Repairing Two PSU60B Power Supplies from a Mini PBX

Today's article is about repairing—or perhaps restoring—two power supplies from a mini-PBX. They're quite unusual, with some rather unconventional circuit designs, so this will be a long and fairly detailed article. I hope it will be useful for beginners and interesting for experienced repair technicians.

A reminder: to keep up with new posts on the blog, I recommend following me on Instagram, where I'll post notifications about every new article and maybe write about something interesting from time to time — link to my account

You can also subscribe to updates and new articles on the Telegram channel — https://t.me/KirichBlog

Once again, I'm repairing power supplies sent to me by one of my readers. This is another clear example of when repairing a power supply makes economic sense. In this case, the power supplies are part of a mini-PBX, and as you can imagine, simply replacing them isn't an option. The PSUs are tied to the system both mechanically and electrically, and they're also quite old, so even just finding replacements is a real challenge. And even if you do find replacements, they'll be equally old units pulled from another system. Replacing the entire mini-PBX would be not only expensive, but labor-intensive too—which ultimately makes it expensive as well.

There are two units, identical in every way and showing the same symptoms: they won't turn on. The backstory is simple enough: a power outage, a voltage surge, and now we have two black bricks.


The case is made entirely of black anodized aluminum, with a plate on the bottom—what's now fashionably called a "baseplate." There's a label with the specifications and a serial number; the last two digits differ. So these power supplies clearly came as a pair and are identical.


We remove the top of the case and find a pair of handsome units. You can see they've been through a lot, but they still look pretty impressive.


A closer look.
The bottom of the case—the chassis—also serves as a heatsink, although some of the heat is transferred to the top as well. There's a substantial mains filter, input capacitors, a transformer, plus a couple of additional transformers and heatsinks.


At the input, there's a mains filter and, notably, two fuses—one on the neutral and one on the live wire. Interestingly, both fuses are intact in both PSUs.


1, 2. There are also two varistors at the input, rated at 470 and 1800 V. They're intact too.
3, 4. The input capacitors have been replaced, but they clearly look used. Both PSUs have 470 µF, 250 V capacitors.


Now we're gradually getting to the interesting circuit details. First, you can see two high-voltage power transistors pressed against the heatsink. Another component is visible on the heatsink to the left: a triac. Next to it, a resistor in a ceramic case is wrapped in heat-shrink tubing.
This is why the fuses didn't blow. Instead of the usual input thermistor, there's a 10 Ω resistor that gets shorted by the triac once the PSU starts up. The failure probably went something like this:
1. A high voltage appears at the input.
2. The PWM controller tries to start, and the high-voltage transistors break down.
3. Since the PSU hasn't started yet, the current-limiting resistor acts as a fuse.
4. The input circuit opens, and the transistors are shorted.

The only curious thing is that the transistors are rated for 500 V and the varistor for 470 V, yet the transistors failed while the varistor is intact. My guess is that this happened because the PWM controller was trying to start and the voltage spikes exceeded 500 V. Under steady-state conditions, the transistors would probably have held up, especially given the circuit design on the hot side, which I'll discuss a little later.


One unusual feature is that the PWM controller doesn't drive the transistors directly, but through a driver—and a rather clever one at that. What's more, one of the transistors is also driven through an isolation transformer.


Before I even received these power supplies, I took a look around online to see if other repair technicians had run into similar problems. It turned out there were not only mentions of the issue, but also a short repair article. By the way, many thanks to the author for the schematic of this power supply.



But judging by the article itself, the author isn't particularly knowledgeable, or belongs to the old-school Soviet-era repair technician camp. Let me explain.
The author writes (despite having the schematic in front of them):
1.
It’s like an ATX power supply only in that it also runs on mains voltage and has a 5 V output. That’s it.

2.
What is there to figure out? The schematic is right there.

3.
There’s no “standby supply” here at all. This is the transformer providing galvanic isolation for the upper transistor, which is clear even without the schematic: there’s only one PWM controller, with no built-in high-voltage switch, and there are no others nearby.

4.
This isn’t a “startup” capacitor. Yes, it essentially enables startup, but it keeps working afterward too, so it’s simply the filter capacitor for the PWM controller’s supply.

