Transforming TPA3110 BD into ABD modulation.

[Hayk]

This thread is off spring from,

The class ABD

www.audiosciencereview.com

This is the first of series of measurments and demos compared to TPA3255.
The pictures below shows the differential output for 1w and 15w @8 ohm to exhibit the cleanness of the filter and the linearity at zero crossing, which is the major enhancement over BD modulation.

 

[Hayk]

This is the residual of 1khz 15w @8 ohms with the output -input to reduce the fundamental 100 times, 100mv rms from 10v rms. This magnifies the distortion 100 times to be visualized. Notice, besides the switching noise, it is perfect sine wave not any trace of harmonics.
Next is 6khz 15w but magnified 19 times. I need to delay to subtract further.

 

[Hayk]

Frequency response.
1khz square,
-3db 78khz
Max peak 67khz +5db

 

[Hayk]

The Character of the filter is different than simulation because of common mode choke. Originally, it should be 2 wires in hand wound. The one I found, see picture, had thin wires not suitable for 2A. The mutual inductance of separated windings 1mH each can reach 2-3uH. I will see how to modify to get the resonance damped.

 

[Hayk]

The simple solution, is to add anti resonance snubber.
I tried just using my piff o meter (nose), 220nF+10ohms, it works.
The final value should be derived upon well assambled circuit.
The peak occurs now at 62khz with 2.2db.

 

[Hayk]

This is the beloved TPA3255.
1KHZ, 2W 15W, square wave and the un usable residual that needs to be filtered to exploit.

 

[Hayk]

The offset loss should occur at 14v. Do you believe that an FFT can see
on this noisy wave, a tiny voltage loss for short period.
6khz 15w.

 

[Hayk]

I wanted to see the noise of TPA3255. I filtered with 2.2k and 100nf the output to get rid of the 453khz carrier. See in the picture below the strange random pulses.

 

[Hayk]

I will show you by recordings how the sound of AD modulation degrading when crosses the offset level. With 48v, 450khz,10uH inductor, the current in the inductor becomes DC at 3w@4ohms and 6w@8ohms, if in BTL mode and double watts at PBTL. see the curves below.
For the sake of my ears and that of the neighbors, I branched BTL and paralleled 8 ohms the output. I'll make a series of recordings at below 1w and above 3w to show you how the sound gets rough. If used at high powers, as crossover distortion, it becomes insignificant, this why class D amps are tested either low power below offsetting or at very high power to convince the noobs, it is good in all power range.

 

[Hayk]

My first attempt of recording failed. I thought Samsung smart phone would be sufficient, no, very bad quality, specially the feature I want to exhibit, the grain in young Barry Manilow's voice gets totally corrupt even at 1w.

After 1:30
I do have a condenser cardioid mic and Behringer UMC202, but it needs time to deploy and prepare editing after each recording.
Another problem is the file size. I tried to upload zipped MP4 of 30s, it was too large for the site to upload.

 

[Hayk]

I understood the Samsung problem. It has 2 mics interfering. Now it is 256kb/s 48khz zipped.

 

[Hayk]

I noticed that my neighbor, 2 apartments away has returned from holiday. He has Samsung refregerator, extremely high EMI noisy. The pulses on #8 is most probably due to his refregerator.

 

[Saponetto]

Mostly interesting.
TPA3255 is nowadays implemented in so many ways that for sure it needs to be better recognized and understood.

 

[Hayk]

What makes class AD modulation go wrong, is the grain in sounds. The TPA3116 in AD modulation goes out track with the sound of washboard. The 3255 does the same but at higher pitch.
B. Manilow's song, I use it since 80's, to test the brightness of the sound system, subjective frequency response, as Jean Hiraga called it. Slight compression of low frequencies, the sound of this song becomes disagreeable. With 3255, it is un listenable, a powerful radio sound. The video below shows exactly.

This phenomenon has nothing to do by the fact that it is classD, it is due to chip character.
I found several grained high pitch voices the 3255 doesn't like that I will demonstrate some of them.

Andrea Bocelli, Claudio Villa, un amore cosi grande.
The list is rather long just to show you that it is not only that particular song.
Now, don't be dissuaded, there are many other music that 3255, because of its brightness, bring to surface, as compressors do, instruments you don't here them with normal amplifier.

 

[Hayk]

I will attempt the next few days to find a means to measure the brightness of an amplifier as this 3255 is exaggerated kind.
I will start by using pink noise or multi frequency with low pass. My tiny, battery operated scope, suitable for floating differential measurment, has FFT, but not calibrated. I will rather use my Rigol's high performance FFT using an isolating transformer 10k-10k and compare the input- output spectrums. Maybe I can see something.

 

[Hayk]

I will not be able to use my Rigol scope. The refrigerator noise on the output of my DAC-Bluetooth, without any input is over 1vp. although the DAC is powered by a battery pack. My neighbor is a kind person but the refregerator doesn't have on-off to ask him shut down. The mains plug is inside a custom made furniture, so the only way is to cut the electricity of the apartment.
I could deploy my cardioid mic with a windows tablet, I will record using Audacity which was unable to configure the Behringer but works with a recent USB 16bit sound card.

 

[Hayk]

By using battery source 220v supply but using the earth as was before, I could make it 10 times better, but still not sufficient for output to input differential measurments.
Bad luck for me.

 

[KL....]

Thank you, @Hayk, to ask, would this architecture (ABD) happen to be utilised in the SMSL AL400 and/or the SMSL AS400?

 

[Hayk]

I don't think so. The 3255 can't modulate in BD. The best enhancement for this chip, is to increase the the open loop gain. The circuit below can do it. The OB and OC are the outputs and +in and -in are the inputs of 3255 capacitors.

 

[Hayk]

I received the 2x30w board and the 24v 4A miniature GaN supply. See the picture compared to 24v 2.1A was 10 years ago.