Why would you put an attenuator before an amplifier?

[MrOmnos]

Just saw a circuit where they had an attenuator before a LNA. To a noob like me, it makes zero sense and seems counter intuitive. Why attenuate a signal if you are going to amplify it but I am pretty sure there some reasoning behind it. Let first see what answers we get and then I will post the circuit.

[pigrew]

I can think of two reasons: help the input matching of the amplifier, and to reduce the input signal power if it is too powerful and above the safe input power limit of the amplifier. Note that many signal generators do not have a huge dynamic range. But, there are downsides such as increased noise due to the attenuator.

Sent from my LG-D850 using Tapatalk

[John Coloccia]

Hard to say. Sometimes, to minimize noise, you take in a signal that's pretty hot and then attenuate it down. That can squash some of the thermal noise in the system. Maybe the signal coming in is just too hot for the front end of the amp? Maybe some designer wasn't sure what signal he's looking at, or makes changes to the attenuation depending on the application. Would be easier to guess with the circuit.

I have a situation now where I need to attenuate the output of an amp, send it through a filter, and then use a recovery stage to bring the level back up. The filter just can't handle the levels coming out of the first amp, and the nature of what I'm doing with the first amp means that it can't be changed to operate at lower levels.  It looks funny, but that's another potential reason.

[Kleinstein]

Some LNAs have a not so good defined (e.g. temperature dependent, nonlinear) input impedance. An attenuator can improve termination and thus reduce possible problems with reflected signals.

[Kalvin]

Here are two articles describing the improved matching using attenuators:

"Fixed attenuators help minimize impedance mismatches":
https://www.minicircuits.com/app/AN70-001.pdf

"VSWR Reduction by Matched Attenuator":
http://www.rfcafe.com/references/electrical/vswr-reduction.htm

[MrOmnos]

Here's the signal path diagram
Operation is for ISM band.


Here is the Max rating of the amp


And here is the Output power curve of the VCO used


So we now know that the amp can handle the input power.

Here are few more graphs

VSWR vs Frequency curve for the LNA


VSWR vs Frequency curve for Attenuator[3dB]


As we can see above, input VSWR of LNA is close to 1 for ISM band operation, so why do you need fixed attenuator. Could you just feed the VCO output to the LNA? How would that affect the performance?

[They are all mini circuit components]

[John Coloccia]

That's just the max input power rating where the thing doesn't break. That doesn't mean that it hasn't already slammed into the wall at 2dBm, just as an example.

[John Coloccia]

Also, if you have a fixed gain with a small signal on one side, and a huge signal on the other, it makes sense to attenuate the small signal unless you have a reason to do it differently. Smaller/cheaper parts potentially, thermal advantages and things like that.

[PA0PBZ]

Also, you are not telling us what the gain of the amplifier is, or what input level the splitter and mixer can handle.

[rfbroadband]

here is the reason:
- dymanic range

+ for lower signal levels you are trying to get as close as possible to the noise floor, thus the attenuator will be set to 0dB and the amplifier will amplify the signal while maintaining good NF.

+ at higher signal levels you don't care about noise, the input signal to noise ratio is good and at higher signal levels, you mainly care about distortion. At higher signal levels the LNA increasingly adds distortion and at some point it would simply go into compression. In order to delay the point at which the LNA adds too much distortion or compresses you increase the attenuation as the signal level increases (called front end AGC). As mentioned the noise penalty introduced by the attenutor does not matter because the signal is strong in this scenario.

I hope this helps.

[whollender]

Looking at the LNA datasheet, it looks like the attenuator is to keep the LNA from getting too compressed and generating a bunch of harmonic products.

The LNA's 1dB compression point happens at an output power of 18dBm.  The LNA has a gain of ~14dB.  The VCO output is around 5-6dB.  Without the attenuator, the LNA would be operating right around its 1dB compression point.  With the attenuator added, the LNA will be operating in a much more linear region.

Your signal flow diagram doesn't show any filtering, but, even with filtering, the additional harmonic content could be a big issue because it is being put directly into the Tx antenna, and being fed into the receive mixer.

Additionally, the attenuator provides a broadband match.  Both the LNA and the VCO will be pretty well matched in the band that they are designed to operate in, but they may have very bad match outside of that band, which could result in other problems if they are hooked directly together.

[MrOmnos]

Looking at the LNA datasheet, it looks like the attenuator is to keep the LNA from getting too compressed and generating a bunch of harmonic products.

The LNA's 1dB compression point happens at an output power of 18dBm.  The LNA has a gain of ~14dB.  The VCO output is around 5-6dB.  Without the attenuator, the LNA would be operating right around its 1dB compression point.  With the attenuator added, the LNA will be operating in a much more linear region.

Your signal flow diagram doesn't show any filtering, but, even with filtering, the additional harmonic content could be a big issue because it is being put directly into the Tx antenna, and being fed into the receive mixer.

