High end Class D amps?

@douglas_schroeder Yes, all technologies follow a price/performance curve. Class D was first proposed in the 1950s but really didn't get going until the late 1990s. These early amps were at the bottom of the curve. The curve itself usually follows the same shape: sort of like a stylized 'S' (called a 'Sigmoid') where the cost (horizontal axis) goes up quickly while the performance does not. But after a while it starts to get figured out, then the performance goes up rapidly while the cost is minimal. Then as the technology matures the curve reverts to its original slope; incremental improvement for lots of money to develop it.

If I had to guess, it appears that we are nearing the top knee of the Sigmoid curve. 10 years ago we weren't there.


5 years ago we realized we needed to be taking class D seriously. We're known for making very transparent tube amplifiers- not the usual sort to be messing with class D. But in November a year ago we were awarded a patent in the field. A smaller player like us usually has troubles doing that in a field of mature technology. That is why I say the technology is nearing or just past the upper knee of the Sigmoid curve.

@mrdecibel We managed to get our hands on an early class D amp 20 years ago. It was hard to imagine what the fuss was about- it was awful! But a lot has changed in the last 20 years and amps have gotten a lot better. But there is still room for change- you can imagine how this is so when a tiny player like ourselves can step in and get a patent! We started work on our class D amp about 4 years ago and within about 6 months it was showing promise. But we didn't start that willy-nilly; by that time there were amps out there that really convinced me that class D was the rising star of amplifier technology, and if you didn't acknowledge that fact you were going to get left behind.

@mrdecibel What a class D amp potentially brings to the table is the ability to be neutral in a way that most solid state amps of traditional design cannot: they can be designed to lack brightness and harshness. For this reason alone they can work quite well on horns.


We are still in a period of the Wild West; there are a lot of different approaches to class D implementation and they sound different, so you have to be patient.

Then next you lot may as well just tell us to use this 6000watt!!!!! Class-D Behringer for just $600us.

Let me tell you, that on the Wilson Alexia in an a/b it was trounced by the 40 x times less powerful!!! Gryphon in dynamics and bass control. The Behringer was thin, anemic, weak in the bass and sounded like it was struggling

This post is using a Guilt by Association logical fallacy, which is that since the Behringer amp is class D and struggles, therefore all class D amps will struggle on the same load. Logical fallacies are false by definition.

 It's my understanding that efficiency has been the primary advantage and raison d'etre of class D amps since the get go. Their further efficiencies in weight, size and cost are just peripheral advantages. 

One advantage of class D amps that isn't mentioned so much is how they make distortion, and what sort of distortion they make. The distortion is mostly based on the linearity of the encoding system and deadtime used in the output section. A good deal of this distortion tends to be lower ordered harmonics. A second advantage is that if you use a self-oscillating circuit (which is an idea that's been around for about 20 years or so) then it is possible to put far more feedback in the design than is possible (for the most part) with traditional solid state designs. This is how the Purify module is capable of such low distortion. This also results in an output impedance that is so low its pretty difficult to measure. **That** simple fact means that the amp can act as a voltage source into speaker loads that are quite a bit less than one ohm, even if its power supply can't support doubling output power at full power into such a load. But if the amp is playing at a few watts (IOW: normal listening volumes) it will do so with ease.

Moto Guzzi's feel very similar to the BMW horizontal flat twin vibrations, both are no-where near as smooth as he V Twin Ducati.
 

For along time the 916 was my dream bike. Then a friend let me ride it and I never wanted to ride one again. It vibrated more than my Guzzi, the seat was hot and I nearly threw the thing on the ground trying to turn into an entrance ramp at slow speed; I was unprepared for its limited turn radius.


Actually you can get the Laverdas to be quite smooth (but not as smooth as a 90 degree or 180 degree twin). The trick on the triples is a computer controlled timing advance. It also helps to weight the ends of the handlebars. With the twins they are actually well-behaved on account of the pistons being 180 degrees apart when firing, and a well-weighted flywheel. Despite all that, they hold up extremely well.

