General Purpose LNA Box

2025

Low Noise Amplifiers (LNAs) allow the signal to noise ratio (SNR) of low power signals to be preserved as they pass through lossy parts of a receiver chain. Whether an LNA will benefit a particular receiver application depends on the noise received by the antenna and the losses in the receiver. An LNA will only help if its noise figure (NF) is lower than the receiver itself, and it will only help if the received signal has a low noise floor. You will not generally see a benefit from using an LNA at low frequencies where galactic noise far exceeds the thermal noise floor — this is why AM radios can get away with using very lossy antennas: You can add loss after the antenna without significantly reducing the SNR because the received noise floor is much greater then the thermal noise floor.

Assembled LNA Box

I built this box as a rugged and convenient solution for improving the sensitivity of receivers and SDRs. Without an external LNA, the noise figure of a HackRF One has been measured at 11dB and the noise figure of an Airspy Mini (my preferred SDR for frequencies in its supported range) has been measured as 8dB. With an external LNA, you can bring the system noise figure much closer to the noise figure of the LNA (the math for this is explained in the next section). By using an LNA part footprint which supports three different LNAs optimized for different frequencies and by including standard filter footprints, this general purpose LNA offers good performance for a variety of applications from 400-4000 MHz depending on the LNA installed.

Design files can be found on Github.

Theory

The performance of a receiver can be characterized with noise temperature and with noise figure. Noise figure is the ratio between the SNR at the input of a device and the SNR at the output of a device in terms of decibels. Noise factor is the same ratio but in linear terms. Noise factor is denoted with F while noise figure is denoted as NF.

\[F=\frac{SNR_{input}}{SNR_{output}}=\frac{S_i/N_i}{S_o/N_o}\tag{1}\] \[NF=10log_{10}\Big(\frac{SNR_{input}}{SNR_{output}}\Big)\tag{2}\]

The noise temperature of a component in a system with a reference temperature of 290 K (room temperature) can be calculated from its noise factor:

\[T_e=290(F-1)\tag{3}\] Noise factor can also, of course, be calculated from its noise temperature: \[F=\frac{T_e}{290}+1\tag{4}\]

So a part with a noise figure of 1dB has a noise temperature of 75 K:

\[T_e=290(10^{1/10}-1)=75 K\tag{5}\]

Normally you can treat attenuation as equal to noise figure for attenuators, but if the attenuator is a different physical temperature, then you need to use this equation:

\[NF_{attenuator}=10log_{10}\Big(1 + \frac{1-G}{G}\cdot \frac{T}{T_0} K\Big)=10log_{10}\Big(1+\frac{1-0.1}{0.1}\cdot\frac{290}{290}\Big)=10log_{10}(1+9)=10 dB\tag{6}\]

The Friis formula for noise can be used to calculate the effective noise figure of a receiver composed of multiple components in series. Take, for example, a receiver composed of a 20dB gain, 1dB NF LNA followed by a 6dB coax cable ending in a receiver with a noise figure of 5dB. The effective noise figure of this receiver system is far less than the receiver by itself.

Noise Cascade Example

\[F_{system}=F_1 + \frac{F_2-1}{G1}+\frac{F_3-1}{G_1G_2} = 10^0.1+\frac{10^{0.6}-1}{10^2}+\frac{10^{0.5}-1}{10^2\cdot 10^{-0.6}}=1.26+0.03+0.09 = 1.37 \Rightarrow NF=10log_{10}(1.37)=1.38dB\tag{7}\]

This illustrates the critical point about noise in receiver systems: the gain of the first amplifier stage of a receiver has the greatest impact on the noise figure of a receiver (except where loss after the LNA exceeds the gain of an LNA; in this degenerate case, that loss dominates the noise figure).

The receiver in this example isn't really the final sink for the signal; that would be the demodulator, but the demodulator is out of scope for this discussion. For the purpose of understanding the above example, imagine that the receiver is composed of a cascade of components such as mixers and filters and that the output of this block is a baseband signal going into a perfect ADC. The degradation in the SNR of the signal from the input to the output of the receiver block is its noise figure.

System Noise

The previous discussion on noise only considers the antenna and receiver in isolation. To calculate the actual SNR of the signal at the input of the antenna, you need to know the noise floor seen by the antenna. This noise floor will not by 290k; it varies with operating frequency and with antenna gain & orientation. As will be shown in later examples, antenna noise is why LNAs are not equally useful in all applications.

