Aleph Clipping Mod

[Allen Newby]

The Aleph controlled current source is a brilliantly simple and effective circuit which provides the active load in the output stage of several Pass designs. The quiescent current in the output stage is set by a single sense resistor and is controlled by a bipolar transistor in a feedback loop. The current is modulated by feeding an audio signal into the base of the control transistor.

The collector load of this transistor is a conventional bootstrap type, with the mid point of two resistors, in series, being fed from the output via a large electrolytic capacitor.

At low to moderate signal levels, the bootstrap node nicely tracks the amplifier output and the gate voltage of the active load mosfet will be a bit higher than the threshold voltage at around 4.5 volts +/- a few hundred mV ac signal. As such, the active load mosfet is operating normally in the saturation region.

If you examine the output voltage, and the gate voltage of the active load mosfet at higher signal levels, you can see that as the amplifier output swings up towards the positive supply rail, Vds falls towards zero whilst Vgs rises way above 4.5 volts.

During these peaks, the active load mosfet is pushed into the linear region and behaves like a resistor of a few milli Ohms. The transitions between normal operation and the lnear region are abrupt and result in instability which takes 100 µS or so to subside. Relative to the 20V nominal mid point, the distortion in the output is noticeable at a positive swing of about 16 volts. (Whether it is audible is another question!?)

A 5.6V zener diode clamp placed across the bipolar transistor prevents the instability:
[IMGDEAD]http://www.pmeweb.co.uk/Audio/images/ZenClipComp_Vo.gif[/IMGDEAD]

The circuit, with zener clamp shown in red:
[IMGDEAD]http://www.pmeweb.co.uk/Audio/images/ZenV4Clamp.gif[/IMGDEAD]

These observations have been with regard to Zen V4 and LTPZen, see thread http://www.diyaudio.com/forums/showthread.php?s=&threadid=9459
but should be applicable to other Aleph CSS designs.

Maybe, the amplifier should not be driven so hard, however, occasional transients may be hard to avoid and the modification is both simple and effective.

For further details: http://www.pmeweb.co.uk/Audio/Aleph.html

 

[Nelson Pass]

Well that's clever. I'm always flattered when people
put thought into one of my circuits.

 

[wuffwaff]

Yes!



this could be the sollution to my Aleph-X problem (distortion starting at around 20V, high current)

Thanks!

William

 

[jupiterjune]

Allan-

Have you done any listening comparisons between the modified and stock circuits? I would be interested to hear your impressions. edit--upon reading this closer, I see this has been implemented in an LTPZen4, perhaps not in an Aleph yet.

I couldn't figure out how the first link you attached related to this?????

I'm always flattered when people put thought into one of my circuits.

We are always flattered and appreciative when you take time to answer our often thickheaded questions!
can't wait to hear about this A40 followup....

JJ

 

[EUVL]

Allen,

I am not sure whether I follow your arguments, so please bear with me for a couple of stupid questions.

You argued that at high voltage amplitudes and swing towards positive rail, the Aleph current source is switched full on. Vgs goes high, and at some point, Vds goes below say 1V, and the current source MOSFET(s) become a voltage controlled resistor (of a few milliohms).

Good arguement. I can imagine that happening.
What I cannot understand is then :

a) why would the "knick" occur at the falling edge of the clipped signal ?
b) how does the ZD cures the problem ?

The ZD limits the voltage Vgs of the current source MOSFET + Id x Rsource. As such, if anything, it limits max current of the current source.

I cannot see how it cures the Vds problem, where you probably have too much rather than too little current (with the MOSFET turning into a milliohm resistor).

Also, if we are in a situation where the load has low impedance (say 2 ohm speakers), then the ZD may clip the max current of the Aleph current source long before a voltage clip.

Maybe you could enlighten me on that ?


Thanks,
Patrick

 

[Nelson Pass]

a) why would the "knick" occur at the falling edge of the clipped signal ?

b) how does the ZD cures the problem ?

The ZD limits the voltage Vgs of the current source MOSFET + Id x Rsource. As such, if anything, it limits max current of the current source.

I cannot see how it cures the Vds problem, where you probably have too much rather than too little current (with the MOSFET turning into a milliohm resistor).

Also, if we are in a situation where the load has low impedance (say 2 ohm speakers), then the ZD may clip the max current of the Aleph current source long before a voltage clip.

The capacitance of the overdriven circuit will keep it from recovering
rapidly from the clip, so you get the overhang.

True, this diode will also tend to limit the + output, and it might
need value adjustment depending on the Vgs of the output
device and the current into the load.

Of course, stock Alephs aren't usually driven into 2 ohms, as they
have poor performance when they leave Class A.

 

[Allen Newby]

Thanks to all for their comments.

JJ

The first 'Zen' amp I built was the standard ZenV4. I noticed the clipping anomaly but didn't do anything about it.

Next I started playing around with a differential input stage and came up with LTPZen. Not surprisingly, since both amps employ the Aleph CSS, this also had the anomaly, so in its latest incarnation of LTPZen, I put in the 5.6V zeners. However, because, I used the same amp case, PSU, watercooling etc, I have not done a proper listening comparison. To be honest, with the volume turned up that far, I don't think you would hear the difference!

Patrick,

As ever, I think Nelson is right. The plot below shows what the active load mosfet has to cope with:

[IMGDEAD]http://www.pmeweb.co.uk/Audio/images/ZenClip.gif[/IMGDEAD]

Choosing a zener which kicks in just above the 'normal' range does not stop the active load mosfet going into the linear region. It does however, not 'turn on' quite so hard and because it limits the gate charge, the recovery is much faster.

I have not looked at low impedance loads. I have a huge 8 ohm resistor made from stainless steel wire, wound onto a short length of drainpipe!