Improved Diamond Buffer Design

[jam]

Any iidea abour R10/11?

Jam

[aparatusonitus]

Originally posted by h_a
You want temperature tracking for a stable bias. The resistor can't achieve that.

Originally posted by janneman
The diode-transistor is current operated. Current operated diodes are linear (Iin = Iout).

Ah, I see, thanks...but what about R10/11?

[lineup]

About R10/R11.
I really not know. I haven't even simed to see if those change anything.
Personally I have not used such resistors for diamond.
I do not think Walt Jung used them.
Maybe he just put them in, because they are usually found in power follower stages.
Often with a capacitor in parallell.

[homemodder]

Originally posted by jam
homemodder,

Sorry, Iwas under the impression that JLH used a single feedback path. It should sound very good as you say.
I wonder how it would compare to the jfet diamond buffer (non-feedback). This might be a good comparison between feedback and non-feedback circuits (I should say non-loop feedback before I get into trouble with the definition of feedback).

Regards,

Jam

Hi Jam

A comparison would be interesting, As i have both the JLH and the bjt buffer I showed, I tried a comparison yesterday, buffer between cdplayer headphone out and headphones displaying around 350 ohms load. The bjt buffer was brighter but lost some in fullness and warmths, the JLH being the opposite. Comparing both to the standard headphone output the bjt buffer was closest to wire with gain. Its a bit of a unfair comparison though as my JLH is running with cheap low gm 2n fets and performance could be much improved with the japanese fets. I tried the same test with seinheisers hd600 which is a even heavier load, but the JLH was at clear disadvantage here. As for the diamond circuit I cannot say, will have to build one.

I dont have access to my desktop at the moment so I cannot have a look at that JLH circuit as originally designed, but Im pretty sure JLH used a pot where I have the resistors and was taking the output from the swiper, so that would make it the dual feedback circuit. I used resistors because I hate pots, they terrible.

Concerning current mirrors, out of curiosity I strapped on some different current mirrors with the bjt buffer. Once again the cascode CM was better, helped with the warmth and fullness of the sound maybe overdoing it a bit. Overall they were all better than std mirror except the widlar, i could hear no difference. I will sim to see the what parameters change with these different mirrors.

Sorry guys for the OT.

[TimS]

Here is a high performance buffer design by Andrea Ciuffoli.

http://www.audiodesignguide.com/Head...dphoneamp.html

Cheers

Tim

[TimS]

I simulated the ACD design and it is pretty good, it has a nice soft clipping near the rails. It will probably have a nice warm sound.

[h_a]

Thanks for the link!

This super-linear circuit certainly looks interesting; however it adds complexity. Again the question wether simpler is better or more complex with a tenfold decreased THD is better.

Have fun, Hannes

[TimS]

Looking through the article on Andrea's site and was amazed at the complexity of the Pioneer design.

Though after simulating the design it is not actually a buffer, it does have gain, but still an interesting design nevertheless.

Cheers

Tim

[peranders]

Originally posted by TimS
I simulated the ACD design and it is pretty good, it has a nice soft clipping near the rails. It will probably have a nice warm sound.

The warm sound comes from even harmonics and it sounds hardly good when the amp is driven into clipping. Since there are BJT's involved you'll have most likely odd harmonics.

[TimS]

I agree Peranders,
the sudden onset of hard clipping can be very bad for your speakers, but with moderately loud music there will be transients that will briefly clip an amplifier.
But I believe the clipping performance of an amplifier (especially power amplifiers) is a important criteria for it to sound good:
Very Good:
Soft Clipping
Good:
Hard Clipping - Until it clips then it sounds bad.
Bad:
Signal Inversion when clipping
Very Bad:
Unstable when clipping

I was comparing the clipping performance of the ACP circuit to a balanced valve circuit. (The attached waveforms)
Valve amplifiers tend to sound louder than what they really are because of the soft clipping but nobody complains about, what must be, the large amounts of distortion.
So I believe the ACP amplifier will sound good while clipping too.

Also I am not totally sold on the idea that the warm valve sound is solely caused by second harmonic distortion.
I am starting to believe that humans don't hear sound like an FFT spectrum based instrument but I believe they hear the shape of a sound.
This is why some measurements don't always correlate with what people hear because the measurements are done in the wrong domain.
The flip side to spectral analysis (which states that any waveform is made up of a mixture of sine waves) is wavelet analysis (which states that a waveform is made up of a mixture of wavelets) and I believe the latter would be better at describing how a human hears, but if there was a suitable measurement tool.

But don't get me wrong the FFT is a very useful tool, a linear amplifier is a very good amplifier, but the linearity and flat and wide frequency response aren't the only design boxes that need to be ticked for your design.

Cheers

Tim