Crossover Distortion, the truth

Disabled Account
Joined 2012
But in a larger sense none of these circuits like NP's stasis, Hawksford, and other error correction schemes have much of a presence in current DIY. Many folks are perfectly happy with a Self blameless with modern devices.

:) :cool: I very much agree.

I have promoted this for a long time. especially when trying to get ultra low distortion at all freqs and levels.

Damir (DIY) took up the challenge and his circuit uses EC and CFA and has my complete "wish list" of performance.

It was also suggested as solution for test equipment needing ultra low distortion.

It just seems the practical way to go and once a few more people do it, might become mainstream.


THx-RNMarsh
 
Last edited:
Sorry Richard, just small correction. I have designed some power amps with EC, but this one does not use it. It uses unique compensation, I called it OITPC (Output Inclusive Two Pole Compensation). It is TPC but with two input legs, one capacitor connected to the VAS output, other one connected directly to the output. In this way I was able to get more than 80 dB of Loop Gain (feedback) at 20 kHz and ULGF moved behind 3 to 4 MHz.
In this way THD was even better than if I used EC (EC complicates the compensation).
I may opinion if one wants to design non global NFB amp then EC is needed.
Damir
 
I may opinion if one wants to design non global NFB amp then EC is needed.
Damir

That was my impression essentially gobs of feedback. An EC output outside the global feedback is an interesting exercise but it would be difficult to expect ppm distortions everywhere. Once you're past -80-90db low order across the board my opinion remains that no compelling evidence has been presented that the distortions alone have any bearing on the sound.

BTW it's hard to make a general purpose amplifier (robust against difficult loads at any closed-loop gain) with output included compensation, but they do exist.
 
Last edited:
Member
Joined 2011
Paid Member
BTW it's hard to make a general purpose amplifier (robust against difficult loads at any closed-loop gain) with output included compensation, but they do exist.

A lot of the discrete transistor opamps from your BurningAmp talk, had output included compensation. To be fair, none of them could deliver even 1 watt into an 8 ohm load. And I spoze they don't meet your "general purpose" requirement since the user had to change the values of the compensation components, to run at low closed-loop gain.
 
Last edited:
A lot of the discrete transistor opamps from your BurningAmp talk, had output included compensation. To be fair, none of them could deliver even 1 watt into an 8 ohm load. And I spoze they don't meet your "general purpose" requirement since the user had to change the values of the compensation components, to run at low closed-loop gain.

I was referring to some IC op-amps with no external comp pins. There are a few but I forget which. The ADSL drivers don't count because that is a specific captive application and the design can include the recommended external circuit. The point is it includes the crossover inside the loop but the crossover is also responsible for output impedance modulation so reactive load drive becomes more of a problem (the fact that an EF looks inductive at high frequencies does not help either).
 
Had built the undermost circuit. Inserted 2 mF at the output and driven any fullrange, > 90 dB;-)))

Gegentaktverstarker
 

Attachments

  • GTV415.npn.pnp.gif
    GTV415.npn.pnp.gif
    10.1 KB · Views: 276
There are valid reasons on why we need GNFB
It all depends on the speed of the negative feedback, on the signal delay, in short, on the group delay. If the group delay is less than 100 ns (and preferably less than 50 ns), then Feedback effectively copes with switching distortions and there is no need to use class A. In reality, the vast majority of amplifiers produced by the industry have a group delay from 200 ns (at best) to 2 μs or more .
 

Attachments

  • RL1993-11_Bode.png
    RL1993-11_Bode.png
    37.7 KB · Views: 128
  • RL1993-11_20kHz-spectr_vs_Bias.png
    RL1993-11_20kHz-spectr_vs_Bias.png
    38.6 KB · Views: 134
  • RL1993-11_10kHz-distortion_vs_Bias.png
    RL1993-11_10kHz-distortion_vs_Bias.png
    19.1 KB · Views: 131
Let's measure the crossover distortion in a near-reality mode. To do this, we mix a 20 kHz signal with a 100 Hz signal in a ratio of 2:3 and measure the distortion products at different quiescent currents: 8, 32 and 63 mA.
 

Attachments

  • RL1993-11_Bias=8mA_Cross-Over.png
    RL1993-11_Bias=8mA_Cross-Over.png
    18.2 KB · Views: 85
  • RL1993-11_Bias=32mA_Cross-Over.png
    RL1993-11_Bias=32mA_Cross-Over.png
    15.7 KB · Views: 91
  • RL1993-11_Bias=63mA_Cross-Over.png
    RL1993-11_Bias=63mA_Cross-Over.png
    15.2 KB · Views: 82
  • Like
Reactions: 1 user