regular electrolytic capacitors in chip type packages.

I'd replace with something like Panasonic FC or Nichicon PW (both low esr but not very low to cause problems), yes they're through hole but you can lay them flat on the board and bend the wires to solder them on the contacts where the old capacitors were situated.

Could replace 47uF 10v with any 10v to 35v rated electrolytic .. keep in mind that a 47uF 10v capacitor in that size would have had 1-2 ohm ESR in the first place. Don't go overboard, bigger capacitors (in volume) will have lower esr so while you shouldn't restrict yourself to 10v capacitors which may be more expensive as people don't buy them you also shouldn't replace with 47uF 100v or something like that.

[Per Hansson]

Re: Help to identify surface mount capacitor

Old thread revival time
I identified the caps: they are Nichicon UWG series.
I replaced all electrolytic caps including those SMD's with Panasonic FC.
Actually I'll append Elfa orderno's for the replacements below.
Anyway sadly it didn't help.
I now have 2 such servos, they both report too low bus voltage compared to reality:
It starts off at 170v and goes up to around 220v as the servo heats up.
But the actual bus voltage is constant at 330v as it should be from rectified 230v mains.
And if I heat up the servo with a heatgun the monitored bus voltage rises to the actual correct value as measured by my DMM.
Sadly I have not found a schematic and it's a pretty complex device so I have not been able to pinpoint which component it is that likes to be heated.
I'm afraid that it might be something underneath the IGBT module...
Attached some pictures of mine.
Here is some company that repairs them and have posted photos with the IGBT module removed:
http://purelove34.tistory.com/5621

Original:				Elfa		
Capacit	Volt	ESR	Size		Replacement	Quantity
Nichicon GU:
330uF	400v	0.15	35  x 30	167-04-632	3

Nichicon GQ:
68uF	400v	0.25	22  x 25	167-04-625	1


Nichicon PJ:
3300uF	10v	0.045	16  x 20	300-13-687	1
1800uF	10v	0.057	12.5 x 20	300-13-677	1
330uF	35v	0.096	10  x 20	300-13-895	1
82uF	35v	0.31	6.3 x 15	300-13-829	1
82uF	25v	0.47	6.3 x 11	300-13-783	1

Nichicon UWG (SMD):
47uF	10v	1.0	6.3 x 5.4	167-30-997	1
33uF	25v	1.0	6.3 x 5.4	110-69-970	3

[R_J]

Re: Help to identify surface mount capacitor

Is it possible that just the circuit that monitors the voltage is at fault? or does the actual voltage change when checked with a meter, if the actual buss voltage is 330 why does the software report 170 - 220, I would look for a component that is connected to the buss voltage that is used for monitoring, maybe like a resistor divider network?
If heat changes it, try isolating what part you are actually heating up.

[Per Hansson]

Re: Help to identify surface mount capacitor

Of course R_J: That's what I tried to say in my post.
Only the circuit that monitors the bus voltage is at fault.
(The bus voltage itself is 100% stable).
As it is I have not found that component yet.

[Wester547]

Re: Help to identify surface mount capacitor

Could they have been Nichicon UWF or Nichicon UWT too? UWF and UWG are “relatively” low impedance SMD parts, and UWT is a general purpose series. Both the UWF and UWT series also match the markings and case sizes of those Nichicon SMD caps.

[Per Hansson]

Re: Help to identify surface mount capacitor

Yes it's certainly possible.
I'll desolder the remaining ones on the other servo and see how they measure up.
So far I have values ranging from 4.2 to 40 ESR for those caps
Replacing the caps did not affect the bus voltage reading though so I think it's unrelated...

That said on capacitor C6 I saw the voltage seemed to vary linearly with the reported bus voltage as I heated the servo up.
So that might be at least connected to the bus voltage sense circuit.

0.1300 = 270v
0.1495 = 250v
0.1590 = 220v
0.1400 = 208v
0.1300 = 205v
0.1155 = 200v
0.1000 = 192v

Here is something interesting I found to read, anyone ever came across an optocoupler that has become heat sensitive?
https://www.edn.com/design/analog/43...re-coefficient

[R_J]

Re: Help to identify surface mount capacitor

I guess I misunderstood, just asking but does a working unit report the full 330 volts?