5.
Judging by the schematic, it’s not just T4 but T3 as well; the circuit is quite interesting, but that’s beside the point. Once again, it’s called a “startup” capacitor. To be fair, though, the author deserves thanks for listing the capacitor values, since the power supplies I got had already been repaired.

6.
Replace just one of the three connected in parallel? What, you can do that? :)

7.
What can you tell from checking whether a power supply turns on with no load?

Yes, of course, you could ask why anyone needs to figure out what the engineers have designed if all they want is to solve the problem.

I don't want to accuse anyone, but before you start repairing a power supply, you should at least have some understanding of how it works and what its typical quirks are—especially since, judging by the article, the person isn't doing this on their own, but most likely at some kind of service center. And, by the way, we'll come back to how service centers work.


Here are the secondary transformers. You can also see the main channel's diode assembly (on the left) and the transistors on the heatsink.


The 48 V channel (the main one here) has 6+1 470 µF, 63 V capacitors. It looks like Samsung WL capacitors were originally installed; the service center replaced them with the same type, but from the RD series.
The second photo shows the PWM controller for the +5 V channel. The -5 V channel is also implemented here, but +5 V is the main channel, while -5 V is regulated by a linear regulator.


There's also a small submodule that regulates the current and voltage of the battery charging channel. The ATS is designed to connect to a 48 V battery.


1. There are two 3 A fuses on the output. They're not ordinary fuses: each filament has two little blobs on it. I'd guess they're slow-blow, though usually the blob is in the middle of the filament. Perhaps this design ensures the circuit breaks reliably at 48 V, but I can't say for sure.
2. The output heatsinks carry the linear regulator, the emergency power transistor, and a pair of synchronous rectifier transistors for the +5 V channel.
3, 4.. There are quite a lot of capacitors here too—I'd even say more than enough.


At first, everything looks more or less decent—until you turn the board over. Then you see the full "beauty" of the previous service center's work. Judging by the sticker, the repair was done on March 5, 2018.
Yes, I was genuinely surprised. I could maybe understand some rosin residue, but the board is smeared with some kind of mystery flux that has hardened over time.


The diode bridge in the high-voltage section of one of the power supplies had been desoldered, and a resistor had been soldered directly across the capacitor terminals for some reason (I'll explain later why it was done that way). There's flux on the connector contacts too; you can see that the wires were soldered on for testing.
There are also cracked solder joints in places on the board. I'm not sure it's easy to see in the photo, but these are the leads of a high-power transistor in the high-voltage section, and you can see the ring-shaped cracks.

I should say a few words about the board itself. First, it's made of phenolic paper laminate, but the traces are an absolute disaster. At first I wanted to compare it with what was used in Soviet electronics in the 1980s, but no—theirs was better. This is genuinely worse than cheap Chinese electronics from the 1990s. The traces come off not just from the soldering iron, but from so much as a dirty look. Here are a few examples:
1. I removed the submodule as carefully as possible: first I wetted all the contacts with decent solder, then heated them with a soldering iron with a wide tip and soldered it back in. The board swelled a little in that spot, and one pad came off—fortunately, it wasn't connected to anything.
2. I noticed a 1 W resistor mounted crooked on the board. I tried to bend it into place and tore off a copper area.
3. What's more, one submodule was loose. The traces on the module itself had clearly come off, which is why I desoldered it.
4. Some of the small capacitors were already hanging by their detached traces.
5. The diodes and transistors along the edges of the board, designed to be clamped against the heatsink, also bend easily and take the traces with them.

The board is laid out so that when you pick it up, all those side-mounted diodes and transistors end up right under your hand. I had to stay alert and only grab it where I knew there were no components.
During the repair, I figured out why one submodule had come loose: it's taller than all the other components, so when you put the board down component-side up, it rests on the submodule. Move the board even a little, and you move the submodule too. I had to put something underneath so the board wouldn't rest on it. In short, I’ll put it this way: you need to be extremely careful working on this power supply. This isn’t one of those modern circuit boards.

I found some interesting resistors on the board. Early on, I tried to find the primary-side current-sense resistor, but there was no familiar-looking resistor, and the ones that were there looked a lot like capacitors.
There are three of these resistors on the board: in the primary circuit, in the 48 V channel, and in the secondary supply circuit. I don’t remember the first one’s value, but it’s connected in such a way that it’s unlikely ever to burn out.
By the way, you can see that the leads are different lengths, just like in ordinary cheap electronics. Some components are genuinely soldered in any old way...


