USB tester, or attempt number two

Not long ago, I did a review of a USB tester with a certain “feature.” The comments on both the review and the video sparked quite a heated discussion, and today I’ll try to explain what was going on.
So, scandals, intrigue, investigations.


Today, in fact, I’ll be reviewing two testers at once. The answer to the question about “Chinese math” will come at the very end, so if you want to know the result right away, feel free to skip to the end of the review.

I ordered several USB testers. Well, “several” means a lot of 10, and the price was reasonable, so I ordered them. When they arrived, I thought I might write a review.
Of course, plenty of people will say that all these “testers” have already been reviewed a million times, and they’ll be right. But somehow I can’t remember a single review that checked not only the basic specs, but also whether the device correctly calculates the capacity that has passed through it. Understandably, who would think that a device could measure current and voltage correctly but get the math wrong? It’s elementary-school arithmetic.
So I ended up putting together another short review, where I’ll show a couple of familiar USB testers but put them through a more thorough check.


The order arrived fairly quickly, neatly packed in a cardboard box.


As ordered, it’s a lot of 10. The eleventh tester and the matchbox are shown “for good measure.”


The stated specs are shown here. Interestingly, there’s no accuracy specification anywhere. So, formally, it would be difficult to make a claim to the seller even if there were a problem. In a way, that’s fair, since calling this device a measuring instrument would be a bit of a stretch.


It looks exactly like the previous tester; the only difference is the case color. Though even that isn’t much of a difference, since the seller offers these testers in two versions: dark and light.


It’s just as awkward to take apart: the plastic is quite brittle, and the clips are tight. But here you can see at least one more difference: the display is OLED, not LCD.


The display is a little larger, and the circuit board is blue rather than black.


But much bigger changes are visible on the back of the board. First of all, the build quality is clearly better, and there’s no flux residue on the board either.


Since the OLED display used here doesn’t need a separate controller, the board looks less crowded.


The controller’s marking has been rubbed off, though I don’t think there are many possibilities here; it’s probably some kind of STM.
Like the previous version, the device is powered through a linear regulator.


But there’s a much more important difference near the output connector. First, the shunt resistance is 25 mΩ, not 50 mΩ, so the voltage drop across it is also half as much: at 2 A, it’s 0.05 V instead of 0.1 V.
Also, this version uses an op-amp to amplify the voltage from the shunt, which is much better than feeding the signal directly to the microcontroller.
The thing is, it’s difficult for a microcontroller to measure such small voltages, which usually means reduced accuracy, especially near zero.


Most of the board is taken up by the ground plane, but the power trace is also quite wide. The specs claim a current of up to 5 A. In my opinion, that’s very optimistic. Besides, I don’t see much point in currents above 3 A, since problems will start with the USB connector overheating, not with losses in the traces.
Still, large copper areas and a wide power trace can only have a positive effect on performance.


As I mentioned above, it has an OLED display. That’s good in its own way, since it can fit more information on one screen than a cheap LCD. But there’s a downside: the brightness is noticeably lower. When connected, it also creates a gradual power-on effect that takes about 0.5 seconds—the screen seems to “light up.”
The dark plastic also has a noticeable effect: without it, the screen is brighter, though not dramatically so. The difference is more apparent in the photo than in real life.


When connected in series (with no load), some testers showed roughly the same reading, while a few strayed from the pack. For example, one showed 5.09, while the testers after it showed 5.12.
I should mention right away that although the display has a resolution of 0.01 V, readings are actually shown in increments of 0.03–0.04 V, as you can see in the photo.


Voltage measurement accuracy comparison. Since I was interested in the relative readings, I didn't measure the power supply voltage; it was enough to know that it wasn't fluctuating.
The spread is about 0.07 V, with the readings distributed fairly evenly.


But one rather odd unit turned up, with a mind of its own. Even with no load, it displays a current of 0.01 A and accordingly counts both mWh and mAh.
I also found out that this model has no “minimum threshold”—or, more precisely, it is 0.01 A, not 0.2 A as on the previous model I reviewed. Whether that's good or bad is hard to say; it depends on the situation. In my opinion, it's better when it starts counting almost from zero.


The full test showed that most of the “testers” read the voltage slightly low. The ninth “test subject” is shown in the photo below.


The tester's own current consumption is almost independent of the input voltage and is about 7.5 mA. This current isn't included in the measurement, so it doesn't affect the result.


Tests at 1 A and 2 A. The paired photos below are in the same order as the “test subjects” shown above.
Overall, most stay within one count in the last digit, which is very good, but a few units are less accurate. The worst is unit number 5, which is off by up to 5% at 1 A.


Current measurements from 100 mA to 3 A. Up to and including 1 A, the current was increased in 100 mA steps (100, 200, 200); above that, in 200 mA steps (1.2, 1.4).
Test subject number five, the worst one. As you can see, it reads 30–50 mA below the actual current across the entire range.


Same test, but with tester number nine, one of the most accurate units in the batch. Its readings differed from the actual current by 10–20 mA, consistently across the whole range: it reads low at lower currents and high at higher currents. Note the accurate reading at 100 mA.
Most of the testers in the batch produced similar results.


And, of course, a test of capacity measurement accuracy, as well as how much the readings drift as the unit warms up.
Load current: 2 A; duration: 1 hour. It should therefore display 2000 mAh, but since the count is based on the built-in meter, the testers' “math” is checked against the measured values.

