# Hunting For Part Numbers: Synchronous Buck Converter in QFN-20(3x3)

What's this "Mini560" synchronous buck converter module hiding from us?
Shouldn't these small boards be a serving suggestion for the particular part?
Let's find the datasheet for this chip.

The sections below are mostly an info dump in chronological order - feel free to scroll around and look at the pictures or read in sequence for the story. Enjoy!


## Marketing Lingo

One table found on the internet lists nominal output, measured input and presumably measured output I current and voltage values to calculate the conversion efficiencies:

These modules come resistor-programmed for various output voltages, and values can be changed afterwards by swapping 0603 1% resistors ([source](https://raspberrypi.web.za/shop/buck/dc-dc-buck-converter-mini-560-12v/)):


On [Amazon](https://www.amazon.de/Converter-Spannungswandler-LAOMAO-Ausgang-Netzteil/dp/B0B92ZDK6T) they're described as follows.

Overview:
- High Efficiency Step Down DC DC Step-Down Converter Synchronous Rectifier Voltage Regulator Power Supply Board Mini560 5V

Technical data:
- Model: Mini560 (Small DC-DC Synchronous Rectifier Down Module)
- Input voltage: 7 V - 20 V.
- Output voltage: 5 V.
- Switching frequency: 500 kHz.
- Conversion efficiency: 99% (maximum). 🤔 *(x) doubt*
- Operating temperature: -40 to 85°C.
- Type of module: non-isolated downward regulator.
- Rectifier method: synchronous rectification
- Module size: 29 x 18 x 5.4 mm.

## Package And Footprint

This is easily obtained ground truth and an indispensable clue warranting the purchase of a sample.

Hot air removal of the IC exposes an ugly QFN-20(3x3) exposed pad (EP) package with bridged pads on the PCB and bridged pins in the leadframe. Immediate suspects are: Texas Instruments (Ti) and Monolithic Power Systems (MPS).


Connections on the PCB

- 2-3-4-5
- 6-19-20
- 7-8-18-EP

Leadframe bridges:

- 2-3
- 4-5
- 7-8

Assumption: Leadframe revisions, package drawings and recommended footprints do not necessarily coincide at all times. We could be looking at an older or newer version or a replacement for an discontinued part. That's why it's always good to collect as much information as reasonably possible, and to search for multiple pictures of the same device or component.

### Ti Packages

Texas Instruments are known for some very unique packages (read: unnecessarily specialized, extra effort to create and maintain libraries, find replacement parts).

On the bright side, they offer a product-by-package search (assuming you know what to look for):

https://www.ti.com/packaging/docs/searchproductbypackage.tsp

#### RJE0020A

Save for the leadframe bridges, this looks like a close match. The short-hand designation is "RJE", and it's not always clear which one will be shipped.

For some reason, the EP centroid is off-center by 60 µm | 18 µm.


- [x] QFN-20 with exposed pad
- [x] 3.0 x 3.0 mm
- [x] visually symmetric
- [ ] leadframe bridges

#### RJE0020B

And then there is the crooked "B" variant:


- [x] QFN-20 with exposed pad
- [x] 3.0 x 3.0 mm
- [ ] visually symmetric
- [ ] leadframe bridges

Since the datasheet isn't too accurate about the exact type, it is not always clear which one you're getting. This aspect can be clarified with store pictures to some extent.

#### Matching Parts


The list of 20-pin RJE package devices is short, and only two are specified for 20V operation.

- [TPS51396A](https://www.ti.com/lit/ds/symlink/tps51396a.pdf) (24V)
- [TPS51397A](https://www.ti.com/lit/ds/symlink/tps51397a.pdf) (24V)
- [TPS568230](https://www.ti.com/lit/ds/symlink/tps568230.pdf)
- [TPS56C230](https://www.ti.com/lit/ds/symlink/tps56c230.pdf)
- XTPS56C230

As it turns out though, 20V often falls between the maximum recommended input voltage, and the absolute maximum rating (regular use which should always be avoided).

XTPS56C230 should be equivalent to TPS56C230.

## Ti Candidates

| Part | Vin,max (V) | f_sw (kHz) | V_ref (V) | I_noload (µA) | Vout,max (V) |
| --------- | ----------- | -------------- | --------- | ------------------ | ------------ |
| TPS51396A | 24 | 600, 800, 1000 | 0.6 | 90 | 7 |
| TPS51397A | 24 | **500**, 800 | 0.6 | 110 (no switching) | 5.5 |
| TPS568230 | 18 | 600, 800, 1000 | 0.6 | 105 (no switching) | 7 |
| TPS56C230 | 18 | **500** | 0.6 | 400 (no switching) | 5.5 |

These parts are also pin compatible.

### Closer Selection

If one believes the advertised 500 kHz frequency, this narrows down the selection to

- TPS51397A (24V max. - plausible)
- TPS56C230 (18V max. - **out of spec**)

Sometimes cheap modules go (way) beyond the recommended operating conditions, and it turns out that TPS56C230 can tolerate up to 20V on Vin. Maybe that's just for bragging, since the listed test conditions only include 12V and 18V input values?


