Leaving Class A? Martin Colloms vs Nelson Pass

[BrianEno]

Knowing that the subject of Class A biasing is an important issue at this forum and at PassLabs, I conducted an experiment with my Class A amps, the Threshold SA/1's.
Before that I read some interesting stuff like the article 'Leaving Class A' from Nelson Pass about the benefits of enriching the output current of amps with a high bias current. Very convincing stuff indeed.
There is a thread on this forum called: Leaving Class A, and the article can be downloaded from the PassLabs website of course.
On the first page a list of current amplifier models is presented to show at which wattages the specific models shift from Class A to Class B.
The XA160.5 for instance has a nominal rating of 160 Watts, is biased with 3.2 Amps and leaves Class A at 328 Watts peak.
According to this article Push-pull amplifiers (like the XA160.5) operate in Class A mode at twice their biascurrent. So the figure of 328 Watts peak is obtained by: square(3.2 x 2) x 8 = 327.68.
The RMS value is half of that namely 160 Watts.

But their are different opinions how to establish the amount of Class A output for a given biascurrent.
In the Hifi News and Record Review of August 1991 the English reviewer Martin Colloms (He also reviewed equipment in Stereophile) states in a feature review about a Threshold SA3.9E amplifier that: "..semi-class-A mode, biased to a standing current of 2A. This allows for a fine class-A output of 16W. As in recent Krell reviews, I dispute Threshold's qualified claims for class-A operation."
Martin Colloms makes the following equation: square(2A x squareroot2/2) x 8 = 16 Watts instead of the claimed 60 Watts Class A for this particulary amp.

I know Martin Colloms is a major fan of Krell designs but that does not make his claim less credible.

So to test both claims I conducted an experiment with one of my SA/1's
According to the manufacturer the SA/1 should put 160 Watts Class A power into a eight Ohm load.
Setup:
- One 10A Variac (230 Volt line)
- An ammometer.
- Two Multimeters
- Two huge wirewound resistors (12 inch in length) each rated at 5 Ohms with several tappings for lower values.
I configured the resistors in series to get a 7 Ohm value.
I checked that with one of the mm's of course.
Connected one of the MM's in True RMS mode with the load for reading the outputvoltage and the other MM with one of the 1 Ohm emitterresistors in the amp.
Before that I measured twenty of them and they all lie within a 5% range.
The one I took measured 1 Ohm and closed the top with the leads connected to it as specified by Threshold's standard amplifier bias procedure.
I connected the 7 Ohm load to the outputconnectors switched the amp on and brought up the AC line to 220 Volts as the original AC Voltage for this amp.
I waited three hours before the AC power intake settled at 1.2A and the mm over the 1 Ohm resistor read 75 mV. That means a 3A bias (40 output devices times 0.075/1 = 3) which I have checked before at one of the railfuse connectors (2 x 1,5A).
The 75 mV and 3A settings are on spec for this amp.

Then I proceeded with playing a 1 kHz signal through it and level it with the volumeknob of my preamp.
First I turned up to the -10 dB setting on the panelmeter of my SA/1. The ammometer registrating the AC Voltage and the MM registrating the voltage over the emitterresistor showed their initial values. Then I turned up to -6dB and nothing happened either. Then at -3 dB both values increased just a little bit indicating that at that point the amp was leaving Class A.
At the 0 db setting both readings definitely increased significantly.
I went back to the - 3dB setting and measured over the 7 Ohm load a 16 Volts output.
This means with Ohm's Law an output current of 16/7 = 2.3 A, and an poweroutput of approx. 36 Watts.

Of course I was somewhat disappointed that my 160 Watt Class A amp could only boast 36 Watts of Class A power for real.

Has Martin Colloms a legitimate claim to dispute the Class A claims of Threshold and probably also the ones made by PassLabs?
He would state that the 3.2A standing bias for the PassLabs XA160.5 translates in a 40 Watt Class A output. My experiment would confirm that specification it seems.

Other thoughts?


By the way; The amp could crank out 40 Volts easily though at 8 or 4 Ohms. At those readings you could clearly hear the resistors producing the 1 kHz feeding signal.

[Nelson Pass]

If your amplifier shows a 3 amp bias, then it will enter
cutoff at 6 amps peak, which into 8 ohms is 6 X 6 X 8 =
288 watts peak which becomes 144 watts rms with a
sine wave. That you measure slightly more draw at -3dB
is not particularly important.

[jacky]

Brian, if you measured 75 mV over the 1 ohm, the standing current is only 75mV/1ohm *20 = 1.5A not 3A since the 40 output devices are splitted into 20 on the positive and 20 on the negative power supply rail. This explains why in your experiment your amp is leaving class A at about 36 W.
Martin Colloms' calculation is only true for single ended design while Nelson Pass' calculation is true for push pull design.
HTH

[Zen Mod]

while Papa is known well as notoriously not caring for numbers and specs ( at least publicly and in his own Service Manuals - that all probably because all job is already done ,when SM must be made) ...... I also have slight impression that he made few more amps than Mr. Colloms .....

in any case - conducting your own tests is positive thinking in work ......


edit:

everything is in perspective ....... Papa mentioned one crucial word "cutoff" ;

for him - that's where amp leaves A and goes to B

AB is probably .......... who cares

[BrianEno]

Gathering from your replies it's only a matter of definition?

