[[email protected]]

Hello All,

I am dealing with a conundrum about distortion measurements in an
amplifier which is serving as a DUT. The architecture consists of a
two op-amp configuration that allows a single-ended input and BTL
output.
The issue arose with the difficulty of using the Audio Prescision
System II in developing a PSRR curve. Instead, the setup with the AP
consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@ sweeping 10 Hz - 20
kHz

The AP was configured to measure THD+N. So the idea is to see how the
THD+N of the amplifier is affected by power supply ripple rather than
your traditional PSSR curve.

My conerns about this test is that it is really a poor Intermodulation
Distortion test. SMPTE, DIN, CCIF from what I have researched do not
use a sweeping frequency; rather 2 tones of specified frequency. So
the test methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a setup to
measure Intermod). Am I "correct" in my thinking here?

I would also like to know if you can predict THD+N from an Intermod
measurement and vice versa.

Now, I will probably be told go Google, but I have been having a hard
time researching this topic. I am very green to audio. I hope this
is the right group to ask!

Thanks in advance.

[Don Pearce]

On Fri, 29 Feb 2008 01:28:40 -0800 (PST), wrote:

Hello All,

I am dealing with a conundrum about distortion measurements in an
amplifier which is serving as a DUT. The architecture consists of a
two op-amp configuration that allows a single-ended input and BTL
output.
The issue arose with the difficulty of using the Audio Prescision
System II in developing a PSRR curve. Instead, the setup with the AP
consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@ sweeping 10 Hz - 20
kHz

The AP was configured to measure THD+N. So the idea is to see how the
THD+N of the amplifier is affected by power supply ripple rather than
your traditional PSSR curve.

My conerns about this test is that it is really a poor Intermodulation
Distortion test. SMPTE, DIN, CCIF from what I have researched do not
use a sweeping frequency; rather 2 tones of specified frequency. So
the test methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a setup to
measure Intermod). Am I "correct" in my thinking here?

I would also like to know if you can predict THD+N from an Intermod
measurement and vice versa.

Now, I will probably be told go Google, but I have been having a hard
time researching this topic. I am very green to audio. I hope this
is the right group to ask!

Thanks in advance.

You appear to have got the whole thing confused.

First the PSRR test. Connect a single source in series with the power
supply. A 1:1 transformer is a good way to do this. Simply connecting
it "to" the power supply just shorts out the source oscillator. You
don't measure distortion, just how much signal appears at the output
of the amplifier. The PSRR is the ratio of the source signal level to
the output signal level. Don't feed any signal to the input of the
amplifier.

Now for distortion. Both harmonic distortion and intermodulation are
simply manifestations of the same thing. Since there are two signals
involved in the intermod test you can expect the measured result to be
a little different to THD, but not very. In general an amplifier that
has good THD specs will also be good for IMD. The big advantage of the
IMD test is that it yields useful results all the way up to the top of
the measurable band.

d

--
Pearce Consulting
http://www.pearce.uk.com

[[email protected]]

On Feb 29, 2:41*am, (Don Pearce) wrote:
On Fri, 29 Feb 2008 01:28:40 -0800 (PST), wrote:
Hello All,

I am dealing with a conundrum about distortion measurements in an
amplifier which is serving as a DUT. *The architecture consists of a
two op-amp configuration that allows a single-ended input and BTL
output.
The issue arose with the difficulty of using the Audio Prescision
System II in developing a PSRR curve. Instead, the setup with the AP
consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@ sweeping 10 Hz - 20
kHz

The AP was configured to measure THD+N. *So the idea is to see how the
THD+N of the amplifier is affected by power supply ripple rather than
your traditional PSSR curve.

My conerns about this test is that it is really a poor Intermodulation
Distortion test. *SMPTE, DIN, CCIF from what I have researched do not
use a sweeping frequency; rather 2 tones of specified frequency. *So
the test methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a setup to
measure Intermod). *Am I "correct" in my thinking here?

I would also like to know if you can predict THD+N from an Intermod
measurement and vice versa.

Now, I will probably be told go Google, but I have been having a hard
time researching this topic. *I am very green to audio. *I hope this
is the right group to ask!

Thanks in advance.

You appear to have got the whole thing confused.

First the PSRR test. Connect a single source in series with the power
supply. A 1:1 transformer is a good way to do this. Simply connecting
it "to" the power supply just shorts out the source oscillator. You
don't measure distortion, just how much signal appears at the output
of the amplifier. The PSRR is the ratio of the source signal level to
the output signal level. Don't feed any signal to the input of the
amplifier.