[Per Hansson]

Re: Help to identify surface mount capacitor

I'm sure it does, actually I don't have a working unit to try on.
Mainly because the communication cable has been, how do I put this, manufactured somewhat in a hurry by yours truly
Mitsubishi cable MR-CPCATCBL3M has been ordered though, but with the Chinese new year I don't expect it here quite yet.
And then I will need to go to a customer with an actual working machine to verify it on

Well, actually I figured out how to do it without the cable as can be seen in this picture, but that was just a few days ago...

[Per Hansson]

Re: Help to identify surface mount capacitor

I'll desolder the remaining ones on the other servo and see how they measure up.
So far I have values ranging from 4.2 to 40 ESR for those caps

New best ESR value is 2.8 for a used SMD capacitor.
Heating that up I got it down to 0.4 however.
That said I've never seen a datasheet that shows ESR / impedance heat correlation curve at 105°C down to -20°C which most caps are rated for.

Here is something interesting I found to read, anyone ever came across an optocoupler that has become heat sensitive?
https://www.edn.com/design/analog/43...re-coefficient

I found the culprit, I used -50°C freeze spray to look for the faulty component.
As I cooled down O15, a HP 7800 optocoupler the bus voltage reading dropped over 50 volts!
Sadly it's located underneath the IGBT module so I had to desolder it.
I don't have a new one to test with yet, and there is only one on each servo and it's broken on both it seems...

For anyone looking closely these pics are taken of the servo I recapped fully.
On it I used Sanyo WX 50v 100uF as replacement for the Nichicon PJ 82uF caps, otherwise it matches the Elfa replacement list I posted earlier.
The optocouplers are made by HP and Avago interchangeably due to the latter buying out the former, the part number is HCPL-7800
And the current owner seems to be Broadcom, maybe I should go for an upgrade to the "A" version?
https://www.mouser.se/ProductDetail/...7QwLI2575o8%3d

[Per Hansson]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

I ordered the "A" version of the 7800 optocoupler linked above from Mouser, actually it was the only one in stock.
It worked a treat, the bus voltage reading is now correct

[Behemot]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

Interesting, opto's are usually not checked, considered OK. Wonder what happened, if it degraded due to diffusion…

[Per Hansson]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

Started work on the other servo.
Under O15 there was the marking HCPL-7840 #56Y
Quite surprising as there was installed an Avago 7800 optocoupler there.
On the servo in my previous post the marking is for a HCPL-7800 500 and a HP 7800 was installed as mentioned earlier.
(I lifted two pads as I tried to cut the old optocouplers legs instead of desolder them, but I fixed it up ok anyway).
I also attached two high resolution pictures of this second servo, front & back without the caps installed, except for the new surface mount caps I have installed).

[Per Hansson]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

Yea Behemot it's a bit interesting failure mode.
Maybe it's being driven too hard or improperly in this implementation.
But due to where it's mounted it's not really very easy to test that.
And if it lasts another 15 years I'm happy

Of course a photocoupler has both a LED that will degrade, and also a phototransistor that will degrade.
The correlation is described in this article, that said I think a high-end device like this:
Which has a DAC on both the primary and secondary side will behave extremely well over it's lifetime.
So the failure could be caused by something different all together.
As a fun test I wired it up to 5v, and used simply the diode test mode on a meter to send it a signal.
I watched the output on a scope: it was a nice well-behaved waveform.
Then I heated it up and the waveform only became smaller, not distorted in any way.
So it's really a quite linear failure mode, at least it makes the troubleshooting much more convenient
Oh, and the second servo is now completed and seems to behave equally well too

[stj]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

opto failure used to be common, specially in philips tv's
people used to overdrive the led and the output dropped over time.

[Per Hansson]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

Yea, in this case that's not really possible though.
The HCPL-7800 is not driven directly, as I mentioned it's a digital device with a DAC on both sides.
The advantages is that it's a very well behaved device.
It would still be possible to misuse it though, the DAC on each side requires a 5v supply, it's derived in this implementation from a 17v supply i think, with a zener diode to get it down to 5v.
But then the sense pins should ideally be driven with max +/- 200mV, it can operate up to +/- 2v but with reduced accuracy.
In not sure if the problem is there maybe: I could see how an overvoltage during motor breaking would increase the bus voltage for example...

[Leopard]

Re: Mitsubishi Melservo MR-J2S-100A alarm: AL.E9

Hi Per,

Sorry about the thread necro, but anyway:


We are slowly starting to get failures of these Melservo units at work.

What were the initial error codes & symptoms?

We have two units which show AL.33 (overvoltage) and two units which had a hard time getting to a working state, initially showing AL.16 or AL.E9. Once they warmed up, they would work normally, but this took longer and longer as months passed.