chuck wrote:

Owen Duffy wrote:

On Fri, 14 Oct 2005 00:58:35 GMT, Owen Duffy wrote:



It is the ratio of signal to noise and distortion, and it is measured
by setting up a test where the receiver produces output from a SSG
(typically for a 1KHz audio output) and notching out the 1KHz output
to measure the noise and distortion wrt the filtered 1KHz output.



That should be: It is the ratio of signal and noise and distortion to
noise and
distortion, and it is measured by setting up a test where the receiver
produces output from a SSG (typically for a 1KHz audio output) and
notching out the 1KHz output to measure the unfiltered output wrt the
noise and distortion.

I should also have mentioned the EIA test requires the receiver be set
to rated output with 1mV RF input at 60% of rated modulation, then the
RF output reduced to find the input level for 12dB SINAD.

You could measure it with soething like a HP334A Distortion Analyser,
but it is pretty tedious if you are trying to find the RF input for a
particular SINAD. Hence you see boxes that have an AGC controlled
amplifier deliving a constant voltage to the filter block. I have a
Motorola one (R1013A) that works ok, there were also Sinadders. Even
more convenient are the ones integrated into a communications monitor.
I don't believe these boxes do true RMS measurements.

An alternative if you have a standalone SSG and want to do SINAD
measurement is to use a PC sound card and software that does an FFT
and calculates the SINAD (using true RMS measurement). Spectrum Lab
does it, its free, but it is such a flexible / general tool, it may be
a bit daunting to get it working. The SpectrumLab menu "Quick Settings
/ Rx Equipment Tests / SINAD test" is a quick path to setup... but it
is still a quite complex package. A whole lot better than the style of
a HP334A though! SL is at http://www.qsl.net/dl4yhf/spectra1.html .

Owen
--
Good info, Owen. I think the EIA test procedures really have FM or AM in
mind, rather than SSB or, what is exactly the same for SINAD purposes,
CW. The 60% figure just doesn't apply to SSB or CW. You would simply use
an unmodulated signal generator with the frequency offset to produce a 1
kHz tone in the receiver's audio output, preferrably centered in the
receiver's passband. Then a measure of rms af voltage at the receiver's
output with and without the 1 kHz filter would be made.

We don't hear much about SINAD testing procedures for SSB and CW. Even
the ARRL's test procedure manual glosses over the procedure for other
than FM.

Chuck
NT3G

SINAD measurements were cooked up because the audio quality of an FM
receiver depends both on the demodulator and on the noise
characteristics of it's front end. Just having a noise figure for an FM
receiver is pretty useless. Keep in mind that an FM receiver is usually
rated as "X dB SINAD for y microvolts input".

For SSB and CW, on the other hand, the noise is purely additive so all
you need to know is the receiver noise figure. Once you know that
(assuming that it's not a really strange radio) you know everything
about it's performance. Given the noise figure in dB you can easily
calculate the 12dB SINAD should you be so inclined, as well as any other
signal vs. noise figure you should want. You have a good reason to
believe that the noise is white so you can even take an SSB receiver and
calculate the noise figure of the thing after you tack on an audio
bandpass filter for CW. This is _not_ the kind of thing you could do
with FM.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Owen, (& crb)
Your words are contrary to the way we measured it (Motorola). You say
...."It is the ratio of signal to noise and distortion,..", but we measured
not just the signal, but everything for the "top" of the ratio (which is
used more like a reference as it is the more stable as signal level varies).
I believe your last line saying wrt to the filtered tone supports this. It
implies that the tone is (bandpass) filtered for one of the measurements and
we don't do that.

Joe's is a bit better of an explanation (the RMS meter and quantities
ratioed).

A meter, a pure 1kHz tone modulated signal generator and a 1kHz notch is all
that is needed. What happens if you don't have a "real" RMS meter? I don't
know.

73, Steve, ,K.9;D'C'I



"W3JDR" wrote in message
news:ljM3f.39197$q81.11651@trnddc06...
It is the ratio of (Signal + Noise+Distortion) to (Noise+Distortion) as
measured at the receiver audio output. It is measured using an RMS-reading
AC voltmeter , typically with a 1Khz modulation tone on the signal applied
to the receiver under test.

