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?: LMZ14203, LMZ36002, LM46002
What were the biggest problems encountered?: No real scope test points
[Edit 12/7/17 - fixed the name of the EVM, for some strange reason I called it the TPS56C512 in the first version]
First of all, many thanks to Element14 and Texas Instruments for sponsoring this road tests. This is now my 3rd road test with a TI power supply board - this makes some interesting comparisons. Here you can see the LM46002 on the top left, the tested TPS56C215 on the top right and the LMZ36002 on the bottom.
Each of the boards follows a different convention and philosophy with regards to form factor and layout. The edge connector on the LMZ46002 is not really of use on an EVM (and its not even labeled). The LMZ36002 was the easiest to test, since it has large test point hooks for the scope and the DVM, and comes with pin headers to directly connect the scope probe for the important signals. The latter one is what I'm missing on the TPS56C215 EVM - this makes testing noise more difficult. A jumper for voltage selection would also be nice, as are holes for standoffs (otherwise you lay down the EVM naked on the table and just hope that no bare wire is hiding underneath). TI should really introduce some standard here.
When you look at the actual solution size instead of the boards, they all need about the same space for the converter and the required components. The LMZ36002 is maybe a tad smaller due to the integrated converter, but not by much. From the delivered output power the TPS56C215 and the LM46002 are nearly the same: 12A@5V vs. 2A@28V (the LMZ36002 comes in last with 2A@7.5V). But the LM46002 and LMZ36002 have a really wide input voltage range of up to 60V, which can be quite handy (e.g. in automotive situations where large voltage transients might occur). This higher input voltage also needs bigger capacitors (size-wise, not capacity) to withstand, making the boards larger.
The EVM is configured for 1.2V output voltage, with a current limit of 12A. The switching frequency is set to the maximum of 1.2MHz, which should result in fast transient response and low noise. With low load the converter is set to pulse skipping (DCM) mode, and the startup time has been set to 6ms (up from the 1ms default).
The board is largely unpopulated, but handling 12 amps needs large ground planes and wide traces. There are test hooks for all interesting signals, but they are quite small (although large enough).
For the tests I used my Tektronix TBS1202B-EDU as usual. For looking at the noise figures I used resistors from 140 ohm down to 0.1 ohm as load. The latter one (a Vitrohm KH 206-8 series) needed active cooling - its specified with 4W but 1 amps through are more than twice of that. This even managed to unsolder one of the wires:
For transient response testing I used my makeshift dynamic load generator with the same resistors. This time the 0.1 ohm one was OK, with a duty cycle of about 50%, but the switching FET also needed some cool air (it has an internal resistance of 60 milli-ohm which means its also converting 6 watts into heat):
For these tests I was using my normal lab power supply as power source (its based on a LM350 and can deliver up to 2.5 amps which is OK at 12V input). For looking at the output ripple I powered the EVM from 8V AC, rectified by a LT4320 active bridge rectifier:
The LT4320 drives 4 N-FETs (the two small SO8 packages) so the losses are quite small - no heat is wasted in the bridge.
So lets see how the TPS56C215 behaved in my tests.
First I looked at noise and ripple. The TPS56C215 data sheets it expects a clean and stable power supply, so thats what I used. Under 10 amps load there is a slight ripply visible, but the noise is quite low:
The ripple is clearly caused by the switching, but 20mV is not that much (and can probably be eliminated by local filtering). With a 1A load its much less, and looking at a 10mA load shows that the TPS56C215 starts skipping pulses (so it reduces its switching frequency):
(Note the different time base - the switching frequency is now about 10kHz).
Next test was the load transient response. This is especially interesting when there are large changes in the output current, so lets look at the results when turning a 10A load on and off:
There is a noticeable drop of about 75mV when the load current rises, but it gets regulated back to just a 20mV drop after 20µs. During load turn-off the transient is about twice of that, so lets look at a zoomed in version of that:
The really large spike right at the beginning is due to the scope probe connection. Since there is no pin header available I needed to use the ground clip which results in some ringing due to the inductance (which can also be seen in the blue waveform which is the load switch N-FET gate voltage). After that the overshoot is about130mV which is acceptable.
I'm quite happy about the low drop of voltage even under high load. As a comparison, the LMZ36002 output voltage dropped by about 50mV with a 1A, so the 20mV of the TPS56C215 are quite an achievement. Also, the response time to load transients is really fast - its just two to three switch cycles for load turn-off.
The next test was looking at the output voltage when starting into a connected load. With a 0.1 ohm resistor connected I connected the input voltage:
Nice, smooth ramp-up with no over-shoot at all (the scope measured 0.006% and this is well in the range of the ripple voltage). I expect the same in a configuration where the rise time is just 1ms, which makes the TPS56C215 unproblematic even in configurations with a large load connected directly.
My last test was looking at how well the TPS56C215 filters out ripple on the input voltage. For that I powered the EVM from a capacitor-filtered bridge rectifier, without additional regulation. With 8V AC input the board should draw about 2A, which results in a significant ripple voltage at the input (I'm using a 4700µF cap). The scope was now triggered on the input voltage, and I used the averaging function to filter out the noise:
So the input ripple is 2Vpp, and on the output the ripple is barely visible. Its there, but since my scope cannot go better than 20mv per division I can only estimate it (looks like 2mV, maybe 3). So no worries here.
For completeness sake I also looked at the voltage waveform at the switch node of the converter:
On the left with a 10A load, on the right a 1mA load. Note the different time bases - with a low load the TPS56C215 goes into pulse skipping (aka discontinuous conduction) mode. Apart from that there is nothing special to see here.
When you need a buck converter for really high loads, the TPS56C215 is an excellent choice. Especially its performance with high load transients is remarkable and make it well suited for such applications. The only drawback is the limited input range when compared to other switch mode converters, but this is to be expected when you want to have such large out put currents. (Actually 17V is quite high when compared to other chips which mostly accept only up to 6V or 5.5V input).