When did class D start to not suck?

[restorer-john]

It was actually a pretty great amp and quite reliable, unlike some other amps of that or this day. IIRC it had better output protection than most other amps.

It sure was Don. Comprehensive protection and a front end mute that operated in <1uS!

It's funny that people who haven't worked on, owned, or used these amplifiers really have no idea how good they were...

 

[Head_Unit]

Hypex UcD was the first one I know of that was designed to deliver flat frequency response regardless of load. That would have been 2005ish?

Alpine did this on the car side, by extending the feedback loop out past the passive output filters. Around the same time or maybe earlier. I *think* using ICE modules inside? Or maybe in-house...I gotta ask my buddy who spearheaded that development.

 

[Hiten]

I just found my sonic impact amp yesterday in a box with assorted playstation accessories. I look forward to hooking it up again. It really sounded pretty good within its limitations. I only bought it out of curiosity at the time.

I had tripath TA2024 amp module. It sounded nice. Power was low though.
regards.

 

[Mnyb]

It sure was Don. Comprehensive protection and a front end mute that operated in <1uS!

View attachment 225729

View attachment 225730

It's funny that people who haven't worked on, owned, or used these amplifiers really have no idea how good they were...

Of topic :
Fyi output stage looks like a third of VFD drive (frequency converter) for AC motors I work with drive systems, there are differences current sensor is in the output leg and is usually a hall sensor , there is usually no lowpass filter only a du/dt filter .
An they are not fet or vfet today , modern drives uses IGBT's (Insulated gate bipolar transistor ) and pure PWM is rarely used nowadays.

Really capable transistors for this application came in the 90's smaller motors was transistor driven in the 80's i reckon before the 90's a large 690volt industrial motor usually used GTO thyristors ( a component that wont cut it for audio btw )

Nowadays ordering a drive with transistors for your 3 megawatt 690v 3 phase motor is quite possible

I'm not very knowledgeable on the actual development of switching transistors , but given the vintage of the SONY it looks like they took advantage of then recent improvement in the transistors of the time ?

I think power suplies amplifiers and motor drives and other tech with switchmode operation PWM and similar ,was application in waiting for better transistors ?
Enginners wanted to do this since forever , but needed better components for viable commercial products.

 

[Zaireeka]

When did class D start to not suck?
When audiophiles finally realized that the D didn't mean digital... ​

 

[JimFarrell29]

Having found my Sonic Impact I just had to set it up with an old pair of Music Fidelity Merlin speakers (4 ohm). You know it's really not that bad. A bit harsh and glary on top, and shy in the bass ( though that could be the speakers which are single driver ). But not at all bad.
Here's a couple of pics for the nostalgic among us.

 

[steve59]

I listened to a little Lumin 60 watt class D integrated with built in dac that retailed for $2k that I couldn't tell apart from the most expensive separates the shop had to offer. The speakers were persona 7f and I was quite familiar with them and how sensitive to source they are. That was 2 years ago and the first time I couldn't tell the class D from the other components on hand. Not to say class D was always inferior as I've heard a/b amps that didn't pair well with the persona's either.

 

[Hiten]

Having found my Sonic Impact I just had to set it up with an old pair of Music Fidelity Merlin speakers (4 ohm). You know it's really not that bad. A bit harsh and glary on top, and shy in the bass ( though that could be the speakers which are single driver ). But not at all bad.
Here's a couple of pics for the nostalgic among us.

may be 2 way combined with woofer will tame down some harshness.

 

[TimF]

Spectron amplifiers claim to be class D. They go back quite a few years and there are still some of them out there. There is no way to get them serviced so they are dying out one by one. I owned their Musician II amp and I know it was very good. It is just that the design and electronics were dense and complicated--I defy anyone to reverse engineer one of these amps. There is just no way to keep such complexity running long term. I don't recommend buying one but I heartily recommend listening to one if you get the chance. They are heavy. When were they developed? Maybe around 1998 - 2000. Stereo Times had an article about them in 2002.

 

[Offler]

Just after I purchased DBR-62s and I still did not decided what amplifier to get, i was told that:

a) I want Class D because of lower power consumption
b) It will introduce hum or distortion no matter what, mainly on 75% or higher load.
c) Because "Class D" stands for "digital" the sound will leave and enter digital domain multiple times, effectivelly damaging sound.

Only first claim proven to be true for my amplifier (NAD d3020v2 with Hypex UcD102 module). Class D is considered to be analogue way of amplification, THD is the best near 95DBFS at headphone output (better at speaker out) about 0.0017%.

And this happened a year ago. Then I found out that those modules were available for over decade, the DAC used was around for 20+ smthin years as well. So ... targetting quality between CD and DVD quality (96-102 dB, at THD lower than 0.005%) with this type of product was available for 10+ years, but people are being very skeptical mainly because of how Class D was used in cheap products.