System Noise Temperature Reference Locations

When calculating the performance of a receiver system, you must refer all noise to some point in the system. That point can be the aperture of the antenna, the output terminal of the antenna, or the input to the receiver.

There is some international standardization on this topic. Per ITU Recommendation P.372-17 "The only appropriate reference point for the overall operating noise for a radio receiving system is the input of an equivalent loss-free receiving antenna. (The terminals of this lossless antenna do not exist physically.)" This means that the input to the antenna should be the reference point for noise. Annoyingly, the recommended calculation for antenna G/T (the overall figure of merit for antenna performance) is performed at the receiving equipment input port. It is easy enough, though, to transfer the antenna noise temperature and gain to this point.

ITU Recommendation P.372-17 provides this formula to calculate the system noise factor at the antenna aperture:

\[f=f_a+(f_c-1)+l_c(f_{tl}-1)+l_c l_{tl} (f_r-1)\tag{8}\]

Where \(f\) is the system noise factor, \(f_a\) is the external noise factor (sky noise), \(f_c\) is the noise factor associated with antenna circuit losses (antenna efficiency), \(l_c\) is the antenna loss (equal to the inverse of antenna efficiency), \(f_{tl}\) is the noise factor associated with transmission line losses, \(l_{tl}\) is the transmission line loss (ratio of input power to output power of the lossy line), and \(f_r\) is the internal noise factor of the receiver.

This can also be expressed in terms of effective noise temperature instead of noise factor:

\[t_{sys}=t_a+t_p(l_c-1)+l_c t_{tl}(l_{tl}-1)+l_c l_t t_r=t_a+t_p(\frac{1}{e_a}-1)+\frac{1}{e_a} t_{tl}(\frac{1}{e_{tl}}-1)+\frac{1}{e_a e_{tl}} t_r\tag{9}\]

Where \(t_{sys}\) is the system temperature, \(t_a\) is the external noise (sky noise), \(t_p\) is the physical temperature of the lossy components of the antenna, \(e_a\) is the antenna efficiency, \(t_{tl}\) is the physical temperature of the lossy transmission line, \(e_{tl}\) is the transmission line efficiency, and \(t_r\) is the equivalent temperature of the receiver.

The multiplication of the individual temperatures by the losses of various stages is a result of referring the noise to a particular point in the system (in this case the antenna aperture). For a much more complete derivation, see this presentation from McMaster University also cached here.

To translate the system noise temperature to the input of the receiver (which is useful for calculating G/T), just multiply \(t_{sys}\) by \(e_a e_{tl}\):

\[t_{sys}^R=t_{sys}e_a e_{tl}=t_ae_a e_{tl}+t_p(1-e_a)e_l+t_{tl}(1-e_{tl})+t_r\tag{10}\]

\(t_p\) (Antenna Physical Noise) and \(t_{tl}\) (Antenna Feed Losses)

The radiation efficiency of an antenna and its physical temperature define the physical noise of an antenna.

\[t_{p-eqv}=t_p(\frac{1}{e_a}-1)\tag{11}\]

So an antenna with a radiation efficiency of 90% and a physical temperature of 290k will have a physical noise temperature of:

\[t_{AP}=290(\frac{1}{0.9}-1) = 32 K\tag{12}\]

Loss in the transmission line after the antenna port and the temperature of that transmission line also contribute to system noise temperature.

\[t_{tl-eqv}=t_{tl}(\frac{1}{e_{tl}}-1)\tag{13}\]

\(t_a\): Sky Noise

Physical objects have an equivalent noise temperature (also known as brightness temperature) based on both their physical temperature and on their emissivity.

\[T_B=\epsilon T_P\tag{14}\]

Where \(T_b\) is brightness temperature, ε is the emissivity, and \(T_p\) is the physical temperature.

Consider an infinitely large 1000 K copper plate facing an infinitely large plane of RF absorber cooled by liquid nitrogen to 77 K. The emissivity of the copper pate will be very low (perhaps 0.01) and the emissivity of the absorber will be high (near 1).

In this situation, a receiver with a direction antenna pointed at the metal plate would see the sum of the power emitted by the plate \((0.01\cdot 1000=10 K)\) and the reflected power of the absorber \((0.99\cdot 77 = 76 K)\) for a total of 86k.

