Power Supply Troubleshooting - Complex Power Supply

[NanoHawk]

I am working on a power supply for an embroidery machine that I own.  I'm unable to obtain the schematic as the manufacturer will not share it.  They no longer make or repair the units.  There are some 3rd party folks that claim to repair but they are obscenely expensive (1/2 the cost of a used card regardless of what is wrong).  Used units go for $2K or so.  The form factor is a circuit board with a heat sink on the bottom.

The power supply is in a computerized (CNC) Embroidery Machine made by Tajima, a TMFX just in case anyone has a schematic and wants to be an angel.

The issue is a missing 12v leg which seems to only power the CPU which is an NEC embedded computer. 
I have managed to get a high level wiring diagram from the US Distributor and some test voltages that allowed me to confirm what I had suspected.
The power supply outputs are +5, -5, +12v, and some high voltage 120v/80v stuff.

After I confirmed the 12v was putting out approx .5v DC I used my bench power supply to provide 12vdc.  The computer came online and the machine will operate normally. 
I did notice a small amount of AC voltage on the 12v line as well.  Approximately 150mVAC.    The computer appears to be the only thing using 12v in the system, or at least the only thing I can find so far.

I pulled the circuit board once and tested all the diodes and used an ESR meter to test the capacitors.  The capacitors are mostly Nitsuko which seems to be a niche Japanese manufacturer.  There are a couple of Nichicons and a Kemet with what appears to be a 3/12 date code on it.  That suggests it's been repaired once.  All of the capacitors had very low ESR but tested good according to my meter when I tested them in-circuit.  There aren't any visually obvious issues with the capacitors.  i.e. no leaks or blown tops. 

I've also checked the 5 fuses on the board and they all test good with my multi-meter.  I'm using a semi-portable Fluke 37.

Because I'm getting the other DC voltages I don't think the high voltage side is the issue.  If it was I wouldn't have any other low voltages.

I've been able to trace the 12v circuit back a couple of components on the board using the test points (called CP1, CP2) and then checking the diodes I figured out which output diode is on the 12v circuit. 

My next step is to pull the board and put 12v on the line side of the output diode to continue tracing it and testing components. 

What else should I be doing?
Are there other ways to test capacitors that I should consider? 

Based on the age (1994), and that 3 capacitors appear to have been replaced, and that I have some ripple AC current I'm thinking I have a bad capacitor somewhere.  Although it could be a BJT/MosFET (I don't know which one they are using) or a bad transformer coil.  I want to make sure I rule out the capacitors first.

[fanOfeeDIY]

Hi,

It may not be the same power supplies, I remembered there are videos of repaying switching power supplies of Japanese CNC machines.

https://youtu.be/pffOJdCQ7kw

https://youtu.be/mwl74jk8ZyU

[bdunham7]

The first suggestion I would have is to take the very best hi-res photos that you can of both sides of the board and post them here at whatever the maximum quality and size the forum will permit.  I'm pretty sure it can be fixed if is just a power supply.  The trick is to avoid blowing up the rest of your machine!  And post that 'high level' diagram as well.

[TheMG]

Some pictures of top and bottom of board would be helpful in figuring out the overall topology.

Bad capacitors won't cause only one output of a SMPS to disappear completely. You would get a ton of high frequency ripple on that output, but you would still measure a substantial DC voltage, definitely more than 0.5V.

Assuming there is no additional switching or regulation circuitry and the 12V is just another transformer tap -> rectifier -> filtering and out, I would start off by checking for any cracked solder joints, then check the rectifier diode(s), and also do a continuity check of the 12V transformer tap/winding (I have seen on some SMPS where the winding breaks off at the pin of the transformer).

[NanoHawk]

@fanoffeeDIY - I saw the first one, and I'll watch the second one this evening.

