Class D amplification - TAS perspective

MartinLogan Audio Owners Forum

Help Support MartinLogan Audio Owners Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

twich54

Forum Administrator
Joined
Nov 25, 2005
Messages
6,935
Reaction score
278
Location
SE Pa
just got finished reading the recent article in TAS on Class D amps and it appears that while strides have been made, overall their inherent design has a ways to go. Are there members here who have and or are presently using any of these types of amplifiers with their M/L's ?? My only experience with switching amps was that God awful Bob Carver "cube" thing from the early / mid eighties.
 
Although I haven't read the article itself, I've read a number of threads about it on other audio forums. It seems the round-table consensus, if any, was that Class D switching amps excel for the lower frequencies (i.e powered subs/woofers), but are still problematic at the higher end. Looks like ML got it right with their implementation of "Ice-Power" amps in the Summits and Vantages!
 
Beat_Dominator said:
The cube was the m400, deffinately not ranked high on sound quality ;)

They rarely worked long enough to listen to IME.
 
twich54 said:
just got finished reading the recent article in TAS on Class D amps and it appears that while strides have been made, overall their inherent design has a ways to go.

I've read it, too. I think that the answer might be more complex than that. First to begin with, I'm not sure that they burned in the amps sufficiently - a lack of treble from a class D amp is a typical sign of not enough burn-in, which can take as long as 400 hours. (I know what it took to my Ps Audio HCA-2 to reach peak performance and I am burning in another PS Audio right now, a GCC-250, on which I miked the frequency response every day, the histograms are quite eloquent). In my opinion, the experts at TAS are simply getting a different sound than what they are used to. And they are used to a sound which is coming out from amplifiers that add or subtract colorations. Amps like the hypex-based CIAudio monoblocks, with a ruler-flat frequency response regardless of speaker impedance, don't add or subtract anything: they just amplify. Of course, getting out exactly what was recorded is not always what one might like to hear :)
 
Lugano, You make some good points, escpescially on the "extraction" bit !

I wonder though, when a Mfg is requested to provide a "sample" piece of gear for evaluation wouldn't it have made sense to have had the burn-in period already performed ??
 
twich54 said:
I wonder though, when a Mfg is requested to provide a "sample" piece of gear for evaluation wouldn't it have made sense to have had the burn-in period already performed ??

It makes sense, but if you unplug a D-Class amp from AC, box it and ship it, the time it reaches its destination you have to start over...
 
lugano said:
It makes sense, but if you unplug a D-Class amp from AC, box it and ship it, the time it reaches its destination you have to start over...

Understood, but the "burn-in" would be done and the amp would reach it's sonic best within minutes to a few hrs , not 400.
 
That is my opinion as well.

Dusty of CIA suggested that his amps may improve over a period of up to 300 hours but said also that they are fine after 24 hours or so of being powered up continuously. They need to be connected to power, capacitors charged and all components at operating temperature. That makes more sense than a burn-in requirement, IMO.

The ICEpower based Rotel 1091s which I tested for a few days did show improvements in the upper decade as well as the bass within the first day. They did not improve (much?) from there.

Meanwhile the D-200s had been disconnected for 2 days. When I swapped the Rotels for the D-200s the latter sounded better right away, and more so after a few minutes. Going back to the Rotels and checking again after two days confirmed this picture.

As far as I am concerned those dozens and hundreds of hours of burn-in are not required.
 
Regarding the state of Class D.

ICEpower 1000ASP based amps like the Bel Canto REF1000 have received extremely favorable reviews. Some reviewers went as far as calling the midrange "tuby"... They liked them because they sound different. Lugano is probably right that the guys at TAS don't like the difference.

I found that in an analytical sense the 1000ASP based Rotel 1091 did not sound that different from my D-200s. The latter seem to breathe and sparkle, however, in a way that the Rotels could not match. "Tubey"? I guess not. Revealing, yes.

I did not even think that these amps sound so different from the analog norm. The contribution of the pre-amp to the improvement of my system was at least as pronounced. Since my pre-amp is a passive attenuator it should not add or remove anything. Maybe that is why the two digital amps did not sound so different.

If, on the other hand, the amps had been driven by pre-amps that work well with a warm amplifier because they add treble, for instance, then I would expect that in combination with a Class D amp the result would be called bright and harsh, or "digital". Maybe something like that contributed to TAS's findings. Having no access to the TAS article I must say that I am guessing here.
 
lugano said:
... (I know what it took to my Ps Audio HCA-2 to reach peak performance and I am burning in another PS Audio right now, a GCC-250, on which I miked the frequency response every day, the histograms are quite eloquent)...
How interesting. I missed that bit during my first read. This sounds like hard evidence. Could you post those histograms please?
 
ralflar said:
This sounds like hard evidence. Could you post those histograms please?