This may seem like an impractical and academic example, but it is actually close to the situation seen in microwave radiometry of the surface of the earth by satellites. At microwave frequencies, the sky has an equivalent noise temperature as low as 2.7 K (the cosmic microwave background). A satellite observing the ocean will see a brightness temperature much less than the ocean's physical temperature as a result of the ocean's low emissivity and the reflection of the cool sky. Microwave radiometry can be used to discern ocean salinity, soil moisture content, and sea ice extent. Land, having a drastically different emissivity from the ocean, show up rather distinctly even on communications systems not designed specifically for microwave radiometry.

A directional antenna pointed at the sky will, similarly, see a much lower brightness temperature than it will pointed toward the ground. Beyond radio astronomy, this has implications for satellite communications: The noise that your signal is sitting on top of is far lower in absolute terms than the noise seen by a terrestrial receiver, so the performance of the first stage LNA in a satellite ground station is more critical than a terrestrial radio receiver.

Below a few GHz, atmospheric noise, solar noise, and galactic noise greatly increase the noise floor above the 2.7 K cosmic microwave background frequency. At 100MHz, for example, sky temperature can vary between 500 and 500,000 K depending on the gain of the antenna and where it is pointed.

Antenna External Noise from ITU-R P.372-17

As a result, you won't benefit significantly from adding an LNA to a 100 MHz FM radio receiver (especially one near a city). An isotropic antenna would see half of its sphere of coverage (the sky) at 40,000 K and half of its sphere of coverage (the ground) at more than 300k (higher depending on emissivity). Averaging between the two, you get a sky temperature of 20,150 K. Assuming that the antenna and transmission line before the receiver is lossless, the system noise temperature will be the sum of the antenna external noise temperature and the receiver noise temperature. If your FM radio receiver has a noise figure of 10dB = 2600 K, then the noise power at the receiver can be calculated by

\[P_n=k\cdot t \cdot =1.38 \cdot 10^{-23}\cdot 22,760 \cdot 106,000=3.3\cdot 10^{-14}W = -104.8 dBm\tag{15}\]

Where \(k\) is Boltzmann's constant, \(t\) is the temperature, and \(B\) is the bandwidth. The mono and stereo FM signals occupy 106kHz of bandwidth which is the bandwidth also used for noise power. Depending on the exact architecture of the receiver, it may be sensitive to noise in a somewhat wider bandwidth. If the received signal is -90 dBm, then the SNR is 14.8dB.

If you add a 20dB gain 0.5dB = 35 K noise figure / temperature LNA in front of the receiver, the receiver noise temperature is reduced to:

\[F_{receiver}=F_1 + \frac{F_2-1}{G1}=10^{0.5/10}+\frac{10^{10/10}-1}{10^{20/10}}=1.12+\frac{10-1}{100}=1.21 \Rightarrow T_{receiver}=290\cdot(1-1.21)=61 K\tag{16}\]

With this receiver noise temperature, the noise power at the receiver reduces from 22,760 to 20,211 K = -105.3 dBm for an SNR of 15.3 dBm. So adding a low noise figure LNA only improved the SNR by 0.5dB. Even beyond the monetary cost, that 0.5dB does not come for free. Adding 20dB of gain means that the dynamic range of the receiver has been reduced by 19.5dB. You can detect 0.5dB lower power signals now, but high power signals can saturate the receiver. Its also possible that enough power will be present from the antenna that the LNA produces intermod products from various strong received signals. If those intermod products fall in the receive bandwidth of the receiver, the SNR can be significantly degraded.

These concerns are less applicable to a satellite application where all signals are weak and directional antennas attenuate spatially diverse interferers.

Attenuation in the atmosphere is also a component of sky noise. Losses in the atmosphere are proportional to the path length through the atmosphere which is a function of elevation angle — you will never measure the 2.7 K cosmic background at an elevation angle of 80 degrees due to attenuation in the atmosphere.

\[T_B=\epsilon T_P\tag{17}\]

System G/T

The system Gain to Noise Temperature ratio (expressed in units of dB / K) defines the overall sensitivity of a receiving system. It must be calculated at a particular point in a receiver. The ITU recommendation S.733-2 calculates it at the input to the receiver which is what I will use here.

\[\frac{G}{T}=G_{ant}+G_{tl}-10log_{10}(t_{sys}^R) dB/K\tag{18}\]

Example 1: 3m Diameter Television Receive Only (TVRO) Parabolic Dish

Assumptions: antenna gain is 39dB at 4GHz, antenna radiation efficiency is 95%, antenna aperture efficiency is 50%, feed line losses are 0.1dB \((e_{tl}=10^{-0.2/10}=0.977)\), and the receiver noise temperature is 30 K.

