TI SWIFT™ Power Module EVM - Review

Table of contents

RoadTest: TI SWIFT™ Power Module EVM

Author: hlipka

Creation date:

Evaluation Type: Development Boards & Tools

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?: TPS56C215, LMZ Simple Switcher modules

What were the biggest problems encountered?: null

Detailed Review:

First, let me thank Texas Instrument and Element14 for providing yet another interesting road test. Although I already reviewed several power module and switching converter EVMs, each one poses its own specific challenges. And in addition I always learn some thing an improve my techniques and tools.

This time the challenge was that the TPSM84A21 has a rather low output voltage (with at most 1.35V) while delivering up to 10A. This means quite low resistances are needed - for full current at 1V as low as 0.1 ohm. Thats the ranged used by electronic loads (at least home-made ones) for their current sense resistor.

The TPSM84A21 power module

When you ever build a switch-mode power supply, you will know that this is quite a challenging task. You need to choose the right controller and switches  (or use one with integrated switches), then need to select the correct inductor (needs to be at the correct inductance and needs to handle the current), and last but not least the proper capacitors at input and output. When you get all of these correct, and the power supply does what it should, designing the PCB is the next hurdle. If this is not done properly, in the best case you just generate a lot of noise (which can radiate into other devices, or disturb your actual circuit). In the worst case it will just not work.

To make all of this easier, switch mode power supply modules were developer. Texas Instruments for example has the Simple Switcher Modules for several years now. They combine the controller, switching transistors and inductor in one module - but they still need external capacitance (although only in the range of 10µF).

With the TPSM84A21 (and A22) TI developed modules that are completely self-contained. They not only have the inductor in the module, but also all the needed capacitance. This reduces your design effort to basically zero. You need to choose the correct voltage set resistor and then place the module on your PCB - much less precautions are needed for this module when compared to a regular SMPS.

Most likely it will even be smaller than any other solution. Look at how much board space the module needs when compared with all the other TI buck converter boards:

(Nitpick: although the TPSM84A21 EVM matches at least one of the other boards in form factor, TI managed to place the connectorsw to different sites, creating yet another version of the EVM PCBs. It would be nice if they could agree one a common version, like all the other manufacturers do)

When compared directly, one can see that the TPSM84A21 module is about the same size as a LMZ12001 Simple Switcher module (including its leads), but doesn't need the external components:

With its 4MHz switching frequency it promises a good efficiency, low ripple and a fast load step response. Its only drawback is the low output voltage. But since many of the power hungry devices (such as FPGAs) nowadays only need such low voltages it will not matter for the intended use. There is also the TPSM84A22 module which has an output voltage of up to 2.05V.

Test setup

Correctly measuring a switch mode power supply is not a trivial task. The high switching frequencies and sharp signal edges mean that you easily introduce additional disturbance and noise when measuring. TI has a nice blog entry explaining this in more detail and also shows how to do this correctly. Keysight has some videos about it, but they are more broad and cover the full measurement of AC-DC power supplies. Nonetheless I found this one quite interesting.

So I'm using my probes in 1x mode, add a 50 ohm terminator to the scope input (the Keysight DSOX1192G unfortunately only has a 1M input) and probe directly, without the ground lead of the probe. Fortunately the TPSM84A21 EVM comes with probe headers that can be used for this purpose.

I'm using my PCBite PCB holder for the board, and hold the probes with some Dial Indicator holders (which I bought at ebay). This allows to position the probes where I need them, and then make sure they will not move.

If you want to get probing quality one would need to sacrifice a BNC cable and solder it directly onto the board, using the pads provided for additional capacitance.

For noise measurement I use fixed value power resistors (I tested with 10 ohm, 1 ohm and 0.1 ohm), since this is the easiest way (and doesn't need any additional space on the bench). For load step response normally one would use a electronic load. But mine can only handle about 4A at 1V (its load resistors are not small enough) I could not use it. So I went with my DIY load generator which just switches between two power resistors, using a power MOSFET.

Originally the generator was using a BUZ10, but its Rds was too high - 60 milliohm is quite near the actual load resistance (which is 0.1 Ohm). So I replaced it with a PSMN1R5-40 with only 1.6 milliohm Rds so it now has much less influence on the measurement (and also doesn't even get warm when switching 10 amps). This generator also creates very short rise times, so its perfect to test the transient response.

Since there are already other road tests looking at the efficiency of the TPSM84A21, I decided to do any measurements here. It would be pointless given the great work of Frank Milburn and Jan Cumps. Suffice to say that the module did never get hot during my tests (especially when compared to the load resistors...).

