ASSESSING THE PERFORMANCE OF THE RSP RECEIVER SYSTEM IN HF

[glovisol]

ASSESSING THE PERFORMANCE OF THE RSP RECEIVER SYSTEM IN HF

The maximum achievable signal-to-noise ratio (SNR) in any receiving system is fundamentally limited by the receiver’s noise floor. In low-frequency receiving systems–typically frequencies at and below High Frequency (HF)—the realized SNR is often limited by external background noise, which includes galactic noise, atmospheric noise, lightening, and man-made noise. In these low-frequency receiving systems, it is often assumed that the mismatch loss (τ) associated with high voltage standing-wave ratio (VSWR) is not significant because it equally attenuates both the received signal power and the received noise power.”

Do you wish to quickly determine your receiver Noise Figure? Do you wish to know how your antenna is performing with respect to local or far away noise? Would you like to know whether your location is really suitable, or not suitable for HF reception? You would, but until now you did not have the terms of comparison, now you have them, just read on…..

In this thread we shall provide methods and data useful in determining antenna and receiver performance in the real world. The following parameters will be easily found:
- Receiver Noise Figure from receiver’s Noise Floor with 1 Kohm and with 50 Ohm input termination.
- Antenna Noise Figure from receiver’s Noise Floor when connected to your antenna.
- Your Antenna noise predicted performance dependent on the characteristics of the receiving site
environment.

Today we are bombarded with horror tales about man made noise lurking around and within our antenna space, we read many articles on so called (….science fiction…) low noise antennas, but we have no terms of comparison to really know and understand if the noise floor we are experiencing and which limits our weak signal receiving capabilities is the best we can have in our location, or if there is, at least theoretically, room for improvement. In other word we need to precisely assess the performance of our receiving system not in qualitative, but in quantitative terms.

Today it is easily possible for the radio operator to achieve this level of knowledge because we have at our disposal:
- The RSP class of communications receivers, which provide accurate & calibrated noise floor data.
- ITU Recommendation ITU-R P.372-13, which provides real world noise data and noise level approximation
parameters.

Tabular information will be provided first to provide simple and easy assessment without the impediment of mathematics. In later posts the mathematic gears behind the show will also be described.

1. RECEIVER NOISE FIGURE FROM TERMINATED NOISE FLOOR

Table 1 shows HF noise floor readings of my RSPduo receiver, S/N 180602D532 with the HIGH Z input terminated with a 1 KΩ resistor and the coaxial inputs terminated with a 50 Ω resistor. All operating parameters are also tabulated for easy reference.

General measurement conditions are shown, so that they can be actually reproduced for other receivers. The RF gain shown in the last column for the two coaxial inputs is the lesser of the two gains found, when different. It is important to do the noise floor measurements using the same SR and DEC values, otherwise inconsistent data will be obtained.

Data in Table 2 has been calculated to quickly determine the receiver’s Noise Figure in each band for any of the three inputs and for a given detection bandwith.

EXAMPLE 1, using noise floor data (Table 1) found for my own RSPduo.
Our receiver, on the 50 Ω terminated coaxial input reads a noise floor of – 135 dBm on Tuner 2 on the 160 m Band ( 1,950 MHz). With a detection bandwith of 2200 Hz (SSB) in Table 2 we read a Noise Figure of : (6.6+4.6)/2 = 5.6 dB.

EXAMPLE 2, using noise floor data (Table 1) found for my own RSPduo.
Our receiver, on the 1 KΩ terminated balanced input reads a noise floor of – 140 dBm on Tuner 1 on the 70 m Band ( 7,200 MHz). With a detection bandwith of 1800 Hz (SSB) in Table 2 we read a Noise Figure of : 1.4 dB.

All in all my receiver shows excellent performance, with the worst value of Noise Figure found on Tuner 2, coaxial input, at 3.750 MHz, with a noise floor of -125 dBm corresponding to a Noise Figure of 16.4 dB. As we shall see, the value of the receiver's Noise Figure, even at this level, impacts little on the overall system Signal to Noise Ratio, because unfortunately the antenna noise figures are much higher due to noise received.

In the next post we shall analyze antenna noise behavior using antenna noise floor data.

