Texas Instruments-DCDC Power:FPGA & Data Converter - Review

Table of contents

RoadTest: Texas Instruments-DCDC Power:FPGA & Data Converter

Author: Fred27

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?: Recom RS-483.3SZ/H3 DC/DC converter

What were the biggest problems encountered?: Soldering extra output capcitor on a board with a large ground and power plane and no thermal relief.

Detailed Review:

TPS62135EVM-698 overview

The TPS-62135EVM-698 is an evaluation module for TI's TPS62135 step-down buck converter. The evaluation module's guide ways it takes a 3V to 17V DC input and outputs 3.3V (or optionally 5V) at up to 3.5A. You're probably wondering how it steps a 3V input down to 3.3V but all will be revealed. [Spoiler: It doesn't]

 

Unboxing

There wasn't a lot to the unboxing of this TPS62135EVM-698 evaluation board. As usual, it came in a box about 5 times the size it needed to be, but inside was just the single PCB. It has a 0.1" pitch header on one side for input and on the other for output. There are some jumpers and test points and that's it. Well - what more do you need?

 

Initial impressions are that the TPS61235 itself is tiny - an odd 11-pin VQFN package measuring only 2mm x 3mm - and is dwarfed by the inductor next to it and a large (7343-43) input capacitor. I wondered if this tiny chip could handle 3.5A and initially assumed that it must be just "supervising" rather than passing that sort of current. A quick look at the schematic shows that apparently the current does pass through.

A very quick explanation of how a buck converter works

Many of you may know exactly how a buck converter works, but just in case I'll give a very simplistic overview. A linear regulator can be represented as a variable resistor connected in series with your load, and a feedback mechanism to constantly adjust this resistance to give the right voltage at the load. Linear regulators have many good qualities, but can waste a lot of power in this "resistor". Switching supplies improve efficiency by rapidly switching the power on and off. The output is monitored and PWM duty cycle is adjusted to keep the output voltage where it needs to be. The output is then smoothed by an LC filter formed by an inductor and capacitor on the output. (I know, I know - this is really oversimplified.)

 

Evaluating the evaluation module

First things first, I'm going to quote from the evaluation module's User Guide - "The EVM converts a 3-V to 17-V input voltage to a regulated 3.3-V or 5-V output voltage that delivers up to 3.5 A". How does a buck converter do that? Of course it doesn't. What the guide should say is that the TPS62135 takes between 3 and 17V and outputs between 0.8V and 12V. The evaluation module has the output user-selectable to be 3.3V or 5V and requires the input to be greater. To be fair though, I measured the output at 3V (without load) and it tracked it to within a few mV. Much as you'd expect, really. The converter will be permanently on and should track the input apart from a minimal drop due to the resistance of the power FET and inductor. (The photo shows my PSU at 2.9V but with the same meter I measured it at 3.005V.)

[photo]

 

Onboard connectors

The board has a few connectors and test points Firstly let's deal with the connectors. J1 is the input and J2 is the output. The both obviously have ground (connected between the two) and sense points to allow accurate measurement of the voltages. At lower currently these separate sense points aren't really necessary, but once you crank it up to a few amps then the voltage drop across any test leads could be significant.

 

J3 allows for easy measurement of the SS/TR voltage. It also has an input for the tracking voltage, but this must be enabled by populating some resistors if you want to use this feature. I would guess that most people wouldn't so I won't dwell on this.

 

J4 allows access to the power good (PG) output. This is an open-drain signal indicating that the desired output voltage has been reached. This is particularly useful if you use the soft-start feature of the device as this can take the place of any voltage supervisor you might otherwise have to add to your design.

 

Onboard jumpers

JP1 allows access to the EN (enable) pin, which as you'd guess enables or disables the device. Not a lot to say about this. Ideal if you will be disabling the device and want to check how it behaves.

 

JP2 allows you to switch force the device to stay in PWM mode (jumper set to PWM) or to use power save mode (jumper set to PFM which is the default) where it switches between PWM and PFM modes as appropriate. More on this later.

 

JP3 allows the user to switch between two output voltages. For the EVM these two outputs are set to 3.3V and 5V, but they are user configurable with resistors.

 

Evaluating the TPS62135

I'm not a huge fan of QFN devices. I use the usual hobbyist approach of a toaster oven modified for reflow soldering. I'm more than happy with 0.5mm pitch LQFP packages, but I find QFN so much harder to work with and to inspect. I can only imaging that the three long pads underneath the device make this harder still. Especially as they are perhaps the most important ones and handle the power. Take a look at this land pattern and tell me it doesn't fill you with dread! I suppose for small space-efficient designs that are automatically PnP'd it's a good idea. I'm not sure how hobbyist-friendly it is though.

General usage

In common with most buck converters, the device works by comparing the scaled output voltage with a reference voltage - in this case 700mV. The output is scaled with resistors R1 and R2. forming a voltage divider between Vout and GND. According to the data sheet these resistors should be below 400k and pass a current of at least 2uA. The formula for calculating appropriate values is given as:

R1=R2×(Vout0.7V 1)

If you're using the pin-switchable output voltages then you need to add R3. R3 is effectively paralleled to R2 to change the voltage divider, so you need to determine the appropriate values that when paralleled give you and "equivalent R2" to satisfy the formula. The evaluation board use R1=560k, R2 = 150k and R3 = 232k (all 1% tolerance). R2 and R3 paralleled give 90k. These values can of course be changed if you're happy replacing a few 0603 components.

