0:00:00.000,0:00:19.290
36C3 preroll music
0:00:19.810,0:00:26.710
Herald: Okay, that was fast two minutes.[br]The next talk is going to be on DC to DC
0:00:26.710,0:00:39.640
converters by Zoé Bőle. She's an[br]enthusiast for open hardware and fan of
0:00:39.640,0:00:50.010
DIY and has been working on the topic of[br]DC to DC converters for a long time and I
0:00:50.010,0:00:58.900
have to keep on talking now because it[br]seems that her computer is not really
0:00:58.900,0:01:07.420
communicating with the presentation[br]device. We do have a picture but we don't
0:01:07.420,0:01:29.490
get it moving.[br]troubleshooting whispers in background
0:01:29.490,0:01:32.759
Herald: While they're still having some
0:01:32.759,0:01:38.950
issues up here I might remind you that it[br]is very helpful if you take your trash
0:01:38.950,0:01:44.969
with you and now please welcome Zoë and we[br]are ready to inaudible.
0:01:44.969,0:01:47.589
Give her a warm hand!
0:01:54.503,0:01:59.740
Zoé Bőle: Hi, my name is Zoë and this is
0:01:59.740,0:02:05.520
the first time I'm standing here in a[br]chaos stage so I'm a little bit, like,
0:02:05.520,0:02:13.030
anxious but I'm here to talk to you…[br]applause
0:02:13.030,0:02:23.110
Zoé: I'm here to talk about these DC[br]converters and the talk's called "DC/DC
0:02:23.110,0:02:28.100
converters and everything you wanted to[br]know about them" but it's unlikely I can
0:02:28.100,0:02:36.260
fit everything into a 50 minute talk, so[br]it's like not everything, but my goal is
0:02:36.260,0:02:42.540
to provide you some starting points and[br]give you an overview and hopefully if you
0:02:42.540,0:02:50.150
already worked with DC/DC's then you're[br]also not gonna be annoyed and not gonna be
0:02:50.150,0:02:57.840
bored. Before I start with the DC/DC[br]topic, I would like ask you to be
0:02:57.840,0:03:04.159
excellent to each other and this is not[br]related to my talk but I hear people
0:03:04.159,0:03:08.939
starting clapping when someone broke the[br]bottle accidentally and I think it's super
0:03:08.939,0:03:18.829
not cool. Yesterday I saw someone breaking[br]down in tears because they just broke a
0:03:18.829,0:03:23.700
bottle and everybody was clapping and[br]paying attention to them
0:03:23.700,0:03:31.959
and that was like harassment. So, please[br]don't be the one who starts clapping. But
0:03:31.959,0:03:42.019
also I'm not here to forbid you to clap[br]and… just know what's happening. So,
0:03:42.019,0:03:49.010
brief introduction to DC/DC's and why:[br]quite often you need different voltages
0:03:49.010,0:03:56.459
than what you have available. For example[br]you have a microcontroller or you have an
0:03:56.459,0:04:02.260
FPGA and you work with the battery[br]then you need to provide a different
0:04:02.260,0:04:14.950
voltage for that circuit and the trivial[br]solution is to just use two resistors and
0:04:14.950,0:04:21.160
make a voltage divider, but this is[br]totally unsuited for power delivery
0:04:21.160,0:04:27.600
because as you start loading the output,[br]the output voltage starts dropping. Also,
0:04:27.600,0:04:36.930
this circuit dissipates power even if[br]there is no useful load on the output. So
0:04:36.930,0:04:44.400
this is only useful for signals and to[br]have some kind of feedback and regulate
0:04:44.400,0:04:52.389
the output to a desired level you can use[br]an LDO, which is the same thing but you
0:04:52.389,0:05:01.470
control, one resistor – very simplified –[br]to always keep a desired output voltage.
0:05:01.470,0:05:08.780
Of course this can only go lower than your[br]input and your efficiency is limited to
0:05:08.780,0:05:17.383
the ratio of the output voltage and the[br]input voltage, and this is even in ideal
0:05:17.383,0:05:26.400
situation. So, instead of burning up power[br]in your converter you can just use
0:05:26.400,0:05:31.740
switches and this is the idea behind[br]switching supplies that you use a switch
0:05:31.740,0:05:36.200
element which is either [br]fully on or fully off.