Yes, I agree. I should have been more clear. There have been several
methods used for PSRR: (1) using the AP as the generator connected via
a capacitor or (2) a transformer, or (3) a simple power amplifier as
the generator and it is connected in a summer configuration (AC+DC) to
the power supply of the DUT. In all three methods the AP was used as
the input analyzer. No input was applied to the DUT; the inputs were
AC coupled to ground. A DC bias to Vcc/2 ensures a DC return path so
this AC coupling to ground is no problem.
However, the results were not yielding expected values so another
approach was used, and that approach was to inject a ripple onto the
power supply rail with a fixed (1Vp@1kHz) input signal and measure the
corresponding THD+N. I am curious to know if measurements taken at
the output would be a valid THD+N measurement or rather it is a
(poorly implemted) intermod test since there are two sources being
mixed together.


Now for distortion. Both harmonic distortion and intermodulation are
simply manifestations of the same thing. Since there are two signals
involved in the intermod test you can expect the measured result to be
a little different to THD, but not very. In general an amplifier that
has good THD specs will also be good for IMD. The big advantage of the
IMD test is that it yields useful results all the way up to the top of
the measurable band.


Well, what I wanted to know is my test setup was as follows:

Source 1 on the input = 1Vp@1kHz
Source 2 on the supply rail = 100mVp@(10Hz to 20kHz)

The resulting output is measured using a typical THD+N distortion
analyzer. IIRC this consists of a notch filter tuned to the
fundamental, then a bandpass limiting filter, and a true RMS
detector. As above is this:

1) A valid THD+N measurement? I am being told that it is.
2) The above setup is *similar* to several Intermod standards, but not
exact. I look at the above setup as a poorly implemented
Intermodulation Distortion test. So when I have my Audio Precision
setup to measure THD+N then I suspect the measurements are
meaningless.

You say that *in general* if THD+N is good, then so is intermod. Ok,
so it is a Rule of Thumb. But can one be converted to another?

[Don Pearce]

On Fri, 29 Feb 2008 03:37:58 -0800 (PST), wrote:

On Feb 29, 2:41*am, (Don Pearce) wrote:
On Fri, 29 Feb 2008 01:28:40 -0800 (PST), wrote:
Hello All,

I am dealing with a conundrum about distortion measurements in an
amplifier which is serving as a DUT. *The architecture consists of a
two op-amp configuration that allows a single-ended input and BTL
output.
The issue arose with the difficulty of using the Audio Prescision
System II in developing a PSRR curve. Instead, the setup with the AP
consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@ sweeping 10 Hz - 20
kHz

The AP was configured to measure THD+N. *So the idea is to see how the
THD+N of the amplifier is affected by power supply ripple rather than
your traditional PSSR curve.

My conerns about this test is that it is really a poor Intermodulation
Distortion test. *SMPTE, DIN, CCIF from what I have researched do not
use a sweeping frequency; rather 2 tones of specified frequency. *So
the test methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a setup to
measure Intermod). *Am I "correct" in my thinking here?

I would also like to know if you can predict THD+N from an Intermod
measurement and vice versa.

Now, I will probably be told go Google, but I have been having a hard
time researching this topic. *I am very green to audio. *I hope this
is the right group to ask!

Thanks in advance.

You appear to have got the whole thing confused.

First the PSRR test. Connect a single source in series with the power
supply. A 1:1 transformer is a good way to do this. Simply connecting
it "to" the power supply just shorts out the source oscillator. You
don't measure distortion, just how much signal appears at the output
of the amplifier. The PSRR is the ratio of the source signal level to
the output signal level. Don't feed any signal to the input of the
amplifier.

Yes, I agree. I should have been more clear. There have been several
methods used for PSRR: (1) using the AP as the generator connected via
a capacitor or (2) a transformer, or (3) a simple power amplifier as
the generator and it is connected in a summer configuration (AC+DC) to
the power supply of the DUT. In all three methods the AP was used as
the input analyzer. No input was applied to the DUT; the inputs were
AC coupled to ground. A DC bias to Vcc/2 ensures a DC return path so
this AC coupling to ground is no problem.
However, the results were not yielding expected values so another
approach was used, and that approach was to inject a ripple onto the
power supply rail with a fixed (1Vp@1kHz) input signal and measure the
corresponding THD+N. I am curious to know if measurements taken at
the output would be a valid THD+N measurement or rather it is a
(poorly implemted) intermod test since there are two sources being
mixed together.


It depends on the amplifier, the frequency and all sorts of stuff.
What should happen is that the ripple signal you put on the power
supply simply appears at the output in diluted form. If the amplifier
is not designed well, it is possible that the ripple will instead
amplitude modulate the audio you feed in. If that happens you will see
the signal frequency plus two more sidebands at signal +/- ripple
frequency. But that has nothing to do with the actual intermod spec of
the amplifier per se. It is really better to keep these measurements
apart.