@bdunham7 - I'll post those photos when I pull it out tonight or tomorrow.  In the meanwhile I did shoot a YouTube video of my first round.  From what I can tell there are at least 3 layers.  I originally thought it was a 2 layer board, but I saw a few things that suggested another layer.  The video has shots of both sides.  Not ideal high-res shots.  I'll post some more detailed photos that I took at lunch.

I recommend scrubbing through this to find the board shots.  The first part is how to get the card out and I generally shoot my videos from the "hang out while I do this" perspective.



@TheMG okay.  that's an interesting idea to look for the rectifier diode and 12v transformer tap.  It still involves working my way back across the board.  If I was doing the design I'd have had a 24v rail and then stepped it down to 12v and 5v with an inverting 5v converter.  I'm by no means an expert, so whoever did this knew what they were doing.

The two large caps are on the high voltage end.  The pigtailed connectors bring in 120v and distribute the 80v and 120v circuits to the servo drivers.  If I remember right it measured 80vac.

The attached PDF is pretty crummy... but it's all I've got.  Page 2 is the high level wiring diagram.

I think it's also worth noting that I keep the machine plugged into a TrippLite UPS unit.  The whole machine only draws 445watts at 120vac.  This is primarily for surge protection purposes.  The machine was operating fine.  One day I turned it on and the CPU wouldn't come online, which is when I knew I had an issue.  No smoke or other surprises.  I did have a nearby lightning strike that caused some induced current damage in other equipment (a laptop power brick, a NAS power brick, and a communications circuit in an HVAC controller quit working).  It was not a direct hit and I couldn't find any physical damage.

[NanoHawk]

I posted all the photos with my notes about each one to my own website.  Here is the page.  This overcomes the 5MB limit here on the forum.  I'm surprised there isn't a resize function on the image upload.  New phones and tablets increasingly take larger photos.  https://www.nanohawk.com/tmfx-power-supply-notes/

Part of my goal in this project is to help others with similar machines.

The "service companies" have the cards priced at 40% of the value of the machine and 50% of that for repair, regardless of what's wrong.  Irritating to say the least.

I've priced all the caps using Nichicon from Mouser and they are all under $5/ea.

The large 470uF 250V caps have been difficult to find in the same form factor and rating.  Digikey and Mouser don't have exact replacements, but I don't think those are bad.  All of these caps are likely to be beyond their lifetimes regardless of if they are 5K, 8K, 10K, or 15K hour run times.  Being 26 years old even with moderate use they would be aged out. 

UPDATE: I pulled the card and took photos of the front and back and regions on each side using my iPad 2020.  Posted to my website on the TMFX page to work around the image size limit on the forum.

In case anyone is wondering, Tajima makes very good embroidery machines and this time frame was their "golden age" with machines that do a great job, are durable and were assembled in Japan with dedicated, qualified people.

[schmitt trigger]

Thinking outside the box......

If you have confirmed that an external +12V supply allows the machine to operate correctly, and with the low cost of these generic supplies..... in case you can't fix the board, would you consider this "solution"?

I know, I know, it is like cheating a little bit. But if you find yourself painted into a corner, it could be a possibility.

[NanoHawk]

Absolutely.  That's my Plan D.  No way am I going to spend $800 or $2000 on a 12v repair if I can substitute a $50 Meanwell 12vdc power supply. 

That will hurt my resale value eventually though.  So I'd like to fix it and not have a frankenstein machine.  There is also a distinct possibility that whatever went wrong will slowly take out the rest of the power supply.

I'm of the impression that this fix is less than $50 in parts and a little bit of solder time.  I'm treating it as an unplanned educational opportunity. 

[drvtech]

If I'm interpreting the pics on your website correctly, there are a couple of black wire links under the output caps for the 12V supply. These look as they have been added by someone who has inadvertently ripped the core out of a couple of through plated holes while removing the caps for replacement. I would remove these caps (more carefully than the last bloke!) and check what tracks go to these pads on the top side and, if you can see them, on the inner layer(s). On that subject, this is probably a 4 layer board - 3 layers would be unusual. We can see the bottom inner layer in the bottom views, you may well be able to see the top inner layer when you remove the caps.