Really more digs?

Ducatis like the 900 you had look fabulous; there's a few of those here in town. Apparently every single one of them emerged from boxes, and as you pointed out, your 750 Sport (also a fabulous looking bit of Italian dreamwork) is in boxes as I write this. That's where most of them are.


OTOH most of the older Laverdas you see in town have been in service (not boxes) since they were built (to be clear, there aren't nearly so many around). Ducati certainly is one of the few to make the Desmo drive work, but if its about riding across the country (here in the US that can be 1000s of miles) I'd prefer to do it on a Laverda even though some of their parts are harder to find (they vibrate more too), simply because it will hold together (I had both the 750SF2 and a Jarama 1000 triple). Ducatis and Moto Guzzi as you know share the same top end angles between their cylinders so vibration harmonically cancels; i.e. 'naturally balanced'.

Ha! Photos or it didn't happen. If he wants a classy Italian early 70s motorbike that he can actually ride without fear of it stranding him, maybe a Laverda instead of a Ducati. A Laverda from that period is far more reliable, far more likely to still be running this whole time, whereas a Duck has probably spent most of its time in boxes, moving from garage to garage.


I sold off my motorbikes except for my Moto Guzzi LeMans (which will probably go on the chopping block this spring). I was getting too fat so I switched to bicycles.

I’ll state once again as yet NO!! Class-D (and that also goes for my choice Class-D Technics SE-R1) will perform as good into a pair of Wilson Alexia and such, with their EPDR bass loadings like a hi current Gryphon Antillion etc etc can.

Just look up the 'on' resistance of any GaNFET. This will be the value of the output impedance of the amp unless it runs feedback- in which case it will be lower. So explain how an amp with an output impedance of 10 milliohms or less will care one bit about a 0.85ohm dip in a loudspeaker.

Well this is an excellent example of how George has no idea what he's talking about. Sheesh.

George, perhaps you could come up with some evidence of your preposterous claim?? You know what they say- photos or it didn't happen.

Combined, these two give’s you what’s called EPDR, (Equivalent Peak Dissipation Resistance) and that can give the amp a very nasty load impedance to look at.
Good big current bi-polar amps don’t mind seeing this
.
Class-D tends to quickly drop off it’s current starting under 4ohms down to 2ohms, look at any Class-D "independently measured" spec and you will see that, they never come close to doubling wattage from 4 to 2 to 1ohm, many even go backwards, showing severe current limitations into low EPDR impedances, this is limiting bass performance and becoming like a tone control, rolling off the bass instead of staying flat to 20hz.

This statement is blatantly false. Class D amps are limited in their ability to play low impedance (and severe phase angles) in the same manner as traditional solid state, although in practice a class D amp usually has a much lower output impedance than most traditional solid state amps.


Be-it class D or traditional solid state, if the output power is dropping into lower impedances, this is entirely due to the power supply and not the topology of the amp. At anything less than full power, any class D amp made in the last 5 years will act as a nearly perfect voltage source with any conventional speaker, with loads as low as 0.5 ohms no worries (since their output impedance is likely less than 100th of that). 

Atmashpere, how would one ever test that variable independently outside of other amplifier qualities? Plenty of seasoned audiophiles and reviewers alike that would argue subtleties in how a woofer is damped is second only to an amps ability to provide current into load variance in determining the best match.

@petland  Actually its pretty easy to determine if a woofer or other driver is overdamped with a fairly simple circuit. An amplifier does not need to be involved. This was demonstrated by the chief engineer of Electro Voice back in the late 1950s (it was EV and MacIntosh that led the way using the voltage rules for driving speakers). The physics of course hasn't changed since then.


Here's D.J. Tomcik's article, in case you are interested:

https://pearl-hifi.com/06_Lit_Archive/07_Misc_Downloads/1385_Critical_LS_Damping.pdf

Bass which unlike most people I find typically to NOT be a strength of most class D because it is almost always over damped for the modern speaker, is stellar on the DS225, fast and articulate but with some body like bass actually sounds.