[glovisol]

2. PREDICTED ANTENNA NOISE FLOOR

It is general knowledge that when you connect an antenna to the receiver’s input the receiver’s output noise significantly increases because it detects the noise picked up by the antenna. This noise is the composite result of the following:
- Galactic noise from the sun and the stars.
- Atmospheric noise.
- Noise due to Nature’s electrical activity, such as lightning.
- Man made noise coming from all local human activities.
Barring the contribution of Nature’s electrical activity, as local rainstorms and lightning are sporadic events, the already mentioned ITU Recommendation ITU-R P.372-13, based on actual noise tests and measurements, determines four (4) categories of noise environments, with decreasing values of noise levels as follows.
CITY
RESIDENTIAL
RURAL
QUIET RURAL

The Recommendation provides a formula with variable coefficients to predict the Noise Figure for an antenna located in any given area. These predictions do not provide firm and absolute values, but statistical data which indicates, for any of the mentioned noise areas, the most likely level of Noise Figure for an antenna installed on the same area. Furthermore the Recommendation also gives the Decile Statistical deviations which provide a statistical range within which the antenna Noise Figure is most likely to be, depending on the type of area it is in.

Table 3 below not only provides the antenna Noise Figure in dB calculated for every HF Ham band, but also translates the data into Noise Floor in dBm for all categories of noise environments and additionally gives the Decile Deviations. Table 3 also gives two values of Galactic Noise, one more pessimistic, calculated with the Recommendation’s formula and coefficients, and the other (in blue colour) calculated with the older CCIR formula. Supposing you have an antenna located in an area absolutely devoid of atmospheric and man made noise, then your antenna output noise would be near the value indicated by one of the two tabled Galactic Noise Floors.

Table 3, together with the fully calibrated RSP receiver Noise Floor dBm display, is an invaluable tool that lets you immediately determine the noise characteristics of your antenna and the quality of the area in which your antenna is operating. To fully qualify a Receiving System, several tests on as many days will be required, but nevertheless this method allows quick and easy determination of system noise parameters. An example will demonstrate the value of this operating procedure.

26/09/18 - Examples below have been amended and measured values now correct.
 
EXAMPLE 3
On Sept. 23, 2018, at 15:00 GMT, our RSP receiver, connected to a Beverage antenna with N-S orientation and located in a rural area, provides the following measurement data.

F = 1.950 MHz / SR=3.52 / DEC=16
IF AGC= ON / RF AGC=OFF/30
GAIN=65.7 dB
NOISE FLOOR = -110 dBm
Looking at Table 3, we see that this value is even better than the best in the 1.8 MHz row, even considering the “old” galactic noise value of -104.5 dBm plus the rural decile deviation of -3.5 dB for a total noise floor of – 108 dBm. On the other hand the antenna is only 135 m long and therefore it is possible that the low level of received noise is due to excessive antenna loss at this frequency.

F = 3.600 MHz / SR=2.2 / DEC=4
IF AGC= ON / RF AGC=OFF/30
GAIN=57.2 dB
NOISE FLOOR = -93 dBm
Looking at Table 3, we see that this value belongs to the “Rural” location: -94.2 (Noise Floor) +3.5 (decile deviation) for a total of -97.7 dBm.

F = 7.200 MHz / SR=2.64 / DEC=8
IF AGC= ON / RF AGC=OFF/30
GAIN=42.8 dB
NOISE FLOOR = -105 dBm
By looking at Table 3, we are in the “Rural” ballpark.

F = 14.200 MHz / SR=3.08 / DEC=8
IF AGC= ON / RF AGC=OFF/30
GAIN=54.0 dB
NOISE FLOOR = -110 dBm
The “Rural” classification is again confirmed by Table 3.

F = 21.275 MHz / SR=2 / DEC=4
IF AGC= ON / RF AGC=OFF/30
GAIN=71.3 dB
NOISE FLOOR = -114 dBm
Table 3 confirms the “Rural” on 15 m as well.

These measurements and comparisons obviously cannot be made when in presence of rainstorm/lightning noise, but apart from this precaution, several measurements over a significant period of time will provide a firm assessment of Receiving System noise performance. The above example, if confirmed with several tests, would evidence that my receiving system and Beverage antenna are installed in a quiet area, leaving me little or no room for improvement. If you are in doubt about the noise performance of your antenna, or suspect your area to be too noisy, then this procedure will immediately give you an idea of what the ideal situation should be.

Mathematics will be covered by the next post.

[glovisol]

3. MATHEMATICS

The following development is based on the already cited Recommendation ITU-R P.372-13.

The noise power generated by a resistor at the standard reference temperature of 17°C (To = 290°K) over a receiving system noise bandwith b = 1 Hz is calculated by the expression:

Pn = k*To*b [W/Hz] (1) where k = 1.38E-23 is the Boltzmann constant:
Pn = (1.38E-23)*290*1 = 4.002E-21 [W/Hz] = 4.00E-18 [mW/Hz]

But Pn can also be considered as the available noise power from an equivalent lossless antenna.
If we have a receiver free from spurious responses (e.g. receiving a single frequency), the system (antenna + receiver) Noise Factor, coming from a number of different noise sources, is given by:

fa = Pn/(k*To*B) (2) if numerator and denominator of this expression are equal, fa = 1 and the Noise Figure:

Fa = 10*LOG10(fa) (3) becomes equal to zero (minimum possible Noise Figure)

We can express thermal noise power in dBm as follows:

Pd = 10*LOG10 (Pn) [dBm] (4)
P’d = 10*LOG10(4.00E-18) = -173.98 dBm this is the value of the minimum system Noise Floor possible.