Now for distortion. Both harmonic distortion and intermodulation are
simply manifestations of the same thing. Since there are two signals
involved in the intermod test you can expect the measured result to be
a little different to THD, but not very. In general an amplifier that
has good THD specs will also be good for IMD. The big advantage of the
IMD test is that it yields useful results all the way up to the top of
the measurable band.


Well, what I wanted to know is my test setup was as follows:

Source 1 on the input = 1Vp@1kHz
Source 2 on the supply rail = 100mVp@(10Hz to 20kHz)

The resulting output is measured using a typical THD+N distortion
analyzer. IIRC this consists of a notch filter tuned to the
fundamental, then a bandpass limiting filter, and a true RMS
detector. As above is this:

1) A valid THD+N measurement? I am being told that it is.

No. Remove the power supply signal if you want a valid THD+N
measurement.

2) The above setup is *similar* to several Intermod standards, but not
exact. I look at the above setup as a poorly implemented
Intermodulation Distortion test. So when I have my Audio Precision
setup to measure THD+N then I suspect the measurements are
meaningless.

To measure intermod characteristics, both signals should be applied to
the input. Only put signals on the power supply if you are
specifically doing PSRR tests.

You say that *in general* if THD+N is good, then so is intermod. Ok,
so it is a Rule of Thumb. But can one be converted to another?

Probably, but I haven't done the maths so I can't say for sure.

d

--
Pearce Consulting
http://www.pearce.uk.com

[Eeyore]

wrote:

The resulting output is measured using a typical THD+N distortion
analyzer.

This is NOT PSRR.

Graham

[Arny Krueger]

wrote in message


I am dealing with a conundrum about distortion
measurements in an amplifier which is serving as a DUT.

What is the power level?

The architecture consists of a two op-amp configuration
that allows a single-ended input and BTL output.

The issue arose with the difficulty of using the Audio
Prescision System II in developing a PSRR curve. Instead,
the setup with the AP consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@
sweeping 10 Hz - 20 kHz

Just guessing here, but I doubt that the AP will adequately drive your load
without some additional boosting.

There is also an impedance-matching problem as power supplies usually have
very low impedances.

The AP was configured to measure THD+N. So the idea is
to see how the THD+N of the amplifier is affected by
power supply ripple rather than your traditional PSSR
curve.

A power amplifier may have several volts of various kinds of signals on its
power supply terminals when it is being driven hard. There will be some
power line related trash and there will be some signals related to the test
signal due to the finite source impedance of the power supply.

The relevant parameter is called power supply rejection. A good power amp
will usually reject 60 dB or more of any signal that tries to creep into the
output from the power supply. One usually measures power supply rejection
by injecting a test signal into the power supply terminal(s) of the
amplifier, and measuring the corresponding signal at the amplifier's output
terminal with the amplifier loaded as it is typically used.

The usual way to measure power supply rejection is to inject a test signal
into the power supply terminal of the amplifier. A series transformer
secondary winding or isolating resistor is used to keep the power supplies
source impedance from shorting out the test signal.

Usually, the test signal has to be robust - another amplifier may need to be
added to the output of the test signal source in order to provide enough
stimulus to get a good, clear reaction from the amplifier.

My conerns about this test is that it is really a poor
Intermodulation Distortion test. SMPTE, DIN, CCIF from
what I have researched do not use a sweeping frequency;
rather 2 tones of specified frequency.

Using sweeping tones is a more modern, and more revealing test than any
traditional IM tests based on fixed frequencies.

But, you don't necessarily need to do a distortion test, all you really need
to do is a test for spurious signals with the amplifier idle.

The spurious signal at the output of the amplifer is a direct test for power
supply rejection.

Testing for increases in distortion while the amplifier is amplifying a test
signal is a second-order test.

Usually, direct tests are more revealing and produce results that are easier
to understand. I don't doubt that the second order effect is there, but it
should be pretty weak.

So the test
methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a
setup to measure Intermod). Am I "correct" in my
thinking here?

No.

I would also like to know if you can predict THD+N from
an Intermod measurement and vice versa.

As others have pointed out, the cause of THD and IM is the same thing -
amplifier nonlinearity. The swept 2-tone is probably the more comprehensive
type of test. What it tells you is nonlinearity as a function of frequency,
and that is of the essence.