@petland any damping factor over about 20 or so is going to be too much for almost any speaker made. You won't be able to hear a difference between 100 or 1000 if that were the only variable....

By the way, if you are using a toroidal transformer and you use metal plates or bolts to secure it to the chassis....you are closing down the sound. You want your toroidal transformer off the chassis on a piece of wood an mounted using tie wraps and glue or a wooden plate on top and teflon screws......way better sound.

Just a FWIW: the problem you are addressing here is the mounting. Most toroid transformer manufacturers supply mounting hardware, and if you use it the transformer will run hotter than it should. This is because the mounting bolt is steel, its magnetic and while a toroid is supposed to contain its magnetic field very effectively, in practice it does not, so that bolt becomes a magnetic short. If you install one just for fun, you'll see that the bolt is running hotter than the transformer. The solution is really very simple: just use a stainless non-magnetic bolt and no worries. And does this sound better? Yes it does- we discovered this phenomena about 25 years ago and its a nice tweak to any amplifier- the sound will 'open up' a little and of course the transformer will run cooler and make slightly more voltage- both audible and measurable!

Yes, but none yet "so far" that have taken the full major advantage of the 1.5mhz switching speed on offer when GaN Technology is used

Actually GaNFETs and MOSFETs are both capable of switch far faster than the switching speeds seen in all class D amps including Technics. The problem isn't how fast the devices are! Its how much noise is created (or not) by the switching at power; how much RFI is created, how much of this noise is then affecting other parts of the amplifier circuit, and then there is the issue of the 'body diode', which is where GaNFETs actually have the advantage over MOSFETs. This gets a bit technical, but has a lot to do with the coil in the output filter, which stores energy and then has to release it. Essentially its what is called a 'flyback' voltage (named after flyback transformers in old analog TVs), working in a manner similar to that of a spark coil in a car. In a nutshell, the output device can behave oddly- you tell it to turn off and it won't, partially due to capacitance effects on the gate of the device and partially due to the body diode of the device, which is a primitive diode formed in the passive layers of the device when its made (and I am vastly over-simplifying in this regard!!).  The filter coil flyback voltage can be effectively used to kick the GaNFET into its off state. That's a more important reason to use them than their speed, since MOSFETs have made so many gains in the last few years.


An important reason to keep the switching speed a bit lower is dead time. Dead time is invariant with frequency- no matter how fast you switch or how slow, the minimum amount of dead time you will need stays the same. So as you switch faster and faster, the output device is waiting for a greater and greater percentage of its ON time. This causes distortion to go up. Its true that Technics has a faster switching time, but it does not seem to be doing them any good IMO as they are paying the price in several different ways- increased cost without increased performance due to the associated issues with speed.

Problems with any amps shouldn’t be fixed by using even more feedback as a fix, it’s a added by the best designers to clean thing up a little and usually just local not global in an already well designed amp.

All the greatest amp designers say it. An amp should be reasonably good spec’d, and to use it just as a clean up tool.
And the preference with the best is to use "local feedback" only by the best designers. Not global as being talked about here, and then even including the Class-D output filter in the global loop, what next throw the speaker wire and the speaker in the loop as well, like Trio/Kenwood tried all those years ago, utter sterilized disaster sound.

The first paragraph is true. So is the 1st sentence in the 2nd paragraph. Right after that things go off a cliff...


The fact of the matter is design-wise you start with a class D circuit with no feedback and get it to sound decent and perform fairly well too. Then you add the feedback, but in this case so much that the amp goes into oscillation. With a normal amp you could never get away with this. But with a class D amp, the oscillation is then used as the switching frequency, so you put a few frequency sensitive components in the feedback network so that when it goes into oscillation, it really can find only one solution- one frequency- at which it oscillates. The switching frequency itself is then used as one of the inputs to the amp (the other being the actual audio) but is converted to a triangle wave for comparison to the audio signal.