First you measure the audio signal out of the receiver using the AC RMS
meter. Then you apply a notch filter at the modulation frequency and
measure
the residual noise+distortion, again using the RMS AC voltmeter. SINAD is
the ratio of the two measurements.

Joe
W3JDR



"crb" wrote in message
...
Is it only valid for AM and FM measurements?

I know its receiver sensitivity.

Is it Signal divided by Noise with distortion??

12 dB SINAD means what? The signal is 12 dB greater than noise and
distortion or
is it more complicated than that?

On Fri, 14 Oct 2005 13:08:22 -0500, "Steve Nosko"
wrote:

A meter, a pure 1kHz tone modulated signal generator and a 1kHz notch is all
that is needed. What happens if you don't have a "real" RMS meter? I don't
know.

Hi Steve,

You don't need a "real" RMS meter. The expressed requirement for a
pure 1kHz tone provides the necessary sine wave shape such that it
simply becomes a matter of scale calibration. If you had said a
square wave 1KHz tone (nothing pure about that), then you would have
to dig deep for a "real" RMS meter. That too, could be scaled, but I
wouldn't count on it because it would be a rare amplifier chain that
could faithfully keep it square - and the notch would inject it into
the measurement as distortion and noise.

73's
Richard Clark, KB7QHC

Richard,
What you said is largely accurate, however at low S/N ratios, or where the
distortion becomes comparable to the signal level, the reading of the
composite signal (signal+noise+distortion) with anything other than an RMS
meter could produce erroneous results.

Joe
W3JDR

"Richard Clark" wrote in message
...
On Fri, 14 Oct 2005 13:08:22 -0500, "Steve Nosko"
wrote:

A meter, a pure 1kHz tone modulated signal generator and a 1kHz notch is
all
that is needed. What happens if you don't have a "real" RMS meter? I
don't
know.

Hi Steve,

You don't need a "real" RMS meter. The expressed requirement for a
pure 1kHz tone provides the necessary sine wave shape such that it
simply becomes a matter of scale calibration. If you had said a
square wave 1KHz tone (nothing pure about that), then you would have
to dig deep for a "real" RMS meter. That too, could be scaled, but I
wouldn't count on it because it would be a rare amplifier chain that
could faithfully keep it square - and the notch would inject it into
the measurement as distortion and noise.

73's
Richard Clark, KB7QHC

On Fri, 14 Oct 2005 13:08:22 -0500, "Steve Nosko"
wrote:


Owen, (& crb)
Your words are contrary to the way we measured it (Motorola). You say
..."It is the ratio of signal to noise and distortion,..", but we measured
not just the signal, but everything for the "top" of the ratio (which is
used more like a reference as it is the more stable as signal level varies).
I believe your last line saying wrt to the filtered tone supports this. It
implies that the tone is (bandpass) filtered for one of the measurements and
we don't do that.

It was a shabby description in my first post Steve, I wrote filtered
instead of unfiltered, had the "wrt", terms back to front at the and,
as you quoted signal to (N+D) when it is the total to (N+D). Wasn't
really worth 2/10, was it! I re-read it when I came back to add some
detail re a software approach and redrafted it.

My recollection was then when the scheme was introduced, one used a
Distortion Analyser, and I can't remember the early HP instrument, the
HP334A I mentioned in my follow up was a newer one. They are tedious,
whereas the R1013 or Sinadder or much more convenient, and the
integrated ones (like in the R2000) are much better.

Once you sort the issues of getting audio samples to a PC sound card
without hum and clipping, that approach can work well, and Spectrum
Lab works well... just it is a multipurpose tool for a simple job.
There are probably other software tools that are more targetted and
simpler to use.


Joe's is a bit better of an explanation (the RMS meter and quantities
ratioed).

Yes, but I don't believe most of the instruments acutally incorporate
a true RMS meter. I set about measuring the difference about six
months ago when I was doing calibration / validation measurements for
FSM (http://www.vk1od.net/fsm/) using custom software against the
R1013, and R2000, and a couple of HP334As. The error in using a
rectifier-average responding meter (as is typically done) is small
relative to the variance of such readings because of the variance of
the noise component.


