Large OLED Display or Building a Powerful Power Supply, Part 2

This review is a continuation of the previous one, but since doing two reviews in a row about the same product would be somewhat illogical and inappropriate, at least the opening will focus on one of the interesting parts I bought for the future power supply.

I bought the display at a store I walked into, but the store also sells online, so technically it qualifies under the site's rules.

While planning the design of the future power supply, I was already deciding which indicator to use.
There were plenty to choose from. At first, I considered these options:
1. Keep the original one. — small and very basic.
2. An indicator using VATN technology. — great, but not very budget-friendly, at around $10.
3. A vacuum fluorescent (VFD) display. This would be fantastic, but it costs even more and is even harder to find, since they mostly turn up on eBay. The main advantages are wide viewing angles and a vintage look. Most are 2002 format, while I needed 1602.

When I got to the store, I decided to buy an indicator using VATN technology after all. These displays have higher contrast and are much better than the familiar LCDs.

But when I pictured how it would look on the front panel, I realized it wasn't right for me—it was too small.
I sketched out the appearance on paper, so I didn't even unpack the indicator I'd bought.
There was only one solution: install a large display. In other words, the same format—two lines with 16 characters per line—but with a larger diagonal.
Besides, after reading the forums more carefully, I learned that VATN displays aren't quite as perfect as they seem. They have good contrast and viewing angles, but they still have the drawbacks of LCD displays, which is what they are, after all.
For example, the viewing angles are wide, but the brightness shifts and the image doesn't look as good anymore.
Still, compared with a regular LCD, the difference is clearly in VATN's favor.

There were several options here, too.
1. A relatively cheap, large LCD. — really bad; the image is awful.
2. The same display using VATN technology, but in a larger size. — it turned out these do exist, but they're almost impossible to buy. I could only find one for sale somewhere I couldn't order from, and even then it was special order.
3. A VFD display. — easier to buy than VATN, but outrageously expensive, and only slightly easier to find than a large VATN display.
4. An OLED display. — This one looked great in every respect except the price. Actually, there was one catch; I'll get to that later.

The cheapskate in me resisted for a long time, kicking and screaming: $35 for a display is a lot.
But a quote from Back to the Future kept going around in my head—
Marty McFly: What, you built a time machine… out of a DeLorean?
Emmett Brown: I figured, if you're gonna build a time machine, why not do it with some style?
First, here's how the four types of displays differ from one another.


Once the cheapskate in me gave in, I decided to look for this display at other stores first. Unfortunately, they weren't available in China or on eBay. The selection of OLED displays is pretty pathetic: tiny little displays the size of a postage stamp, maybe a bit bigger, and that's IT. The largest ones are about the size of a standard 1602 display, but that's as big as they get; I couldn't find anything larger, and even those weren't cheap.
To my great surprise, I found the display at the same store where I'd bought the previous VATN one. The price was high, but definitely lower than elsewhere.
I took the previous display and the receipt to the store, and they exchanged it without any trouble, with me paying the difference, of course.
But since I'd researched it online beforehand, I already knew about a quirk of the controller in these displays.
Technically, the display doesn't work as a 1602 but as a matrix display, and its controller emulates the familiar HD44780 controller, though not always quite correctly.
You can make a small change to the device's program to fix the error, and then everything will be OK. But I couldn't modify the firmware, so I explained the problem in advance and warned them that I'd try it and return it if it “didn't work out.”


The display really is large.
Its dimensions are 122 x 44 x 10 mm. The display model is WEH001602BGPP5N00001, and here's a link to the datasheet.
There is one catch when connecting it.
The pin numbering on the familiar display installed on the converter board goes like this—
1, 2, 3,,,,,15, 16.
The new display has slightly different pin numbering:
14, 13, 12,,,,,,,,2, 1, 15, 16.
Keep this in mind when connecting it.


Since the return was agreed on the condition that the display not be soldered, I used a “no-solder” method to connect it.
But everything went great—well, almost great.
The thing is, when the cursor is displayed on the screen, it flashes for a fraction of a second in random places wherever the cursor can appear.
At first I suspected the display was incompatible with the board, but then I realized this was a case of a good thing turning bad.
The display is very “fast,” with a response time of about 10 µs, several orders of magnitude faster than a regular LCD. If you look closely, you can even see the cursor flicker slightly on an LCD; it’s just that the inactive positions don’t have time to show because of the LCD’s high persistence, while they do on the OLED. I wouldn’t say it’s a serious problem; it’s just noticeable in certain modes.

