"Ian Smith" wrote in message
om

"Arny Krueger" wrote in message
...

These really aren't digital amps. They are switchmode amps. The
buggar in just about all of them is the analog low-pass filter at
the output, which causes the output impedance to rise over the top
octave or two of the audio band. Therefore, they are somewhat
sensitive to speaker impedance variations in the same range.

They also tend to behave vastly different when they clip. Some pick
up a lot of high frequency trash in the last dB of dynamic range
before clipping.

I wonder what a digital amp is if it isn't a switchmode amp?

You've missed the point.

Yes the
output filter increases the output impedance, which can be assisted
with a zobel network.

A zobel is a parallel network. In order to reduce the output impedance of an
amp substantially, it has to substantially short out the output of the amp.

They can still achieve lower output impedance
than a lot of valve amps, so what's the bother?

Nice job of damning switchmode amps with faint praise.

The way they behave when they get near clipping is up to the designer,
just like an analogue amp.

I can't imagine a designer willfully designing in the sort of bizarre
behavior I've seen in some cases.

Some digital amps can do some things better than many analogue amps.

You've got the whole rest of the post to list what they are.

They can take a CD's input digitally,

Ironically, very few of them implement digital inputs, and many that do, do
so with a regular DAC chip.

they can get better SNRs and full power THDs.

Really? Got a real-world example?

"Arny Krueger" wrote in message
Tripath :
These really aren't digital amps. They are switchmode amps. The buggar in
just about all of them is the analog low-pass filter at the output, which
causes the output impedance to rise over the top octave or two of the audio
band. Therefore, they are somewhat sensitive to speaker impedance variations
in the same range.

Tripath class 'T' is quite a bit different from a PWM switchmode amp :
http://www.tripath.com/downloads/an1.pdf

"Instead of using PWM, Tripath Class-T amplifier processors use
proprietary algorithms and techniques to create the modulation that
drives the switching transistors. A Class-T amplifier's processors
learn the characteristics of the output transistors. Then, based on
the analog input signal, they switch the output transistors with
exactly the right timing to eliminate Class-D PWM problems:
transistors not being perfect switches, ground bounce, output
transistor mismatches, dead-time distortion and residual energy from
the oscillator in the audio band. The result is a high power
efficiency, audiophile-quality audio amplifier – a Class-T amplifier.
If one were to compare the waveform before the output filter of a
Class-D PWM amplifier to a Tripath Class-T amplifier, some significant
differences would be evident. The waveform for a Class-D PWM amplifier
would be a pulse-width varying digital signal at the fixed, 100-
200kHz, frequency of the triangle wave generator. The waveform for a
Tripath Class-T amplifier would be a complex digital waveform of
varying frequency. A Class-T amplifier
switches the output transistors in a fashion similar to spread
spectrum technology, at a varying rate up to 1.5 MHz and averages
600kHz to 700kHz.

They also tend to behave vastly different when they clip. Some pick up a lot
of high frequency trash in the last dB of dynamic range before clipping.

May want to check this out :
http://www.tripath.com/downloads/an6.pdf

"The output characteristic shown in Figure 1 is normal for all Tripath
amplifiers and is a feature of the Tripath unique Digital Power
Processor (DPP). The near rail to rail swing of the amplifier output
results in overflow conditions within the DPP. These overflow
conditions are periodically reset resulting in the observed output
characteristic. As the occurrences of this overflow condition increase
(by increasing the input signal level), the DPP intelligently outputs
an indication on the OVERLOADB pin. Further signal increases will
result in higher distortion as the amplifier is pushed towards
clipping.
System designers may use this OVERLOADB signal to actively reduce
pre-amp gain to avoid clipping. Alternatively an indicator may be
designed to notify the listener that clipping is occurring."

In article , Mike
wrote:


Tripath class 'T' is quite a bit different from a PWM switchmode amp :
http://www.tripath.com/downloads/an1.pdf

"Instead of using PWM, Tripath Class-T amplifier processors use
proprietary algorithms and techniques to create the modulation that
drives the switching transistors. A Class-T amplifier's processors learn
the characteristics of the output transistors. Then, based on the analog
input signal, they switch the output transistors with exactly the right
timing to eliminate Class-D PWM problems: transistors not being perfect
switches, ground bounce, output transistor mismatches, dead-time
distortion and residual energy from the oscillator in the audio band.

The last phrase is an interesing claim. (See below) I wonder what frequency
range they regard as the "audio band"? With people pushing SACD and DVD-A
I'd assume they'd have to assign a higher frequency for this nowdays than
20kHz...

[snip]

The waveform for a Tripath Class-T amplifier would be a complex digital
waveform of varying frequency. A Class-T amplifier switches the output
transistors in a fashion similar to spread spectrum technology, at a
varying rate up to 1.5 MHz and averages 600kHz to 700kHz.

This is the potential problem that concerns me. The method means that power
is nominally present in a 'wash' of anharmonic components - mostly I
suspect in the few 100 kHz to few MHz region. This must be blocked from
reaching the speaker. Hence the method requires an output filter that will
pass the "audio band" with no significant loss, and no significant effect
upon the amps output impedance, yet strongly reject the ultrasonic wash.

The few measurements I have seen make me suspect that the existing amps
have an output impedance in the 5 - 20 kHz range high enough to alter the
resulting frequency response in way that depends upon the speaker impedance
properties. Thus the result may be like a conventional amp with an
excessively large output inductor. The standard network I've seen quoted
for use with these amps (a inductor-cap-inductor 'tee') may give a resonant
peak at HF into some loads as well.

The measurements I have seen, and the data I have seen from the makers do
not clarify the actual total ultrasonic wash power level that reaches the
speakers under normal conditions of use. I'd love to know this as there are
concerns if it is too high. This is complex as it is signal and loudspeaker
dependent. Hence being OK with test sinewaves into an 8 Ohm load doesn't
really cover it adequately in my view.

With sources like SACD and DVD-A the source may also be supplying
significant amounts of signal 20kHz. I wonder if this may find it way into
the power amp, and if so, how it might then interact with the switching
patterns the power amp employs. Again, I have seen no measurements or
analysis of this, but I wonder if it might lead to effects in the 'audio
band'...

Hence my personal view at present is that 'digital' power amps are a neat
idea, but I am not clear that - as yet - I'd prefer one to a conventional
design. There are too many questions which I've not yet seen answered.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html

"Andy Evans" wrote in message


Digital amps are coming down in price steadily - HiFi+ reviews the
latest at £800. all Tripath based so far. I heard one of the
iterations at the HiFi show a couple of years ago. Limited programme
material, but sounded bloody good. So exactly how high-end are they
right now? Can anybody that has heard the latest crowd give any sonic
comparisons with audiophile SS and valve amps? Andy

These really aren't digital amps. They are switchmode amps. The buggar in
just about all of them is the analog low-pass filter at the output, which
causes the output impedance to rise over the top octave or two of the audio
band. Therefore, they are somewhat sensitive to speaker impedance variations
in the same range.

They also tend to behave vastly different when they clip. Some pick up a lot
of high frequency trash in the last dB of dynamic range before clipping.