And now we get to the actual diagnosis and repair.
As I wrote, the PWM controller doesn’t drive the transistors directly, but through a driver. That suggests the PWM controller itself is intact, so diagnosis comes down to checking the driver.


And here we discover not only that a relatively rare driver is used, but also some peculiarities in this power supply’s circuit design.
First, the driver: it’s a TC4428. There are several in the series: the 4426, 4427, and 4428. The first two differ in whether the output signal is inverted, while the third has two different channels, one non-inverting and one inverting.


Many manufacturers make drivers like these, but this unusual configuration means they’re not in much demand, so they’re rarely available for sale. International Rectifier also makes a similar driver, but it’s a little different: its inputs are inverted rather than its outputs. So one driver is controlled by a low signal and the other by a high signal. And since this power supply’s circuit design is quite unusual, it’s best to look for the same part that was originally fitted.


I started checking with the driver, since the transistors were definitely shorted.
It’s very straightforward: apply 15 V from a power supply to the PWM supply capacitor, then first check the driver outputs. One should be at 0 V, the other at 15 V. In practice, you only need to check the lower driver, because if it’s working, the other one is too.
First I checked the output of the lower-switch driver, then applied 15 V to its input. The voltage appeared at the driver output too, so everything was OK.
Warning: don’t apply a signal to the upper-switch driver input, because it will pass through the regulator and diode to the PWM controller output.


A few notes on this power supply’s circuit design:
First, it’s quite unusual.
Second, the current-sense resistor isn’t in the transistor circuit, but in the transformer circuit.
Third, at first it might look like a half-bridge, but it isn’t. As far as I can tell, it’s a flyback converter.
Fourth, the transformer is isolated by a capacitor.
Fifth, the transistors don’t just turn on alternately; one of them is always on.
And sixth, the dead time is generated using RCD networks. There’s a capacitor at the input of each driver: one charges quickly but discharges slowly, while the other does the opposite. Zener diodes are also added.


Interestingly, the secondary converter uses the same circuit, except for the driver, which is built from discrete transistors.
Why it was done this way remains a mystery to me. Maybe someone has come across this and can explain? I got the impression that the engineer knew only one circuit design and used that.
And what about the solution with two transformers?


Let’s set the schematic aside for now and continue with the diagnosis. At this point, it was clear that at least the transistors and the current-limiting resistor had failed.
But we know this is an old power supply, and capacitors are usually the first things to die. Besides, they’ve been replaced here before, so let’s check their condition.
First, the capacitors in the main 48 V channel.
1. 3.2 Ω—that’s rather high.
2. For comparison, a capacitor from the same manufacturer with the same capacitance but rated for 100 V has an ESR of 52 mΩ.
3. Out of curiosity, I measure all the remaining 5 (6) capacitors together on the board and get a fairly decent result: 124 mΩ, well, at least compared with 0.5–1 Ω.
4. I desolder just one capacitor, the one after the inductor, and, as expected, the result changes noticeably. This capacitor usually lasts longer than the others, so it was affecting the test result.


Here’s a closer look at the filter capacitor section. How can anyone make such a mess of a circuit board?


But that’s not all. Look closely and you can see solder blobs stretching onto the adjacent traces.
Worse, solder even got onto the top of the transformer, where it solidified....


1, 2. I desoldered a couple more capacitors. As expected, their readings were all over the place, but they’re all going to be replaced anyway.
3, 4. At the charger channel output there was a WL-series capacitor, and it had clearly been desoldered from somewhere. The whole thing was a mess: 50 Ω and 20 µF. And since it was in an unloaded circuit (no battery connected), they hadn't even tested it—they'd simply moved it from one spot to another.


After looking at this, I decided to replace all the capacitors. Most of them needed replacing anyway, and the cost of the rest wouldn't make much difference.
There was a problem, though. I wanted to use low-impedance Samwha WL capacitors, which were what had originally been installed, but it turned out they... wouldn't fit. Yes, that's the paradox.
The explanation is simple: a 470 µF, 63 V RD-series capacitor is 12 mm in diameter. A capacitor with the same ratings from the WL series, desoldered from the board, has exactly the same diameter, but is taller.
When I tried to find the same capacitors, it turned out that WL-series capacitors are 16 mm in diameter and 20 mm tall everywhere—that is, they have the same volume, but are now wider and taller, so they don't fit on the board. The only suitable alternative was low-impedance Panasonic capacitors, which had the same diameter but were 30 mm tall. They're good capacitors, but at seven times the price, and with this many of them, that put me off.
In the end, I had to settle for ordering 470 µF, 63 V Samwha RD-series capacitors.