Additionally, the attenuator provides a broadband match.  Both the LNA and the VCO will be pretty well matched in the band that they are designed to operate in, but they may have very bad match outside of that band, which could result in other problems if they are hooked directly together.

Why would a simple sinusoidal wave have additional harmonic content?

[whollender]

To start with, the output of the VCO isn't a simple sine wave.  If you look closer at the datasheet, you'll see a plot of harmonic amplitudes vs tuning voltage.

And, even if the VCO had a perfectly clean output spectrum (no harmonics at all), the distortion in the LNA will create harmonics.  The amplitude of the harmonics depend on the amount of distortion, which will be pretty high once you hit the 1dB compression point.

At the 1dB compression point, the amplitude is getting close to the point that the amplifier can't generate any more output power, no matter the input power, and starts to clip the output waveform, resulting in high levels of harmonic distortion.

[MrOmnos]

To start with, the output of the VCO isn't a simple sine wave.  If you look closer at the datasheet, you'll see a plot of harmonic amplitudes vs tuning voltage.

Yes, just noticed it. This is a diagram from MITs ISM band synthetic aperture radar. The theory here is:
A FMCW (here up-chirp) is generated using a ramp generator as the input of the VCO. So, at an instant Tx frequency might be lets say 2 GHz. This Tx signal is reflected off of a target and received by the Rx antenna. Some time t passed during the period transmitted signal reached the target and got reflected. During the time t the transmit signal has changed from 2 GHz to lets say 2.00004 GHz (because the signal is ramping) so now if we mix (down converte) the current Tx signal 2.00004 GHz and received signal 2 GHz  we get an IF of 40 kHz which can be easily sampled and processed. Here IF will contain the distance information. Greater IF means greater target distance. For direction of motion of the target CW is transmitted and then received Doppler shifted signal is mixed with the CW and IF determines if it is moving towards or away.

For harmonics we will have the following conditions when mixed (I don't know if I am doing this correct)
1) Harmonic frequencies of Tx at an instant t will get mixed with received signal (which was fundamental Tx at some instant <t) and produce Ghz signal which will not be picked up by the Preamp of the processing system. No worries [Harmonics mixing with fundamental frequency]
     i.e. 4.00008-2 = 2.00008 GHz

2)Harmonic frequencies of Tx at an instant t will mix with the reflected Harmonics transmitted at some instant prior to t which will produce imaginary targets at twice the distance of the real target. [Harmonics mixing with themselves]
   i.e 4.00008-4 = 80KHz

So, how do you solve the imaginary target problem?
Add a filter which lets in 2-3 GHz signal?

[whollender]

So, how do you solve the imaginary target problem?
Add a filter which lets in 2-3 GHz signal?

Yep, that's the idea.  You generally always want to filter out signals that aren't in your band of interest to avoid noise and interference issues.

You can determine your filtering needs by looking at the relative signal strength of your wanted signal and the one generated by the harmonics.  At a minimum, I'd suggest one prior to the Tx antenna (to prevent any harmonics from radiating) and one before the receive LNA (to prevent out of band signals from interfering or even overloading the LNA).

[KE5FX]

Another reason is isolation.  Sometimes you want low gain (or none at all) in the forward direction but lots of loss in the reverse direction.  You might see an attenuator before or after an amplifier in that case, depending on the signal levels and the need to preserve SNR. 

Isolation amps and/or circulators are common at the output of microwave oscillators, because variations in the load VSWR can influence the output frequency.  The combination of a pad followed by an amplifier provides the oscillator with a very stable load.

[David Hess]

Another reason is isolation.  Sometimes you want low gain (or none at all) in the forward direction but lots of loss in the reverse direction.  You might see an attenuator before or after an amplifier in that case, depending on the signal levels and the need to preserve SNR. 

Isolation amps and/or circulators are common at the output of microwave oscillators, because variations in the load VSWR can influence the output frequency.  The combination of a pad followed by an amplifier provides the oscillator with a very stable load.

You just barely beat me to this.

Receivers often add an RF amplifier before the first mixer just to prevent radiation of the first local oscillator and any mixing products.  The RF amplifier will compromise the dynamic range unless an attenuator is included.  In lower frequency bands, the increase in input noise will not even be a factor.

[Zero999]

I can think of two reasons: help the input matching of the amplifier, and to reduce the input signal power if it is too powerful and above the safe input power limit of the amplifier. Note that many signal generators do not have a huge dynamic range. But, there are downsides such as increased noise due to the attenuator.

We do this at work. The signal generator has a resolution of 0.01V and only a 10V range which is too course for many applications. One thing we do is use it as the signal source for an audio amplifier, driving a shaker for vibration testing. The audio amplifier will clip if the full 10V is put into its input so an attenuator is added between the signal generator and the amplifier.