Richtek Synch. Step-Down Converter EVB_RT6204GSP - Review

Table of contents

RoadTest: Richtek Synch. Step-Down Converter EVB_RT6204GSP

Author: hlipka

Creation date:

Evaluation Type: Evaluation Boards

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?: LMZ21700, LMZ36002, LM46002, LMZ14201

What were the biggest problems encountered?: null

Detailed Review:

First, many thanks to Richtek and Element14 to give me the chance to test this evaluation board! And especially thanks to both for not only sending me the board, but also an additional tool to test it properly.


When I received the package, my first thought was "oh no, they did it again - why do they always pack these small board in such big boxes?":


But when I got to see the packing slip I noticed that it mentioned two boards, so I got curious. This is what I found:


The RT6204 was what I expected, but a "Load Transient Tool"? This doesn't sound like an evaluation board. Well, this is what it says: a tool to generate transient load on a power supply (or any DC-DC converter). Very handy, and it looks


(the tester is the board to the right) much nicer than the make-shift load generator I have used in the past for my road tests:


So that was a really nice surprise and a nice gesture to make testing easier.


The transient load tester

So before starting with evaluation the board, I did have a deeper look at this tester. You always should know your tools!

First thing to notice: there isn't any real data sheet or a specification for this tool. The user manual has some information throughout, but its spread out and even hidden in image descriptions:

  • 1W power for the load resistors
  • switching MOSFET with 30V and 4.5A - and the web site lists this tool as being suitable for power supplies up to 5A
    (although the data sheet for the FET lists 5.7A for less than 5s load)
  • no spec for the minimum resistance of the FET (I looked up the manufacturer data sheet for it, which says 33mOhm)
  • switching times of 500nSec
  • and the specification for the base load resistor is just on the web site (its a 10W resistor, and of unknow origin)

Although you can find anything when looking at the web site and the user guide, this really should be consolidated. A short paragraph in the user guide listing the maximum values and the normal working parameters would be nice.


I like that the tester comes a battery box attached at the bottom, and also has proper standoffs so it doesn't lay around on the work bench. It would have been nice if it had a real power switch instead of of using a jumper, though (especially when you want to turn it off quickly for some reason).

Note that the 10 ohm resistor is actually too low to be used with the RT6204 - with the configured 5V output it would already draw 500mA which is the maximum current - so you cannot draw more power. So I used a 24ohm resistor (actually two 12ohm ones in series) for the transient testing (see below).


So I had look at my scope and went to see how fast it really switches. 500nSec are quite slow compared to dedicated FET-Driver chips. I configured the tester to use the 22ohm resistor, connected ground to my power supply and then another 22ohm resistor between the 12V line and the load input of the tester. So the tester then toggles the voltage on its load resistor between 6V and 12V (or the voltage on the second resistor between 0 and 6V), and we can look at its rise and fall times:


(For triggering I connected the pulse output of the tester to the second scope channel, which is why you see the 244Hz trigger frequency).  So the rise and fall times are quite a lot better than specified, with 110nSec and 280nSec. Not to bad, and certainly usable for testing power supplies. The small overshoot is due to the inductance of the scope probes ground connection, and the length of the wires used.


After that, I repeated the same test with my own load generator, which results in these traces:


The rise and fall times are much lower now (10nSec and 30nSec), which unfortunately results in much more overshoot from the scope probes (the ground lead is too long for such sharp edges). So its one order of magnitude faster, which will put even more stress on the power supply.

So maybe Richtek can use a FET driver to further improve their tool. I also could add a second driver  board it, and then had the choice between faster and slower pulses.


For the rest of my testing I decided to go with Richtek's transient load tester. Its easier to use than mine (less fiddly setup), and probably the road test was intended to use it. (But if someone would like to get some scope shoots with the faster load pulses, feel free to ask).


Testing the RT6204

Back to the actual device to be reviewed - the Richtek RT6204.


The RT6204 is a buck converter, meaning it converts a high input voltage to a lower one, but with as few losses as possible. It is extraordinary in that its allowed input voltage can go as high as 60V. This makes it suitable for application with either high input voltages (like telecom systems) or where large input voltage spikes can occur (like cars or industrial systems). Its also quite rigged in that its protected against shorts on the output (via current control) and thermal shutdown. A soft-start capability prevents too large inrush currents (which might blow fuses), and it shuts down when the output voltage gets too low (to prevent damage from brown-out situations). Its specified for 500mA output current, but the current limit will start not before 600mA.