A meter, a pure 1kHz tone modulated signal generator and a 1kHz notch is all
that is needed. What happens if you don't have a "real" RMS meter? I don't
know.

Yes, I didn't mention that the 1KHz tone needs to be relatively low
distortion. For measuring 12dB SINAD, the demand is not onerous, but
it is important. The frequency of the tone is important as the notches
in semi automatic instruments are typically +/10Hz or so, so one needs
to verify that the SSG modulation oscillator is close enough. More
importantly when testing an SSB receiver (where you use a CW carrier),
that the carrier is kept "on frequency" for a 1KHz beat note.

Owen
--

On Fri, 14 Oct 2005 20:28:17 GMT, "W3JDR" wrote:

Richard,
What you said is largely accurate, however at low S/N ratios, or where the
distortion becomes comparable to the signal level, the reading of the
composite signal (signal+noise+distortion) with anything other than an RMS
meter could produce erroneous results.

Hi Joe,

In the practical world of SINAD (having tuned a number of GE and
Motorolas), one is not very interested in how poor your set is, but
rather meeting a service standard (that 12 dB which is as arbitrary as
any).

I doubt if many of the current generation of commercial surplus
equipment comes with a stock tester employing what would have been an
expensive converter chip to insure RMS measurements. I come by that
assessment by noting those I used employed standard meter movements.
The first RMS meters I calibrated in the mid 70s came from Fluke (just
up the highway), and the components of that circuit were scrubbed of
all identification numbers or cast in epoxy. Such was the cachet of
being hi-priced, and having others try to break into the market with
knock-offs.

My Radio Shack multimeter makes that claim (ca 1995) and if memory
serves, that Micronta's "True RMS" was barely capable of poor voice
grade bandwidth. This was 20 years after Fluke, costing about as much
(economic inflation), and not performing as well (technical
deflation).

73's
Richard Clark, KB7QHC

On Fri, 14 Oct 2005 10:50:03 -0700, Tim Wescott
wrote:


bandpass filter for CW. This is _not_ the kind of thing you could do
with FM.

Indeed Tim. The complications as I see it in predicting the SINAD for
FM when you know the receiver NF is that most narrow band
communications applications of FM are closer to PM because of the
pre-emphasis characteristic.

FM + 6dB/octave preemphasis over the entire modulation passband is PM.
(PM is where the modulation index (dev/fm) is independent of fm (the
modulating frequency)).

However, the receivers in my experience are over de-emphasised (at the
top end) presumably to get better SINAD. In fact, I think
specifications of the de-emphasis curve are commonly stated along the
lines of +1 to -3dB of -6dB demphasis relative to 1Khz over 300 to
3000Hz. This accomodates a over de-emphasis at the high end for little
loss in intelligibility and a dB or so improvement in sensitivity
figures.

Filter / demodulators and CTCSS IM also contribute to distortion
products significantly.

The result of demod distortion, PM with a slope across the passband,
and uncertain high pass filtering to accomodate CTCSS makes prediction
of S/N out from C/N in a bit of a guess in FM comms receivers, not
nearly as accurate as you suggest for linear receivers.

Owen
--

Richard,

I understand the historical difficulties of making accurate RMS
measurements, however I didn't know the original post only solicited ways to
make the measurement with "current generation of commercial surplus
equipment ". My intention was to point out some measurement nuances that
might not be obvious at first glance.

Recently, it has become quite easy to do true RMS measurement at audio
frequencies using DSP techniques. In fact at audio you can even do an
accurate RMS measurement in DSP using a PIC microcontroller to sample the
signal and perform the calculations.


Joe
W3JDR


"Richard Clark" wrote in message
...
On Fri, 14 Oct 2005 20:28:17 GMT, "W3JDR" wrote:

Richard,
What you said is largely accurate, however at low S/N ratios, or where the
distortion becomes comparable to the signal level, the reading of the
composite signal (signal+noise+distortion) with anything other than an RMS
meter could produce erroneous results.