Sure I can. Unfortunately I have only 4 positions to save presets, but you can see below how it looked like on the first day and the 2 days after it. The 4th position was used and replaced every day thereafter, so I can only show you how it looks today, after 6 days of nonstop music playing, loud daytime and soft nighttime. Keep in mind that what you see on the histogram is NOT the measured response, it is the inverse of it, i.e. what the equalizer recommends in order to level up pink noise. It goes like this: the eq outputs pink noise and measures first the left channel, then the right channel. It establishes an optimal equalization (shown in the graph) which should offer the needed correction for a flat response. Of course, it will never be flat, as my panels also need replacing, but you get the picture. There are 2 overlapping histograms, one channel is brighter than the other.
 

Attachments

  • gcc250.jpg
    gcc250.jpg
    44 KB · Views: 174
Thanks for this. Now let's see if I get this right:
  • Most corrections were needed in bass and lower midrange.
  • The two channels' responses needed a few days to converge, again mostly in the left half of the spectrum.
  • As they converged over time, fewer corrections were needed for a flat response.
  • The left/right channel difference are suprisingly large during the first three days in that lower region.
  • White channel was pretty good after day 2. Blue channel bounces around like a foal on the first day of spring.
  • The treble region needed more boost during day 1. From then on the right half of the spectrum (can't see the frequencies in the images) appears to have shaped up.
Did I interpret the plots right? It would be good to know how big the error bars are in those plots. You wrote this is with pink noise, so how do the measurements apply to listening to music?

I like how you blended the edge of the photos into the white background. Which image processor do you use?
 
Well, actually the first day it was a mess. Corrections were needed everywhere and the 2 channels were each on its own trip :)

For a better picture of the scale and frequency points, look at the following picture:

http://img328.imageshack.us/img328/401/technics806626hl.jpg

As of today, the correction is needed only in the highs, the last bit of the histogram is showing a + 6 db centered around 16 kHz. There is also a strange +4db around 160 Hz, but it's not so important. The important thing about all these histos is to show that burn-in is necessary to a big extent, and the treble region is last problem which will be solved by burn-in. And it also applies to music listening: the treble is there, but it's distant, hidden by the mids (which are stellar, btw). I'll let it cook further and we'll see how it looks after one more week. In any case, it's more than listenable now, it's almost enjoyable.

For graphic pic processing I use Photoshop. You can get that effect easily - just select a 10 pixel feather value in the selection's parameter, select, copy and paste into a new, blank document.
 
I am using a pair of Nuforce ref9's paired with a Meridian 603 pre and the Aerius for some time now.

It is true that they need some burn in...mine took approx. 100 hours.
But after that the tonal balance was very good; I do not agree that the mids and highs are of lesser quality compaired to the lows.

Over time I have upgraded several compontents, and they every time manage to portray the improvements very well.
In fact , I just finished improving the X-over filters with some high quality caps, and the smoothness and depth improved further, which means to me that the amps were still not performing at their best..eventhough it's incredible already.

They sound very powerfull with lots of air and soundstage and are very precise without lacking in the musicality department.

I know that I will not part from them easily... :D
 
Well, I decided to return the Rotels when they had around 80 hours on them. Their sound did not change after the first day or so, at least not to my ears.

Which goes to show one of the following: (a) I need new ears, (b) I could not hear further improvements because of my scepticism that any would materialize, (c) I subconsciously wanted the Rotels to sound worse because I prefered the better looking and cheaper D-200s, (d) the Rotel 1091s need no extended "break-in", (e) I need a shrink.

I believe neither (b) nor (c) apply, unless (e) is true. ;)
 
lugano said:
...
For graphic pic processing I use Photoshop. You can get that effect easily - just select a 10 pixel feather value in the selection's parameter, select, copy and paste into a new, blank document.
Thanks. I do not have Photoshop but I guess Paint Shop Pro must have a similar filter.

If I had a Mac instead of a Windoze box I'd probably use Photoshop as well. The new Mac Pro looks extremely attractive. I am a Unix person at heart. Must ... resist!
 
Dave,

There is an interesting thread regarding the TAS article in the Channels Islands Audio corner of the AudioCircle (linky). Sure you will find Class D bias but it still is a good read.
 

Latest posts

Back
Top