The RT6204 also goes into a low-power mode with low output currents for better efficiency (which is done by pulse-skipping on the switch node), but the data sheet doesn't go into further details on that. It just states that pulse skipping will happen "With Vfb 1% higher than Vref, the COMP voltage will be clamped at a minimum value and the converter will enter into pulse skipping mode." but it doesn't state when this will actually happen (and doesn't provide formulas to calculate this switch-over point).


The evaluation board is quite small, and one could easily incorporate this in a real projetc. It even comes with mounting holes - something which is easily forgotten. Too bad that it doesn't come with screw terminals for input and output, that always makes working with it easier. For testing I missed additional ground points near the signals, and especially I missed pin header connections for the high frequency signals (input, output and switch node). Using the normal scope probe with its long ground tail always introduces ringing and noise which can be avoided by such connections.

The EVM is configured for 5V output, and its not easy to change. There is not selection possible,  and the resistors that need to be changed (R2 and R3) are not easily accessible. The user guide doesn't tell anything about the actual circuit, and what the non-populated components can be used for (e.g. the additional diode to further reduce losses). Thats something I would like to see.


I had two instances where I accidentally connected the input voltage in reverse, and the board survived both of them (during the first one I blew the fuse on the current meter). So its at least a little bit rugged with respect to that, but you should not count on it.


What I have seen during my measurements (and its even shown in the data sheet) is that the input voltage rises a little bit (about 50 to 100mV) when the output current drops below a certain threshold (probably when it goes into pulse skipping mode). You need to look for when designing your own circuit, and should not go to near to the limits of your powered devices.


Efficiency measurements

First I started with looking at the conversion efficiency. This is always interesting, since it also decides whether you need additional cooling or not. While doing so I noticed that the supply current was the lowest with input voltages between about 13 and 15. It started to rise again with lower input voltage, which is in contrast to the datasheet, which I found strange. Maybe that's something that the additional diode D1 on the board could solve, but I did not test it.


For the setup I used my small home-made bench power supply to create an input of up to 45V:


I use it for smaller loads, since it can deliver only 100mA, but it doesn't take so much space on the bench. The bigger one on top was used for the lower input voltages since they need more current (at can deliver 2A at 13V).

Then I connected four multimeters to measure all values at the same time:


(Bunch'o'Wires one)

The small DMM one on the left was one of my first ones, but is still works and I check its accuracy regularly.

I use resistors as a load because they are easy to exchange, and give reproducible results (I don't own a digital electronic load yet) without additional adjustment. All the values came into a spread sheet. There I also calculated input and output power, and resulting efficiency.


For the first run I tested with input voltages of 45V, 30V, 12V and 6.5V. Anything lower and the output voltage will drop. I used loads of 10ohm, 140ohm, 1kohm and no load.

The results are what one can expect from a buck converter: the efficiency is higher with larger output currents and lower input voltages, since then the switching losses and the quiescent current are not so big compared to the output power. I uploaded the complete spreadsheet with all measurements so I don't repeat them here. With 12V input the efficiency reaches 90%, and in the best case of 6.5V input it can go as high as 94%. In all my tests the RT6204 stayed cool to the touch even with 500mA output, so there is no need for additional cooling. And apart from a big ground plane and some via stitching the EVM doesn't do anything special. (But with 2.5W output power, and 90% efficiency there are only 250mW to dissipate)


Next I looked at one specific efficiency curve of the data sheet, namely the one for 12V input and 5V out. Here I used my electronic DC load (with manual control) to measure efficiency at a larger span of currents, in much smaller steps.


(Bunch'o'Wires two - even more this time)


Starting with the highest load, I measured these values:

VinIin (mA)VoutIout (mA)Pin (mW)Pout (mW)Efficiency

The input voltage changes a little bit each time because of the burden voltage drop on the current shunt resistor, so I did regulate the power supply to have nearly 12V at the board.

These results are quite exactly what the data sheets states - the efficiency is at 90% or better for a large output current range. and only drops with current lower than 100mA.


