Power Integrations RDR-928 60W USB -C PD 3.0 PSU Reference Design RoadTest

Table of contents

RoadTest: Test and Review the P.I. 60W ultra-compact USB PD 3.0 Charger RDK

Author: Gough Lui

Creation date:

Evaluation Type: Power Supplies

Did you receive all parts the manufacturer stated would be included in the package?: True

What other parts do you consider comparable to this product?: Nil - was not able to find a similar integrated opto-less solution to the InnoSwitch4-CZ from another vendor.

What were the biggest problems encountered?: Design minimises board area, leaving little space to mount or secure board. Mains connection via test loop contacts.

Detailed Review:

Power Integrations RDR-928 60W USB -C PD 3.0 PSU Reference Design RoadTest

By Gough Lui – February 2025

The USB-C connector has finally become the new “universal” connection for both data and power. Thanks to the Power Delivery (PD) standard, a USB-C port can easily provide power to run a variety of electronics devices including smartphones, tablets and even laptops. Standard power ratings include 15W, 27W, 45W, 60W, 100W in the standard power ranges and up to 240W in extended power range. Such feats can only be achieved through negotiation of power delivery using the PD protocol encoded as BMC signals travelling on the dedicated CC1 and CC2 lines. Such negotiations allow the source and sink to come to agreement as to what PDO is to be selected and to ensure that any connected cable is capable of carrying the requested power through the verification of an eMarker chip. Consumers appreciate the ability for a single reversible connection to carry both power and data, although due to the proliferation of different capabilities and power ratings between different chargers, there may be some compatibility issues.

Given the universality of USB-C, many devices no longer include a charger and consumers frequently will purchase their own aftermarket units for convenience. As a result, often consumers will prefer adapters with higher power ratings to allow for use with a wider variety of devices, but also demand them to be small and light-weight to save space in a backpack. As consumer electronics are very cost-focused, consumers also look for low prices. With such conflicting design constraints, it’s not surprising to see some small chargers get very hot, stop producing their rated power and in some cases, even “explode”.

As a well-equipped engineer who has tested different USB-C PD chargers in the past, I’m familiar with the Power Integrations InnoSwitch-range of controllers with primary-side sensing. If you have a collection of USB-C PD chargers, there’s a decent chance it might even have one of these Si-based or GaN-based chipsets inside it. They have often performed well in testing while reducing BoM and design complexity.

As a result, I am very fortunate to have been chosen by element14 and Power Integrations to have an opportunity to test their RDR-928 reference design for a 60W USB-C PD 3.0 power supply.

Unboxing & Feature Introduction

image image

The reference design comes in a board with a colour cardboard “sash”. Documentation can be accessed via the QR code at http://www.power.com/rdr-928, but at present, it responds with a 403. As a result, I relied on the documentation provided on the RoadTest page consisting of the Reference Design Report only. Unfortunately, no board design files appear to be available at this time and there is no reference to the board on the site at this time.

image image

Inside is a natural cardboard-box with the board packed inside a static shielding bag. There was no padding, just the board loosely flying around. But it seems to have survived.

image

One of the key features of this reference design is the “tight” assembly that minimises footprint and volume. This is important, especially when considering that consumers prefer smaller and lighter chargers for better portability. The provided documentation gives full assembly instructions for the board including the custom-wound transformer, which is nice.

image

However, at a glance, it is clear that this board is focused on minimising board area, but this means it is not quite a practical design for implementation or testing. For example, there is no mounting points or edge rails to allow the board to be nicely mounted for testing.

image

Likewise, power input is via two test loop contacts rather than a more familiar figure-8 lead connector.

image

The USB-C port also sits on a mezzanine board that rises out of the base in a precarious way that may not be durable enough for a practical charger. But as a demonstration of what is possible, it is still quite neat.

image

It is clear that attention has been placed to ensuring sufficient insulation between primary and secondary to ensure user safety. This does involve the use of some card and tape.

image

The design uses a mixture of Aishi and Ymin electrolytic capacitors for the high voltage DC rail. Both are fairly well-known capacitor brands which I’ve not had trouble with. They appear quite frequently chosen for such designs because of their temperature tolerance, lifetime and leakage characteristics. The output capacitors are solid electrolytic, which should offer even better lifetime.

image

The board is double-sided populated, two layer, with many of the semiconductors mounted on this side. The guide seems to claim the copper thickness is 0.040 inches, but this appears to be a typo because that would correspond to 30-oz. Perhaps it’s meant to be 3-oz.