0:05:36.200,0:05:41.340
And if it's fully on then there[br]is no loss on a switch and if it's
0:05:41.340,0:05:46.300
fully off there is no current flowing[br]through it so there is no loss either.
0:05:46.300,0:05:51.560
There are some practical problems[br]with this approach but
0:05:51.560,0:06:01.639
but this works for LEDs and heaters if your[br]switching frequency is high enough.
0:06:01.639,0:06:08.930
To think of a DC/DC converter is a[br]box with four terminals. It has an input
0:06:08.930,0:06:15.750
side and an output side. Right now I'm[br]talking about buck step-down DC/DC
0:06:15.750,0:06:21.711
converters which are non isolated. This[br]means the ground in the input side and the
0:06:21.711,0:06:31.719
ground on the output side are connected[br]together inside and this limits certain uses.
0:06:31.719,0:06:37.940
Also you should not connect these[br]DC/DC converters in series, so if you have
0:06:37.940,0:06:44.070
a block like this and you think "oh, I[br]could use two or three of them and just
0:06:44.070,0:06:50.080
connect them in series to give it higher[br]input voltages," that's gonna blow up very
0:06:50.080,0:07:03.340
quickly. A block looks like this on a[br]screen and might look like this in reality.
0:07:03.340,0:07:13.296
Let's take a look inside. So all of these[br]DC/DC converters consist of a power stage,
0:07:13.296,0:07:23.620
a control system, and the feedback.[br]The feedback is there to provide a
0:07:23.620,0:07:31.169
regulated output regardless of the[br]operating conditions. So what's inside a
0:07:31.169,0:07:38.460
power stage? To have a deeper look inside[br]we can consider this asynchronous buck
0:07:38.460,0:07:46.099
converter where a switching element – a[br]MOSFET – is controlled by an analog and
0:07:46.099,0:07:53.490
digital circuitry. Feedback is provided[br]from the output voltage and we see a diode
0:07:53.490,0:07:59.490
in the middle which I'm going to talk[br]about soon. You also see two capacitors on
0:07:59.490,0:08:04.430
the input side and on the output side,[br]which are also very important.
0:08:04.430,0:08:07.740
More about them later.
0:08:07.740,0:08:14.790
Let's consider the first situation:[br]the switch is on – this is
0:08:14.790,0:08:26.289
so-called "the on state" – and this forms[br]a loop from the input to the output. The
0:08:26.289,0:08:36.780
input capacitor we can neglect and in an[br]ideal situation the output capacitor is,
0:08:36.780,0:08:46.410
um, I will talk more about more about the[br]output later.
0:08:46.410,0:09:07.910
pause
0:09:07.910,0:09:14.443
All right, I don't wanna make this into a[br]lecture and everybody is sleeping in
0:09:14.443,0:09:19.530
and the fun part will start very soon.
0:09:19.530,0:09:26.580
So, this DC/DC has two states: either the [br]switch is on or switch is off. Right now
0:09:26.580,0:09:30.880
the switch is on and you see that the [br]current can flow from the input through
0:09:30.880,0:09:41.430
this inductor to the output. The inductor[br]resists the change of current. It's like
0:09:41.430,0:09:49.339
pushing a heavy mass and once it starts[br]moving it wants to keep moving. That's why
0:09:49.339,0:09:57.610
in this "on" state the input current flows[br]through the inductor and starts to
0:09:57.610,0:10:06.370
increase while it's also flowing to the[br]output. Then the converter turns the
0:10:06.370,0:10:14.860
switch off, which comes to the off state,[br]and now the diode comes into play, which
0:10:14.860,0:10:22.940
will keep the current recirculating. In[br]this "off" state there is no current from
0:10:22.940,0:10:29.390
the power from the source to the output,[br]but the output is still powered from this
0:10:29.390,0:10:40.720
decaying magnetic field through the[br]inductor. Sometimes you hear about
0:10:40.720,0:10:48.779
synchronous DC/DC converters where this[br]diode is replaced by another switch.