Tester number five, the worst one.
The result is close to the actual value, but since the tester reads slightly low, it should have displayed 1970 mAh. Instead, it showed 1903 mAh, which is tolerable, in principle.
Interestingly, the capacity in mWh is also low: instead of 9593, the tester showed 9266. The built-in timer runs a little “slow,” but not enough to account for this result.


Tester number nine, one of the most accurate.
During the test, the readings “jumped” between 2.01 and 2.02 A. It's simpler to assume the current was 2.15 A and the voltage was 4.87 V.
The mAh reading is also slightly low: 1956 instead of 2015. The mWh capacity shows the same pattern.
Interestingly, this unit's timer runs a little “fast”—by 11 seconds per hour.


And just as an experiment, I tested the ninth unit at 1 A. The result was 3% below the calculated value: it showed 970 mAh instead of 1000. The mWh measurement has the same error.


Just to avoid any misunderstanding, all tests were performed using the same settings and the same equipment as last time. The experiment is 99% repeatable.


And now for the “scandals, intrigue, investigations” I promised at the very beginning.

After all the review materials were ready, I received several more testers of a different model. This particular type of tester sparked quite a heated discussion, and I simply couldn't ignore it. So I put the review on hold for a bit and decided to test this one too.
People wrote to me that it was all because of the “wrong” power supply, an incorrect test using an electronic load, a large measurement error, and so on. But the review was called “USB Tester, or the Subtleties of Chinese Mathematics,” so that’s exactly the math we’re going to check—and find out “who’s to blame” while we’re at it.

I bought these testers from another seller, product link, for $2.80 each, excluding shipping. Since I ordered more than one, though, the total came to only a little more.


Taking the tester apart revealed just one difference between the old and new versions: the current-sense shunt is rated at 25 mΩ in the new one, versus 50 mΩ in the old one. All the other components are identical.


You can even spot the difference without taking the case apart.


First, the basic tests.
Voltage measurement accuracy. The results are certainly worse than for the version shown above, but not by much.


On the plus side, its current consumption is lower: 5–6 mA versus 7.5 mA.


Current measurement results are noticeably worse, though. As usual, I tested at two current levels, 1 and 2 A, and the photo clearly shows that the readings are off.


I tested across the full range, from 100 mA to 3 A, using the same procedure: in 100 mA increments up to 1 A, then in 200 mA increments.
Because this model’s microcontroller measures the voltage across the shunt directly, the measurement error is large at low currents. For example, a current of 100 mA doesn’t show up at all, while 200 mA is displayed as just 60 mA.

As I mentioned above, when the shunt is connected directly, the controller has trouble measuring the small voltages coming from it—and there you have the result.


But what I was most interested in was the accuracy of the capacity calculation. In fact, I wasn’t so much interested in the accuracy itself as in whether it differed from the previous unit.
And there is a difference, a pretty clear one. But first, the bad news: the current readings climbed as the tester warmed up, and by noticeably more than on the version from the previous review. For example, this unit initially displayed 1.99 A, but after 20 minutes it was already showing 2.13 A.
The capacity measurement, on the other hand, is completely accurate. The calculated result should have been about 2130 mAh; it displayed 2109 mAh, which is quite close to the calculated value, since the measured current was still a little lower before the tester warmed up.


All right, let’s try another unit. Its ammeter readings were much too low to begin with, but drifted high as it warmed up.
The result should have been 2090–2100 mAh; it displayed 2067 mAh. Here, too, we need to account for the fact that the current measurement was lower when the unit was cold.


Now let’s go back almost six months, to when I tested the same type of tester.
The current measured by the “tester” was around 2.10–2.14 A, so it should have counted no more than 2140 mAh—probably closer to 2100 mAh. But the result was almost 2500 mAh.
That’s exactly what I was trying to get across to readers and viewers, but unfortunately not everyone got it :(

So what’s the cause?
The cause is simple: a software bug. You might ask what the software has to do with it, since we clearly saw a different shunt on the circuit board.
The thing is, when the shunt is connected directly, with no dividers or amplifiers, changing the shunt also requires changing the program to account for the different shunt resistance and, consequently, the different voltage across it.
But apparently they changed more than just that, and used a different mAh calculation algorithm as well.
So I can say with 99% confidence that if your shunt is marked R025, the current measurements will be worse, but the tester will calculate capacity correctly. If it’s marked R050, it’ll be the other way around: current measurements will be a little more accurate, but the capacity calculation will be completely off.
I suspect one manufacturer stole the source code from another, but changed the shunt rating. They corrected the section of code that measures current, but introduced a bug somewhere in the capacity calculation. And so “we’re left with what we’ve got.”


A video where I talk about the tester from the previous review.


That’s all from me. Time to wrap up with a few brief conclusions.

Of the two models shown in the review, I preferred the first one. It has a better circuit design, its readings don’t drift with temperature, and it’s more accurate when measuring low currents. Still, there’s an element of chance involved—for example, you might get a unit that displays something even when the current is zero. The test also showed that the measured capacity is about 3% too low, both in mAh and in mWh.

The second tester measures capacity correctly, but because of its lousy circuit design, when measuring current you can get anything but the actual value. And there’s no way to calibrate it, since this error is baked in at the “genetic” level.

And for my own part, I’ve figured out not only why the previous tester measured capacity incorrectly, but also found a difference between the models that lets you tell a tester with sound math from one with “Chinese” math.

That’s all from me. I hope this review was useful and answers some of the questions raised in the comments on the previous review.