image image

This is a very slim, surface-mounted bridge rectifier. There are also some other supporting components – here’s a Schottky and a MOSFET.

image

The main switching controller, the InnoSwitch4-CZ, specifically the INN4073C-H182, a zero-voltage switching flyback controller rated at 60W (enclosed) to 75W (open-frame) with output voltages ranging from 3-24V with an input of 85-265V. It features a switching frequency up to 140kHz with integrated Fluxlink feedback control and operating temperature up to 150 degrees C. Protection features include overtemperature, open SR FET detection, input overvoltage, input undervoltage, output short-circuit and output overload.

image

The partner to the switching controller is the ClampZero IC, specifically CPZ1062M. Its role, specifically, is as an active clamp that serves to increase energy efficiency by recycling wasted leakage inductance energy.

image

Finally, is another innovation, the MinE-CAP (MIN1072M) which is an IC that switches in and out capacitances depending on the instantaneous line voltage to allow for a reduction in the need for large high-voltage bulk capacitances by ensuring sufficient capacitance at lower voltages where the capacitance is really needed. It’s definitely an interesting approach to minimising space consumption as high voltage electrolytics are bulky.

This brings on the obvious question – is it small compared to current market products? Well, here’s a comparison with an (unreviewed) Novoo 67W GaN charger.

image

In terms of footprint, the RDR-928 is a little larger.

image

But vertical height is shorter on the RDR-928 so volume is probably somewhat similar although the Novoo does have three ports. As a result, it would seem the RDR-928 is a competitive design for size, even though the report and PCB seems to date this design back to 2021.

Performance Testing

This chapter focuses on the results of a battery of tests run using a variety of test equipment to verify the performance of the RDR-928 reference design as initially proposed in my RoadTest application. As this is a power supply intended for running USB-C loads rather than a lab power supply or one intended to run highly sensitive loads, I did not perform any transient load tests, so apologies if you were hoping for that – it was not part of my proposal.

USB-C Output Modes

Testing of USB-C output modes was performed using the Fnirsi FNB58.

image

The RDR-928 appears to support USB-C PD 3.0 at 60W (as claimed) with four PDOs. Aside from this, it does not appear to support Apple fast-charging (hence 0.5A indicated) but does appear as a USB battery charging standard dedicated charging port of 1.5A. Most of the other legacy fast-charging protocols are not supported. Support for QC 4.0 is indicated as it overlaps with USB-C PD, as does PumpExpress.

image

The four supported PDOs are 5V, 9V, 15V and 20V, all at 3A. No support for PPS is available, unfortunately, which will mean that some smartphones will not achieve the fastest-and-coolest charging possible.

This may seem somewhat narrow in terms of compatibility compared to third-party chargers which can often support a majority of the displayed protocols, however, it is my understanding that USB-C PD compliant supplies are discouraged from supporting such legacy and proprietary protocols to avoid ambiguity and compatibility problems. As a result, this reference design isn’t doing anything wrong – check a branded laptop USB-C adapter and you’ll find the same. But for an end-consumer that might expect a single power supply to fast-charge anything (even through an adapter), this is perhaps not the friendliest configuration.

Efficiency

Perhaps the most important metric of a power supply is its efficiency – that is, how little energy is lost when converting from one voltage to another. To perform this measurement, I used a Tektronix PA1000 Power Analyser to measure the real AC power consumption, a synthetic mains source comprising a pure-sine-wave inverter and variac to produce a stable AC voltage of 230V at 50Hz (unlike utility mains), a USB-C decoy (PD trigger) board to command the supply to produce the 5V, 9V, 15V and 20V that it supports and a B&K Precision Model 8600 DC Electronic Load to apply load to the output and measure output voltage, current and power. A short length (30cm) USB-C to USB-C cable was used to connect the decoy board to the RDR-928 to minimise losses, while four-wire connection was used to the load from the decoy board to compensate for voltage drops on that part of the connection. To connect mains into the RDR, an old IEC connector salvaged from a power supply was used with a short length of wire, soldered to the test loops. A 10A-rated mains cable of 1.8m length (0.75mm^2 conductors) was used to supply power from the power analyser break-out to the board. The test set-up appears below:

image

Measurements were made by scripted control of the instruments using pyvisa. Specifically, the test stepped through the load at 1mA increments until the output voltage fell below the requested voltage by at least 2V to capture over-current protection threshold. At each increment, a total of 32 readings are taken from both power analyser and load, and averaged, to reduce noise in the output curves. This test is repeated at each of the output voltages offered by the power supply. The recorded data is analysed in Excel and the self-consumption of the decoy board is compensated for.