0:10:48.779,0:10:58.000
In that case efficiency's increased since[br]the voltage drop across MOSFET is lower
0:10:58.000,0:11:04.506
than the forward voltage of the diode. In [br]this case, as you can see, current is still
0:11:04.506,0:11:10.184
being delivered to the output. And this is[br]the big advantage of the buck converter
0:11:10.184,0:11:16.760
that in both an "on" and an "off" state [br]the output is sourced with current.
0:11:16.760,0:11:25.149
What the output capacitor[br]does there is it provides the difference
0:11:25.149,0:11:33.070
between the inductor current. On the lower[br]end you can see the inductor current.
0:11:33.070,0:11:39.320
As the switch is on it ramps up and as switch[br]is off it ramps down, and in the middle
0:11:39.320,0:11:48.930
you see this line which is the output[br]current. So you see these triangles and
0:11:48.930,0:11:57.040
this is what's provided by the output[br]capacitor. Alright, so this is an actual
0:11:57.040,0:12:01.621
part without the simplifications and I[br]would like to talk a bit about the
0:12:01.621,0:12:08.010
reference voltage and how that works. So[br]this device creates an internal 0.7V
0:12:08.010,0:12:18.149
reference and you can program the output[br]voltage by choosing R1 and R2 on the
0:12:18.149,0:12:27.360
left side so at your desired output[br]exactly 0.7V will be at this voltage
0:12:27.360,0:12:36.726
divider and this converter will keep[br]regulating to reach the state.
0:12:45.190,0:12:53.351
If you're looking for DC/DC converter to[br]your next project then you might see
0:12:53.351,0:12:58.152
a bunch of parameters and [br]I'm gonna talk about those.
0:12:58.152,0:13:04.243
So first you see a 3.3 volt 2 amp [br]converter. What does it mean?
0:13:07.065,0:13:12.389
This depends on how and [br]who specifies that output
0:13:12.389,0:13:19.110
because someone says it's two amps if it[br]can provide two amps for a second and
0:13:19.110,0:13:25.220
someone says it's two amps if it can[br]continuously provide the two amps even in
0:13:25.220,0:13:32.670
a warm environment, so it's important to[br]talk about if it's a peak or continuous
0:13:32.670,0:13:40.399
current rating. Then there is this so[br]called "output ripple." You saw that
0:13:40.399,0:13:48.760
switching action going on and off and that[br]will create a ripple on the output voltage
0:13:48.760,0:13:58.009
so it won't be 3.3V it will be oscillating[br]around that. This can be as low as a few
0:13:58.009,0:14:07.740
microvolts and as high as a few volts,[br]depending on the parameters. Also there is
0:14:07.740,0:14:18.040
a voltage accuracy: maybe it's labeled [br]as 3.3V but actually it's 3.5 or 3.0.
0:14:18.040,0:14:30.420
Load regulation: it's maybe 3.3V when it's[br]unloaded and as you increase the output it
0:14:30.420,0:14:37.889
starts changing the output voltage. There[br]is the line regulation which means the
0:14:37.889,0:14:46.570
input voltage has influence over the[br]output, which is undesired. Then there is
0:14:46.570,0:14:52.800
this maximum input voltage rating. Let's[br]say this converter can tolerate seven
0:14:52.800,0:15:00.260
volts on its input so you think "oh let's[br]just hook it up to USB, that's 5V, right?"
0:15:00.260,0:15:12.399
Yes, but no, because when you use cables[br]and non-ideal conditions, you can create
0:15:12.399,0:15:24.190
transients which overshoot the voltage [br]possibly way above this maximum rating and
0:15:24.190,0:15:33.180
this can lead to very nasty surprises.[br]Because sometimes they fail short, which
0:15:33.180,0:15:38.330
means they connect their input[br]directly to their output.
0:15:38.330,0:15:45.360
In this case the device you connected to[br]the converter might also go up in flames.
0:15:45.360,0:15:50.886
So mind the transients and always[br]have some margin between
0:15:50.886,0:15:57.824
your desired input voltage and the[br]maximum the converter can tolerate.