Hi Joe,

In the practical world of SINAD (having tuned a number of GE and
Motorolas), one is not very interested in how poor your set is, but
rather meeting a service standard (that 12 dB which is as arbitrary as
any).

I doubt if many of the current generation of commercial surplus
equipment comes with a stock tester employing what would have been an
expensive converter chip to insure RMS measurements. I come by that
assessment by noting those I used employed standard meter movements.
The first RMS meters I calibrated in the mid 70s came from Fluke (just
up the highway), and the components of that circuit were scrubbed of
all identification numbers or cast in epoxy. Such was the cachet of
being hi-priced, and having others try to break into the market with
knock-offs.

My Radio Shack multimeter makes that claim (ca 1995) and if memory
serves, that Micronta's "True RMS" was barely capable of poor voice
grade bandwidth. This was 20 years after Fluke, costing about as much
(economic inflation), and not performing as well (technical
deflation).

73's
Richard Clark, KB7QHC

Originally, two way shops set FM rigs up for 20 dB quieting, on a volt
meter (crank up the signal until the AUDIO VOLTAGE, UNMODULATED, was
1/10th the voltage of a no signal audio output . Tho, for the most
part, this works well, there are constraints on sensitivity, because of
bandwidth concerns, and , as bandwidth is halved, the signal improvement
is 6 dB (quadrupled). Sinad is Signal/Noise /signal/(noise+distortion)
and in fact, in recent times , devices that will measure it are built
into many pieces of test equipment (IFR meters comes to mind), also
look for an outfit called "SINADDER" . The main thing is that it adds
a "Bandwidth" component to the sensitivity equasion. It is measured
with a 1 KHz tone, at (in FM), 3 KHz deviation- and the smaller the
signal that is detectable , with this constraint, the more sensitive
the reciever is considered to be! This also works at SSB/AM. Tho, it
is true that this measures Sensitivity, it includes a BANDWIDTH
component, that a (noise figure/ quieting) would NOT consider (at least
fully!) Hopefully, this is helpful-- Jim NN7K


For SSB and CW, on the other hand, the noise is purely additive so all
you need to know is the receiver noise figure. Once you know that
(assuming that it's not a really strange radio) you know everything
about it's performance. Given the noise figure in dB you can easily
calculate the 12dB SINAD should you be so inclined, as well as any other
signal vs. noise figure you should want. You have a good reason to
believe that the noise is white so you can even take an SSB receiver and
calculate the noise figure of the thing after you tack on an audio
bandpass filter for CW. This is _not_ the kind of thing you could do
with FM.

On Sat, 15 Oct 2005 00:40:34 GMT, "W3JDR" wrote:

I understand the historical difficulties of making accurate RMS
measurements, however I didn't know the original post only solicited ways to
make the measurement with "current generation of commercial surplus
equipment ".

Hi Joe,

That was interjected by me, knowing the market of the past several
years being flooded after trunk systems began replacing older service.

My intention was to point out some measurement nuances that
might not be obvious at first glance.

Useful information, that.

Recently, it has become quite easy to do true RMS measurement at audio
frequencies using DSP techniques. In fact at audio you can even do an
accurate RMS measurement in DSP using a PIC microcontroller to sample the
signal and perform the calculations.

Yes, the miracle of Moore's law. 20 years ago I was with HP, here, to
help them introduce their 100KHz real-time dual channel audio spectrum
analyzer. That was a tremendous effort with a million lines of Pascal
code and 5 years in the making when most HP instrumentation hit the
market in 18 months from inception.

I got to know the range of FFTs under some of the most brilliant minds
on the topic. One, Nick Pendergrass, went on to teach at an eastern
university.

Today, it is an underclass topic, probably occupying no more than 6
weeks of instruction coupled to other interests. Still and all, I see
considerable errors of omission in the discussion. Such errors often
make the difference in delivering a serviceable performance compared
to that which is 100 times better (actually a million times, but few
could get their imagination around a number that big so I understate
it).

73's
Richard Clark, KB7QHC