Note that this setup contains exposed live connections at mains voltage and must not be touched while energised! I did consider enclosing the power supply while testing it, however, decided against this as it would potentially affect the thermal performance of the solution and make thermal imaging impossible. I do acknowledge that practical power supplies sold to consumers will be in a suitable enclosure, however, the RDR is set-up in a way that isn’t entirely practical as a product either so I didn’t feel too bad about this (e.g. no mounting points, USB-C port on a mezzanine board that doesn’t feel durable enough for rough handling, no power inlet jack or interface to pins).

image

Looking at the wide efficiency versus output current curves, we can see that efficiency rises quickly at low loads. Higher efficiencies are reached at higher output voltages.

image

Zooming in, I measured a peak efficiency at 5V of about 86.7%, at 9V of about 89.9%, at 15V of 92.3% and at 20V of about 92.8%. It should be noted that these efficiency figures aren’t without error due to AC measurement uncertainties, instrument accuracy at low currents and the lack of periodic instrument calibration. In my estimations based on the instruments’ capabilities, a discrepancy of 1% in my measurement would be expected.

Comparing this to the figures in Section 12.2.4 of the RDR-928 report, they indicate at 5V, a peak efficiency of 90.57%, at 9V of 92.53%, at 15V of 93.53% and at 20V of 93.82%, in all cases at 100% load.

Part of the reason for the difference would be due to the ohmic losses in the short segment of USB-C cable, which is likely the reason for the curves “sloping off” as load current increases. In the 15V and 20V cases, the measurements seem close enough to be in agreement in my opinion. However, the values for 5V and 9V seem to be significantly below expectations and this may be because they were measured with the PA1000’s more sensitive 1A internal shunt rather than the 20A internal shunt. While the higher power levels necessitated the use of the 20A shunt to accurately capture peak values, the smaller power levels could benefit from increased accuracy of the 1A shunt. However, this 1A shunt does have a higher resistance and does contribute some losses to the measurement as well. This may be the reason for the discrepancy, or perhaps the calibration on my 1A shunt is a bit off. Nevertheless, the efficiency values are excellent overall.

image

The same data, but instead re-plotted versus power output.

image

Voltage Stability & Over-Current Protection

This test was performed at the same time as the previous, and as a result, relies on the same test setup. As the four-wire sensing only extended to the USB decoy (PD Trigger) board, there was still an uncompensated voltage drop in the short 30cm USB-C to USB-C cable, the connectors and the decoy board itself. As a result, a small “slope” to the voltage curves under load is expected, but should be constant across all voltages if it is due to cable resistance (an ohmic loss).

image

At 5V output, the initial output was about 5.14V and at 3A, this fell to about 4.91V which is a very acceptable voltage stability. Over-current protection triggered at 3.32A. The computed resistance is 77.5mOhm which is easily explained by cable and connectors.

image

At 9V output, the output began at 9.19V and fell to 8.965V at 3A. Over-current protection triggered at 3.32A. Computed resistance is 76.2mOhm, a very similar value.

image

At 15V output, the output commenced at 15.35V and fell to 15.01V at 3A. Over-current protection triggered at 3.31A. Computed resistance is slightly higher at 95.6mOhm, but still within the realm of cable and connector losses.

image

At 20V output, the voltage commenced at 20.32V and fell to 20.1V by 3A. Over-current protection triggered at 3.32A. Computed resistance is 82.5mOhm, similar to the other readings.

Based on these results, the output of the RDR-925 appears stable across the different voltage outputs under load. It seems the output voltage is set slightly above the nominal value to compensate for cable losses, but there wasn’t any clear evidence of any negative-internal-resistance style sloped voltage compensation scheme as some more fancy retail adapters sometimes have. This is a perfectly good result nonetheless and is safe for connected loads.

Output Endurance & Thermals

To ensure that the power supply is capable of sustaining the rated power output, a full load (20V/3A) is presented to the supply while running on the bench for an hour. Power supplies which are not properly designed to account for thermals will usually “fall-back” in voltage or shut-down entirely during such tests.

image

The RDR-928 had no issues completing the test. The voltage started at 20.146V, taking a dive to around 20.01V initially before rising slightly to settle around 20.1V, possibly due to the thermal coefficient of feedback resistances. This deviation is very small, all things considered.

image

Thermally, the board did get somewhat toasty – the top side measured a peak of 91 degrees C via Topdon TS001 thermal imaging camera using the default 0.95 emissivity value. This appears to be near the middle of the board, specifically at the transformer. Perhaps winding and magnetics losses are responsible for this – after all, that transformer has been minimised in terms of size!

image

Looking at the chips at the underside, most of them are comparably “cool as a cucumber”, with the exception of the main Innoswitch4-CZ which is doing the hard work and on the underside of the transformer. Peak temperature is 88.1 degrees Celsius on the chip itself.