The OLED display draws about 40 mA, but unlike other types of displays, its current consumption depends on how many segments are lit. The more segments are lit, the higher the current draw.
The display can be powered with 3.3–5 V.
When I connected it with a long cable, an interesting effect appeared: the display powers up gradually, like a vacuum tube.

The color is reminiscent of the good old VFD


But the viewing angles on this display are incredible. I’m willing to forgive its high price and flickering cursor for them. Only a VFD can compete, and even that might not win. And the contrast: bright characters on a completely black background, with no filters of any kind.
I couldn’t find an angle at which the image disappears. Either it’s readable, or it simply “vanishes over the horizon.”
You can see the matrix a little in the last photo.


After finishing with the display, I decided it would be better to also add a filter for the interference generated by the converter’s PWM.
Although the manufacturer claims relatively low ripple, I decided to improve the design because I consider this converter board more of a “semi-finished product.”

So I decided to make an output ripple filter.
I wasn’t aiming to build a large filter, even though I have all the necessary components at home; I wanted to keep it simple.
I made the filter using this circuit:


1.2 For this, I took a couple of toroidal cores from ATX power supplies (radio amateurs usually have plenty of these lying around).
3. I chose cores about 28 mm in diameter and unwound all the windings.
4. Since I don’t have much thick wire at home, I simply straightened the wires I had removed.


.
.


The first inductor is wound with two wires, each 1.4–1.5 mm in diameter (7 turns); the second with a single wire about 1.7 mm in diameter (two windings of 4 turns each).

I made several more circuit boards, but I’ll get to those a little later.


Since I’ve moved on to the controls, I’ll say a few words about the encoder I used.
I bought a PEC16-4015F-N0024 encoder made by BOURNS; here’s the link to the datasheet.
I also got a large knob with a diameter of 30 mm; a smaller one just wouldn’t look right.

This encoder is fairly large, but it was simply convenient in terms of price (less than a dollar) and shaft length.
And it also had a threaded bushing, which is very handy, since the one I found at home didn’t have threads. Neither did the one installed on the circuit board, for that matter.


Next came the front-panel prep. Nothing unusual here: I printed out several versions, checked how they’d look in real life, and picked the most acceptable one.
The front-panel design is very similar to that of my previous power supply, but this time I put the LEDs to the left of the display, which seemed more convenient to me.


I drilled the holes and cut out the openings, adjusting their sizes as I went to get them right.
I had to do some fitting because I was afraid of making the openings for the display and button too large. I only had one shot at it; a second attempt would have cost $30, or I’d have had to put up with a wobbly button or a misaligned indicator.


I ran into another snag along the way. The encoder worked well, but it had no detents, so the knob turned smoothly. That's fine for a volume control, but not for an instrument where it's handy to count changes by clicks without looking at the screen.
Anyway, I stopped by the market and bought another encoder from the same manufacturer, but this time from a different stall. It cost noticeably more—another $2.50 added to the bill—and on top of that, the sellers had swiped the mounting nut and washer.
I'd bought a counterfeit TOP250Y at this same stall once before (it was in one of my reviews). How can they do that? I only noticed when I got home, but since the previous encoder had a nut, I just let it go. They can keep it as a souvenir.


Since the display wasn't exactly cheap, I decided to give it a little protection.
First, I applied thin double-sided tape around the edges.


This photo isn't really relevant to the review; I just liked the shot because it shows the display matrix clearly.


After that, I cut a small piece of clear plastic left over from some packaging—maybe a headset, maybe an external hard drive enclosure.
I stuck the resulting little window on with double-sided tape, after carefully wiping it down and removing all the dust.
You can see some minor scuffs, and the contrast has taken a hit, but it's now much harder to damage the display.
It looks worse with the flash than it does in real life.


While assembling it, I decided not to attach anything to the front panel with through-holes, so I mounted the display on spacers cut from the plastic left over after cutting out the opening for it :)
Why throw away something you can put to use? I did have to add some washers about 0.5 mm thick; after that, the display surface was flush with the front panel.


1. The keyboard board was a little trickier.
I made four standoffs from the same plastic scraps. Each one had three layers, but they still weren't quite tall enough. Offcuts from the power supply enclosure came to the rescue. I'd cut them out to move the enclosure closer to the rear panel, which is why I said not to throw them away—they might come in handy :)
2.3.4 When I routed the output filter board, it completely slipped my mind that I'd bought a fuse and wanted to add protection diodes.
I had to add all of that to a board that would be screwed onto the output terminals.
There's no point drawing a schematic: the output is connected through a fuse, and a pair of KD213 diodes is connected in parallel across the output on the power supply side.
The purpose of this circuit is to blow the fuse if a battery is connected with the wrong polarity.
There's already a protection diode on the converter board, but it seemed a bit underpowered, so I decided to add another one in parallel.
I used standard 6.3 mm terminals for the fuse, soldered directly into the board.