Soft start

The supplied capacitor C3 is used to set the soft-start or ramp up time of the device.It has been set at 3.3uF for a fairly fast start time of 920us. The value of the capacitor required is determined by the following formula - with a minimum of 150us. Once again replacing an 0603 component is needed.

Css=Tss×2.5uA0.7V

There's loads more that this device can do, but rather than go into detail here, it's all very well documented in the datasheet.

 

Trying it out

The first thing I decided to test out was to compare it with an off the shelf DC/DC converter I bought recently. I connect the evaluation board and a Recom RS-483.3SZ/H3 to the output of a TENMA power supply. Neither outputs had any load or capacitor on the output. I set the scope to trigger as the PSU passed 3.3V and measured the output of both the Recom and TPS devices.

 

The first 2 screenshots show how the TPS (pink) and Recom (blue) compare. The TPS started up 1ms after the trigger and hit a peak of 4.8V before settling down. The Recom took 12.6 ms and hit a peak of 6V

A closer look at the TPS shows a nice clean ramp up over that 1ms and a nice clean output. Nowhere near the switching noise you can see on the Recom.

More power

OK so time to ask a little more of this device. Supplying voltage with no load is easy, but lets actually ask for some power. The first thing I had to hand was a  35W12V GU53 incandescent lightbulb, so I thought I'd use that. A nice resistive load that draws about 1.3A and gives a gentle glow when supplied with 3.3V. Here's what happens when I hook this up and switch my 16V supply on. You can see that the input supply doesn't rise as quickly once the TPS starts drawing some power, and the output has a little more of a ramp up. The output eventually settles at an average of 3.52V with about 400mV of ripple. Hmmm. A little high and a little noisy. And what is that spike just as the input crosses 3.3V and the scope triggers. Wow. Off the screen and at least 14V!

I admit I haven't got an output capacitor other than the 22uF ceramic that already populated, but that doesn't seem great. There are 2 further unpopulated pads for C8 and C9 but to quote the guide "these capacitors are not required for proper operation but can be used to reduce the output voltage ripple and to improve the load transient response". My go-to microcontroller input decoupling capacitor is a 10uF tantalum so I stuck one of those on the board.

 

oops!

After finding little difference and some very inconsistent results I realised a couple of things. Firstly, that powering the PSU on could give varying amounts of noise and ramp-up time. Secondly, that the ground connections on the scope weren't ideal and were the cause of most of the noise. I almost removed the above screenshot as reflects badly on me and is not representative of the device I'm testing. However, I decided to leave it. It's a good reminded that if things don't look right, make sure you double check your measurements and methods.

 

This is how it really looked. I'm now measuring properly and powering up by reconnecting a lead. I've also added a trace (dark blue) for the "power good" output. Even with the extra capacitor the noise is still 320mV (between 3.28 and 3.6V). Not brilliant, but certainly OK for the intended purpose.

Trying some noisier input

I'm planning to try the device out with a noiser supply. I've got a SMPS that in combination with the Recom device managed to destroy a TI CC3200 Launchpad. I'll add that here once I've had a chance, but I thought I'd get the review up for now.

 

Conclusion

This is a nice evaluation module. There's not much too it but there doesn't need to be. Everything you'd need to evaluate the device is there. TI provides excellent documentation on both the evaluation module and the IC.

 

The IC itself is very impressive. It does an excellent job with reasonably low noise for a switcher and good efficiency. The only reservation I'd have is that the package doesn't make it particularly easy for hobbyists. If you'll be ordering your board from one of those Chinese PCB houses, applying solder paste with a syringe/toothpick and reflowing it in a modified toaster oven, then perhaps it's not the device for you.

Anonymous
  • It does give an "achievement unlocked" type of feeling when it works

     

    Below are two ICs similar as the one on your board (that's how they looked like after being removed with hot air).

    I never thought that replacing them would be feasable, but it is.

    For mental preparation, I look at this (one of my favourite videos about FPGA soldering):

  • I'm just about OK with "standard" QFN (although they're not my favourite). I feel this one with the longer pads underneath is just asking for solder bridges you just can't see or correct without removing the IC and starting again.

     

    I suppose I should try before assuming the worst. However, if I had a choice between this and an easier package then I'd probably not risk it.

  • Thank you for the review!

     

    the QFN packages are ok fore home soldering. I was reluctant to work with them at first.

    When one such IC failed on a design I was reviewing, I gave it a go, and it's not that difficult.

     

    The first ones I used hot air only and rework took longer than I felt comfortable with. But it worked.

     

    Now I'm pre-heating the PCB  with a Tenma device I got from Shabaz. In combination with hot air it works like a charm. I've soldered similar devices than the one on this board, with a few big pads and many tiny ones.

  • "I almost removed the above screenshot as reflects badly on me ..."

     

    Don't worry, we've all been there. [If only I'd been taught such useful lessons back when I studied electronics. They couldn't even teach me how to design things, let alone the practicalities of how to test them.]

     

    What does the noise on the output look like? [Put it on ac and wind it up so we can see it properly.]