0:15:57.824,0:16:06.800
Then you might see 95% efficiency[br]and that's also question at
0:16:06.800,0:16:18.380
which load because at maximum specified[br]load it will be lower, and at lower/less load it
0:16:18.380,0:16:23.810
will be also lower, so there is this[br]efficiency peak. That marketing people love
0:16:23.810,0:16:29.170
to specify. There's also this so called[br]"quiescent current" which means your
0:16:29.170,0:16:38.720
converter draws current from your input[br]even when there is nothing on its output
0:16:38.720,0:16:45.831
and if it runs from a battery this can[br]drain your battery in days or weeks, so
0:16:45.831,0:16:53.910
you must pay attention to this. And there is[br]this other factor called "switching
0:16:53.910,0:17:00.600
frequency" so how fast, how often the[br]internal switch changes state, but this
0:17:00.600,0:17:06.290
might not be a constant value, especially[br]with the previously mentioned quiescent
0:17:06.290,0:17:16.960
current feature, the converters that excel[br]at having a low quiescent current don't
0:17:16.960,0:17:23.640
have fixed switching frequency, so you[br]might have noise at different frequency
0:17:23.640,0:17:35.040
bands and disturb your circuits or radio[br]noise. Let's talk about a few features,
0:17:35.040,0:17:43.130
you might want to look for. "Enable": enable[br]functionality. This is very useful to
0:17:43.130,0:17:50.600
easily disable your DC/DC converter and [br]without having to interrupt either the
0:17:50.600,0:18:01.900
input side or the output side. Let's say[br]you have a 20 amp output converter – you
0:18:01.900,0:18:09.080
really don't want to switch the 20 amp[br]with a mechanical big switch. Instead
0:18:09.080,0:18:15.650
of that you have a logic input to[br]your DC/DC converter with which you can
0:18:15.650,0:18:23.440
turn this completely off. Then there is so[br]called "undervoltage lockout": you might
0:18:23.440,0:18:31.750
want to prevent it from running below a[br]certain input voltage to prevent draining
0:18:31.750,0:18:40.180
your battery too deep and turning it[br]completely off. There's "power good" that
0:18:40.180,0:18:45.800
can provide information to your processor[br]that the output voltage is in regulation
0:18:45.800,0:18:51.470
and stabilized. So if you hook up the[br]"power good" output to let's say a reset
0:18:51.470,0:18:58.680
line or "enable", then you can be sure[br]that the output voltage is always stable
0:18:58.680,0:19:07.270
and your processors are not going to go[br]into glitch. Overtemperature shutdown is
0:19:07.270,0:19:16.050
very common these days and that makes[br]these tiny converters almost
0:19:16.050,0:19:22.460
indestructible because if they get too hot[br]they just turn off completely before they
0:19:22.460,0:19:30.770
get permanently damaged. Efficient[br]standby: this is the so called low
0:19:30.770,0:19:37.760
quiescent current option. That means if[br]your output is off, your processor is
0:19:37.760,0:19:45.750
sleeping, then it willl reduce switching[br]action to reduce switching losses and
0:19:45.750,0:19:51.900
might only draw a few micro amps or even[br]nano amps. Very important for battery
0:19:51.900,0:19:59.130
powered applications. Then you might see[br]overcurrent protection which makes the
0:19:59.130,0:20:04.800
output very robust. You can even make a[br]short circuit and the overcurrent protection
0:20:04.800,0:20:17.420
will limit the output current to this[br]value and this prevents damaging of the
0:20:17.420,0:20:24.440
converter and also damage to the cables[br]and switches if they are rated to
0:20:24.440,0:20:33.000
withstand the overcurrent protection[br]limit. Now let's talk about noise. The
0:20:33.000,0:20:43.740
output ripple is not exactly noise. Output[br]ripple is there because the output
0:20:43.740,0:20:55.160
capacitor is non-ideal and usually this[br]this is very low on a properly designed
0:20:55.160,0:21:02.440
converter but if you measure the output[br]you might see spikes on the output and
0:21:02.440,0:21:14.870
that's not ripple. That's conducted EMI[br]because on that inductor the windings are
0:21:14.870,0:21:21.780
coupled very closely, there is some[br]capacitive coupling between the wires, so
0:21:21.780,0:21:28.800
the digital on-off action from the[br]switches will propagate to some extent to
0:21:28.800,0:21:35.250
the output. This is attenuated by the[br]capacitors but they cannot be completely
0:21:35.250,0:21:42.030
filtered off and you will see the[br]switching frequency and even upper
0:21:42.030,0:21:50.790
harmonics of it but this can also be[br]filtered. There is also radiated EMI,
0:21:50.790,0:21:56.980
which comes mostly from the switching node[br]and capacitive coupling to the ground
0:21:56.980,0:22:05.920
plane, and also the inductor – if it's not[br]shielded then a magnetic field can also
0:22:05.920,0:22:15.960
radiate out and cause interference. On[br]this picture what you see is that gray
0:22:15.960,0:22:27.280
block, that's a shielded inductor, and the[br]two blue connectors at the end of this PCB
0:22:27.280,0:22:34.750
are screw terminals. I personally advise[br]against using this style of screw
0:22:34.750,0:22:41.700
terminals because the wires can easily[br]slip out, make a short, or you don't
0:22:41.700,0:22:52.900
notice that they are not connected, so I[br]prefer a different style of connectors.