[[email protected]]

On Feb 29, 5:05*am, (Don Pearce) wrote:
On Fri, 29 Feb 2008 03:37:58 -0800 (PST), wrote:
On Feb 29, 2:41*am, (Don Pearce) wrote:
On Fri, 29 Feb 2008 01:28:40 -0800 (PST), wrote:
Hello All,

I am dealing with a conundrum about distortion measurements in an
amplifier which is serving as a DUT. *The architecture consists of a
two op-amp configuration that allows a single-ended input and BTL
output.
The issue arose with the difficulty of using the Audio Prescision
System II in developing a PSRR curve. Instead, the setup with the AP
consisted of two signal sources:

Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@ sweeping 10 Hz - 20
kHz

The AP was configured to measure THD+N. *So the idea is to see how the
THD+N of the amplifier is affected by power supply ripple rather than
your traditional PSSR curve.

My conerns about this test is that it is really a poor Intermodulation
Distortion test. *SMPTE, DIN, CCIF from what I have researched do not
use a sweeping frequency; rather 2 tones of specified frequency. *So
the test methodology is fundamentally in error: source frequencies
used, and the test setup (THD+N measurement rather than a setup to
measure Intermod). *Am I "correct" in my thinking here?

I would also like to know if you can predict THD+N from an Intermod
measurement and vice versa.

Now, I will probably be told go Google, but I have been having a hard
time researching this topic. *I am very green to audio. *I hope this
is the right group to ask!

Thanks in advance.

You appear to have got the whole thing confused.

First the PSRR test. Connect a single source in series with the power
supply. A 1:1 transformer is a good way to do this. Simply connecting
it "to" the power supply just shorts out the source oscillator. You
don't measure distortion, just how much signal appears at the output
of the amplifier. The PSRR is the ratio of the source signal level to
the output signal level. Don't feed any signal to the input of the
amplifier.

Yes, I agree. *I should have been more clear. *There have been several
methods used for PSRR: (1) using the AP as the generator connected via
a capacitor or (2) a transformer, or (3) a simple power amplifier as
the generator and it is connected in a summer configuration (AC+DC) to
the power supply of the DUT. *In all three methods the AP was used as
the input analyzer. *No input was applied to the DUT; the inputs were
AC coupled to ground. *A DC bias to Vcc/2 ensures a DC return path so
this AC coupling to ground is no problem.
However, the results were not yielding expected values so another
approach was used, and that approach was to inject a ripple onto the
power supply rail with a fixed (1Vp@1kHz) input signal and measure the
corresponding THD+N. *I am curious to know if measurements taken at
the output would be a valid THD+N measurement or rather it is a
(poorly implemted) intermod test since there are two sources being
mixed together.

It depends on the amplifier, the frequency and all sorts of stuff.
What should happen is that the ripple signal you put on the power
supply simply appears at the output in diluted form. If the amplifier
is not designed well, it is possible that the ripple will instead
amplitude modulate the audio you feed in. If that happens you will see
the signal frequency plus two more sidebands at signal +/- ripple
frequency. But that has nothing to do with the actual intermod spec of
the amplifier per se. It is really better to keep these measurements
apart.

When you say "if the amplifier is not designed well [...] the ripple
will instead amplitude modulate the audio..." are you impling the
specification at fault here would be the PSRR of the amplifier?
Also, if I see the signal frequencies applied to the DUT and the
sidebands, is this not intermod?
I am not disagreeing here; rather, I want to ensure I understand.


Now for distortion. Both harmonic distortion and intermodulation are
simply manifestations of the same thing. Since there are two signals
involved in the intermod test you can expect the measured result to be
a little different to THD, but not very. In general an amplifier that
has good THD specs will also be good for IMD. The big advantage of the
IMD test is that it yields useful results all the way up to the top of
the measurable band.

Well, what I wanted to know is my test setup was as follows:

Source 1 on the input = 1Vp@1kHz
Source 2 on the supply rail = 100mVp@(10Hz to 20kHz)

The resulting output is measured using a typical THD+N distortion
analyzer. *IIRC this consists of a notch filter tuned to the
fundamental, then a bandpass limiting filter, and a true RMS
detector. *As above is this:

1) A valid THD+N measurement? *I am being told that it is.

No. Remove the power supply signal if you want a valid THD+N
measurement.

Ok, I did not think such a setup was a valid THD+N measurement either;
however, can you elaborate a little on why it is not?
My answer would be that what is being measured are not harmonic
components of the input signal. With the Audio Precision, I believe
you can move the frequency of the notch filter but only one signal
would be removed. Meaning the fundamental of the input signal or the
power supply signal would be removed. Since one signal frequency
remains then this is not a THD+N.