The relationship between Noise Floor, expressed in dBm and Noise Figure Fa, expressed in dB, is given by:

Fa = Pd – (10*LOG10(b)) + 173.98 [dB] (5)

If we know Fa, we can obtain Pd:

Pd = Fa + (10*LOG10(b)) – 173.98 [dBm] (6)

We can also write equation (2) as follows:

Pn = Fa + (10*LOG10 (b)) – (10*LOG10(k*To) [dBW] (7)

with: 10*LOG10(b) = B and (10*LOG10(k*To) = - 204, expression (7) becomes:

Pn = Fa + B – 204 [dBm] (8)
 


The ITU Recommendation gives the following expression to calculate Noise Figure as a statistical median variation:

Fam = c – d*LOG10(f) [dB] (9) where f is the operating frequency expressed in MHz and c & d are constants which depend on the man made noise existing in different environments, as described in the previous post.

Tables 2 and 3 have been constructed as follows.

Table 2 uses (5) to calculate receiver’s Noise Figure in dB from measured Noise Floor in dBm.

Table 3 uses (9) and gives constants c and d to calculate predicted Antenna Median Noise Figure Fam for every environment and every frequency band. Then uses (6) to calculate predicted Antenna Noise Floor, which we can now compare to actual receiver’s readings.

Finally the ITU Recommendation gives a table providing the decile deviations of man made noise, as described in the previous post and presented in Table 3.

The other expression for Galactic noise (Table 3 in blue colour) is given by:

198.6 + (186.7-(20*LOG10(f*10^6)) + (10*LOG10(b)) [dBm] (10)

In the next post we shall see how to use the above expressions and tabled data.

[Roger]

I tried replicating your results in my RSPduo and my noise floor is much worse. You got -140 dBm and I get -126.9 dBm. I have two RSPduo and I get the same results on both. Could you check yours again on 20M and maybe post a screen shot? Thanks.

Edited to add the following. My test was done with the SMA antenna ports terminated in 50 ohms. The Hi-Z was terminated in 1000 ohms.

[glovisol]

Hi Roger,

Since you do not specify whether or not your RSPduo is connected to an antenna, I presume it is not connected in your measurement and that the balanced HI Z input is terminated with 1 KOhm. My noise floor readings are taken, as to be expected....on the noise floor, by reading the mobile cursor set on the border. Even your PICT shows - 140 dBm.

[glovisol]

Roger's observation is correct.

The noise floor shown as a baseline on the receiver's dispaly is the dbm value of internal or received noise calculated on the total FFT bandwith: on Roger's screen this is 47 Hz on 3.08 MHz. The value shown on top of the display: - 126.9 dBm is the value of the noise level over the detection bandwith, in this case 1800 Hz. This can be easily proven by applying the data and expressions given in my posts. I shall now proceed and correct the examples given.

[glovisol]

Since it is impossible to amend my #1 post, I upload the corrected PART 1 below. I have also amended the measured values in PART 2 above.

1. RECEIVER NOISE FIGURE FROM TERMINATED NOISE FLOOR

Table 1 shows HF noise floor readings of my RSPduo receiver, S/N 180602D532 with the HIGH Z input terminated with a 1 KΩ resistor and the coaxial inputs terminated with a 50 Ω resistor. All operating parameters are also tabulated for easy reference.

General measurement conditions are shown, so that they can be actually reproduced for other receivers. The RF gain shown in the last column for the two coaxial inputs is the lesser of the two gains found, when different. It is important to do the noise floor measurements using the same SR and DEC values, otherwise inconsistent data will be obtained.

Data in Table 2 has been calculated to quickly determine the receiver’s Noise Figure in each band for any of the three inputs and for a given detection bandwith.

EXAMPLE 1, using noise floor data (Table 1) found for my own RSPduo.
Our receiver, on the 50 Ω terminated coaxial input reads a noise floor of – 120 dBm on Tuner 2 on the 160 m Band ( 1,950 MHz). With a detection bandwith of 1800 Hz (SSB) in Table 2 we read a Noise Figure of 21.4 dB.

EXAMPLE 2, using noise floor data (Table 1) found for my own RSPduo.
Our receiver, on the 1 KΩ terminated balanced input reads a noise floor of – 128 dBm on Tuner 1 on the 40 m Band ( 7,200 MHz). With a detection bandwith of 1800 Hz (SSB) in Table 2 we read a Noise Figure of 13.4 dB.