On the subject of why the output is zero rather than just high frequency a.c., it is conceivable that you won't get any output at all if the caps are disconnected. In this circumstance the PSU management IC may well go into overvoltage shutdown as it will be seeing large positive pulses which are unconstrained by the caps.

Another possibility is that the feed to the caps from the rectifier diode(s) is open circuit because of the damaged plated through hole.

Overall topology of the PSU does indeed look as though it has multiple switch mode modules rather than a single switcher with lots of taps. (Just as you implied when you mentioned multiple transformers).

[NanoHawk]

The black wires are actually resistors.  I think they are bleed resistors from the caps.  They have shrink wrap over them.  It's sort of an odd thing on such a well done board.  I was told about them in advance and advised to leave them alone unless I replaced a failed cap above them, in which case they would usually have blown.

My plan of attack at this point is to inject 12v on the line side of D23 and then look for it with a multimeter.  As I figure out where the circuit goes I'll test each component I find.  If someone has a better idea I'd be open to it.  Tracing without a schematic is not how I prefer to work on things.

I agree, a 4 layer board makes more sense.  I only could observe 3 layers.  The high voltage end is what clued me in to their being a 3rd layer.  Being a specialty Japanese board nothing would surprise me in terms of it being unusual.  What I don't understand is why the supply wasn't more modular with a 120vac/80vac unit and then that supplying a more standard switcher with switching step-downs.  Of course, this is a 1993 design that probably has some legacy influence present from prior designs.

Tajima was widely copied by other manufacturers.  I'm not sure if they copied the power supply setup, but I know the rest of it has been knocked off by Happy and other brands.  It may be worth my time to see if I can figure out what they were doing at that era.  While the mechanical parts vary a little from one machine to the next, the Tajimas used pretty much one control system in their systems.

[NanoHawk]

I spent some more time looking at it after dinner tonight.

I have a e-cap with suspect bottom leakage.  I see some crud around the base of the cap.

When I measure it with my ESR meter it has extremely low resistance.  Less than 1 ohm.  That doesn't make sense to me.

Where can I find data sheets for these Nitsuka (sp?) caps?  They don't seem to make this type of capacitor anymore.  I've looked on their website.

First photo is the suspect area... second is the side of the capacitor in question.  I'd appreciate some help in deciphering the series, etc. 
I can tell it's a 105 degree cap.
I think it's an MR series
I'm used to looking at the values... in this case 25v 2200uF

I don't know what the N988 means.

Thanks

[bdunham7]

When I measure it with my ESR meter it has extremely low resistance.  Less than 1 ohm.  That doesn't make sense to me.

Where can I find data sheets for these Nitsuka (sp?) caps?  They don't seem to make this type of capacitor anymore.  I've looked on their website.

Unless the cap is very small, I would expect an ESR of under an ohm.  That's not low at all.  I just test-installed 2 well-used Sam Young brand 1000uF/50V caps that are about 1/2 inch diameter and 1 inch length--so not very big--and they were ~0.06 ohms.   Your power supply is broken--the type of things  you are looking at would not stop it from working. 

[NanoHawk]

Thanks for confirming that values below 1 ohm are acceptable.  I don't know what I don't know.  The MESR-100 that I picked up has a chart of values for "typical" e-caps.  The values on it are significantly higher.  I initially attributed that to these being made by a specialty manufacturer.

Both of the capacitors in that location test out at .009 which seems rediculously low.  I thought perhaps that a capacitor with too low of values could be bad.  As I stated earlier, I'm treating the repair of this power supply as an unplanned educational opportunity.  Instead of giving my money to the Pirates of the Embroidery Repair business I'm diverting some of it to support my electronics habit and repair this power supply.

I still would like to find a datasheet for these capacitors. 