2) The above setup is *similar* to several Intermod standards, but not
exact. *I look at the above setup as a poorly implemented
Intermodulation Distortion test. *So when I have my Audio Precision
setup to measure THD+N then I suspect the measurements are
meaningless.

To measure intermod characteristics, both signals should be applied to
the input. Only put signals on the power supply if you are
specifically doing PSRR tests.

You say that *in general* if THD+N is good, then so is intermod. *Ok,
so it is a Rule of Thumb. *But can one be converted to another?

Probably, but I haven't done the maths so I can't say for sure.


Not a problem.

Let me give you a little bit more background on this issue. The
reason I am asking these questions deals with the difficult I have had
with a PSRR test. I completely understand that the current set up
(input signal + signal on power supply) is *NOT* a PSRR test. Rather
it is a substitute, and some of the senior apps say that seeing a "THD
+N" vs. Power Supply Frequency curve would serve customer's better.
Simply put, "We cannot give you a PSRR curve, but there is this THD+N
curve vs. Power Supply Frequency" which we think is better. I, the
junior engineer, am having trouble justifying this hence this
discussion thread.

[[email protected]]

On Feb 29, 6:54*am, "Arny Krueger" wrote:
wrote in message



I am dealing with a conundrum about distortion
measurements in an amplifier which is serving as a DUT.

What is the power level?


5V supply, 1W into 8 ohms.

The architecture consists of a two op-amp configuration
that allows a single-ended input and BTL output.
The issue arose with the difficulty of using the Audio
Prescision System II in developing a PSRR curve. Instead,
the setup with the AP consisted of two signal sources:
Input signal source: 1Vp@1 kHz
Source connected to the power supply rail: 100mVp@
sweeping 10 Hz - 20 kHz

Just guessing here, but I doubt that the AP will adequately drive your load
without some additional boosting.

There is also an impedance-matching problem as power supplies usually have
very low impedances.

Good point. Here is a link to a Texas Instruments application note
that I was using to connect the Audio Precision to the power supply
rail of the DUT and the DC supply:

http://focus.ti.com/lit/an/sloa068/sloa068.pdf

Pages 17-20 and specifically 19&20 dealing with the impedance issue.

The AP was configured to measure THD+N. *So the idea is
to see how the THD+N of the amplifier is affected by
power supply ripple rather than your traditional PSSR
curve.

A power amplifier may have several volts of various kinds of signals on its
power supply terminals when it is being driven hard. There will be some
power line related trash and there will be some signals related to the test
signal due to the finite source impedance of the power supply.

The relevant parameter is called power supply rejection. A good power amp
will usually reject 60 dB or more of any signal that tries to creep into the
output from the power supply. *One usually measures power supply rejection
by injecting a test signal into the power supply terminal(s) of the
amplifier, and measuring the corresponding signal at the amplifier's output
terminal with the amplifier loaded as it is typically used.

The usual way to measure power supply rejection is to inject a test signal
into the power supply terminal of the amplifier. A series transformer
secondary winding or isolating resistor is used to keep the power supplies
source impedance from shorting out the test signal.

Usually, the test signal has to be robust - another amplifier may need to be
added to the output of the test signal source in order to provide enough
stimulus to get a good, clear reaction from the amplifier.

Good points, and I agree. The desire is to test PSRR. As I mentioned
to Don, the decision was to scrap PSRR tests because of my initial
difficulty in getting expected measurements. I still need to verify
the test setup, and Audio Precision settings.
So my colleagues decided to scrap the traditional PSRR test, and
substitute it with a THD+N vs. Power Supply Frequency curve saying
that this is more useful information. I have a hard time accepting
this and even moreso now that others (Don) also agree that it is *not*
a valid THD+N measurement; e.g., input frequency + power supply
frequency.

My conerns about this test is that it is really a poor
Intermodulation Distortion test. *SMPTE, DIN, CCIF from
what I have researched do not use a sweeping frequency;
rather 2 tones of specified frequency.

Using sweeping tones is a more modern, and more revealing test than any
traditional IM tests based on fixed frequencies.

I have not seen a specification that states this, but I can certainly
see that this would be a more revealing suite of tests.

But, you don't necessarily need to do a distortion test, all you really need
to do is a test for spurious signals with the amplifier idle.

The spurious signal at the output of the amplifer is a direct test for power
supply rejection.

Testing for increases in distortion while the amplifier is amplifying a test
signal is a second-order test.


There is little doubt in my mind that distortions would worsen as a
ripple signal is applied to the rail as a signal is applied to the
input. However, it is not a valid PSRR test nor a valid THD+N test as
it has been confirmed, so what kind of test would it be? Would you
rather see this test (whatever test it really is beyond the non-
linearity of the device) instead of a PSRR graph?

