The fact that noise is largely random in nature is precisely "here or there"; that is what distinguishes it from distortion in terms of theory, measurement and listening. Distortion, by definition, is signal-related. Noise is not. There is a small overlap if noise is signal-modulated. In order to understand the join, you first have to understand the dichotomy.
The pro world sees things differently it seems - from the Rane Pro Audio Reference, http://www.rane.com/par-d.html:
Now it is me who asks for the link . But not any link to THD+N vs. frequency plot. No need to see noise higher than distortion at low power.
So what is better - the design that has such low non-linear distortion that it rises from noise no sooner than somewhere at 10W, with noise level about -100dB re 2.83V unweighted over 20kHz.
OR
The design with noise level about -80dB re 2.83V unweighted over 20kHz, much higher non-linear distortion that dominates from some 100mW above noise, but is claimed to have superbly clean first watt?
Yes, thank you. But the 2nd is the one that would have a THD+N plot rising with amplitude and at a first look it would have lower distortion at low power.
Pavel, I'm guessing tht these days you don't see much of mainstream audio inside.
MOST of it is underbiased, and especially the HT, 5 or 7 channel gear. With those flimsy "heatsinks", actually aluminium plates with soldered-on fins about 0.5 mm "thick", how else could it be?
Of course, it also depends on what one consideres as undrbiased, but to me, circa 25 mA per single tranny is definitely underbiased.
But they did drive the prices down ...
A pint of cider would be lovely, two even more so, but only assuming we have like 2 hours to discuss matters. Before lunch, that is.
I would think that if we start being the problem, I've done my homework. I see it this way - with 2.2W/4 Ohms, even someone with 4 Ohm speakers will likely spend most of his listening time in pure class A domain. I've read somewhere that in reality most of us listening in our rooms spend over 90% of that time using 1W or less. In my case, even assuming the minimum impedance of my speakers, which is 6.5 Ohms, I have like 3.5W of pure class operation, which I assume will raise the ante from 90% to over 98%.
This more or less corresponds to Otala's 600 mA of bias in his old amp of 1974. He postulated that the class A region of operation should extend to -17 dB of nominal power, so his amp ran at 2.9/5.8W into 4/8 Ohms in pure class A. In its day, it simply sounded WAY better than most, a relatively low power level of 25/50W into 8/4 Ohms notwithstanding.
As for the sound turning "muddy", rest assured, exactly the same thing will happen to a BJT output stage if you overdo it. This is the point at which you need to turn it down a bit, because you won't be gaining anything, but will start losing definition and focus. Because there is no single set limit to the emergence of this effect, I always say 110-130 mA per output trannie is about right, but it does vary, and you do have to listen for it yourself.
We should always bear in mind that this is a class AB topology we are using to force it into a higher bias regime. Therefore, there will be a limit.
I agree, but I thought we were speaking about high end here, not about commercial low-fi. I use as a rule of thumb 25mV per Re for BJT output devices, that means for 0.22 ohm Re the idle current per pair is about 113mA. My old amp digged out from amplifier grave box has 2 pairs of MJL21193/4, so it is biased at about 226mA total.
Well, not really high end only, rather high end as well.
Could I trouble you with a request? The amp you mentioned, could you post its schematics? I am very curios.
Also, perhaps you could post your new ->BJT<- work, or give a link for it? If you have any, of course, you may have gone over to the MOSFET camp.
There's also an excellent treatment of the subject in Bob Cordell's book, which boils basically down to what jcx posted above.
Jan
It is more or less a predecessor of this
PA4 power amplifier
but only a functional sample with 2 pairs of output devices, only 2 x 27V supply voltage, small heatsink etc. This first sample had sub-optimal pcb and it was completely re-designed, regarding ground returns of supply rail bypass 220uF capacitors.
Please feel free to visit my photo gallery here
Power amplifier 2 x 300W - My Photo Gallery
Dvv . I think a 130 mA current gives about 0.27 watts class A or did I misunderstand JLH ? As you use triple outputs that is about 0.8 watts @ 8R ( 0.4 W @ 4R ) . That would be very OK .
My little amp's 0.7A to meet a 8 watt criteria . 1.15 A in Self's book is a comfortable 20 watts .
My little amp's 0.7A to meet a 8 watt criteria . 1.15 A in Self's book is a comfortable 20 watts .
Last edited:
130mA of output bias for a peak ClassA output of 260mApk does indeed give 0.27W into 8r0 of maximum ClassA.
Using a doubled output stage give 4times the ClassA power.
Using a 3pair output stage gives 9times the ClassA power.
using a 10pair output stage gives 100times the ClassA power.
9times 0.27W is 2.4W of ClassA
Using a doubled output stage give 4times the ClassA power.
Using a 3pair output stage gives 9times the ClassA power.
using a 10pair output stage gives 100times the ClassA power.
9times 0.27W is 2.4W of ClassA
the Cordell feedback thread had a number of pages on Baxandal "harmonic multiplication"
I did my usual simmed demonstration for a 10% "pure square law" gain device:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-287.html#post1875719
here is that sim reworked with ~ 1% open loop 2nd order distortion, and with 40 dB loop gain negative feedback
you can see the new 3rd harmonic in the yellow fft harmonic peaks, it is created by the negative feedback around the "pure square law" behavioral I source
@ -108 dB this new distortion product would be at or below the human hearing threshold in a anechoic chamber for most loudspeakers < mid 90s dB SPL/W sensitivity at the "First Watt" drive level under discussion
the added E1 gain block in the feedback circuit only uses Av=20 because I credited the feedback loop with the gain of 5 from the output device since the closed loop gain is 1 for the feedback circuit
I did my usual simmed demonstration for a 10% "pure square law" gain device:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-287.html#post1875719
here is that sim reworked with ~ 1% open loop 2nd order distortion, and with 40 dB loop gain negative feedback
you can see the new 3rd harmonic in the yellow fft harmonic peaks, it is created by the negative feedback around the "pure square law" behavioral I source
@ -108 dB this new distortion product would be at or below the human hearing threshold in a anechoic chamber for most loudspeakers < mid 90s dB SPL/W sensitivity at the "First Watt" drive level under discussion
the added E1 gain block in the feedback circuit only uses Av=20 because I credited the feedback loop with the gain of 5 from the output device since the closed loop gain is 1 for the feedback circuit
Attachments
Last edited:
- Status
- Not open for further replies.