In the meanwhile, my next step is to put a tracer voltage at low current on the 12v rail and see what components are involved.  As I find them I can test them and eventually that should lead me to something that is not working properly.  Without a schematic I don't see any other way to do this.  I eventually expect to reach a transformer, but I don't know which of the transformers, or which of it's taps is involved at the moment.  Each transformer has multiple taps with the exception of the one near the two caps that I was asking about.

[bdunham7]

Both of the capacitors in that location test out at .009 which seems rediculously low.  I thought perhaps that a capacitor with too low of values could be bad.  As I stated earlier, I'm treating the repair of this power supply as an unplanned educational opportunity.  Instead of giving my money to the Pirates of the Embroidery Repair business I'm diverting some of it to support my electronics habit and repair this power supply.

I still would like to find a datasheet for these capacitors. 

In the meanwhile, my next step is to put a tracer voltage at low current on the 12v rail and see what components are involved.  As I find them I can test them and eventually that should lead me to something that is not working properly.  Without a schematic I don't see any other way to do this.  I eventually expect to reach a transformer, but I don't know which of the transformers, or which of it's taps is involved at the moment.  Each transformer has multiple taps with the exception of the one near the two caps that I was asking about.

That attitude is why my garage is filled with tools to do everything from cutting concrete slabs to accurately measuring HVAC duct pressure. 

Trust me on this--a datasheet for those capacitors would be the least useful thing in the world for you right now.  I haven't had time to look at the photos (and, TBH, I wish they were better) but instead of 12 volts, I suggest you just use a continuity tester and a strong light and try and determine which transformer and which high-side drivers are involved with 12 volts.  If you can connect it to power while it is out on a bench, you could also use voltage testing to trace the other good power supply voltages to determine which components are associated with +5 and -5 and so on, and rule those out. 

Also, if you could start reading the part numbers of the large semiconductors and annotating your pictures with those, it might make it easier for someone to guess the power supply topology.  I don't know the detailed history of low-to-mid power SMPS design, so I couldn't guess what was likely to be in a 1994 Japanese design.

[NanoHawk]

Most of the power semiconductors are wedged against the heatsink at the bottom of the board.  I'll see what I can see and annotate them. 

My guess is that it will be modular and routed as simply as possible to conform to the form factor.  I can see some signs they were bumping up against the number of layers.  There area  few jumpers here and there that jump across traces.  Overall it looks extremely well engineered.  Exactly what I would expect from a Japanese design.  They take such pride in their work, both in design and in fabrication.  I have no doubt that if I had a schematic I could troubleshoot it and keep it going without much hassle.

[asis]

Hello,

If you want a circuit or at least a fragment of it, indicate the types of all microchips and optocouplers.
Also, do not be too lazy to dismantle the heatsink to record the types of MOSFet's and rectifier bridges and half bridges.
The quality of the photo does not allow this.

[NanoHawk]

Had to get around to finding some new test leads.  I swear I own 30 of them but they all ran away or hid in boxes.    Sorry it took time to get to and re-post.

As for taking apart the heat sink.  NO.  Not until I have a reason to.  This is a difficult to get proprietary board with no schematic.  To buy the board would be over $1800 USD if I can find one.  As only one voltage output is not working the risk of doing damage by taking apart the heatsink is too high.  Right now I can simply bypass the 12v rail with an inexpensive power supply if I need to.  If I damage the board I lose that option as a backup.  That's a hack though and I want to figure out what is going on.

I spent some time analyzing the board tonight with my multimeter and a bench power supply putting out 3v and .1A DC as a test current. 

It looks like someone took several modular, independent circuits and married them on a circuit board for production purposes.

The high voltage side outputs 120vac and 80vac (servos).  The machine is 240v compatible so I suspect something on the board handles stepping down from 240v to 120v, but that's not my focus right now.  Could also be a different PCB for 240v.  There are two RBV-1506 bridge rectifiers on the high voltage side. (see image)

There are 3 PCB Mount stepdown transformers that appear to be configured in parallell mode for 115v to their target voltages. 
The board outputs +5, -5v, 12v, and 26v. 
I have no idea what 26v is being used for, but it's not my concern at the moment.