Remembering the connections on the PCB, let's compare:

- 2-3-4-5
- 6-19-20
- 7-8-18-EP
- 9, 10, 16 are NC

vs. datasheet:


It's a pretty good match. It's very unlikely that other parts accidentally use such a peculiar pinout. Right?

### Some Problems Emerge

The elephant in the room is probably Vout,max. Since all devices are synchronous buck converters, the switch node (SW) must tolerate Vin + excess ringing voltage (22V abs. max, transient spike). This does not present an inherent limitation to <= 7V out. Additionally, the feedback is adjustable via voltage divider, so only SW sees the output voltage, and only during discontinuous conduction mode (DCM) as V_SW decays towards Vout.

Iteratively, one gets deeper and deeper into the workings of newly found components, a trade-off between spending too much time on irrelevant datasheets and missing a crucial detail.

TPS56C230:


The SW voltage is monitored via three comparators, but they do not seem to concern themselves with overvoltage.

With an oscillator and counters, one could also measure the on-time and period, or on-time and time to zero current (in DCM). Thanks to internal components, the timing could be known. For buck converters, the duty cycle relates to the output voltage as stated below (as the inductor voltage is proprtional to the slope of the current and the flux / energy changes in each steady-state cycle are balanced):

([source](https://www.powerelectronicsnews.com/power-supply-design-tutorial-part-1-2-topologies-and-fundamentals-continued/): PowerElectronicsNews)

Confession: At this point and depth of reading, it's not clear to me why there out to be a hard limit to the output voltage.
Options to further clarify:

- buy TPS56C230 and TPS51397A, rework Mini560 PCBs and test
- read datasheets thoroughly
- more online search
- measure actual PCB in operation

## Measurements

### Operating Frequency And Light Load Behavior

Between GND and the input-facing terminal of the inductor, V_Sw can be measured when the output is loaded:


Remarks:
- period is roughly 2 µs with some jitter.
- there are some irregularities / potential subharmonic oscillations?
- light-load causes pulse rate to gradually decrease, no hard fixed frequency (good for efficiency, weird in the spectrum)
- [x] 500 kHz operation

### FB Comparator Reference Voltage


- [x] FB is tied to pin 14
- [x] 0.60 V

### No-Load Current

- No LED, no load: 12V in, 2.53 mA in, 9V out, 0mA out
- output driven above setpoint: 12V in. 1.931 mA in, 9.2V out
- powered-down output resistance measurement with both polarities: 15.28 kOhm, 15.20 kOhm, resulting in -0.6 mA out at 9.2V

The no-load operating current seems a bit excessive, and certainly nowhere near the 100-400 µA listed above. Sporadic switching at no-load cannot explain this, and we haven't investigated the remaining components on the PCB, of which there are few. Pin 13 EN leakage can't explain it although it max see only less than 6V (assuming the 5.1k resistor (512) would force current down that pin, dropping 6V would mean 1.18 mA - not recommended in any way. Don't do it!).

### Package

Taken from another online seller who pixelated the watermark of another store, hehe:


The laser marking (assuming these are genuine parts) does not seem to be compatible with the blank space on the left. The characters are just too close to the Pin 1 dot:


Ditto for TPS56C230:


But maybe these are not genuine parts? Let's continue.

### UVLO

Here's where the real plausibility problems start.

- UVLO trip: 2.76 V (**not 3.7 +/- 0.1 V**)
- UVLO recover: 2.91 V (**not 4.3 +/- 0.1 V**)

## It's Not A Ti Part

So far, we have

- matching footprints
- plausible PCB layout
- a very peculiar pinout
- 500 kHz switching frequency (typical, excluding light-load behavior)
- 20 V operating voltage when stretching it, but plausible operation at 18 V
- FB voltage of 0.6 V on the correct pin

But

- minor package variations (leadframe with additional bridges where the pads are already internally connected and connected on the PCB)
- suspicion that the maximum output voltage max exceed the specified range
- incompatible UVLO behavior (extended operating voltage range).

The last point rules out the Ti parts, since they all have thresholds around 4.2-4.3 V and 3.7-3.8 V.

What's more, LCSC has pictures of the underside, showing RJE0020B leadframes which are nothing like the specimen above.


## Component Identification

Based on the type of PCB, another independent search (shout-out to [MacroWorld21](https://twitter.com/MacroWorld21)) found this:


Sporting a JoulWatt JW5069A:


Crummy JW5069A [datasheet preview](https://www.sekorm.com/doc/1768802.html):


We can compare it to [JW5068A](https://arduino.ua/files/1903051031_JoulWatt-Tech-JW5068A_C371422.pdf) (LCSC C371422):


They seem to share the same leadframe, pinout and similar specs. What's funny though: They managed to center the EP. I think that's just them flexing here...