0:22:52.900,0:23:00.350
It's good to know about non-ideal[br]components. The capacitors that are used
0:23:00.350,0:23:10.120
have a so called DC bias. These multi-[br]layer ceramic capacitors are very
0:23:10.120,0:23:17.720
sensitive to the DC voltage across the[br]terminals and if they are rated, let's say
0:23:17.720,0:23:24.830
20 microfarads, at the rated voltage they[br]might lose up to 90 percent of their
0:23:24.830,0:23:30.661
capacity. So you always have to pick a[br]capacitor that's rated to a higher voltage
0:23:30.661,0:23:38.330
than what your output is to compensate for[br]this effect, and you also need to put more
0:23:38.330,0:23:47.770
capacitors at your output than what you[br]would think in an ideal situation. Mind
0:23:47.770,0:23:55.001
the transients! As I said, if you plan to[br]hotplug, connect to live wires to your
0:23:55.001,0:24:01.290
converter, you have to keep in mind the[br]inrush current. Those capacitors, when
0:24:01.290,0:24:11.750
they are fully discharged and you connect[br]that to the input, then they will try to
0:24:11.750,0:24:21.490
charge to the input voltage as fast[br]as the cabling lets that happen, and the
0:24:21.490,0:24:28.980
cables have inductance which will store[br]energy and overshoot the input voltage.
0:24:28.980,0:24:39.350
When fiddling with MOSFETs, don't forget[br]the ESD protection. MOSFETs are very
0:24:39.350,0:24:49.790
sensitive at their gate, because the oxide[br]layer is so thin that even 20V voltage is
0:24:49.790,0:24:57.450
enough to break it down, and a 20V ESD[br]strike is something you probably don't
0:24:57.450,0:25:08.920
even notice, but it can damage the[br]MOSFETs. And ever avoid the 7800 series
0:25:08.920,0:25:17.481
LDO, because it's a very old part and I[br]still see it in new designs, while there
0:25:17.481,0:25:25.930
are much better ones with better[br]regulation, less quiescent current, and
0:25:25.930,0:25:39.990
it's also an LDO, so it's like just[br]marginally related to DC/DCs. If
0:25:39.990,0:25:46.300
you make your own DC/DC converters instead[br]of buying one, you should read the
0:25:46.300,0:25:57.670
datasheet and follow the instructions[br]because the manufacturers give you a
0:25:57.670,0:26:05.340
proven tested layout, which is typically[br]good advice to follow and you should only
0:26:05.340,0:26:25.900
deviate from that if you know what you're[br]doing. I'm sorry
0:26:25.900,0:26:48.300
pause[br]Alright, that mostly concludes what I was trying
0:26:48.300,0:26:56.810
to talk about and now it's time for your[br]questions.
0:26:56.810,0:27:02.073
applause
0:27:02.073,0:27:09.030
Herald: Now there are two microphones one[br]there and one over there, usually they
0:27:09.030,0:27:18.720
are… ah, here comes the light. Are there[br]any questions? How about the signal angel,
0:27:18.720,0:27:25.190
does the internet have any questions? The[br]internet doesn't have a question but
0:27:25.190,0:27:29.110
here's one up front.[br]Q: What would you recommend instead of
0:27:29.110,0:27:35.650
screw terminals?[br]Zoé: That's a very good question and that
0:27:35.650,0:27:42.290
really depends on the application. You can[br]have different kind of screw terminals,
0:27:42.290,0:27:56.370
which use either crimped therminals on the[br]cable so you have a cable shoe.