The gain calculation is easy. The gain number from the manufacturer should be inclusive of the feed line losses to the mounting flange of the LNB (low noise block downconverter), so the gain for this G/T calculation is 39dB. Since this is the gain at the receiver, we must calculate temperature at the input to the receiver.

The system temperature at the receiver is:

\[t_{sys}^R=t_ae_a e_{tl}+t_p(1-e_a)e_{tl}+t_{tl}(1-e_{tl})+t_r=2.9*0.95*0.977+290*(1-0.95)*0.977+290*(1-0.977)+30=2.7+14.5+6.7+30=53.9 K\tag{19}\]

So the G/T of this system is:

\[\frac{G}{T}=G_{ant}+G_{tl}-10log_{10}(t_{sys}^R) =39-0.1-10log_{10}(53.9)=21.6 dB/K\tag{20}\]

Example 2: A Cell Phone

The purpose of this example is to show that G/T can easily be negative. Assumptions: The noise figure of the cell phone is 9dB (per the 3GPP LTE spec), 3dB of that is duplexer losses (considered as feed-line losses here) and the antenna is 63% efficient (2dB loss), the operating frequency is 0.9 GHz, and the antenna is isotropic (0 dBi). Since the antenna is isotropic, it sees the average of the ground and the sky which is ~650 K in a city at this frequency.

\[t_{sys}^R=t_ae_a e_{tl}+t_p(1-e_a)e_{tl}+t_{tl}(1-e_{tl})+t_r=650*0.63*0.5+290*(1-0.63)*0.5+290*(1-0.5)+290\cdot (10^{(9-3-2)/10}-1)=205+54+145+438=842 K\tag{21}\]

An isotropic antenna with antenna efficiency of 63% has a gain of 0.63 in linear terms, so the G/T of this system is:

\[\frac{G}{T}=G_{ant}+G_{tl}-10log_{10}(t_{sys}^R) =10log_{10}(0.63)+10log_{10}(0.5)-10log_{10}(842)=-2-3-3=-8 dB/K\tag{22}\]

There is a good online calculator here if you'd like to calculate G/T for yourself.

Noise Figure Degradation from Blockers

The noise figure of a particular LNA is a function of a variety of factors including frequency, biasing, physical temperature, and received power. The impact of received power on noise figure is why you will often see filters before an LNA. Signals other than the signal of interest are referred to as blockers.

Blocker signals degrade receiver noise figure in several ways:

This 2002 QEX article by Ulrich L Rohde covers these factors in depth, and this RF Cafe article has several informative figures to help understand how these factors are interrelated. Image rejection is covered in This Analog Dialogue article.

Note that it is generally the case that blockers are only a problem when they far exceed the power of your desired signal.

To mitigate blockers close in frequency to your signal of interest, you need to increase the linearity of the LNA (more power), increase the dynamic range of the receiver (more cost), and use a lower phase noise LO (more cost). If a blocker is sufficiently far from your signal of interest, though, you can attenuate it with filtering. What constitutes close and far is set by the filters and your required tuning range. Another option for reducing power at the receiver is to use a very narrow band antenna; if out of band power ian't received by the antenna in the first place, it doesn't need to be attenuated by a dedicated filter

Whether the filter should go before or after the LNA depends on the expected received power. In a device with multiple transceivers the the Amazon Project Sidewalk Development Board, you could end up jamming bluetooth while transmitting at 900 MHz and vice versa if the receivers aren't protected against strong out of band signals. In quiet RF environments, you will get better performance without a filter, though, because filters add losses. In many cases, the frequency selective nature of matching circuits in receivers along with the antenna's limited bandwidth is sufficient to avoid desense from strong out of band blockers.

If you are trying to receive weak signals in a city with a log periodic yagi antenna (quite broadband) with an SDR (no frequency selective filtering), you're probably going to want a filter before the SDR and before the LNA if you are using an external LNA.