[Don Pearce]

On Sun, 2 Mar 2008 17:54:16 -0800 (PST), wrote:



It depends on the amplifier, the frequency and all sorts of stuff.
What should happen is that the ripple signal you put on the power
supply simply appears at the output in diluted form. If the amplifier
is not designed well, it is possible that the ripple will instead
amplitude modulate the audio you feed in. If that happens you will see
the signal frequency plus two more sidebands at signal +/- ripple
frequency. But that has nothing to do with the actual intermod spec of
the amplifier per se. It is really better to keep these measurements
apart.

When you say "if the amplifier is not designed well [...] the ripple
will instead amplitude modulate the audio..." are you impling the
specification at fault here would be the PSRR of the amplifier?
Also, if I see the signal frequencies applied to the DUT and the
sidebands, is this not intermod?
I am not disagreeing here; rather, I want to ensure I understand.


In a well designed amplifier, the gain is set by two resistors, and
nothing else. If the design is poor - a lack of open loop gain for
example - some of that gain figure will depend on device current,
which is determined by the rail voltage. This will allow ripple on the
rail to modulate the audio signal, rather than simply appearing as a
low level interference. And yes, if you see the sidebands, that is
intermod.

measurement.

Ok, I did not think such a setup was a valid THD+N measurement either;
however, can you elaborate a little on why it is not?
My answer would be that what is being measured are not harmonic
components of the input signal. With the Audio Precision, I believe
you can move the frequency of the notch filter but only one signal
would be removed. Meaning the fundamental of the input signal or the
power supply signal would be removed. Since one signal frequency
remains then this is not a THD+N.

THD+N stands for total harmonic distortion plus noise. What you are
measuring is total harmonic distortion plus noise, plus some random
other signals you are choosing to inject. Stop injecting those other
signals and you will have a THD+N measurement.


You say that *in general* if THD+N is good, then so is intermod. *Ok,
so it is a Rule of Thumb. *But can one be converted to another?

Probably, but I haven't done the maths so I can't say for sure.


Not a problem.

Let me give you a little bit more background on this issue. The
reason I am asking these questions deals with the difficult I have had
with a PSRR test. I completely understand that the current set up
(input signal + signal on power supply) is *NOT* a PSRR test. Rather
it is a substitute, and some of the senior apps say that seeing a "THD
+N" vs. Power Supply Frequency curve would serve customer's better.
Simply put, "We cannot give you a PSRR curve, but there is this THD+N
curve vs. Power Supply Frequency" which we think is better. I, the
junior engineer, am having trouble justifying this hence this
discussion thread.

Well, it isn't better. It is a totally unrealistic and meaningless
measurement. PSRR is a parameter you need during design and
development. It is the parameter that allows you to design your power
supply - size of decoupling caps etc. It has no place in a finished
amplifier spec because the customer has no access to, or ability to
inject signals into, the power supply rails. All the customer needs to
know about is the resulting level of hum (and that is generally
nothing to do with PSRR, but usually sub-optimal design or layout).

As for a PSRR curve - what for? 50 or 60Hz mains is the customer's
interface with the power supply. He doesn't connect signals there. As
I say, it is necessary to know and understand the PSRR of the
amplifier (internal component) as a DESIGNER. A PSRR spec for the
amplifier (whole system) is irrelevant to the customer.

So measure the hum, measure the THD+N and provide individual figures
for them. That is what everyone else does, for good reasons.

d
--
Pearce Consulting
http://www.pearce.uk.com

[[email protected]]

On Mar 3, 1:08*am, (Don Pearce) wrote:
On Sun, 2 Mar 2008 17:54:16 -0800 (PST), wrote:

It depends on the amplifier, the frequency and all sorts of stuff.
What should happen is that the ripple signal you put on the power
supply simply appears at the output in diluted form. If the amplifier
is not designed well, it is possible that the ripple will instead
amplitude modulate the audio you feed in. If that happens you will see
the signal frequency plus two more sidebands at signal +/- ripple
frequency. But that has nothing to do with the actual intermod spec of
the amplifier per se. It is really better to keep these measurements
apart.

When you say "if the amplifier is not designed well [...] the ripple
will instead amplitude modulate the audio..." are you impling the
specification at fault here would be the PSRR of the amplifier?
Also, if I see the signal frequencies applied to the DUT and the
sidebands, is this not intermod?
I am not disagreeing here; rather, I want to ensure I understand.

