RoadTest: SIMPLE SWITCHER®: LM46002PWP EVM by TI
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?: LT1074, LMZ12003H
What were the biggest problems encountered?: The EVM user guide is showing only the most basic information. Anything above that needs to be gathered from the (barely readable) schematic or by experiments.
When I applied for this road test, my intention was to use (or at least try to) this board as pre-regulator for one of my bench power supplies. But I realized that with the delay for the Sudden Impact challenge I probably won't finish that before that deadline for this review. So I decided to write down the real review part, and write about that project later on.
The board itself is larger than I expected. I have already a TPS92551 board (a power LED driver) which is significantly smaller, so I was a little bit surprised. There might be difficulties to fit it
into the enclosure for my power supply.
There is much unused space on top and bottom, so the board could have been smaller. But I guess its a standard form factor for TI eval boards. When looking at the bottom side, there is an interesting trace running all the way to the bottom edge, without an apparent need. I needed to look up the data sheet to find out that its the output voltage sense connection (running to the feedback input) which is recommended to run away as much as possible from the high-frequency parts.
I like that the board comes with screw terminal installed, but unfortunately they are not really marked. One needs to look at the test points near Vin and Vout to see which pin is which (and GND is not marked at all). On a side note - why aren't these test point installed with hooks like the other ones (there is one for GND, but not for the other ones)? I wanted to connect a scope probe there, because its kind of difficult to do so with just a screw terminal...
What I also missed are labels on the edge connector on the bottom side. None of the pads here is marked, and even in the user guide for the EVM they are shown only (in small print) in the schematic. On the top side edge connector there are two additional unlabeled pads that are used for the sense voltage (for both input and output).
Otherwise the board looks really nice. The layout looks clean and organized, and the main switcher is rather small (about 3 by 3 cm) There is an additional footprint if you want to add another input filter capacitor, and on the bottom side there is even room to to place a LC filter for the input (then the Vsupply input needs to be used instead of Vin). Too bad the user guide doesn't even mention all of this. It doesn't even mention that there is a jumper that can be used to set the output voltage to 5 instead of 3.3V.
After unboxing the board (which came in a rather large package, its about 6 times the size of the board itself) I just tested that it really delivers the promised 3.3V on the out (and tested the 5V setting too). Next step was to look at the actual performance of the module. For that I used my TBS1052B-EDU scope (I had not received the TBS1202 version at that time), powered it from my lab power supply and devised a small test circuit.
I looked at three different characteristics:
The ripple without or with load (I used a 6.2 ohm resistor) is really small:
Even in higher resolution it still is below 10mV:
(but the frequency measurement is clearly bogus)
I also tested the FFT mode here, but there is nothing significant to see:
So everything is fine here. The output signal is clean, and with a small LC filter probably can be used directly even for more sensitive analog circuits.
Out of curiosity I looked how the voltage at the switch output looks like. Even with my 0.5A load the duty cycle is low:
There is a some overshoot and some ringing visible when the switch turns on, so here is the zoom view:
This might be due to the ground connection of the probe, but it is much more pronounced when no load is connected:
To verify the behavior during power-on, I set the scope to trigger at about 2V. The result (without load) looks promising:
But its strange that the ramp-up is so long - the output voltage should rise much faster (the LM46002 is specified with a soft startup of 4.1ms, so it should be less than one division). After some data sheet reading it turned out that the EVM is configured to a slower ramp-up of about 20ms - another feature thats not mentioned in the user guide.
The next test is then with load, but it's not much different:
Still no overshoot (but with this slow ramp-up there is none to be expected). The small transient right at the beginning is most likely from me attached the test lead plug into my power supply (since I want to test the EVM behavior only, and not the startup behavior of my lab supply).
The same behavior can be seen when just connecting a the load resistor with a test lead to the EVM:
Since this obscures the true behavior I set up a small circuit to turn the load on and off:
I used my small crystal test signal generator to generate a 83Hz square wave that drives a N-FET. The FET in turn then connects the load-resistor to the EVM. The generator then doubles are trigger signal source (since in the best case there is no transient response visible at all). Note that I did not go to such great lengths in term of setup and measurement as Jim Williams explains in its AN104 about load transient response testing for voltage regulators - first I don't have all the equipment and second my goal was not to completely characterize the LM46002.
The result of my setup looks quite nice:
(the blue line is the test signal, the yellow one the output voltage)
There is a short drop of about only 160mV. And since I measured not directly at the output, part of it is voltage loss in the wire.During disconnect of the load it looks a little bit different, there is significant overshot:
But that ringing is quite short (just 1µs) and the 350mV are not very severe (and should be handled by any decoupling cap in an attached circuit).
So all in all the performance looks quite good to me, performance-wise the LM46002 performed very well. The chip even didn't get hot during my test (finger-tested), in contrast to my load resistor.
The EVM exposes all connections that are needed for using it as a tracking pre-regulator, but it needs some preparations. The top feedback resistor (Rfbt) and the capacitor across if (Cf) need to be removed. J1 that was used for setting the output voltage to 5V can be used to access the feedback pin when its resistor (Rfbb1) is shorted out. Dave Jones explains in EEVBlog #329 how a tracking pre-regulator works (http://www.eevblog.com/2012/08/07/eevblog-329-tracking-pre-regulator-ltspice-simulation-part-2/) and the presented idea should also work with the LM46002.
Despite my criticism I actually like the board. It performs quite well and seems to be a good choice for use in lab power supply. The data sheet is quite detailed and has all the needed information (at least so far). Unfortunately that isn't true for the EVM documentation. It really explains only the most basic stuff, and doesn't even mention the basic configuration of the LM46002. The schematic thats part of it is too small (and gets blocky when zooming in). So its quite difficult to see how the EVM works. This guide is the area where TI needs to invest to make that EVM a really good one.
I'm looking forward to using it in my power supply project, this will probably (hopefully) another blog post.