I know the difference between a complementary design (Push Pull) with PNP and NPN output devices sharing the load and single-ended ones. I know what cutoff and saturation means.
The meterreadings at the front of the amp are not really important here. I believe those Modutecs are calibrated for peaklevels.
I also know that the 75 mV reading at the emitterresistors counts for 1.5 Amps at the Positive railvoltage (PNP devices) and another 1.5 Amps at the Negative railvoltage (NPN devices). That constitutes 3A bias in total for one channel.
When the PassLabs XA160.5 Class A Push Pull design is specified to have a 3.2A bias per channel does the mean, along the lines of Jacky's reply:
1. That this only counts for the negative or positive supply;
2. or both of them (2 x 1.6)?

If the first case is is true I should increase my bias for the SA/1 to a reading of 160 mV to get a 3.2A bias each for the Negative and Positive powersupply, 6.4A in total for having a 160 Watts RMS Push Pull Class A Poweramp at eight Ohms.
In the second case the XA160.5 would be a 80 Watts RMS Class A poweramp.

So at the moment my SA/1 is underbiased to be a 160 Watts Class A amp.

With 3 Amps bias in total in my case can sombody tell me then what is happening at the measured outputvoltage of 16 Volts I over the 7 Ohms dummyload when from that moment on the voltage over the emitterresistor is steadily rising (> 75 mV) as the intake of AC Power is?
What shift is going on from that moment?
I interpreted as that the particular powertransistor stops amplifying the full sine wave and from that point onwards only increases the negative or positive portion of the cycle depending if it is a NPN or PNP device.
But that is then probably not the point the amp seizes to be a Class A design?

Can somebody explain to me at what point the amp shifts from Class A to Class B?

[Zen Mod]

Originally Posted by BrianEno

....
I also know that the 75 mV reading at the emitterresistors counts for 1.5 Amps at the Positive railvoltage (PNP devices) and another 1.5 Amps at the Negative railvoltage (NPN devices). That constitutes 3A bias in total for one channel.
..........

nope - bias/shmias is sum of current going from positive to negative side of PSU; so , in this case you have just 1A5 of Iq

ground is just reference point for amplifier ..... well , sort of .
amplifier doesn't care for ground , energetic wise - in steady condition , without signal .

so - that's just engineer's choice where other side of loudspeaker will be connected ..... example : in bridged amps loudspeaker isn't connected in any way or means to gnd ...... but to just opposite potential point , looking from one amp/half hot output to another amp/half output

leaving class A point - I'm pretty sure it's somewhere where one side ( either positve or negative ) is choked/closed/non-conductive ........

cutoff .

(first morning coffee still not finished , English isn't my native ..... just few reasons why I didn't/can't write all this in more clearer way )

[BrianEno]

Originally Posted by Zen Mod

nope - bias/shmias is sum of current going from positive to negative side of PSU; so , in this case you have just 1A5 of Iq

So that means 1.5A quiescent current.
According to Nelson Pass that would mean for a PushPull design:
square(2 x 1.5) times 8 = 72 Watts Class A peak, 36 Watts RMS?

To get to 144 Watts Class A RMS one would need a double amount of Iq:
square(2 x 3) times 8 = 288 Watts Class A peak and 144 Watts RMS.

I wonder how hot my SA/1 becomes when I turn up the bias to 150 mV on the 1 Ohm emitterresistors to get the 3A iq.
So when I measure 3A at the negative railfuse holder (fuse removed of course ;-)) then it should be 144 Watts Class A RMS?

It now idles at 43 degrees Celcius and according to the Threshold Standard Biasing Procedure of 1989 a SA/1 'non optical bias' version should idle at 42 degrees Celcius. My roomtemperature is a normal 20 degrees by the way.

Originally Posted by Zen Mod

(first morning coffee still not finished , English isn't my native ..... just few reasons why I didn't/can't write all this in more clearer way )

Know what you mean. Apart from the coffeine, tt's a major disadvantage in this technical world if you're not a native English speaker.

[Zen Mod]

like someone said - "don't use force ..... use bigger hammer"

seems that your Treshold is well biased ; according to Papa - it's hardly possible that you're hearing "BANG" in moment when amp is coming to B class .
your intellectual side can be satisfied with that math above ...... and I hope that your speakers are satisfied with Treshold ...... doesn't matter which type of Watts are coming out

[Nelson Pass]

I think the clear indicator is that your amps are underbiased,
since you are drawing 1.2A*220V out of the wall.

If you count the number of devices along the top of the
heat sink on both sides, you should divide 3.16 by that
number and that will tell you the bias figure you probably
want.

At the same time you should check the heat sink temperature
so that it doesn't exceed about 55 deg C. over the course of
a couple hours.

[BrianEno]

Originally Posted by Nelson Pass

I think the clear indicator is that your amps are underbiased,
since you are drawing 1.2A*220V out of the wall.

If you count the number of devices along the top of the
heat sink on both sides, you should divide 3.16 by that
number and that will tell you the bias figure you probably
want.

At the same time you should check the heat sink temperature
so that it doesn't exceed about 55 deg C. over the course of
a couple hours.

Two transistors on the top rows are the Stasis devices does that make any difference? So 18 transistors are the currentdumping devices.
Okay I know what you mean. 3.16 Amps must be shared by 20 devices meaning 0.158 Amps per device with a 1 Ohm emitterresitor means 0.158 mV.

I used the attached specifications from Threshold self were in a hot state on a 120 Volts line the SA/1 should have a AC intake of 2.2A. That means 1.2A for a 220 Volts line I reckon.
I thinks it's to conservative probably.

I'll try your settings and hope the thermistors on the Stasis devices will allow that.