To save on shipping and cost, I ordered everything from one place, the Kharkiv-based Kosmodrom. The total is different here because I took the screenshot almost two weeks later.


In principle, I could have gone to their store, but it was easier to order delivery to a parcel locker. I also wanted to see how their online store works these days.
It's all pretty straightforward: add everything to the cart, where you can see the quantities and total. The process is more convenient at Voron, for example: you can change quantities right there instead of deleting and re-adding an item with a new quantity. Also, Kosmodrom doesn't show stock levels. I think they could learn a thing or two from Voron's ordering process.
Then I entered my details and submitted the order. I got an email confirmation right away, followed shortly by an invoice after they checked availability. I paid, and the next day I got a notification that the payment had gone through.
Here's how it went: I placed the order on Tuesday, it was processed on Wednesday, but wasn't shipped until Saturday—not very fast. But when I placed a second order a week later (ordered Wednesday, processed Thursday), I realized they probably spend the week putting orders together and ship them on Saturdays.

In short, their service could use some work. But Kosmodrom's strengths are its selection and the fact that the parts are genuine branded products in the vast majority of cases. Their closest competitor in that respect is probably Radiomag, but they have their own quirks.

The invoice looked like this. I also ordered a few other odds and ends for myself.


The order shipped on Saturday. On Sunday evening, my wife and I went for a walk and picked it up along the way. The photo shows the parts for the power supplies. Everything was packed neatly, each item in its own bag with a number matching the invoice.



Time to get started on the repair. First, I desolder everything from the board that I'm going to replace, but before that I always strongly recommend photographing every step.
By the way, I didn't know the values of some components because they had already been replaced. I used these:
1. High-side PWM controller supply capacitor — 100 µF, 35 V
2. Low-side PWM controller supply capacitor — 47 µF, 35 V
3. Small capacitors on the submodules — 1 µF
4. Small capacitors on the main board — 2.2 µF
5. Main channel capacitors — 470 µF, 63 V.
6. Capacitor in the charging circuit — 680 µF, 63 V (originally 470 µF)
The rest can be seen in the previous photos; I used the same values as before.



And of course, I clean the underside of the board. I should probably swear here, because I had to clean it with acetone. Alcohol? Flux remover? Forget it—acetone barely made a dent in this nasty stuff. I really had to scrape it off.


From the very beginning, I was also puzzled by why one of the resistors had been moved to the underside of the board. I thought it might be to keep its heat from affecting other components, but the explanation turned out to be much more mundane: the power supplies originally had a routing error. I noticed it when I started examining the board more closely before modifying it for a circuit with a single high-voltage capacitor.

Yes, if you look closely, you can see that originally (highlighted in red) the resistors were arranged like this:
the first in parallel with the lower capacitor
the second in parallel with both capacitors.

That is, The voltage across the upper capacitor was always higher, because in the normal configuration these resistors are needed not only to discharge the capacitors, but also form a voltage divider that balances the voltage across them.

In the end, at the repair shop (judging by the soldering), one of the resistors was moved to the lower position and connected correctly.


Just for the record, I measured the capacitors' parameters. The ESR of the 680 µF capacitors is noticeably lower than that of the same type rated at 470 µF. My original idea was to use 680 µF capacitors, but I didn't want to take the risk; there's no telling how the power supply might behave in the future.


The leads of any capacitors whose lead spacing doesn't match the spacing on the circuit board must be formed to fit. The previous repair shop didn't do this, which, combined with the poor quality of the board, caused some of the traces to lift.
Forming the leads this way keeps them from sinking into the board, but unfortunately doesn't prevent the pads from lifting if the component is tilted. In this case, glue is the only option.


And speaking of “glue,” as I mentioned, I had to desolder one of the boards because it was wobbling suspiciously. The traces were intact, but had partially separated from the board. I resoldered them, pushed the contacts in a little, and secured them with glue. In this case I had to use Kafuter K-5204K; I didn't have anything more suitable on hand. The glue seeped under the contacts, so it holds pretty well.