0:27:56.370,0:28:01.640
Q: Like a ring or something?[br]Zoé: Yes. Because then there is no way that it can
0:28:01.640,0:28:09.610
slip out. For less current you can use[br]dupont connectors, they can take like
0:28:09.610,0:28:22.880
2 to 3A per contact. You know the standard[br]pin header and that kind of thing. And
0:28:22.880,0:28:29.800
there are also latching connectors from[br]molex and other manufacturers. The problem
0:28:29.800,0:28:39.310
is with that you need crimping tools and[br]those can be very expensive. So it first
0:28:39.310,0:28:48.540
makes sense to get those when you have a[br]hackerspace or you can share it with other
0:28:48.540,0:28:52.820
people.[br]Herald: The next question, please?
0:28:52.820,0:28:58.990
Q: Thank you for your talk. On your last[br]slide last point you mentioned stability
0:28:58.990,0:29:07.170
analysis. What is your experience with[br]running such converters in parallel for
0:29:07.170,0:29:13.650
redundancy and how would you do the[br]analysis there?
0:29:13.650,0:29:22.920
Zoé: Running current mode converters[br]parallel is typically okay, but they won't
0:29:22.920,0:29:30.250
do current sharing automatically. So this[br]one converter has a certain output voltage
0:29:30.250,0:29:37.720
set and the other one has little bit[br]different voltage, and that will create a
0:29:37.720,0:29:44.090
difference in their output currents, and[br]there are topologies and there are
0:29:44.090,0:29:50.620
converters which are prepared for parallel[br]operation and they can provide current
0:29:50.620,0:29:56.300
information to all of the parallel[br]converters, and they can ultimately
0:29:56.300,0:30:07.410
synchronize. For stability, that should[br]not influence the stability of it. What I
0:30:07.410,0:30:14.760
should have mentioned is stability[br]analysis because we have a control loop.
0:30:14.760,0:30:22.780
The control loop takes the output and[br]creates a control signal that influences
0:30:22.780,0:30:36.600
the output, but this loop has a delay, and[br]because of this delay, basically you can
0:30:36.600,0:30:47.230
make an oscillator of this, and to avoid[br]that, you can use a network analyzer and
0:30:47.230,0:30:56.860
inject a signal into the converter.[br]Q: Thank you.
0:30:56.860,0:31:09.170
Herald: Yeah, you go ahead, over there.[br]Q: Hi, what would you say the choice is
0:31:09.170,0:31:13.520
between a dis-synchronous mode or[br]a forced-synchronous mode?
0:31:13.520,0:31:33.010
Zoé: That's a very good question. All[br]right so when I talked about this briefly
0:31:33.010,0:31:39.970
and mentioned the synchronous converters,[br]with forced synchronous converters you
0:31:39.970,0:31:49.290
have a control switch and those have[br]typically fixed switch frequency. If the
0:31:49.290,0:31:55.030
output current is zero, then during half[br]of the period current will flow backward
0:31:55.030,0:32:01.750
from the output capacitor to the input[br]side and then the next half period that
0:32:01.750,0:32:06.430
current will flow back from the input to[br]the output, so basically energy swings
0:32:06.430,0:32:15.660
between input and output and this causes[br]efficiency loss, but this also avoids
0:32:15.660,0:32:30.770
operation in discontinuous mode, which[br]reduces ripple and reduces EMI. So it
0:32:30.770,0:32:34.850
depends on your application.[br]Q: Thanks.
0:32:34.850,0:32:38.700
Zoé: You're welcome.[br]Herald: The next question?
0:32:38.700,0:32:48.390
Q: Hi, Zoé, thank you for the talk! I have[br]a question about: you mentioned linear
0:32:48.390,0:32:54.220
regulators at the end, what are they used[br]for in this context?
0:32:54.220,0:33:02.010
Zoé: you mean 7800 series?[br]Q: Yes.