Overload testing

In the last test you can see I already tested with higher output currents than specified, to see how the current limiting might look like. To test this with a bit more detail, I again used the big load resistor supplied with the transient load tester and just changed its resistance. With an input of around 19V to 20V, the current limit kicks in at about 800mA output (at least for my chip - the typical value for that is stated as 830mA with a minimum of 600mA). When reaching that current, it will not switch off but just limit the current so the output voltage drops. Again, these values can be found in the attached spreadsheet.


Noise and transient response

Last part of my test involved using an oscilloscope to look at noise, ripple, and transient response. Here the supplied tester came into play. The first look was to check the output voltages at startup. There should not be any overshoot, and since the RT6204 comes with soft-start we should see a smooth ramp-up:


The EVM is configured to a 13mSec ramp-up time, but here we see only 10mSec. I guess the capacitor is a little bit on the small side of its allowed range. Starting up into a load looks exactly the same, until I connected the 10ohm resistor as a load:


Whoops. This looks as if we either hit the current limit at the output, or an under-voltage condition is detected. It certainly looks strange.


The next look was the voltage at the switching output - there you can see the duty cycle and should also be able to detect the pulse-skipping mode. In addition the yellow trace shows the output voltage:


The left trace is the one with the 10ohm load, where you can easily see the 366kHz switching frequency, and a duty cycle for about 40%. The noise is quite low, but on the scope probes there was some coupling (via the ground connection)

On the right hand side I did not connnect any load, so the RT6204 goes into pulse-skipping mode. The switching frequency is reduces to about 20kHz, and we get a much larger ripple at the output.


Next was looking at only the ripple and noise at the output. Here you should use a scope with 50ohm input impedance (to get rid of any noise coupled in into the leads), and a connection as direct as possible. As already mentioned, the EVM doesn't provide test points that are suitable, so I used the normal scope probe:


The images are, in order, with 10ohm, 140ohm and no load. With high load the output  is more or less flat - the noise floor of my scope with that probe is about 15mVpp. So this looks quite good. As soon as the current is low enough to get into pulse-skipping mode, the picture changes and you get noticeable ripple. This means additional capacitance at the output.


The last test was looking at the transient response.

I did set it up so the normal load was about 200mA (again using 24ohm as load resistance), and then checked with different resistors at the load tester. I started with the 22ohm resistor, which gives 400mA output:


This is quite OK, but 75mV under and overshoot are a little bit high. The reaction time is quite good, the RT6204 reacts after about 10µSec. With the 10ohm resistor, we get at about 650mA:


Even though its already of of the specification range, this is something it should handle. A voltage drop of nearly 170mV is not really acceptable. Also, the time until the voltage drop gets regulated to an acceptable level gets longer.

I used this test to show the behavior during an overload situation, so I also tested with the 4.7ohm resistor, which creates nearly 1.3A current:


Its easy to see how the output voltage gets regulated to a much lower value so the current doesn't go above its limit.


So the transient responses were OK from a timing perspective, but the voltage drop was larger than I expected. I think this might be solved by larger output capacitance, but since there is no footprint at the EVM for that, I did not test it.



The switching frequency of about 350kHz means that the converter has less switching losses (and can more easily achieve better efficiency). But on the other hand it needs a larger inductor and larger capacitors to get the same low-ripple output.

For a size comparison, I took a photo together with some other DC-DC-Converters with similar specifications (about input voltage and output current):


The two board on top don't have an external inductor, they are complete modules. The left runs also with 60V input and delivers 2A - but its about the same size as the inductor for the RT6204. The right one delivers 650mA but can handle only 17V input. This is much less than the RT6204, but the complete solution is smaller than even only the inductor. The biggest solution is the one on the bottom left, being a good deal larger than whats needed for the RT6204. But OTOH it also delivers 2A output current.

So the RT6204 needs to score on its other special features, such as the undervoltage-lockout and the quite helpful output current limiting (and probably its price)



So I have some mixed feelings about the RT6204 and the evaluation board. Its quite small, and nicely layed out, but could be better suited for testing (e.g. pin headers like the LMZ36002 board, or screw terminals). The performance under load was quite good. But with low load the output and the ripple start to rise, so one needs to handle these situations with some care.

The large input voltage range its its safety features make it a good candidate to power e.g. a RaspBerry Zero in a car setup (where it needs to withstand large voltage transients), or in projects with high input voltages (like my 3D printer with its 24V power supply)

Oh, and the transient load tester earned a fixed place in my toolbox, too.