[ARC5] S+N/N ratio results.

[Geoff]

>>I think these radios do just fine with weak signals on 40M
>>or lower where
>> the atmospheric noise is higher.  I put a passive DBM in front of a 
>> BC-453B
>> and it will capture whatever is above band noise. Stability is good too.
>> The issue is CW selectivity.  You get used to what we did as novices, use
>> the analog signal processor between our ears.
>>
>> Dennis AE6C
>>
>
>    I think it depends on the particular receiver. There are other 
> characteristics beside noise that affect the performance at higher 
> frequencies.  One is the losses in the coils and condensers and other 
> insulators. The material of the tube sockets and bases can contribute.
>    Conventional signal to noise is measured as follows:
> Instruments needed; A signal generator with a calibrated attenuator 
> capable of putting out very low levels down to a fraction of a micro-volt, 
> and with very low leakage, adequately shielded cable to attach the 
> generator to the receiver. Some means of verifying the voltage level at 
> the receiver; an audio frequency volt meter. This will be connected at an 
> audio output so does not need high sensitivity.


** Since most audio meters are calibrated for 500-600 Ohms there is a 
mismatch to many radios.
I use an overhauled GR 583A to match most anything and eliminate those 
errors.


>    Most receivers that have a specification for signal to noise also 
> specify a dummy antenna or a source impedance for the measurement. The 
> voltage at the generator terminal needs to be known if the measurement is 
> to be meaningful.


** Receivers with screw terminal or single wire antenna inputs will not give 
accurate results, especially on higher frequencies. The best you can do 
there is use a shielded balun with as close to zero length output leads as 
possible. Even then there is some leakage/stray coupling. I also wrap the 
input coax many times thru a 2.4" 43 Mix toroid to minimize common mode 
leakage. See below for impedance concerns.


Some
> receivers have fairly constant input impedance, some do not but specify a 
> dummy antenna to allow for reproduceable measurements.


** Even those with coax input connectors can have a 2 to 4:1 mismatch if the 
receiver is general coverage. The various dummy antennas do a decent job for 
a simple military type PM and alignment since the manual is matched to the 
required test equipment and the "close enough for government work" rule 
takes precedence.
Getting a perfect SNR reading would require a using a bridge, VNA, MFJ 259B 
or similar, to locate that almost perfect 50 Ohm input impedance spot 
frequency (with or without a balum, dummy antenna or whatever) at which to 
take the measurement. This is a waste of time to most where an "acceptable" 
average is sufficient.

** An alternative, and good idea in all measurements, is to force a 50 Ohm 
input and output with a 6dB minimum loss pad made from carbon film resistors 
if you dont have a commercial version. Any value between 6-10dB is fine and 
you can use an online calculator to determine attenuation using standard 
resistor values.


>    The generator is modulated 30% at (usually) 400 hz. Preferably the 
> generator should be capable of low distortion modulation with low 
> incidental FM so that the sidebands are symmetrical.


** This eliminates the usual "service grade" kits and other low end test 
gear many use and who then take results as gospel and not to be disputed. 
They are also huge harmonic generators to further result in poor accuracy.


>    We will assume the measurement is for a 10 db signal to noise ratio. If 
> for some other value set the levels below accodingly.
>    The audio meter is attached to the receiver.  AVC is turned off and RF 
> gain is set at maximum. This is not changed.

** OK

A reference level is chosen for the noise,
> preferably a level where there is no hum or other audio noise but enough 
> audio dynamic range to allow at least a 10 db increase.


** Since the meter calibration is likely non linear pick a pair of 10dB mid 
range points that is easy to interpolate small changes.



 Set the generator for a low value of output
> with no modulation, say one microvolt.  Tune and peak the receive 
> carefully and set the audio gain for the reference you have chosen to be 
> zero level.  turn on the modulation. The output will go up. Adjust the 
> receiver audio gain for an output exactly 10 db above the noise reference 
> (for a 10 db ratio). Now, turn the modulation off again and read the 
> noise. If not still at zero (probably will not be) adjust the _generator_ 
> output level to bring the noise back to the reference.  Now start the 
> modulator again and adjust the receiver AF gain for ten db above the noise 
> again.  You may have to juggle these two several times to get it to where 
> the level with modulation is exactly 10 db higher than the noise. When you 
> acheive this read the level of the RF output from the attenuator or meter 
> on the generator.  This is the RF level for 10 db Snr.