Now your already fairly linear amplifier has a mess of feedback (over 35dB) too, and the reason for doing this is to allow the amp to compensate for the distortion that is normally generated by the application of feedback. You need over 35dB for this phenomena to occur. This gets rid of the brightness and harshness that is otherwise associated with **all** amplifiers that have used feedback at insufficient levels (which describes nearly all traditional solid state designs of the last 60 years). So its quite unlike that unsuccessful Kenwood system George mentioned above, and yes, you do include the filter in the loop so that the amp can compensate for phase shift induced by the filter.


In a conventional amplifier you have a thing called Gain Bandwidth Product, which is how much gain you have vs how much bandwidth you have. The idea here is that you are going to blow off the gain when you use feedback to reduce distortion (and output impedance). But there’s a sort of zero crossing where the the amp circuit just does not have the bandwidth to support the amount of feedback used. In a nutshell you could say (inaccurately...) that the GBP is a sort of fuel that you use up when you add feedback to the circuit. When its gone the feedback at that frequency is gone too. Put another way, an amp with insufficient GBP will have distortion low at low frequencies and the distortion will rise as frequency rises. This is very audible- our ears are very sensitive at 3-7KHz and they use higher ordered harmonics to sense sound pressure. We’ve been hearing this problem for 60 years and its literally why there are still vacuum tubes around after all this time.


This is why solid state amps traditionally do bass well as there is plenty of feedback to allow them to be low distortion. But at 7KHz there isn’t enough, so harmonics at that frequency aren’t suppressed, and there are actually more harmonics because the feedback is adding them through a process called ’bifurcation’. And so we wind up with an amp that is bright but measures flat. To get around this problem, manufacturers and the test and measurement industry have resorted to a simple technique called ’lying’. The distortion of the amp and the resulting harmonics are measured at 100Hz, where the amp measures quite well. If you measured it at 2KHz you would see something vastly different, which is why that’s never done.


Because a class D amp can have so much feedback, it can measure the same low distortion at 15KHz as it does at 100Hz. IOW the distortion product can be ruler flat across the board. Prior to this technology, the only other way to do this was to have no feedback at all, and use the most linear circuit with the widest bandwidth you could get (200KHz is required to prevent phase shift at 20KHz on that basis). This is all because it really is important to the sound of the amp that the distortion product be both benign (mostly lower harmonics) and ruler flat across the board.

How many class D do I have to listen to? Is there a The class D which disproves the "myth"? Did I just happen to miss the good ones?

Here are some tips. The first thing is avoid amps using switch mode power supplies. The reason for this isn't that SMPSs work, they do and they can be quite low noise. The problem with them is they are cheap to buy off the shelf and expensive to have built custom. Off the shelf units tend to have current limiting protection circuits that cause the amp to suffer a loss of bass impact. So unless you are looking at an enormous amplifier manufacturer, they won't be able to afford to have a supply built custom.


This means that the supply has to be a traditional supply: power transformer, rectifiers and filter caps. The power transformer current and the capacity in the supply can have a profound effect on the performance of the amplifier, since class D amps can go from almost no current draw to really quite a lot at full power. This really means that the supply has to be well filtered, and the power transformer high current, else the power supply will sag at higher power (bass notes).

If the amp is zero feedback look for a filter frequency of 100KHz or higher. This is to reduce phase shift at audio frequencies. If they won't publish details like that, consider looking elsewhere.


If the amp employs feedback, find out how much. If details like that are not published, consider looking elsewhere. The feedback **must** be in excess of 35dB! Any less and the artifacts of the feedback itself will become audible (as they are in traditional amplifiers).


I understand your skepticism! The first class D amps I heard I really thought were a joke. But they have improved over the years, so much that about 5 years ago I personally started to take them seriously, but even then some are terrible (we have a small 30 watt amp we use for testing, but it sounds really boring; I've not looked into it to see why that might be the case, but its pretty obvious to everyone in the shop) and some really work quite well and everything in between.

George is correct. Lack of switching bandwidth is the problem. I know Bruno Putzeys claimed differently in an old interview, but he's mostly a self-promotor so no one should take what he says as gospel truth.