So that's it. Pick one:

- JW5068A: 4V to 26V in, 8A out cont.
- JW5069A: 4V to 23V in, 6A out cont.


## Addendum

Without the full JW5069A datasheet, a little detail was missed:

产品参数
1.输入电压：4V~23V
2.输出电压：0.765V~VIN-3V
3.开关频率：500kHz
4.输出电流：6A
5.效率：高达95％
产品电路示意图

JW5069A: 0.765V minimum output voltage? With FB directly connected to the output, the minimum output voltage is realized, which is defined by the internal reference. The board however exhibits 0.60V at the FB divider. Are there other limitations (nonlinearity?) at play, or is this proof that it's in fact not the JW5069A seen on a related PCB, but JW5068A (who indeed uses 0.6 V ref).

Never be afraid to make simple comparison graphics and draw into them:



green marked pins are fully compatible. BYP can be made to fit with a BOM change since SMD resistors (per Ti) can be removed and an MLCC to GND can be put in place. The only true change, and the only one to be found here, is that pins 11 and 12 are swapped during the migration to the JW part.

This means:
- JW5068A, JW5069A are very likely pin compatible
- JW5069A is essentially pin compatible with the Ti counterpart, that is: 3 our of 4 quadrants with the power path are identical.
- The control / interfacing quadrant of the IC is made to be nearly compatible: ILMT can be tied to GND, setting the most conservative output current limit. FB is a match, BYP is optional (can be connected to VCC without problems but is intended to be connected to the output when the regulator produces >= 5V to lower/avoid internal LDO losses)
- Only difference: where the Ti parts use pin 11 for soft-start (which is optional!) and pin 12 for *enable*, this pair is swapped on the JW part, and external SS is omitted altogether.
- One can modify production gerber files that use a Ti component by adding a bridge across pins 11-12 and the board will accept the JW part.

### Plot Twist

When searching for this module with "JW5069A", we get the actual picture without the laser etched package. Is this evidence that the 0.765V spec cited in a store is an error or "guaranteed operating regime"? If the part on the sample measured is indeed JW5069A, then both chips would use 0.6V references (and reasonable for parts of the same family).


And if that picture isn't enough: JW5068A is a "better" -as in higher power, greater voltage range - device. For cheap modules, replacing JW5068A with JW5069A would probably mean saving a few cents, so it would be done without hesitation.

let's wrap this up with an excerpt from the JW [2024 Selection Guide](https://www.joulwatt.com/index1.html?re=support-news-detail&id=39):


Both use 0.6V as a reference.

### One More Thing

As if we hadn't looked at this board long enough now, Here's [a video](https://www.youtube.com/watch?v=y2YlCr8ZseQ) dating back to 2020 that shows proof with an incomplete laser etch that at least the earlier versions used JW5068A.


I don't know in which order they were released, but now the answer to "which one is it?" is "Yes". Probably earlier models used the higher power version JW5068A, but then they figured they could get away with less. The laser etching also helps conceal this downgrade.

## Conclusion

Sometimes one walks right into the answer, sometimes one skips the right clues. Some manufacturers are very systematic with their package naming, others not so much (JoulWatt calls it "QFN3X3-20", which does not declare any of the special features).
This time, the clue came from the screenshot of an identical module. Sometimes one finds PCB photos with a similar part that has matching circuitry around it. Such deeper forensic investigations can take much longer, and occasionally one may need to purchase and analyze more assemblies to have enough evidence.

The story told herein is one of compounding evidence systematically ruling out the initial candidates. Advances were possible though disassembly of a sample PCB, as well as electrical measurements to interrogate the device characteristics.

The findings along the way however also tell a different story of [import substitution industrialization](https://en.wikipedia.org/wiki/Import_substitution_industrialization) and a longer-term battle over [standardization](https://www.kan.de/en/publications/kanbrief/2/21/china-a-developing-global-power-in-standardization). JoulWatt seems to have been founded in 2013, and setting up domestic production of [trailing edge](https://www.forbes.com/sites/timbajarin/2021/05/13/why-trailing-edge-semiconductor-manufacturing-matters/) products and specialty processes nevertheless takes time. In 2019, a major company [suspended its JEDEC membership](https://asia.nikkei.com/Spotlight/Huawei-crackdown/Huawei-s-voice-in-future-tech-standards-restricted) as a lever against US trade restrictions. And here we are, wondering why there is a better-spec'ed drop-in replacement for less than half the price in a mildly deviating package.

## Outlook

One may still decap the JW5069A and have a look at the die. Since it's essentially pin-compatible with a selection of Ti parts, minor details regarding mode selection may be absorbed into the BOM, and with the specs exceeding those of the Ti parts on paper, one can expect some actors to sell them as counterfeit Ti products. For such a counterfeit parts market, one just needs to have a batch packaged with EN bonded to a different pad.
To explore the realities of that, one can buy several of the Ti parts (sold as TPS51396A, TPS51397A, TPS568230, TPS56C230) and inspect the dies to see if some of them just so happen to match JW5068A or JW5069A.