"Iain M Churches" wrote in message

"Arny Krueger" wrote in message
...
"Iain M Churches" wrote in message



Hi Arny. Good to be discussing something of mutual
interest with you. I have rather lost track of whether
you though Zo and DF were important or unimportant
factors in the evaluation of a valve amplifier.

Can you please confirm your position?

I think that Zo is an important parameter for every power
amp that ever drives a load whose impedance is
frequency-dependent.

I made up a resistive box of 4R and 12R in series, with
a switch across the 12R to short it out. I then
measured the voltages across R1 and R2 (calling them V1
and V2)

(snip. my methodology)

That's one way to do it. Pretty classic and of course
very good as far as it goes.

Please elaborate. If there are shortcomings in this method
I would be interested to hear of them. I know you have
considerable experience in this field.

The method I use is to measure the output impedance of the
power amp by treating it like it is just another impedance.
I use a second power amp plus an isolating resistor as the
signal source for the measurement. I measure the voltage
across the UUT, as well as the current flowing into it.
SpectraLab software calculates the quotient of the two
signals in the frequency domain in real time to produce a
plot of impedance magnitude and phase versus frequency.

Since the results are based on a quotient, the procedure
works with a wide variety of signals, signal levels, and
other test conditions.

There are a number of pieces of measurement software that
perform similar calculations. Sample Champion from
http://www.purebits.com/ is an relatively inexpensive
package that does a similar calculation as described in this
appliation note: http://purebits.com/appnote16.html .


I know of three static methods (there may be more)
The original classic way seems to be to inject a low
level signal into an amp, and then measure the output
voltage with no load. Then add a variable wirewound
resistor and adjust it until the Vo falls by 6dB. This
resistance is then equal to the
Zo of the amp. This is a simple method, but assumes that
all amplifiers are stable into an open circuit.

I am told that various companies measure Zo at 400Hz
700Hz or 1kHz.

That's typical. Those frequencies aren't far apart for
an amplifier test. If you want to see Zo vary, go to the
frequency response extremes.

I did that, but into my 4 and 16 Ohms resistive load.
There was very little variance. I feel that the same
measurements into a real-world load would yield
different results (more of that later)

If the amplifier that I use as the signal source has a low
output impedance and is robust, when it sources the test
signal to the UUT though a loudspeaker or loudspeaker-like
load, then the UUT's response to working with the given load
becomes part of the test.

I got very similar figures at all these
frequencies. I also found that there was no variation
with power either which confirms what you have stated.

Thanks for the confirmation, as your methodology is
classic.

Here are examples of some of my Zo measurements:

http://www.pcavtech.com/pwramp/macrot-5000VZ/zout.gif

http://www.pcavtech.com/techtalk/wir...4BSTleft-z.gif

http://www.pcavtech.com/techtalk/wir...ST-left-ph.gif

The methodology I use is dynamic. I can change the
operating conditions and the plot data on my PC screen
in real time. I can test Zo using loudspeaker and
loudspeaker-like loads. I can use actual recordings of
music to drive the test rig. I've collected a lot of
informal data about how Zo varies under different
operating conditions. For good amps, it doesn't.

I know of a static third method, which involves using
both channels of a stereo amp. The dummy load is
connected between the output of one channel which is
driving with a smallish
input signal, and the output of the other channel which
has no signal.

That's the starting point for the procedure I use.
at its input.

I would be interested to hear about your dynamic testing,
Arny. I have access to a broadcast facility lab.

I thought your PC plots were good. I use Excel and enter
the data into two columns and let the software plot the
chart,
in four colours on log paper which looks even better.

I am interested in what you term loudspeaker-like loads.
By this I presume you mean simulated loads. I am in the
process of building one of these from the circuit
published in
Stereophile and designed by Ken Kantor

http://www.stereophile.com/reference/60

It represents a smallish two way speaker. I am interested
also to find a schematic that simulates a large studio
three way design.

Lets's forget Mikey falling down for the moment.
I have renamed this part of the thread.

Cordially,

Iain