For the record, Bruno Putzeys is the one that made self-oscillating class D amplifiers a reality, a practical thing. That's a pretty big deal; if you work out the math for that you are doing quite well. You better be good at calculus with multiple variables! He also is able to show the measurements to back up what he says about how the technology works. Attacking Bruno is really done at your own peril!

By 'switching bandwidth' I think you must mean 'switching frequency' since 'switching bandwidth' isn't a thing. And in this regard your statement is false. Based on the rest of your comments I would venture that you simply haven't heard everything that's out there, and just like traditional solid state amps and tube amps, there is a tremendous variance in class D implementation!


You might think of it this way: digital has come a long ways since 1981 when it first started showing up. Its common now to see scan frequencies of 192KHz; class D amps are commonly switching well over double of that. At the current state of affairs, the practical upper limit is around 600-700KHz before you really start to get into troubles with radiation and oscillation issues- the fact that Technics seems to have gone well past that says a lot about their engineering expertise. But Technics has to switch faster, since (if their claims are correct) their circuit is zero feedback, so they have to raise their filter frequency quite a lot in order to avoid phase shift at audio frequencies. Also for the record, self oscillating class D amps don't have to do this- their filter frequency can be lower since they can run so much feedback that it is able to correct for phase shift. 
@tweak1 Thanks for your offer!

The answer is simple: Gain knowledge instead of made up stories.

These are words of someone either in denial or in product protection mode.

George, as a simple tip: the ability to admit that you were wrong is a strength, not a weakness.


In my quote above I was simply stating a fact. My advice is to study how class D amps work (the theory is simple) and keep in mind that with any current output device (MOSFET or GaNFET; the difference between them these days is slight) the output impedance is going to be lower than any traditional solid state amp, often by a couple of orders of magnitude. As I mentioned earlier, the PuriFi modules (not made by me, so clearly not in 'product protection mode') have output impedances that are tricky to measure as they are about the same as a few inches of 12Ga wire.


So this is not about denial, its physics. When you have an amplifier with an output impedance that low, it pretty well makes no difference what phase angles the speaker has nor its impedance, since both are several **hundred** times more than the output impedance of the amp. This simply means that the amp will be able to make constant voltage into that load.

How can anyone believe that "any" Class-D driving a pair of Wilson Alexia etc etc could possibly do anywhere near the sonic job that an amp like Gryphon Antllion or similar could do.
Especially in the lower bass where the Alexia's are 0.9ohm loading presented to the amp.

The answer is simple: Gain knowledge instead of made up stories.

Now you go find a Class-D that will keep pushing even more current into that kind of loading like a good bi-polar amp can.

You mean like, *most* class D amps?? Because its most of them.

@jaytor  I've found that dealing with George means you have to accept that Dunning-Kruger is at play.

Class-D semiconductor devices operate in saturation and would not be impacted by this.

That, and the fact that the amp has an output impedance of a handful of milliohms. Phase angle isn't going to affect it.

Why would a Class-D amp care about phase angle?

In a nutshell, any good class D amp wouldn't.

 In general, solid-state (ss) amps reproduce deep bass frequencies better than tube amps. A good indicator of an amp's bass reproducing ability is its Damping Factor, which basically measures its ability to control, or start and stop, the bass driver. The higher the number the better its control. Class D amps usually have much higher damping factors than the other amp types; class D amps can be rated in the low thousands, class A and AB are typically rated in the low hundreds and tube amps even lower.

There really aren't any speakers that need more than 20:1 in a damping factor and some need considerably less. On top of that, traditional solid state amps actually have enough feedback at bass frequencies to do the job (whereas they usually don't at high frequencies, which is why they are often bright and harsh).

I always thought pure class A amplification was most ideal. So why manufacture Class D amps?

Class A is all about the output section- keeping its distortion as low as possible and its quite effective. But this says nothing about the input section and driver circuitry, and the output section is still going to make some distortion (just less than if the same circuit were class AB or B).

