0:02 The second group of questions is
0:05 traditionally class D. Is it capable or
0:08 not? And here I picked one picture that
0:10 in my opinion illustrates these
0:14 situations very well. When something new
0:17 appears, well, someone invents something
0:20 and the crowd shouts which a woman can't
0:23 be that beautiful. Such a small woman
0:26 can't run that fast. A woman with two
0:30 children must be fat. hanger burner.
0:33 Jordan Bruno came out with the thesis
0:36 that the universe is infinite. He was
0:38 burned at the stake. I would have given
0:42 up, but he was a tougher guy.
0:44 Why does the market in general have such
0:48 a negative attitude toward class D?
0:50 Because essentially it destroys the
0:52 entire established business and the
0:55 first stage is denial. I just want to
0:58 show you all the drawbacks of class D. I
1:00 had a presentation like that back in
1:03 2008 when we were working on our first
1:06 developments. And let's quickly go
1:09 through all the problems. This picture
1:11 shows almost all the problems of a class
1:14 D amplifier. Class D technology is not
1:16 new, but the industry wasn't ready
1:18 because it's actually difficult to create.
1:20 create.
1:22 There were problems with the input
1:24 circuits with the modulator and then
1:25 there was a serious problem with
1:28 operating in the on mode because the
1:30 power supply couldn't handle it and
1:33 developed specific issues. In this case,
1:35 the MOSFETs didn't switch fast enough.
1:38 So, you had to make a very large reserve
1:39 and then there was a lot of noise
1:41 because the MOSFETs weren't ready for
1:44 this. There was nonlinearity in the
1:46 output filters. There should be
1:48 highquality filters there. There was
1:51 hystericis and we needed to solve it. So
1:53 how should we approach this? We can be
1:56 in this crowd and scream class D can't
1:58 do it because it has sand on top and
2:01 other issues. Or we can see the whole
2:04 perspective of this technology and try
2:06 to solve it. If we look at a story like
2:09 Kodak for example, they had the film
2:12 market. I think they had 90% of the
2:15 market. They had a lot of money. And
2:17 when digital photography appeared, they
2:20 couldn't invest in it and keep working.
2:22 They said that film is the best. And
2:25 where is Kodak now? It's gone. Their
2:27 patents were bought up. LG bought their
2:31 fourpixel patent. And LG's organic light
2:34 emmitting diode TVs are the best. They
2:36 have that patent. Nokia's button phones
2:39 were everywhere. Nokia was a very good
2:41 manufacturer, but they didn't keep up
2:44 with smartphones. What happened? Nokia
2:47 is gone. It is a niche player now. So
2:50 with class D, it's the same thing. We
2:52 can pick a side or watch and decide or
2:55 say that everything is bad. And indeed,
2:58 the tasks that this so-called dead time
3:00 is supposed to solve. It can't be that
3:03 both MOSFETs are open at the same time.
3:05 Yes, such a reserve was needed. The
3:08 bigger the reserve, the more distortion.
3:10 Today it was decided to use base pumping
3:12 when we have what's called a half bridge
3:14 in English. I don't even know how to
3:17 translate that. When we don't have a
3:18 bridge circuit and when we do have a
3:20 bridge circuit, let's say when we don't
3:22 have a bridge circuit, this problem
3:24 exists and [snorts] it can be dealt
3:27 with. There's a specific task set for
3:29 this because these switches work as
3:32 voltage multipliers. But if you have a
3:34 fullbridgeidge circuit, everything is
3:37 solved without any problems. And chips
3:41 like TPA 3,255
3:43 have four channels and the best
3:45 connection is two channels in a bridge
3:48 circuit to get stereo and two channels
3:50 even though there are four channels
3:52 inside the chip. There was noise with
3:55 transistors like this. So they made
3:57 specialized transistors and solved it.
4:00 And finally, the reason why so many
4:02 companies that produced class A and B
4:05 amplifiers are no longer working or
4:08 can't work in the field of class D is
4:10 because it requires completely different
4:13 knowledge. We have very powerful pulses.