The one thing I couldn't find at Kosmodrom was ordinary ceramic resistors, so I had to buy thermistors instead. They cost more, but the manufacturer is Epcos. Rating: 10 Ω; tolerance: ±20%; I max: 5 A.
Of course, with a triac circuit already in place, this is overkill. But buying resistors at another store and paying for shipping again would have cost just as much, if not more.


I decided to use the same transistors as before, IRFP450. If they've lasted this many years, why experiment with replacements? Although I do have some IRFP460s at home.
They look noticeably different from the originals, but I haven't had any problems with power components from Kosmodrom so far, so I think these will last just as long.


I soldered everything back in place and moved on to the preliminary check.


By the way, the board solders perfectly with ordinary solder and flux, without any extra flux. I wash the board afterward anyway, but even without washing it usually stays clean. Why they used extra flux—and such lousy-quality stuff at that—is a mystery. They might as well have used rosin.


Of course, the first power-up must always be done through a light bulb.
I didn't mention above that this was a preliminary check for nothing: in this case, it was done not only without a load, but also with an input capacitor of reduced capacitance. This is to make sure that, in the event of a mishap, the energy stored in the capacitor doesn't burn anything out.


All the voltages were within spec except for the battery charging channel. For some reason, it was only 49 V instead of 54–56 V. And yes, all the power supply's voltages are negative except for +5 V.


The temperature check showed no problems in the primary circuit, but one of the power transistors in the secondary circuit was heating up quickly and significantly. According to the schematic, it's the battery charge regulator transistor. I didn't investigate further right away, especially since the test was done without a heatsink, but I made a note of it—particularly since there was also a problem with the voltage on this channel.


In the first power supply, the input capacitors had a combined ESR of about 500 mΩ. The new capacitor measured 300 mΩ, although, to be fair, I expected it to be noticeably lower.
But there was a worse problem: the maximum allowed height was 40 mm, and I chose the replacement capacitors based on that. Unfortunately, the actual height turned out to be 41.5 mm, and that difference was decisive: the cover wouldn't fit properly, and I didn't want to leave it crooked.


In the end, I had to lay the capacitor down on the board. Yes, it's a clumsy solution, but I was running out of time. I glued it to the board just in case, and wired it so it would stay firmly in place. I forgot to take a photo, but briefly: one lead is soldered with a short piece of 1.5 mm² tinned wire. On the long jumper, I first made a little bend, tinned it to create a stop, then slipped on heat-shrink tubing, bent it into an L shape, pushed it all the way in, and soldered it. This way, I got both a reliable mount and protection against the board traces being torn off.


I didn't replace the input capacitors in the second power supply, though. It turned out that either I measured something incorrectly earlier, or I assumed the new capacitor would have a lower ESR. The combined resistance of the two was about 140 mΩ, while the new one measured 300 mΩ. If the new one had been in the 140–150 mΩ range, I would have installed it simply because it was new, but at 300 mΩ I left things as they were.
The capacitors are used, by the way, but in good condition. I took a photo of the markings just in case.


While cleaning the board, I also noticed soot marks near one of the transistors. I cleaned them off, but when I tried to start the power supply, I heard a hissing sound. I removed the capacitor and found a burn mark underneath it, effectively shorting the drain and source of the upper transistor.
Fortunately, the capacitor itself wasn't damaged, just a little sooty, and I didn't have a suitable replacement at home. I cleaned it, replaced the heat-shrink tubing, and just to be safe, measured its capacitance and tested it at high voltage. Everything was fine.


I cleaned out the burn mark, coated the area with protective varnish, and soldered the capacitor back in.


The second power supply's voltages were all in order: 54 V on the battery charging output. I also soldered in a small input capacitor here for testing, but in series with one of the original capacitors, so it doesn't interfere in this case.


After some searching and comparing it with the working power supply, I found that transistor Q19 in the battery circuit had failed. It's in the middle of the left heatsink.
It had failed completely, right down to the gate.
Interestingly, it had already been replaced. The photo on the left shows the transistor from the second power supply. It's from the same manufacturer, but the markings are clearly different.
I found an IRFZ44N at home to replace it. I suspect the original failed because of dried-out capacitors and increased ripple amplitude.