Standby Power Consumption

Standby power consumption was measured with the Tektronix PA1000 Power Analyser using the PWRVIEW application in the IEC 62301 mode.

image

At 230V input, the tested standby power result was 57.659mW with an uncertainty of 8.7532mW. According to Power Integrations’ report, Figure 9 indicates the test result to be around 54mW, which suggests both measurements agree.

image

At 115V input, the tested standby power result was 52.432mW with an uncertainty of 4.9415mW. According to Power Integrations’ report, their result measured about 49.8mW which is again within uncertainty bounds. This suggests that both measurements agree.

As a result, the standby power consumption of the adapter is excellent, easily meeting EU’s latest rounds of standby power reduction to 300mW slated to take effect in the middle of this year.

Ripple & Noise

While the RDR-928 report offers their methodology for ripple measurement, I did not opt to use it and as a result, the values may not be directly comparable. Testing for ripple and noise was performed with the Rohde and Schwarz MXO4 oscilloscope with the Rohde and Schwarz NGM202 power supply acting as load. The included passive 10:1 probes were used for the measurement, in DC-coupling with offset used to ensure the full characteristic of the ripple is captured. Bandwidth limit was set to 20MHz and HD mode was engaged for the clearest traces.

Testing of the 5V output was performed using a USB-C to USB-A adapter with a 4-wire connection, for the most accuracy.

image

At 5V, under no load, the output shows a sawtooth waveform which indicates an active power-saving mode. Average ripple is 48mV peak-to-peak.

image

At full-load of 3A, the ripple measured an average of 95mV peak-to-peak, which compares well to the 112mV peak-to-peak as indicated by the test report.

Unfortunately, testing the ripple and noise for the other voltages proved problematic, as I needed to use my USB-C decoy board to trigger the voltage outputs, but as the board powers from the supply, it introduces its own capacitance which likely causes the voltage rail to become filtered and for the ripple to be dampened. As a result, these figures may be somewhat lower than the actual ripple output, but I don’t quite have a better way to test this at present.

image

With 9V output at 3A load, the ripple measured 58mV peak-to-peak. This compares favourably to the report’s 99mV peak-to-peak, likely because of the capacitance on the USB-C decoy board. The shape of the ripple is very similar.

image

At 15V output at 3A load, the ripple measured 54mV peak-to-peak, again comparing favourably to the 96mV peak-to-peak on the test report.

image

At 20V output with 3A of load, the measured ripple averaged 60mV peak-to-peak, much less than the report’s 114mV peak-to-peak. This shows just what effect a little bit of capacitance can have.

Insulation Resistance

Insulation resistance was tested across primary to secondary using a Keysight U1461 Insulation Resistance Testing Multimeter. A test voltage of 1000V DC was used and a special adapter connected to all power and USB 2.0 data pins on the USB-C connector.

image

A measurement of 52Gohms was achieved, which is more than sufficiently high for electrical safety. In fact, this reading may be due to leakage in cables or internal to the meter (as I had repaired the OLED display in the past and potentially contamination could have gotten inside), so I wouldn’t consider the fact it isn’t reading off-scale (>200Gohm) to be a defect.

Conclusion

Whether you know it or not, there’s a good chance you’re not far from a Power Integrations-based USB-C charger. This RDR-928 reference design integrates their InnoSwitch4-CZ series controller with ClampZero and MinE-CAP technologies to maximise efficiency and reduce footprint. This reference design comes with a very detailed 76-page test report, however, it seems other online resources are missing at this time. In spite of the design being dated to 2021, its size remains competitive with commercial products of today.

Testing of the RDR showed generally excellent efficiency, at 5V of about 86.7%, at 9V of about 89.9%, at 15V of 92.3% and at 20V of about 92.8%. Efficiency at 5V and 9V were below indicated efficiency, but this is likely a test issue (perhaps due to calibration issues with the 1A AC shunt on my power analyser) as results for 15V and 20V on the 20A shunt measured within 1% of expectations. Voltage stability was excellent and over-current protection is consistent tripping reliably at 3.32A. Ripple and noise results were good and standby power results confirm the results presented in their test report. The reference design under full load does get a little hot, peaking at 91 degrees C in open air, but perhaps not critically considering its GaN credentials.