Design

This is the first of a few "RF Box" projects. I have previously used Hammond 1590BBK boxes for projects, but these boxes are unnecessarily large for simple RF designs, and it is, annoyingly, more than 100mm in maximum dimension which (slightly) increases cost. These "RF boxes" are based around the more appropriately sized Hammond 1550A box but follow the same design principal: The PCB is part of the enclosure and all parts that aren't connectors should be on the bottom of the board inside the box. I used a two layer board from JLCPCB here since the routing is simple. Yes: you can build reasonable performance 2.3 GHz circuits on common two layer stackups. You just need to be mindful of the loss by keeping routing short.

Inside of LNA Box

I experimented with conductive fabric covered foam to make a better EMI closeout, but this is definitely not a production ready solution. The fabric covered foam has a peel and stick adhesive that I connected to the box, but it doesn't hold up well to the shear loading it sees from the PCB pressing onto the side of the foam. This foam isn't really needed; I was just trying it out.

The PCB was assembled at JLCPCB and the only parts that I had to hand install were the SMA connectors, USB connector, and reverse mounted LED.

Part selection

The goal for this project is a good performance general purpose LNA. That means that I want low noise figure (<1dB), good gain (>15dB), high P1dB (>10dBm) to have the linearity to handle moderate blocker signals, reasonable price, and a package that can easily be soldered. The best part I found for this in 2024 was BGU8052 which shares a common footprint with BGU8051 and BGU8053. The different part numbers are optimized for different frequencies but otherwise have identical reference circuits (aside from bias inductor). JLCPCB had stock of BGU8053 which is really only ideal for the 13 CM band, but that is what I had assembled for the first revision of this board. Here's a comparison of the performance of the three parts:

Parameter BGU8051 BGU8052 BGU8053
Recommended Frequency
 0.3-1.5 GHz 1.5-2.5 GHz 2-6 GHz
Gain 18.2 dB 16.8 dB 18.9 dB
Noise Figure 0.44 dB 0.56 dB 0.58 dB
P1dB 16.3 dBm 15.5 dBm 11.4 dBm
OP3 35.3 dBm 30.6 dBm 30.1 dBm
Bias Inductor 33 nH 15 nH 15 nH

The characteristics above are taken from the datasheet assuming a 3.3V supply, 48mA biasing, and with BGU8051 at 900MHz and the other two parts at 2.3 GHz. I selected 3.3V so that I could easily power the box off of a USB port using a cheap linear regulator. Since all modern smartphones can power devices plugged into their USB-C ports, I've found myself using old phones as power banks for these LNAs in the field.

Here's the RF part of the schematic which is a pretty straightforward implementation of the datasheet recommendations:

LNA Schematic RF Portion

All passives are 0603 case size: the width of a 50 Ω trace on this two layer stackup is close to the width of an 0603, so I didn't bother using 0402s.

Filters

Since it isn't always advantageous to put a filter before the LNA due to the filter insertion losses, I left the filter parts separate and selectable with SMA jumpers. I initially didn't like the idea of adding loss as a result of using SMA jumpers, but the losses are actually pretty minimal. A Minicircuits 141-3SM+ 3" SMA cable has a typical insertion loss of less than 0.1 dB up to 5 GHz. There are two filter options: A 915 MHz filter and an unpopulated option with a standard 3x3mm footprint.

The filter can be placed before or after the LNA depending on the noise environment. At home in the city, I generally put the filter before the LNA. If you are downlinking a signal from a cubesat in a mountain valley, you would probably want to put the filter after the LNA.

Result

This is just an implementation of an LNA reference design put in nice packaging. It performs as expected and works nicely along with a filter as a front end before an SDR. Here is a quick gain measurement made with a Nano VNA:

LNA Gain

The plot in blue is the gain with a filter, and the plot in brown is the gain of the LNA with the 915 MHz filter at its output.

Gain Measurement Setup

As expected from the gain measurement, you can clearly see the passband of the filter with a spectrum analyzer. You will see this even with no signals present because the noise floor itself with be amplified by the LNA and then attenuated by the filter outside of the filter's passband.

Observing 23 cm Amateur Radio Band with LNA Box

I attempted to measure the noise figure of the LNA with the method shown here (except I used a nano VNA to get the gain at the frequency of interest) and unsurprisingly measured a noise figure below what this scrappy method is capable of measuring showing that I had at least not severely messed up the implementation of the reference design.