The terminal board layout. It has footprints for two types of diodes: KD213s and diode assemblies in TO-220 packages.
It also has extra pads for connecting the feedback wire.


Since I was following the principle of easy maintenance, all connections were made detachable. To do this, I bought several different types of connectors.
For connecting the display, I used the large ones — link
For all the other connections, the small ones — link.
I also used color-coded wire to make connections easier. I connected the encoder with shielded 4x0.22 cable, since interference on this line could have serious consequences.

I mounted the LED board in the enclosure in a slightly unusual way.
For this, I picked up some special bezels at the market. They are originally intended for decorative trim, but work perfectly for securing a board in an enclosure.
The process is very simple.
Drill a 6.5–7 mm hole
Put a ring on the LED
Insert the decorative part into the enclosure hole
Insert the LED into the decorative part
Push the ring onto the decorative part until it stops, and that's it.


In principle, the indicator board is very simple, and you could even skip making it and just wire the LEDs directly—that's much easier.
But there's one small catch.
The LEDs need to be very bright because the current through them is very low.
And you can't increase this current simply by lowering the resistor value. Not at all.
The only LED whose brightness can easily be increased is the CV mode indicator.
If you lower the resistor value for the output-active LED, the output voltage won't turn on (if I remember correctly).
And if you lower the resistor value for the CC LED, this mode will no longer be shown on the display.

As before, I used LEDs in three colors: red, green, and yellow.
The first two types are easy to buy, but finding a bright yellow LED proved difficult. I don't even remember where I bought one last time.
So I decided to eliminate the problem at its source.
Since the indicator board gets a common wire and 5 V power, I added a transistor and a couple of resistors, which means you can use any LED to indicate that the unit is on.

You could say the power supply is almost finished. There was one minor issue, of course: I had to recalibrate the keyboard, but otherwise everything worked on the first try.
The front panel is still a temporary version, but I decided to improve it a little anyway by moving it into the FDSIGNER program and changing the label fonts.



The fuse board is screwed directly to the output terminals.
I wouldn't call this the best solution, but I didn't have many options.
Keep in mind that power wires should not be too long. I used 6 mm² wire and made it just long enough to let me fold back the front panel and unscrew the board if needed.


It's best to keep the power and signal wires as far apart as possible.
I once ran into a problem when the encoder wires ran alongside the power wires, so it's best not to repeat my mistake.


Right at the end of the build, I set up the connection to a computer.
For wireless communication, I ordered a pair of Bluetooth modules, and for a wired connection I used the adapter that came with the unit.
The USB-RS232 TTL converter uses the common PL2303 chip, so there's not much new to say about it.


The Bluetooth modules were sent to me for review, so I ordered a pair of different ones. But honestly, I still can't tell what the difference is—they look like identical twins.
The first is sold as HC05, the second as HC06.


For some reason, the HC06 module wouldn't "play nice" with me, so I went straight to work with the HC05.
I'll still try to figure out why one of the modules didn't work, even though its LED flashes away quite happily—it just won't respond.

But the modules really do look identical. I even had to mark the second one with a marker so I wouldn't mix them up.
Maybe I'll use it in a review once I figure out what it needs :)


The converter board can work through these modules, but in practice the original software doesn't work through them. I've read that everything works correctly when using a Bluetooth 4.0 module in the computer, but my adapters are version 2.0, and the stock software refuses to work with them.


Nevertheless, I was still able to test them using my own software, even though it wasn't finished (now I'll definitely have to work on it).


But the experiment was actually about something else.
If you look closely at this photo, you can see that both interfaces, USB and Bluetooth, are connected to the board at the same time.
That was the experiment.
The USB connection was set up in the same way as in this review, using a chip that provides galvanic isolation.
Bluetooth, on the other hand, was connected in parallel using two diodes and a resistor.
The idea was to be able to use either interface without switching. And it worked.
You can use either the cable connection or Bluetooth, but naturally only one can be active at a time.


Since the experiment worked, I glued a strip of plastic to the board, insulated it with heat-shrink tubing, and mounted it on the inside of the rear panel. Since the case is metal, this was necessary.