4:15 There are analog circuits. Everything is
4:17 on one board. You need different
4:19 knowledge. Either you hire these
4:22 specialists or you don't have them. And
4:24 it turns out that other players like Ice
4:27 Power, Hypex, companies that work in the
4:29 professional market, these are more
4:32 popular ones are the ones who lead.
4:35 Regarding transistors, a new standard
4:38 has emerged for leadless transistors.
4:41 Low noise, good cooling, new casings,
4:43 and so on and so on. International
4:45 rectifier specialized on it. Then
4:48 infinion bought them. Here's the
4:50 difference between a simple transistor
4:52 and a specialized one in how each
4:54 behaves as a pulse. And we're talking
4:55 about nightmare currents. About
4:58 nightmare currents. Specialized
5:00 transistors appeared where they are
5:02 paired in one package to have very
5:05 precise outputs. Filters have
5:07 hystericis. They're bad there. And there
5:10 was also a problem that filters could
5:13 only be set to one frequency. Today we
5:16 have feedback after the filters. It's a
5:19 complicated thing, but it's been solved.
5:21 We have a clear amplitude characteristic
5:24 for any load. There was a problem. We
5:26 solved it. There were different fite
5:29 issues, and you can see how it changes.
5:32 Distortion at the output only changes by
5:35 selecting the inductor components. But
5:38 this was still 2008, and solutions had
5:40 already been found. There are different
5:42 developers here, but the Japanese were
5:44 the first to solve the inductor problem.
5:47 It's another level. And here I've picked
5:49 out an inductor for you that we were
5:52 also developing back then.
5:54 It's a good manufacturer. It was one of
5:57 the first to offer different solutions.
5:59 This is Ferox core. And as you can see,
6:01 they cut the magnet, preventing it from
6:04 saturating, and solved the hysterosis problem.
6:07 problem.
6:09 Here, we even wound it by hand and
6:11 tested it, but there are none left in stock.
6:13 stock.
6:15 You see there's a cut here. It's filled
6:18 in, but the ferry is cut and it
6:20 immediately loses a huge amount of thermobility.
6:22 thermobility.
6:24 I don't know if that's the right term,
6:27 but one needs more of these windings.
6:30 But they found some kind of option, but
6:33 it was considered cool. We used it in
6:35 our own development.
6:37 So when your mind tunes into finding
6:40 solutions, everything becomes solvable.
6:42 Well, here are different types of these
6:46 inductors. Some are like this. These are
6:49 also wellused, good ones. We see these
6:51 kinds. We see fewer of these. They are
6:54 present in various developments.
6:57 A large air core one basically doesn't
6:59 make sense since these ferrets have
7:01 already solved the whole problem. We get
7:04 distortion through the filters. Three
7:06 zeros after the decimal point.
7:09 Everything is solved.
7:11 Everything is solved. The same applies
7:14 to filter components like capacitors.
7:17 But that was much easier to solve. And
7:19 overall, here's what we get.
7:22 Polyropylene capacitors are the best.
7:25 Polyester ones are also fine. Basically,
7:27 that's not a problem, but you can't use
7:30 ceramic ones. If you see ceramic ones,
7:32 throw it out. Throw away that board.
7:35 I've seen a few. Well, come on. I don't
7:37 really know the theory, but they have
7:39 different problems. I won't get into
7:42 electronics. But the cost of these
7:44 capacitors is relatively high if you use
7:48 ceramic ones. You might save a euro or
7:50 even less, maybe half a euro, but you
7:54 absolutely can't cut costs here. That's
7:57 the worst thing. And regarding gate
8:00 drivers, in 2008 there were about five
8:02 of them. Let's see what's there today.
8:05 Here we have Infinaon.
8:08 It is a German company. They make chips
8:11 for the automotive industry and they're
8:14 very strategic and good developers. As
8:16 you know in cars, microchips have to
8:20 work from minus50 to + 70°.