0:33:02.010,0:33:13.910
Q: Not, the one before, I think.[br]Zoé: Those were very good regulators in
0:33:13.910,0:33:22.130
the 70s and those are linear regulators,[br]and the problem with the 7800 series is
0:33:22.130,0:33:28.540
everybody knows about them because books[br]are full of them but they have quite a few
0:33:28.540,0:33:41.270
milli amps of quiescent current. They also[br]have bad regulation against load and line
0:33:41.270,0:33:48.140
transients and they are not cheaper than[br]much of their alternatives, so there's
0:33:48.140,0:33:56.020
really no reason to use those. You can use[br]for example a DC/DC pre-regulator and then
0:33:56.020,0:33:59.590
an LDO afterward to smooth out the[br]voltage.
0:33:59.590,0:34:07.204
Q: Okay, thank you.[br]Herald: Go ahead
0:34:07.204,0:34:13.490
Q: Thank you very much! My question is,[br]you mentioned the noise coupled via the
0:34:13.490,0:34:18.440
inductor to the output. Which sort of[br]filter do you recommend: differential
0:34:18.440,0:34:24.060
noise or common mode noise, and input or[br]output? Which is most important from your
0:34:24.060,0:34:32.040
perspective?[br]Zoé: So lots of the noise goes actually
0:34:32.040,0:34:40.460
back to the input supply and I said that[br]in an ideal circuit the input capacitor is
0:34:40.460,0:34:46.970
not necessary, but in a real circuit the[br]input capacitor is critical because the
0:34:46.970,0:35:03.520
input inductance is seen by by the switch.[br]If you let me show you. On this chart you
0:35:03.520,0:35:11.660
see the inductor current and the input[br]current, it follows the inductor current
0:35:11.660,0:35:17.720
only during the on phase which means after[br]the end of the on phase and beginning of
0:35:17.720,0:35:25.520
the off phase it falls from maximum value[br]to zero and later on at the end of the off
0:35:25.520,0:35:32.460
phase and the beginning of the on phase[br]the current jumps from zero to the output
0:35:32.460,0:35:42.380
current, and these jumps in the supply[br]current create an awful lot of EMI if the
0:35:42.380,0:35:51.610
input capacitor is not large enough so[br]this is a very critical thing. I saw quite
0:35:51.610,0:35:56.810
a few converters where the input capacitor[br]is under dimensioned and when you run it
0:35:56.810,0:36:03.370
over longer wires with more parasitic[br]inductance, that can create a lot of EMI.
0:36:03.370,0:36:12.090
For ways of reducing the the noise on the[br]output, the best way is to have proper
0:36:12.090,0:36:20.710
filtering capacitors. If you use ceramic[br]capacitors and enough high enough value
0:36:20.710,0:36:33.670
you can get rid of almost all of the[br]noise. I made a design which had microvolt
0:36:33.670,0:36:43.610
noise because I found a capacitor with its[br]resonance frequency exactly at the
0:36:43.610,0:36:48.610
switching frequency, so basically all that[br]noise that was coming from switching
0:36:48.610,0:36:55.260
action was reflected away and higher[br]frequency ranges where it got filtered
0:36:55.260,0:37:05.270
dissipated much faster. You can[br]use PI filters at the output but mind
0:37:05.270,0:37:16.890
that you worsen the transient behavior of[br]your converters. So if your load suddenly
0:37:16.890,0:37:23.060
needs a lot more power and starts drawing[br]more current, then your converter will
0:37:23.060,0:37:35.940
react slower because of the[br]filter you just added. PI filters or RC
0:37:35.940,0:37:38.890
filters if you don't need that much[br]current.
0:37:38.890,0:37:44.620
Q: Okay, thanks.[br]Herald: Okay great I don't see any more
0:37:44.620,0:37:52.540
questions, so everything seems to be fully[br]explained. Thank you and give her a
0:37:52.540,0:37:55.220
applause and good night.
0:37:55.220,0:37:56.670
applause
0:37:56.670,0:38:02.960
Zoé: Thank you
0:38:02.960,0:38:06.840
postroll music
0:38:06.840,0:38:30.000
Subtitles created by c3subtitles.de[br]in the year 2020. Join, and help us!