There are two software options to choose from.
The old one works properly with the 6020 boards, while the new one works with the 6020 board even though it was originally designed only for the 6005 boards.
The whole situation is a bit odd: separate software is released for each board, even though the protocol is essentially the same for all of them.
The only difference is that the software sets a different maximum current for each board. If you connect to a 6020 board using the software for the 6005, you won't be able to set the current above 5.2 A.
But there's another drawback: the current will be displayed as one-tenth of its actual value. This is because the 6005 sets current in 1 mA increments, while the 6010 and 6020 use 10 mA increments.


I connected the remote-sense wires last of all.
The board supports four-wire (or three-wire) load connections, which means it can compensate for the voltage drop in the power leads. This is useful at high currents.
To do this, remove the two solder bridges and connect the connector to the output terminals (or to a point as close to the load as possible).
To reduce interference, I used tightly twisted wire and slipped insulation from a 4x0.22 cable over it. The wires were from that same cable, actually.
Be very careful when connecting this wire. If there's no contact, the full voltage will be applied to the output, regardless of the set value.
If you're not sure, simply don't use this feature of the power supply. The performance will be slightly worse, but it will be safer.


That's it, the power supply is fully assembled. All that's left is to put the cover on.




A few photos of the finished result.




The rear panel is practically empty. I didn't even bother adding labels, since it's impossible to connect anything incorrectly—it's all self-explanatory :)


1. The final step was setting the maximum output power limit to 700 W.
2. This photo shows what happens when the cursor appears in the wrong place. It got into the photo by chance, since it only flashes on for a very short time.
3, 4. Five minutes after switching it on, the temperature reads about 30 degrees, but after an hour to an hour and a half it warms up to 42–43 degrees with no load. Once the fan turns on, the temperature quickly drops back to its previous level.

During testing, I eventually switched the feedback (see above) back to the standard mode, measuring the voltage at the board's output terminals. I did this because an unusual noise appeared at loads above 50–60 W. I'll investigate the cause; for now, I suspect it's because I took the feedback signal after the dual-winding inductor.


A little testing was in order, too.
I was mainly interested in checking the ripple at the power supply output.
The manufacturer specifies 50 mV at 12 V, 8 A, and an input voltage of 54 V.
Everything matched my setup except that the input voltage was 68 V and there was a filter after the board.
1. At the specified settings, the ripple was noticeably higher, even with the filter after the board. I measured about 110 mV.
2. Interestingly, as the output voltage increases, the ripple voltage decreases.
Even more interestingly, the ripple frequency isn't 150 kHz, but about 300 kHz.


Next, I set the current to 10 A (about 50% of the maximum) and tested at 30 V and 40 V.
1. At an output power of about 320 W, the ripple was around 60 mV.
2. Then I increased the output power to 400 W, and the ripple rose to 70–80 mV.
That was the maximum power the electronic load could dissipate, and even then, only briefly.
The ripple seems a bit high to me; there's room for improvement.



Side-by-side comparison of the old and new power supplies.


Well, that's all for now. It's quite possible there'll be a third part sometime in March, where I'll talk about the modifications and rework, but the main part is finished for now. Now the power supply needs to stand the test of time.

During testing, a few additional drawbacks of the board became apparent.
1. Ripple is clearly higher than specified, at least at an output power of around 100 W.
Most likely, this is because although the step-down converter can operate over a wide range, everything ultimately depends on the inductor. Different power levels (and input/output voltage differences) require different inductance values.
2. I couldn't get the four-wire feedback connection working properly.
Voltage regulation was more accurate, but an additional noise appeared (the converter is silent in its normal operating mode).
I think this is because the circuit uses a dual-winding inductor. I'll investigate the problem.
3. The fan. It makes noise the whole time the power supply is running. Something needs to be done about that.
4. It also turned out that overtemperature shutdown does work after all, but since I modified the temperature measurement, it now works incorrectly—that is, in reverse.
So for now, it seems you can either get a normal temperature display or have the emergency shutdown work properly, but then the readings need to be converted into a more understandable format.
It's up to each person to decide. Another option is to modify the circuit so the readings correspond to degrees, but count down from 100 degrees instead.

Otherwise, no problems have come up so far; everything works as planned.
If you want more power, just install a more powerful power supply and either remove the maximum power limit altogether or set it to the limit required for the power supply you've chosen.

This review turned out to be very long. I hadn't expected that at all, but I got so carried away describing the process and wanted to cover so much that it ended up as two reviews instead of one.
I don't think many people will decide to reproduce the whole setup in its entirety, but some of the details may be useful in their own projects and designs—that was the idea, after all.

I think that's everything. I'm sure I've made a bunch of mistakes, so I'd be glad to hear any corrections or questions, or just your comments.
I hope you find this review useful.