8:22 It's a very difficult task. They're a
8:25 good manufacturer and they also produce
8:28 these gate drivers. All the complexity
8:30 we've talked about is already inside the
8:32 chip. If you look at their brochure
8:34 where they talk about their developments,
8:35 developments,
8:37 everything is there. There's a small
8:39 three-phase strip, there's a solution
8:41 for that. There's a bigger one, there's
8:44 a solution for that, too. They combine.
8:46 They do everything. They give you all
8:48 the hints, all their knowledge, offering
8:50 you a whole range of gate drivers from
8:53 low power to high power. And they're
8:56 getting better and better. if there were
8:58 about seven back then. Now there are
9:02 already several pages more and more.
9:05 Voltage is 500 W. That's the range we're
9:06 interested in. We're not interested in
9:09 anything else. The same goes for field
9:12 effect transistors. You can choose any
9:16 for any power. Everything is available.
9:18 They've even grouped them here by the
9:20 loads they can handle, which is mainly
9:24 current. There are different options. So
9:26 the range we're interested in which is
9:29 up to 500 W it's going to get cheaper
9:32 and cheaper. Also for less experienced
9:34 developers all the key references design
9:37 guidelines and reference projects are
9:39 provided where everything is done
9:42 correctly and properly. Regarding a
9:44 circuit board it is a complex thing and
9:47 all the grounding component grouping and
9:52 so on must be done very properly.
9:54 Developers need to know many things on a
9:57 theoretical level, especially everything
9:59 about grounding high current circuits.
10:02 It's not an easy task. It's much more
10:04 complicated than just doing A and
10:06 B-class amplifier. But if you do
10:07 everything properly, then it will be
10:10 fine. These hints are given to
10:12 developers. If you look at the different
10:15 Chinese boards that are produced, they
10:18 take the Gerber files as they are, then
10:20 tweak them a little, remove some
10:23 unnecessary things and might add another
10:26 operational amplifier. But overall, they
10:28 use these reference designs and quickly
10:31 enter the market. Not every developer in
10:34 China is highly skilled. There are many
10:35 developers who can assemble and
10:37 replicate it. To build something like
10:40 this from scratch, you need a top tier
10:42 developer. They must be extremely
10:44 well-trained in theory. And once again,
10:46 I want to return to the key developer
10:48 behind highquality class D amplifiers
10:50 that are now outperforming high-end
10:53 systems. That's Bruno putsis. He worked
10:55 on it from the very beginnings.
10:57 Essentially, he solved many, many
10:59 problems that already existed in
11:01 integrated amplifiers. He's better known
11:03 in the amateur environment since this
11:06 company grew out of the amateur one. But
11:09 he also created the entire Hypex lineup
11:11 and now they have a separate engineering
11:14 company there working right at the
11:17 cutting edge focusing on the last 2% of
11:19 quality with a whole group of top
11:22 engineers. And I recommend keeping an
11:24 eye on them and this company Purify
11:27 reading their blogs. Yes, in English it
11:29 might be a bit difficult but not too
11:32 much. They're popular science so
11:34 everything is understandable. And the
11:37 second one who we know less about there
11:39 are two other developers Patrick
11:42 Bostonramm and Lars Press Peterson. Many
11:44 developers rely on them. That's Ice
11:47 Power, its key developers. And if you
11:50 want to keep up with the latest news and
11:53 developments there, these two guys are
11:56 also top class specialists. They have
11:58 whole teams behind them. It's even bigger.
12:00 bigger.
12:02 But when you read their articles, not
12:04 just in the professional world, these
12:07 are really, really strong guys. They can
12:10 create both chips and specialized
12:12 drivers of the future, specialized
12:15 drivers for powerful amplifiers. They
12:18 work together with Infinine. In general,
12:21 they do a lot of work on the amplifier.
