2017-11-10

Overview of how the ZL8800 works, and how its unique design enables responding to transients in a single cycle.

Transcript

So the ZL8800 features a no-compensation fixed frequency charge mode control. So let's take a look and turn on the device to see how it works. Now, the output voltage sense differentially. It goes back into the device. The first stage we use is a programmable game amplifier, and this is automatically set-up with the device, and we'll compare it against the target output voltage before we go into our ADC. So we're only going to be quantisizing the error voltage. This is a high-bandwidth fast transient response so we're only looking after relative error in order for regulation.

So next stage is we quantisize the error voltage using our high-speed ADC. And from this point on all the way to the gate drivers, we're gonna be in the digital domain. And this gives us a lot of flexibility to use a lot of non-linear loop compensation. In order to achieve this and ensure stability of the system, we use something called an ASCR modulator. That stands for a single cycle response. It's a high-speed, high-bandwidth loop control that's used to sit there and adjust the DPWM. So for any load transient that comes on, it will sit there and look at the amount of charge displaced from the output capacitor and put the probe signal back on to replace it within a single cycle.

The inputs to this block are two. There's a gain setting and a residual. This sets up the overall transient response speed, and a residual is a dampening factor. It sets the response rate of the loop. From that stage, it's directed to DPWM, so it goes out to the gate drivers. And beyond this, the device will also set up adjustable dead time and optimize the dead time if you're using a DPWM driver. The best way to take a look at this is on the bench, so I'll show some scope waveform and how the gain and the residual settings affect the overall loop response.

To really understand the benefit of the ZL8800 and the no-compensation modulator, let's take a look at the performance on the bench. I'm using the dual channel output board for the ZL8800. So this has two independent output voltages, each regulating itself through 1.2V on channel zero and 1.5V right down on channel one. But for this exercise, I'm just gonna use channel zero. So you can see I've got the electronic load connected to this output. For the scope probe output, I'm just going to set there and have a direct connection right across one of the output capacitors so we'd get a fairly clean pickup. I'm wiring with a 12V input and I'm just gonna use an electronic load box to step between 5A and 15A. So we're just gonna apply a 10A load step and look at the transient response.

So if I enable this and take a look in the scope on the performance, what we can see here is I've got a 5A to the 15A load step that's turning blue and then purple is the output voltage waveform. With initial load step of 10A going from a 5A to a 15A level, you can see the output voltage deviation is about 25mV to 30mV. That corresponds to about 2% output voltage deviation. That's great for a just a standard default set-up of a device.

So the initial ASCR gain that's programmed to the device is 256, which is a very low setting. That's fine for almost any set-up that you use. No matter what output capacitance, inductor, that default setting should come up and derive nice stable yet fast transient response. If you want to make improvements, we can make some adjustments to it, increasing the gain levels from 256 to 400, 600, 800, maybe up to 1000 if you really want a high-gain output.

So let's take a look at what it does and how it affects the transient response. So I'm gonna increase it to 400 and as I send the command across PMBus, right away we can see the output voltage deviation drops from about 25mV to under 20mV. So on the scope right now, what you're seeing is output voltage waveform is AC-coupled and it's about 20mV per division. I can keep increasing it from 400, and now I go up to 600 and it gets even better. Now, we're probably about 15mV of deviation for that load transient. We can keep making improvements by continually increasing the gain. So as I go to 800, you can see now we're down to about 10mV deviation.

At this point in time, this is probably about the highest gain we would want to use. Continued gain increases won't really improve the performance of the device, and the reason why is we have the device as fast as possible. The major limitation now is not the small signal response. It's a large signal response. That inductor current has the slew so the only way we can get to go faster is to use a small inductor value or use greater output capacitance. So the only downside now of increasing the gain is we'd just be adding unnecessary jitter into the system without improving the transient response. So typically, this will be more what we consider to be an adequate transient. We're down to a less than 10mV deviation on the output voltage which corresponds to less than 1% of the output voltage, which is a great target to achieve on the power supply.

The one other value that we can change for the charge mode control loop is the residual value. As we talked about earlier, that effectively looks like a dampening factor. It would make an equivalent to an analog controller. We can think of it as the phase margin. The lower the value, you're gonna see a little bit more overshoot, a little more ringing, which just looks like a lower-phase margin for power supply. In most cases, you'll find the value of 90 to be ideal. I wouldn't recommend changing it. But you can tune in the power supply if you want to.

So let's take a look at how the values affect the transient response. The default value of 90 is set up right now with an ASCR gain of 400. If I change the residual value down to 60 and send it across the PMBus command, two things occur. The first is you get a little bit more ringing, a little bit more overshoot on the output, which looks like a lower-phase margin. The other piece you see is the excursion. The initial drop on the transient response has actually come up. You've actually improved the deviation, and that's because you gain a slightly faster response to the initial transient event, which corresponds with a lower-phase margin system. I can keep lowering it further but at some point, you're going to see diminishing returns. So if I drop it down to 30 and send that command across, effectively, there's very little change in the system.

So in most cases, the residual has a slight effect. You can make a little bit of an improvement if you want to dial in the very last part of performance. But the system will always be stable with the default values of 90, and that's probably the great place to start with. The benefit of the ZL8800 is there's no compensation. There's no complex math. It just works.