**  I find it far easier to leave the modulation on and tune the receiver on 
and off the signal since the receiver is likely still drifting on the higher 
frequencies. I first set the noise only reference on the meter.  The above 
procedure is fine for stable receivers but still requires considerable arm 
movement.


  This level is likely to change with
> frequency depending on receiver design. It will also change with IF 
> bandwidth and is meaningless without a statement of the bandwidth.


**  As long as one accurate test is made the data for other bandwidths can 
be calculated.


  It will also change somewhat with the shape
> of the IF filter being less for a given bandwidth if the skirt selectivity 
> is greater.
>    Many receivers give specs for a 10db Snr but not all. For instance, the 
> RCA AR-88 gives one for 6db and one for 20 db.
>    Also note that many receivers give only a minimum RF level for a 
> minimum audio output level.  This is meaningless other than being a simple 
> test of of the receiver's condition.

**  Typical marketing BS

>    Typical 10db signal to noise ratios with typical IF bandwidths will 
> vary from around 5uv to less than 1 uV partly due to the variation in 
> bandwidth.  Older receivers with fixed bandwidths will vary from around 6 
> khz down to around 3 khz.


** More like 10-15 kc at the 6dB points with the typical 2 IF stage, 4 tuned 
455kc circuits receiver. AKA broad as a barn door.


  If the SNR is measured at seveeral
> frequencies over the range of the receiver it will give some idea of the 
> consistency of performance. Very good receivers will not vary much.


** The usual 4-5 band .5 to 30mc receiver can have a fairly wide variation 
between band edges but since they overlap, one or the other will be better. 
Ham band only have tighter tolerance wheras some of those 30 band sets can 
have quite a bit of variation...think Collins.



>    Single conversion receivers may show higher noise at high frequencies 
> because one is seeing the image noise as well as the noise at the main 
> signal tuning. This is probably significant only for single RF receivers 
> with low IF frequencies.


** Single IF's of later years were often at 5 to 9 mc and images werent an 
issue at HF. A 455 kc IF and a 1 to 5 uV input should not have any 
measurable image noise if it has at least one RF stage. Even those popular 
no RF stage entry level sets likely have a 8-10dB image rejection at 30 mc.


>    My AR-88, which is a remarkable receiver for its time, requires less 
> than 1 uV for 10db SNR right up to 30 Mhz using the widest crystal filter 
> setting, about 3 khz.
>    I've measured a number of other receivers this way but have misplaced 
> my notes.  By memory the SP-600 did about 2 uV for the 3 khz bandwidth. 
> It should have done as well as the AR-88 but its possible the bandwidth of 
> the AR-88 is narrower. However, the AR-88 gave 1.8 uV at the 8 khz 
> bandwidth setting so I think perhaps its actually pretty quiet.


** For decent AM all filters should be OFF even tho the numbers often are 
not very complimentary.


>    I stress again that this method does not produce a true noise 
> measurement. Its mostly useful for trouble shooting. A true measurement 
> would measure _Noise Figure_ which is a meaasure of the noise from a 
> receiver or amplifier in comparison to the theoretical minimum noise. That 
> is, the ratio in db of the noise output of a practical circuit to the 
> noise from an ideal noisless circuit. Noise figure is measured using a 
> noise generator.  Noise generators can be made so that the noise output 
> can be known very accurately. The measurement is simply a reading of the 
> output on the generator to the output with a proper termination on the 
> amplifier or receiver.  The ratio in db is the N.F.
>    Hewlett-Packard published an application note, actually a book on N.F. 
> measurement which is available on line from: 
> http://www.hparchive.com/appnotes.htm
>   look for AN-57-1 and AN-57-2  while these are oriented toward microwave 
> applications the theory holds for lower frequencies as well.

> Richard Knoppow
> Los Angeles
> WB6KBL
> dickburk at ix.netcom.com

** Ive been using a NF meter ever since I built a Sylvania 5722 vacuum tube 
diode version in the 60's. These day I use a HP 8970A  with noise head that 
I bought at a belly up company auction in 2002. I used them for so many 
years at various jobs I could likely program it in my sleep (-;
Carl
KM1H