[Dave]

wrote in message
...

Let me say this again, at the risk of overstating the
point: a perfect 400 Hz square wave band-limited to
20 kHz will look exactly like the trace I generated and
probably like the trace you photographed. You
photographed a "perfect" 400 Hz square wave that
was band-limited to 20 kHz.

Okay?

Crystal clear. Thank you. Not so complicated after all.

I see various postings in the audio newsgroups and audio sites such as
audiokarma and audioasylum, showing traces such as these:

http://www.audioasylum.com/cgi/vt.mp...bediy&m=137865

Would some type of commercial signal generator with higher bandwidth have
been used to generate these waves?

[[email protected]]

On Mar 4, 11:57 am, "Dave" wrote:
wrote in message

Let me say this again, at the risk of overstating the
point: a perfect 400 Hz square wave band-limited to
20 kHz will look exactly like the trace I generated and
probably like the trace you photographed. You
photographed a "perfect" 400 Hz square wave that
was band-limited to 20 kHz.

Okay?

Crystal clear. Thank you. Not so complicated after all.

I see various postings in the audio newsgroups and
audio sites such as audiokarma and audioasylum,
showing traces such as these:

http://www.audioasylum.com/cgi/vt.mp...bediy&m=137865

Would some type of commercial signal generator with
higher bandwidth have been used to generate these
waves?

The examples in the URL you posted are typical of what
I would expect to see with non-brickwall bandwidth
limited systems with a fair amount of in-band phase
shift. Both examples the first at 10 kHz, the second at
1 Khz, show exactly the same thing: the leading edges
with their exponential response indicate a significant
high-frequency limit, and the fact that it is non-time
symmetrical indicates non-linear phase behavior. The
10 kHz and 1 kHz rise-times are essentially identical
when you correct for the differences in the time scale.

The 10 kHz looks like its a horizontal scale of 20us/div,
and it takes about 1 division to go from 10 to 90% of
the way, so its rise time there is 20 us.

The 1 kHz is more difficult, because of the slower
sweep, but we can discern that it's about 200uS/div,
and it takes about 1/10 of a division, so it's rise time
is 200uS/10 or, again, 20 uS.

Using the old rule of thumb, where one can approximate
the bandwidth of a system from its rise time, as:

BW = 0.35/Tr

where BW is the bandwidth and Tr is the rise time in
seconds. It's based on some simplifying asumptions,
such as the rolloff is a single pole (6 dB/octave), which
is almost never the case in audio amplifier. In this
particular case, the estimated bandwidth of the
amplifier from my guessing at th traces is about
18 kHz. Again, the simplifying assumptions can result
in errors in this estimate.

The 1 kHz also shows that the top has some downward
tilt, indicating that the system's response doesn't make
it to DC, not surprising assuming what we're looking at is
a transformer-coupled amplifier.

Now, it's likely that these square waves were generated by
some commercial function generator. Generally, the internal
circuitry of these devices has bandwidths on the order of
1 to 10 MHz or greater, they are non-linear in phase and thus
do not generally exhibit the time-symmetrical response of
digitally sample-synthesized square waves.

But, here's the more important point:

THERE IS ABSULTELY NOTHING THE SQUARE
WAVE RESPONSE WILL TELL YOU THAT YOU
WOULDN'T ALREADY LEARN FROM A PROPER
BROADBAND FREQUENCY RESPONSE.

So, if you have, as you already suggested, Right Mark
Audio Analyzer that works with your sound, why not use
that and forget the square wave stuff?

Why then, you might ask, are square-wave measurements
used?

Because way back when, when it was difficult, time-
consuming and tedious to measure frequency response
directly, square waves provided a shortcut to intimating
the amplifier's response, though there was no precise
means of interpreting the results in a straightforward,
deterministic and objective way. Just look at the hedging
I went through trying to derive the bandwidth from the
rise time: in order to precisely derive the bandwidth,
I have to really know the frequency response, and
the frequency response tells me the bandwidth directly
anyway. Once I know the bandwidth, I don't need to
measure the rise time to know the bandwidth, if you
catch my drift.

Use RMAA and be done with it.

[Dave]

wrote in message
...
But, here's the more important point:

THERE IS ABSULTELY NOTHING THE SQUARE
WAVE RESPONSE WILL TELL YOU THAT YOU
WOULDN'T ALREADY LEARN FROM A PROPER
BROADBAND FREQUENCY RESPONSE.

So, if you have, as you already suggested, Right Mark
Audio Analyzer that works with your sound, why not use
that and forget the square wave stuff?


Point taken. I'll grab some 5K pots, use the scope to set the output level
below 2V and see what RMAA has to say. Thanks very much, I've learned
something.

I didn't know about RMAA when I read the post for the first time, as
evidenced by my questions about it and the technical issues involved in
interfacing the computer's soundcard input. I figured square wave analysis
could give me, as you said, "a shortcut to intimating the amplfier's
response".

As an aside, do you happen to know of any PCI signal generator cards on the
market for reasonable prices? I'd like to have a full-function, high
bandwidth signal generator, and if I could do it with a card in a computer
vs. another large box with knobs and displays, that'd be all the better as I
have limited space.

Dave

[[email protected]]

On Mar 4, 2:00 pm, "Dave" wrote:
wrote in message
As an aside, do you happen to know of any PCI signal generator cards on the
market for reasonable prices? I'd like to have a full-function, high
bandwidth signal generator, and if I could do it with a card in a computer
vs. another large box with knobs and displays, that'd be all the better as I
have limited space.

Try a google search for "PCI signal generator". Don't stop there,
try USB stuff as well. A lot more available I suspect.

And you need to define "reasonable price."

[Dave]

wrote in message
...
On Mar 4, 2:00 pm, "Dave" wrote:
wrote in message
As an aside, do you happen to know of any PCI signal generator cards on
the
market for reasonable prices? I'd like to have a full-function, high
bandwidth signal generator, and if I could do it with a card in a
computer
vs. another large box with knobs and displays, that'd be all the better
as I
have limited space.

Try a google search for "PCI signal generator". Don't stop there,
try USB stuff as well. A lot more available I suspect.

And you need to define "reasonable price."

Less than $100 would be ideal. There's not much out there. I found a
couple which looked promising, but they were $800 and up. HP/Agilent have
PCI cards, those cost lots, and standalone used signal generators seem to
start at about $200.