In a well designed amplifier, the gain is set by two resistors, and
nothing else. If the design is poor - a lack of open loop gain for
example - some of that gain figure will depend on device current,
which is determined by the rail voltage. This will allow ripple on the
rail to modulate the audio signal, rather than simply appearing as a
low level interference. And yes, if you see the sidebands, that is
intermod.

measurement.

Ok, I did not think such a setup was a valid THD+N measurement either;
however, can you elaborate a little on why it is not?
My answer would be that what is being measured are not harmonic
components of the input signal. *With the Audio Precision, I believe
you can move the frequency of the notch filter but only one signal
would be removed. *Meaning the fundamental of the input signal or the
power supply signal would be removed. *Since one signal frequency
remains then this is not a THD+N.

THD+N stands for total harmonic distortion plus noise. What you are
measuring is total harmonic distortion plus noise, plus some random
other signals you are choosing to inject. Stop injecting those other
signals and you will have a THD+N measurement.

You say that *in general* if THD+N is good, then so is intermod. *Ok,
so it is a Rule of Thumb. *But can one be converted to another?

Probably, but I haven't done the maths so I can't say for sure.

Not a problem.

Let me give you a little bit more background on this issue. *The
reason I am asking these questions deals with the difficult I have had
with a PSRR test. *I completely understand that the current set up
(input signal + signal on power supply) is *NOT* a PSRR test. *Rather
it is a substitute, and some of the senior apps say that seeing a "THD
+N" vs. Power Supply Frequency curve would serve customer's better.
Simply put, "We cannot give you a PSRR curve, but there is this THD+N
curve vs. Power Supply Frequency" which we think is better. *I, the
junior engineer, am having trouble justifying this hence this
discussion thread.

Well, it isn't better. It is a totally unrealistic and meaningless
measurement. PSRR is a parameter you need during design and
development. It is the parameter that allows you to design your power
supply - size of decoupling caps etc. It has no place in a finished
amplifier spec because the customer has no access to, or ability to
inject signals into, the power supply rails. All the customer needs to
know about is the resulting level of hum (and that is generally
nothing to do with PSRR, but usually sub-optimal design or layout).

As for a PSRR curve - what for? 50 or 60Hz mains is the customer's
interface with the power supply. He doesn't connect signals there. As
I say, it is necessary to know and understand the PSRR of the
amplifier (internal component) as a DESIGNER. A PSRR spec for the
amplifier (whole system) is irrelevant to the customer.

So measure the hum, measure the THD+N and provide individual figures
for them. That is what everyone else does, for good reasons.



Excellent. Thank you for the clarifications.

[Don Pearce]

On Mon, 3 Mar 2008 01:09:20 -0800 (PST), wrote:

So measure the hum, measure the THD+N and provide individual figures
for them. That is what everyone else does, for good reasons.



Excellent. Thank you for the clarifications.

Let me give one caveat before anyone else jumps in. Where an amplifier
has a balanced input (microphone preamplifiers are a common case),
there is a spec that is somewhat related to PSRR, and that is CMRR -
common mode rejection ratio. That defines the ability of the amplifier
to reject external interfering signals which appear on the two input
lines with the same polarity. CMRR will be specified for the better
grade of preamp.

d

--
Pearce Consulting
http://www.pearce.uk.com

[[email protected]]

On Mar 3, 2:29*am, (Don Pearce) wrote:
On Mon, 3 Mar 2008 01:09:20 -0800 (PST), wrote:
So measure the hum, measure the THD+N and provide individual figures
for them. That is what everyone else does, for good reasons.

Excellent. *Thank you for the clarifications.

Let me give one caveat before anyone else jumps in. Where an amplifier
has a balanced input (microphone preamplifiers are a common case),
there is a spec that is somewhat related to PSRR, and that is CMRR -
common mode rejection ratio. That defines the ability of the amplifier
to reject external interfering signals which appear on the two input
lines with the same polarity. CMRR will be specified for the better
grade of preamp.


Let me play devil's advocate one last time.

We agreed that with the ripple voltage on the power supply rail with
an input signal present is not a valid THD+N test. And we also said
that if sidebands are present then there will be intermodulation
distortion.

Now the noise portion of THD+N should include *everything* within the
tested bandwidth once the fundamental is removed with the notch: buzz,
hum, harmonics, noise, etc. Wouldn't the power supply ripple signal
be considered 'noise' or hum (although hum is usually mentioned to be
the mains frequency). So wouldn't this be a valid THD+N test;
although, the additional signal (noise) from the power supply should
raise the noise floor and mask the real distortion of the amplifier.

I have seen an article on Wikipedia saying that THD+N includes
intermodulation distortion, but I will have to think about that
further. It is late here.


Thanks for the help in this matter.