 

[sweetsounds]

I think the art of Class D lies in a sort of pre-distortion of the audio signal.

A class A is the most simple design and it performs superb if you don't screw it up with wrong components.
Class D replaces amplitude modulation with pulse width modulation.

Since the output stage now goes high-low only, there is little time in the energy-wasting linear region of the output stages.

There were reliability issues if the design didn't avoid shoot-through of the output stages.

The early designs had output low pass filters to convert the PWM signal back to amplitudes. These filters could interact with the speakers.
That was fixed with better feedback circuitry and multi stage class D.

Then came noise shaping to get that bang-bang switching noise out of the audio band.

Now finally the phase issues of all of this are claimed to be solved and some new designs liky Purify or nCore claim linear phase (ie they only introduce a time delay to the signal).
I haven't seen measurements here.

There are some small issues left it seems, e.g. somewhat lower voltage gain, which requires a pre-amp stage and not all frequency responses are fully flat.

Net: with lots of know-how on the feedback circuitry and pre-distortion Class-D performs very well.

In subwoofers all of the issues above are less critical.

 

[Offler]

Now finally the phase issues of all of this are claimed to be solved and some new designs liky Purify or nCore claim linear phase (ie they only introduce a time delay to the signal).
I haven't seen measurements here.

There are some small issues left it seems, e.g. somewhat lower voltage gain, which requires a pre-amp stage and not all frequency responses are fully flat.

Net: with lots of know-how on the feedback circuitry and pre-distortion Class-D performs very well.

What amount of delay are we talking about? Because... I run integrated amplifier (DAC+AMP) fed over Toslink, paired with a display which is fed through completely separate display port.

Media Player software indicates that Audio/video tracks are synced up to 30 microseconds at level of operating system, and there is simply no way to notice. Display and AMP are obviously introducing delays of their own, but ... how much it used to be on Class D amps?

With the specific product i have here (d3020v2) the frequency response is flat. It does not even seem there is any custom filter in action.

 

[KMO]

What amount of delay are we talking about?

"1 degree at 20kHz" for the Purifi, which would be under 0.2 microseconds.

 

[KMO]

Note that on the wide-bandwidth view, the phase response isn't linear - in the Purifi there's basically a 60kHz low-pass filter, meaning you do have a distinct phase curve bending down to 45 degrees by the time you get to 60kHz.

But when you zoom in to the audio band, the curve inflection is so far off the graph to the right, that what you're left seeing is basically a straight line from 0 to 1 degrees, which is thus effectively a constant delay for anything in the audio band.

 

[voodooless]

"1 degree at 20kHz" for the Purifi, which would be under 0.2 microseconds.

That's a whopping 0.07 mm difference. Moving your head has more influence, or rather: trying to keep your head still has more influence

 

[KMO]

Well, if you were talking 45 degrees at 20kHz, that would still be "only 3mm", which you could say the same "moving your head has more influence" about.

The non-linearity across frequency could still be an issue though.

It's a bit daft - they started with people worrying about the phase distortion, so they got phase so linear that it's now equivalent to an even-lower delay, but now people are worried about what the delay is instead, because they started talking about delay.

The delay aspect was never significant in the first place.

 

[Offler]

Well, if you were talking 45 degrees at 20kHz, that would still be "only 3mm", which you could say the same "moving your head has more influence" about.

The non-linearity across frequency could still be an issue though.

It's a bit daft - they started with people worrying about the phase distortion, so they got phase so linear that it's now equivalent to an even-lower delay, but now people are worried about what the delay is instead, because they started talking about delay.

The delay aspect was never significant in the first place.

There is a parallel from PC gaming.

There is a claim that Vsync introduces delay which might be crucial for online gaming. This claim was based on a fact that display and GPU "lock" one image for given time (at 60Hz, its 16.6ms, at 120Hz its 8ms etc) and statement that this time is somehow "added" to existing delay.

The effect which most users suffered from when using VSync was real, but it wasnt "added delay". Certain buffers in render stream have overflown and result was inconsistent frame rate.

If there is a statement made about a delay I usually ask how much it is, and you guess right that mostly its insignificant value.

Also the claim about Vsync helped Nvidia sell their Gsync technology, which in fact did not reduced any non-existent delay, but instead reduced graphic buffer overflow to minimum.

So ... if its engineering which creates its own problems to be solved, its usually already good enough.

 

[curiouspeter]

When did class D start to not suck?​

When audiophiles finally realized that the D didn't mean digital... ​

You mean my TDAI is not totally digital?

 

[Rottmannash]

C'mon, everybody knows that class D = digital, class H = H-bridge, class G = good, class AB = absurdly better

....finally an expert to clear this up.

 

[tonycollinet]

You mean my TDAI is not totally digital?

Well, given it has analogue inputs/outputs, I'm going to suggest not.