12:23 This is a class D amplifier, and it's
12:25 gradually becoming smaller and smaller
12:28 for us. We're interested in a peak power
12:31 of 500 W because it can handle any
12:34 speaker, even a very demanding one. And
12:36 as soon as this is thoroughly worked
12:38 out, everything will be fine. We'll have
12:40 an excellent amplifier, an excellent
12:43 amplifier at a reasonable price. And for
12:46 those who shout which, yes, for those
12:49 who shout which and are looking for some
12:50 kind of reason to say that this
12:53 technology is bad, it will come anyway.
12:55 It is so far ahead of class B in all key
12:58 parameters, especially in terms of
13:00 thermal characteristics and so on, that
13:03 there are no alternatives. It's about
13:06 price, power, and quality. And someone
13:08 in the comments wrote that Sukov said
13:11 there's hysterosis, but Sukov is a
13:19 He was a great engineer who published in
13:21 the radio magazine. I also subscribed to
13:24 it and looked at these circuits. I was
13:27 impressed too. But the philosophy here
13:30 is that we are what we are today. What
13:33 was in the past? This isn't the Olympic
13:36 Games. It's the World Cup. And every
13:38 year we start over. In general, Sukov
13:41 should be here. He should be dragged to
13:43 the gallows. And he should shout that
13:45 yes, the class D witness zone has
13:48 hysterosis, but I know how to solve it
13:51 there. If you cut the ferit, you can
13:54 solve it. And then the next hypexex that
13:56 comes along will be, I don't know, a
13:59 Hypexex amplifier with a Sukov Hoffman
14:03 inductor. He is there. And if he's
14:05 shouting from the crowd like there's a
14:07 witch and she's a redhead and young
14:10 people just don't look back, then that's
14:12 it. Old grump. He should be able to use
14:14 his intellect to solve the challenges of
14:17 the future. And for those who are
14:19 seriously interested in this topic, I've
14:22 selected an article written by Bruno
14:26 Putsis. To me, he is a sensei. He has
14:28 truly done a great deal for class D
14:30 amplifiers and for the future. Here's a
14:33 long article. It answers a lot of audio
14:35 file questions in a very competent way
14:38 and so on.
14:40 Everything is there. It's a long
14:42 article. It's published and we'll
14:44 provide a link when we publish this
14:49 video. Here is the author Bruno puts. He
14:51 started as a student and look, he's
14:55 already a mature specialist
14:57 and he did a lot to create those really
15:00 cool class D amplifiers.
15:03 But the second one who is kind of in the
15:06 shadows is Ice Power. I only work with
15:08 Ice Power because they produce
15:11 industrial-grade units. They are very
15:14 well refined and you can buy them at a retailer.
15:15 retailer.
15:18 They've supplied them to several large
15:20 and commercial stores and enthusiasts
15:23 can buy them directly as well. Other
15:25 companies don't do that. So even with
15:27 Hipex, you can't buy the professional
15:29 line directly, but you can buy it from
15:33 Ice Power. I only work with Ice Power.
15:34 But that doesn't mean Hipex is bad.
15:37 Hypex is very good. They are the same.
15:40 They use the same technology, the same
15:42 principles, the same architecture.
15:46 So, well, as for me, my focus is 100% on
15:49 class D amplifiers as the amplifiers of
15:51 the future. I don't want to be like
15:53 Nokia saying that button phones are good
15:56 and then it's over. Well, that's in some
15:59 way my answer to the skeptics of class
16:02 D. I think the biggest risk for
16:04 amplifiers in the class A and B
16:11 amplifier market is the TPA 3255 chip.
16:13 The Chinese have mastered it and you can
16:17 really build a 250 W clean amplifier at
16:20 4 ohms with it. And at 8 ohms, that's
16:23 120 W. And that's the niche where class
16:28 A and B can compete with the TPA 3255.
16:30 But the designs based on this chip are
16:33 smaller in size. The Chinese have
16:36 mastered it and are now flooding the
16:38 market with their developments and
16:41 they're getting better and better. And
16:43 basically, there are no complaints about
16:45 these amplifiers where the power is
16:49 higher up to 500 W. There aren't any
16:52 good Chinese ones yet. Those tasks are
16:54 much more difficult, but up to 100