Class D uses a different means (switching) to make power, and so does not make distortion in the same way. There are basically two means that the amp produces distortion- the first is how accurate the encoding scheme is. The second is caused by the fact that the output transistors take a finite period of time to switch on and off and so to prevent damage to them, a bit of waiting is done to allow one device to shut off before the other is turned on. Otherwise current can shoot through both devices, heating them up quickly towards failure. This waiting time is called 'dead time' and contributes to distortion. The distortion signature is different from traditional solid state and does not tend to be bright. Instead it tends to have more lower harmonics like tube amps, and not surprisingly  can sound a lot like a tube amp, except when the amplifier is overloaded.  Certainly the reduced amount of heat is an incentive- this is the main argument against both tube amplifiers and class A amps and especially class A tube amps  :)

"For the perfectionist audiophile, they will not be in the class A quality, but 99% of the people will not quite notice the difference."

I've heard class D amps that sound as good as some of the best tube amps made. This quote is out of date IMO.

class d greatest strength is outstanding bass, so yes on that front

but dual subs are much better than one... some will even say the best way is a 'swarm of subs' to smooth out the deep bass response, which can be very lumpy and hard to modulate with a single sub in many rooms

I'm running a pair of Swarm subs with a class D amplifier, to break up the standing waves in my room. My main speakers go down to 20Hz so I didn't need 4 subs. Its amazing how well it works- the bass is absolutely consistent everywhere in the room now (a lot of bass was absent at the listening chair prior, with too much bass elsewhere in the room).

This is a smoke screen, they are still "basically" the same, with the same glaring problem, upper mid/highs phase shift, because of the "switching frequency noise filter" on the speaker outputs.

This statement is false- they are not 'basically' the same at all.


Bruno Putzeys has shown that you can run enough feedback in a class D amp that it can correct for phase shift at high frequencies that would otherwise be the result of limited bandwidth. To do this the amp has to have more than 35dB of feedback. Any less and the phase shift will show up. In this way the bandwidth of the amp can be limited to 20KHz and yet no phase shift in the the audio band. Normally you need lots of bandwidth (usually 10x the frequency of concern) to prevent phase shift.


With traditional solid state and tube amps, this much feedback would likely result in oscillation due to phase margin issues. A class D amp can take advantage of this because the oscillation is welcomed- and is used as the switching frequency. This type of class D amp is known as 'self-oscillating'. The pulse of the switching frequency is then converted to a triangle wave, and compared to the input signal to create Pulse Width Modulation (PWM) for signal encoding. Its a neat trick- and then you don't have any of the issues described in the quote above. 

Class-D wattage instead of doubling for each halving of impedance load actually gets goes backwards when they see 2ohms and worse if there is any - phase angle involved, because they can’t deliver the current like a good bi-polar linear Class-a or a/b amp can.

This statement is false. The limitation of a class D amp is similar to what is is with any amp- how big are the heatsinks, how much current can the output devices handle, and how much current is available from the power supply.


The output impedance of any modern class D amp is only a few milliohms (the Purifi module' output impedance is lower than a foot of 18ga wire). So it can double power from 1 ohm to 0.5 ohms as long as the current doesn't exceed the limits of the output section. The real issue is the filter; lowering the load impedance lowers the effective 'Q' of the filter, IOW it flattens out the filter response. So as the load impedance is reduced, the 'residual' (which is the sine wave remnant of the switching frequency) will increase in amplitude. For any speaker this isn't a problem as the residual is very small on any competent class D amp.

One thing I find reprehensible is the way the big names have added all kinds of tweaks to Putzeys’ design (op-amps, power supplies) that are completely unnecessary, or even cause spec deterioration, so they can charge more. 

The reason everyone has different input circuits is because the module itself doesn't have one. Its input impedance is about 2,000 ohms- most preamps and sources would fall flat of their face driving a load like that. When you're designing a module, there is only so much gain you can get out of it, and you have to use an input circuit called a 'comparitor' which tends to have that low input impedance. So some sort of buffer is typically needed. You are correct that this can color the sound if not implemented correctly. Bruno of course has a recommended circuit, but its not cast in stone that it be used.