The configuration of the IC's on the heat sink appears to be a fast recovery rectifier paired with a MOSFET.  One pair that I can read is an NEC S20LC40 and a NEC K703.
Datasheet for the S20LC40 (fast recovery rectifier): https://www.mouser.com/datasheet/2/366/J533_S20LC40UV-271056.pdf
NECK703 Datasheet (MosFET): https://www.datasheetq.com/datasheet-download/231726/1/NEC/K703
Apologies in advance for linking to that awful datasheet-spam trap website.  It's at least an NEC datasheet.  The rectifier seems to have gone to China as a generic at this point.

My hypothesis is that each transformer is driving a power rail as an independent circuit with a common chassis ground.  The only advantage I can see to this is good isolation / stabilization of your voltages.  It's not how I would have designed it, but I'm sure whoever did this has more experience than me.   

I believe the capacitor is fed from the MOSFET and rectifier.  I'm not entirely sure how to check the mosfet and rectifier in circuit.

The next thing that makes sense to me is to get an isolation transformer and a variac and put some AC voltage in the high voltage sign and probe the transformers to see if my hypothesis on the circuit design is right.  i.e. what voltage is going into the transformers and what is coming out. 

Once I identify the input voltage I can turn the voltage up to 120vac to see what the output voltage of each transformer is.  That will tell me which one is 12v (or not) and where my +/-5 and 26v are at.

From there I should be able to disconnect power and use continuity to see which rectifier and mosfet are involved.

I think the circuit design is:
Transformer-> Rectifier -> Mosfet - > electrolytic capacitor -> output diode
There should be some sort of sense/feedback loop for output voltage management.  It may be sandwiched in the heatsink and not visible to me.
I think the electrolytic is there to buffer the switching voltage. 

My questions are:
1) Is there a better way to trace and test this?
2) Am I headed along the right lines of circuit design or is there a different way I should be thinking of?

My expectation is that I will find a component that has failed somewhere between the transformer and the capacitor.  It's possible the transformer has gone out as well. 

NOTE: At the top I mislabeled the electrolytic capacitor as a transformer.  Sigh.  long day.  We'll see who reads the whole post.  lol.

[NanoHawk]

I made a little more progress.

[NanoHawk]

And the flip side which contains the NEC chip that regulates the switching power supply

It's designed for up to 31v output.
Here's the datasheet.  https://pdf1.alldatasheet.net/datasheet-pdf/view/6696/NEC/UPC1905GS.html

So it appears to be an intermediate switching power supply that the 12v is stepped down from.

[asis]

Hi,
First of all, don't force events.
Do not apply any power to the PSU until you have a complete circuit diagram and a complete understanding of what you are doing.
Earlier, I asked you to indicate the types of PC1-PC4 optocouplers and U5, U6 microcircuits. If it is LM393, please confirm it (these are soft start circuits for two channels of PWM UPC1905GS).
Since U3, U4 UPC1905GS and U5, U6 (LM393) are two each, this means that you have 2 main channels, respectively, 4 control MOSFET's. You don't want to define their types yet.
In vain, it could speed things up.
I will also ask you to take a photo in the area U5, U6 so that I can see the resistor values.
In fact, the scheme is simple, you just need to make an effort.
Good.
Let's agree so - I will try to help you as soon as I receive the information I need from you.
You need patience.

Vladimir.

[NanoHawk]

I will post some photos.  Not much hope of reading the chip values.  Tried visible and UV light.  I really need a bench microscope, but that's not this weeks' money burn.    Just added a rigol DS1504Z, variac, isolated power supply and differential probe to my collection.

So far every single chip is NEC on this board.  I can make out NEC on U5 and PC1-4

PC3 drives U5
PC4 drives U6

PC1 and 2 head back to the heatsink, which is also the chassis ground.

[asis]

Hi,
Check, specify.
Across the ocean, unfortunately, I cannot do it.