Why do amplifiers have minimum frequencies?

Question

RF amplifiers, both power amplifiers and low noise amplifiers, are rated with a minimum and maximum frequency. For example, the Analog Devices HMC517LC4 Low Noise Amplifier has a frequency range of 17 - 26 GHz, or the HMC374 has range 300 MHz - 3 GHz.

I understand why there's a maximum frequency: because the underlying components can only switch at a certain speed, because too high of a switching speed can cause ground bounce, and because at too high of a frequency the wavelength is short enough that conductors no longer have uniform voltage.

Why is there a minimum frequency? What happens if e.g. I try to use one of those components for 100 MHz or 1 MHz or even 1 kHz? Why?

I ask because there are plenty of both LNA and PA listed for microwave frequencies, but almost nothing listed for HF or below. What would the impact of using a microwave LNA or PA on HF frequencies be?

Answer 1

Low frequency noise can become fatal to the high frequency noise performance if the low frequency signal is not prevented to be amplified. Low frequency noise can be strong in semiconductors when compared to the wide band noise, but that's mentioned already by others.

How does LF noise become harmful? The high frequency signals are the interesting ones. Shouldn't a filter remove the LF noise? The amp isn't especially linear when compared to what can be achieved with feedback in near-DC opamp circuits. There's some intermodulation (=unwanted mixer operation between the summed signals).

The intermodulation products between the wanted signal and LF noise are as harmful as noise, because they can occur in the same frequency band as the wanted signals. The designers of the LNA IC surely have found that to keep the intermodulation with the LF noise low, the operating frequency range should be limited.

You surely can find low noise amps for low frequencies. They can be constructed without using so special parts that any marketing hype would be useful. The preamp before the mixer stage in an ordinary 88-104 MHz FM radio is a LNA when the idea of its meaning is considered. A very tricky area of low noise amplifying is the making the mic preamp for audio recorders which can capture whispering or a mosquito from several meters (not centimeters) without too much audible hiss.

Answer 2

What would the impact of using a microwave LNA or PA on HF frequencies be?

If it's AC coupled, it has a low frequency cutoff.

Besides that, there's going to be filters and bias networks ; if the output is the drain of a FET that needs to be biased then the output is connected to an inductor which also sets a low frequency cutoff.

If other types of automatic biasing circuitry is employed then... it also has a low frequency cutoff.

Even if the circuit is DC coupled and could in theory have a bandwidth that goes down to DC, besides cost and other issues, an important one is noise. Fast semiconductors tend to be tiny and that implies a high 1/f noise corner. All semiconductors have 1/f noise that goes down at -20dB/decade ; at the noise corner frequency it meets the noise floor, and above that noise density is rather constant and measured in \$ nV/\sqrt{Hz} \$

For the sake of example, I picked a random GHz gain-bandwidth opamp, OPA858, which seems to have a 1/f noise corner frequency around 200kHz.

Lower bandwidth opamps optimized for low frequency noise performance can have a 1/f corner of 100Hz or below (example: ADA4898) which means the extra bandwidth of the GHz opamp comes at the cost of more than 1000x worse low frequency noise.

Answer 3

A while ago I played a little with SiGe transistors such as BFG425W and BFP650. I built some cascaded amplifiers and a noise source. The noise generator had about 5 stages of amplification.

I noticed that the SiGe transies had even more gain below their nominal band of intended use (hundreds of MHz and above). I.e., their gain was growing further, towards DC. Initially, I made the amp's bandwidth reach far towards DC - only to find out that my multi-stage amplifier picked up a feedback oscillation via the power supply rail. Although there were many low-pass filters, somehow the sheer DC gain and some phase shift in the filters were enough to make the thing oscillate at some relatively very low frequency (near DC).

My way to stabilize the amp was: use much smaller AC coupling capacitors between the stages, and also make the negative feedback "tilted" to ramp up the gain towards the upper end of the band (countering the inherent gradual decrease in gain of the semiconductors).