All in all my receiver shows passable performance, with the worst value of Noise Figure found on Tuner 2, coaxial input, at 3.750 MHz, with a noise floor of -119 dBm corresponding to a Noise Figure of 22.4 dB. As we shall see, the value of the receiver's Noise Figure, even at this level, impacts little on the overall system Signal to Noise Ratio, because unfortunately the antenna noise figures are much higher due to noise received.

[Roger]

Since it is impossible to amend my #1 post, I upload the corrected PART 1 below. I have also amended the measured values in PART 2 above.

You can edit your previous posts (after logging in) by clicking the pencil icon as shown below.....

edit.PNG (7.5 KiB) Viewed 33976 times

Hope that helps.... Roger

[Roger]

There are two ways to calculate the 1 Hz. noise floor on an RSP using SDRuno. One is to use the signal power displayed below the S-meter and the other is to use the waterfall display. In both cases the RSP should be tuned to the band of interest, set to CW or SSB mode and the antenna port terminated with a resistive load (50 ohms for SMA ports and 1K ohms for the Hi-Z port).

Both methods required knowledge of the equivalent noise bandwidth (ENBW) of the filters used to measure the receiver noise. Filters do not act like a brick wall passing the desired signal and then infinitely attenuating the signal outside the band of interest. They have some ripple in the passband, rolloff at the band edges and then steeply attenutate the signal as the frequency increases. So if the filter is a lowpass with an 1800 Hz. cutoff it will still pass some signals above 1800 Hz. When measuring noise we need to know what ideal filter bandwidth would pass the same amount of noise. This is known as the ENBW. The diagram below illustrates the concept.

eqnb.PNG (181.06 KiB) Viewed 33963 times

Once we know the signal power from the output of the filter we can calculate the noise power in a 1 Hz. bandwidth by subtracting
10 log (ENBW) from the power measured at the output of the filter.. Now we will look at the two methods of calculating the 1 Hz. noise floor and the Noise Figure of the receiver. I am using an RSPduo on 20M with a resistive load of 50 ohms on the SMA port of tuner 1. RF gain is set to Max and IF AGC is on. Total gain in the receiver is 76.1 dB (according to SDRuno)

water.PNG (450.25 KiB) Viewed 33963 times

Method 1: A bandwidth of 1800 Hz. has been selected but we need to know the ENBW. SDRplay does not publish the filter specs but a reasonable guess might be 2000 Hz. for the ENBW. SDRuno has calculated a noise output power of -126 dBm using this filter. The noise power in 1 Hz. will be -126 dBm - 10*log (2000) = -126 - 33 = -159 dBm. The noise power in 1 Hz. due to random electron movement in the resistor is -174 dBm at an ambient temperature of 17 Celsius (a common test temperature). So the receiver is 15 dB noisier than the resistor alone and this is the Noise Figure under these conditions.

Method 2: This method uses the waterfall to calculate the noise floor. The waterfall is calculated with a mathematical technique called the Fast Fourier Transform (FFT). One way of viewing the FFT is to consider it to be a bank of filters that simultaneously display their output on the waterfall display. The width of these filters is known as the Resolution Bandwidth (RBW) and this is shown on the bottom right hand corner of the waterfall display. The RBW is determined by the Frequency span of the waterfall (385 KHz. in this case) and the FFT size. The user can change this with the RBW control at the bottom of the waterfall display. The ENBW of the RBW filter is determined by what is known as a Window function that is used alongside the FFT. The Window finction is selected in the Waterfall SETTing as shown below. The Sin^3 is used by default. The one we want is Rectangular because it is the only one that has an ENBW = RBW and this makes the calculations easy. The other Window functions are much wider.

window.png (10.99 KiB) Viewed 33963 times

After setting the FFT window to rectangular we set the FFT averaging time to 128 to average many FFT calculations and get a good estimate of the noise floor in the RBW. In this experiment I get -142.3 dBm in a 47 Hz. RBW. So the noise in 1 Hz is -142.3 - 10 Log 47 or
-142.3 -16.7 which is 159 dBm which is the same as method 1 and gives us the same Noise Figure of 15 dB.

[glovisol]

You can edit your previous posts (after logging in) by clicking the pencil icon as shown below.....

edit.PNG
edit.PNG (7.5 KiB) Viewed 18 times

If you look carefully at my corrections, you discover that in general all posts can be amended as you propose and as I always do, SAVE FOR THE VERY FIRST POST of a thread, WHICH THE SYSTEM DOES NOT ALLOW TO BE AMENDED.... I have asked this quirk to be corrected, but to no avail.