Noise figures

My first test looked at the quality of the output voltage. The voltage itself was accurate down to the single-digit mV level, and most likely this depends more on the actual value of the value set resistor than on the module. And with such high load currents the PCB traces also will affect the voltage at the point-of-load.

{gallery} Noise and ripple

Noise at 9.3A output current

Noise at 3A output current

Noise at 3A output current, but with 1 Mohm input termination

3A output level, seeing the regulation in effect

The noise level on the output is really low - even at high load its about 2 mV peak-to-peak. (Note that in the second measurement my scope still was configured for a 10x probe on the second channel, so its showing the wrong settings) The last picture shows the ripple, there you can see the output voltage regulation in action. The ripple here is about 4mV peak-to-peak.

All of this was measured without any additional capacitance at the output or the input. Adding these probably reduces the ripple further.

The yellow trace shows the input voltage. Here one can see the switching frequency, and also the effects of the input current spikes for each switching cycle. With 50 mV in the worst case I have seen this still seems to be acceptable, and one can easily add additional capacitance to reduce any effect to other parts of the circuit when needed.

When you compare the 2nd and the 3rd picture, you see the effect of the correct input termination. The 2nd is with a 50 Ohm terminator, the 3rd without (so using the 1 MOhm scope input). The difference in measured noise is nearly 10-fold!

As a comparison, here is what I measured for the TPS56C215 at 10A (with a 1.2V output):

Quite a difference - the higher switching frequency makes filtering easier, and since everything is in a (relatively) small package there is less change for any unwanted noise to escape.

Transient response

Second item of interest is the load step response. The TPSM84A21 is intended for demanding applications with high current needs, and such application quite often also have fast changes in current consumption. And when load current increases the output voltage must not drop too far, and it cannot overshoot when the load drops.

With its 4 MHz switching frequency the TPSM84A21 should deliver good results, buts lets see. I measured my load generator with about 10 ns rise time and 30 ns fall time, which should be fast enough for the 250 ns switching cycle time of the TPSM84A21.

{gallery} Load step response

Load step 0.1A->9.3A

Load step 0.1A->9.3A zoomed out

Load step 9.3A->0.1A

Load step 9.3A->0.1A zoomed out

The response when the load current increases looks very good - the small drop of 2 mV is most likely just due to the resistance inside of the package and soldering. But it looks much worse when the the load current drops again. After an initial small overshoot of about 15 mV it drops by 100 mV and then overshoots again by nearly 100 mV. This looks quite strange, although it might be caused by the lead inductance from the connection to the load generator. I would need to repeat that with much shorter wiring to be verify.

But the initial response is still very quick, so the initial assumption can be verified (more or less). For comparison, here the scope shots from the TPS56C215 again:

{gallery} TPS56C215 load step response

 

TPS56C215 10A load turn on

TPS56C215 10A load turn off

The undershoot is nearly 75 mV, and the overshoot (when ignoring the bad ringing due to probe connection) is still 130mV . This still OK, but much worse than the TPSM84A21.

Rating

When I rate a category with 5 it means that the device does what I expect, its not bad but also doesn't exceed my expectations. So a 30 would be a perfectly fine score.

  • Product Performed to Expectations: 7 because noise floor and transient response are better than expected, but the overshoot when reducing the load worries me a bit
  • Specifications were sufficient to design with: 6 - the data sheet looks good, there are additional application notes, but nothing else. Nothing to worry about
  • Demo Software was of good quality: there is none, so its a 5
  • Product was easy to use: the EVM gets an 8 here - I like the availability of probe points for a scope probe, makes testing much easier. Also the TPSM84A21 itself is easy to use. Would be a 9 if TI could finally come up with a common form factor and layout for their EVMs...
  • Support materials were available: I did not need any support from TI, so its a 5
  • The price to performance ratio was good: gets an 8 here - although the TPSM84A21 is not the cheapest, it can reduce your BOM size significantly, and also can save you a ton of development time.

So overall its a 39 - exceeding expectation. Gets a recommendation from me.

Wrapup

I like that module. When I had a project in need of such low voltages I would certainly use it immediately. Although it cost about 4 times as much as the TPS56C215, just by reducing the PCB real estate and the BOM size you might save money. And in addition you do not need to care about all the problem switch mode converters can make when you design them by yourself - just place that module on the PCB and you are done.

In addition the TPSM84A21 performs really well. Noise and ripple at the output can be neglected (except when you do precision analog circuits, but then you would not use a switch mode supply near your circuit anyway). Transient response also looks good, the the efficiency figures (from the other reviews mentioned) show that there probably won't be any thermal issues too.

Anonymous