As a PD 3.0 power supply, it supports PD modes only for fast charging, eschewing support for legacy modes. While this may not be so consumer friendly as it may mean that the fastest charging might not be achieved with this solution in all cases, this is actually the expected result from a proper PD supply, as legacy modes can cause ambiguity and compatibility issues and (from what I am told) should not be implemented despite many aftermarket chargers doing so.

Thanks for element14 and Power Integrations for the opportunity to test the RDR-928 design. If you have any questions or comments, please leave a comment below and I will try my best to answer it. If you found the review interesting, insightful or helpful, I’d appreciate if you’d leave a like or share this with someone who might find it useful.

Anonymous
  • Take care Gough.

    I know I have cut way back since I became disabled.

    Just do what you can and what you enjoy.

    Life is too short to stress yourself.

  • Indeed, negative internal resistance in practice can be useful as there is no "far end sensing" wires in a USB cable.

    However, too much negative internal resistance can lead to thick and short leads producing over-voltage when at high loads ... so usually the compensation is conservative. Some chargers I've encountered have a bit of a split - negative internal resistance only kicks in above a certain current level resulting in a voltage "knee" rather than a straight line. This might help some devices that rely on voltage drops to detect the source limit to settle on a higher charging current.

    A charger may just seem to be a simple device, but there's definitely some differences in the way they are implemented.

    - Gough

  • Work on your health - that is a priority. Hope to see you back in action soon.

  • Thanks  .

    I'll be taking a hiatus from the RoadTest program for a bit to try and solve some complex health issues. Hope to return in the future some-time, but just to let everyone know that this may be my last review for a little while.

  • Very true, although I hope they would have opted for a better solution than a slot in PCB that has solder flowed from pad to pad. The mechanical support feels a bit lacking and the forces are directly transmitted.

    Even just using solid angled pins can provide some strain relief. Better yet, a ribbon, wire or connector of sorts may give more accommodation too. People can trip on cables, yank them out or shove them in without much care and you'd rather not have that force act on just solder alone.

  • A good question  .

    In essence, what I am reporting as the computed resistance is the voltage loss per ampere of load, which is essentially the "internal resistance" of a source as  helpfully pointed out. This value is inclusive of the USB-C short cable and decoy board internal losses.

    Low resistances are quite difficult to measure reliably. When we're talking about mOhm levels of resistance, you need to pass quite a bit of current to get a good signal. When the total reported resistance is below about 100mOhm, I don't find it worthwhile to try to measure where the resistances lie.

    The main reason is, if you have tried this yourself (as I have), you will often get very different resistance results for just a cable because of differences in mating plug connector contact shape, condition, force and shell design. The same cable mating to different USB-C sockets often results in 10-30mOhm variations per mating per connection. Plug and unplug the same cable a few times and you will get different results, in part, due to the relatively "loose" specifications for USB ports in general and a lot of variation in compatible designs. It is also not possible to fully eliminate the connector losses of the fixed connection on the RDR board either. 

    I would have computed the cable copper losses but the particular white cable I used does not have the wire gauge information printed on it, so no dice there.

    As I was measuring a relatively low value of internal resistance compared to these variables, I decided it was enough to conclude that it is likely such losses were cable, connector and decoy board related and the output was stable.

    Ideally, I would probably need a USB-C connector with split power pins for proper four-wire connection at the source and a decoy board with its own power supply to negate these losses, but that is sadly something I don't have and I don't feel like building right at this time as I'm still battling a few health issues.

    - Gough

  • Still, would be interesting to know how much of it is cables/connectors

  • This is what chatGPT says: "what is the computed resistance of a power supply?"

    The computed resistance of a power supply typically refers to its internal resistance, which affects voltage regulation and power delivery.

    How to Compute Internal Resistance of a Power Supply:

    1. Using Voltage and Current Measurements
      The internal resistance Rint can be estimated using Ohm's Law:


      image

      where:

      • Vopen = Open-circuit voltage (no load connected)
      • Vload= Voltage when a known load is connected
      • Iload = Current drawn by the load
    2. Using the Thevenin Equivalent Model
      A power supply can be modeled as an ideal voltage source Vs in series with an internal resistance Rint. If the voltage drops under load, you can calculate Rint as shown above.

    3. Typical Values

      • Ideal power supply: Rint=0 Ω (no voltage drop)
      • Real power supplies: Typically range from milliohms (high-quality) to a few ohms (poor regulation)
  • Another very good review Gough.