[Don Pearce]

On Mon, 3 Mar 2008 03:49:24 -0800 (PST), wrote:

On Mar 3, 2:29*am, (Don Pearce) wrote:
On Mon, 3 Mar 2008 01:09:20 -0800 (PST), wrote:
So measure the hum, measure the THD+N and provide individual figures
for them. That is what everyone else does, for good reasons.

Excellent. *Thank you for the clarifications.

Let me give one caveat before anyone else jumps in. Where an amplifier
has a balanced input (microphone preamplifiers are a common case),
there is a spec that is somewhat related to PSRR, and that is CMRR -
common mode rejection ratio. That defines the ability of the amplifier
to reject external interfering signals which appear on the two input
lines with the same polarity. CMRR will be specified for the better
grade of preamp.


Let me play devil's advocate one last time.

We agreed that with the ripple voltage on the power supply rail with
an input signal present is not a valid THD+N test. And we also said
that if sidebands are present then there will be intermodulation
distortion.

Sidebands are an indication that there is intermodulation. But that
doesn't necessarily mean that the amplifier itself will suffer
intermodulation of signals applied to the input - they do not have the
modulation mechanism available to a signal on the power supply. So do
not confuse the true IM spec of the amplifier with any IM you can
achieve through signals on the power line.

Now the noise portion of THD+N should include *everything* within the
tested bandwidth once the fundamental is removed with the notch: buzz,
hum, harmonics, noise, etc. Wouldn't the power supply ripple signal
be considered 'noise' or hum (although hum is usually mentioned to be
the mains frequency). So wouldn't this be a valid THD+N test;
although, the additional signal (noise) from the power supply should
raise the noise floor and mask the real distortion of the amplifier.


It depends on how you are measuring. If your test set simply nulls the
fundamental and measures the residual power, then yes, hum will be
included. Hopefully the design will be such that any residual hum is
well below other, unavoidable noise source. But a modern test set
would use an FFT analyser, which allows separate measurement of all
frequencies, so hum, harmonics, intermodulation and wideband noise can
be identified and measured separately.

I have seen an article on Wikipedia saying that THD+N includes
intermodulation distortion, but I will have to think about that
further. It is late here.


No. THD measurements are made with a single input signal. That means
that there is no mechanism for IM to occur. If you want to measure IM,
you need at least two input signals and tune your analyser to the sum
and difference rather than the harmonics of either.

d

--
Pearce Consulting
http://www.pearce.uk.com

[[email protected]]

On Mar 3, 5:04*am, (Don Pearce) wrote:

Now the noise portion of THD+N should include *everything* within the
tested bandwidth once the fundamental is removed with the notch: buzz,
hum, harmonics, noise, etc. *Wouldn't the power supply ripple signal
be considered 'noise' or hum (although hum is usually mentioned to be
the mains frequency). *So wouldn't this be a valid THD+N test;
although, the additional signal (noise) from the power supply should
raise the noise floor and mask the real distortion of the amplifier.

It depends on how you are measuring. If your test set simply nulls the
fundamental and measures the residual power, then yes, hum will be
included. Hopefully the design will be such that any residual hum is
well below other, unavoidable noise source. But a modern test set
would use an FFT analyser, which allows separate measurement of all
frequencies, so hum, harmonics, intermodulation and wideband noise can
be identified and measured separately.


Well, the Audio Precision 2700 uses a tunable bandreject (notch
filter) during the THD+N measurement. Essentially, it is as I
mentioned earlier. There is some signal conditioning at the input of
the audio analyzer, then the signal is passed to the notch filter, and
then finally to the measuring circuitry.
So, the only signal being removed will be the fundamental of the input
signal. The power supply signal will remain. This should be
attenated by the PSRR of the amplifier, and whatever is left over at
the output will be the distortion of the input signal along with noise
which includes the effects of the power supply signal.

I have gotten a hold of Bob Metzler's 'Audio Measurement Handbook' and
the section on 'Interference Signals Effects on THD+N' reads:

"Since THD+N is instrinsically wideband measurement [...] it is
suscetible to interfering signals. If an interering signal exists in
the DUT, it acts to establish a floor below which measurements cannot
be made. [...] the most common type of interfering signal is AC mains
(power line) hum."

The only difference with my test is that the 'hum' will *NOT* be the
mains frequency; rather, it will be a swept frequency at a fix
amplitude.

So I suspect that the signal I am purposely injecting into the power
supply rail; even though not mains frequency, will raise the noise
floor of the measurement to some degree. The signal will be
attenuated because of the PSRR of the amplifier, but will remain since
the notch filter is only removing the input fundamental. This will
obscure the real distortion of the amplifier as per Bob Meztler.