The transistor that was getting hot is marked in green; the one that failed is marked in red.



Everything worked. While I was at it, I checked the maximum charging voltage: 55.65 V, or 13.91 V per 12 V battery. The adjustment is very smooth, I should note.


Just in case, I took a photo of the power components mounted on the shared heatsink.


After all the work, I cleaned the boards again and coated them with protective varnish.


During the inspection, I also noticed that the silicone insulating caps had not only hardened, but were literally crumbling in my hands in places, so I immediately ordered a new set.


While taking it apart, I also noticed that some of the screws holding the components down were clearly in the wrong places. Of the five screws, four are countersunk and one is a regular screw, and there's a special hole for the regular one in the top cover.


Then came another surprise. The insulators for the TO-247 transistors originally had similar pieces of silicone stuck to them, apparently to improve the clamping. But in practice, if you assemble it as intended, the clamp sits crooked and doesn't hold the transistors properly, so I clamped them down without any makeshift fixes.
Before reassembly, I had to straighten all the power components that would be clamped to the heatsink. This has to be done very carefully: I pressed firmly on the leads from the PCB side to avoid tearing off the traces, and only then straightened them.


You could say the main part of the work is done, and it's time to start testing.


After reassembly, check it again through a light bulb, since there could be surprises.


Testing under a gradually increasing load is essential. I started at 1.5 A and gradually worked up to 3.5 A, then ran it at that current for an extended period. I tested the second power supply mostly at 3.3 A.
Initially, I calculated the maximum continuous power based on the schematic's ratings:
-48 V - 1.5 A
-56 V - 0.4 A
-5 V - 1.5 A
5 V - 8 A
That comes to about 142 W, or 2.95 A, without accounting for the efficiency of the secondary converter.

In fact, according to the label, the maximum power is:
-48 V - 1.4 A
-54 V - 0.45 A
-5.3 V - 1.5 A
5 V - 7 A
That comes to 135 W.

As you can see in the photo, I loaded it at 3.5 A, which at 48 V gives almost 170 W.


Thermal image after a test lasting almost an hour.
First power supply


Second power supply. You can see that the temperature is higher here because I switched to a 3.3 A load almost immediately. Also, the transformer winding temperature of nearly 100 degrees shows that the power supply was running at its limit.


I tested the second unit at 3.3 A not because it was any worse, but because after the thermal test on the first one, I remembered that the output fuse was actually rated for 3 A :)


After testing was complete, I reassembled the units, put all the replaced parts in a little bag, packed everything in a box, and sent it to the customer. You can't see it in the photo, and I forgot to mention it above, but I also removed the 110/220 V switch: it was no longer needed in one power supply, and I removed it from the other just in case.


The person who sent me the units sent a video report after installing them. Here's a screenshot from the video.


And, of course, the conclusions.
Basically, I wouldn't say the repair itself was difficult. In fact, I could have brought these units back to life by replacing fewer components, and they would have worked just as well. The question is, for how long...
The repair turned into a major undertaking because of the “quirks” of the power supplies themselves—their design and circuitry—as well as their extremely poor build quality and the need to fix the previous repair shop's mistakes.
At this point, the power supplies are almost as good as new. “Almost” because it would be better to use WL-series output capacitors rather than RD-series ones, or equivalent parts. As I mentioned, I could have bought Panasonic capacitors, but the 16 main ones alone would have cost around 1,250–1,300 UAH. So I told the customer to bring the units in for preventive maintenance in 4–5 years, if the PBX is still in service.

As for the power supplies themselves: yes, they're truly awful. I haven't seen devices built this poorly in a long time—in fact, I've never seen anything like it, unless you count the blatantly basement-made “Chinese” stuff from the early 1990s. And these are branded units, clearly not consumer-grade equipment. How did this happen? A mystery...

I'm sure many of you are wondering how much this all cost the customer. I won't hide it: it wasn't cheap. The repair of both power supplies cost 3,000 UAH, plus 1,380 UAH for the replacement parts (I didn't charge for parts I bought as spares or didn't use, of course), plus about 100 UAH for shipping. We agreed on a maximum price upfront, and the actual cost came in lower, so I think the customer was happy—and I certainly was too.

That's all from me. I hope this was useful, or at least interesting. This turned into a very long article, but these power supplies are quite unusual—so please forgive me.