[SOLVED] cascading multiple gain stages

I am trying to amplify a 30dB ENR noise source to +30dBm. For my initial stages I am using a MMIC darlington configuration amplifier with a gain of ~19dB (these are input and output internally matched to 50 ohms). I made two pc boards one with 3 stages (54 dB gain) and one with 2 stages (37 dB of gain), each work well independently, but when I cascade them I never get larger than 72 dB of gain instead of the total of 90 dB. It is almost like I am prematurely saturating the amplifiers. I am using low pass filters between stages to attenuate any high frequency oscillations and I don't see any low frequency oscillations when I test with a CW source. I have a power amplifer for the front end to generate the +30dBm but I need to get the initial low power stages working. Is this even possible with darglington gain stages? Should I be using a different configuration? Any suggestions would be helpful.

comsys_engineer said:

I am trying to amplify a 30dB ENR noise source to +30dBm. For my initial stages I am using a MMIC darlington configuration amplifier with a gain of ~19dB (these are input and output internally matched to 50 ohms). I made two pc boards one with 3 stages (54 dB gain) and one with 2 stages (37 dB of gain), each work well independently, but when I cascade them I never get larger than 72 dB of gain instead of the total of 90 dB. It is almost like I am prematurely saturating the amplifiers. I am using low pass filters between stages to attenuate any high frequency oscillations and I don't see any low frequency oscillations when I test with a CW source. I have a power amplifer for the front end to generate the +30dBm but I need to get the initial low power stages working. Is this even possible with darglington gain stages? Should I be using a different configuration? Any suggestions would be helpful.

Click to expand...

I think you need to do a quick power cascade analysis and figure out if you are driving the later gain stages in to saturation (likely). You can use a simple spreadsheet to accomplish this. Start with your input signal level, add the typical gain of the first amp stage to determine the expected output power level, then compare the expected power level to the saturated output power (Psat) spec of the device. In leiu of Psat, you can use P1dB and assume Psat is about 1-2 dB above that (good approximation). Repeat that simple math and comparison for each gain stage in your lineup. I have a sneaky suspicion that your final gain stage devices will need to have a higher Psat/P1dB level.

Another way to confirm that you're starting to compress the signal is to use a signal generator and spectrum analyzer or power meter. Ramp the input power level up in steps and monitor the output power. You should get a 1:1 increase (increase input power by 3 dB, output power increases by 3 dB), until you start to reach the point of compression in one of the stages. Then, you'll be fighting a losing battle; raise the input 3 dB, output only goes up by 2 dB. Eventually it gets worse; raise the input 3 dB, output goes up 1 dB. Finally, the output level will stop increasing because you've reached saturation on at least one of your gain stages.

enjunear said:

I think you need to do a quick power cascade analysis and figure out if you are driving the later gain stages in to saturation (likely). You can use a simple spreadsheet to accomplish this. Start with your input signal level, add the typical gain of the first amp stage to determine the expected output power level, then compare the expected power level to the saturated output power (Psat) spec of the device. In leiu of Psat, you can use P1dB and assume Psat is about 1-2 dB above that (good approximation). Repeat that simple math and comparison for each gain stage in your lineup. I have a sneaky suspicion that your final gain stage devices will need to have a higher Psat/P1dB level.

Another way to confirm that you're starting to compress the signal is to use a signal generator and spectrum analyzer or power meter. Ramp the input power level up in steps and monitor the output power. You should get a 1:1 increase (increase input power by 3 dB, output power increases by 3 dB), until you start to reach the point of compression in one of the stages. Then, you'll be fighting a losing battle; raise the input 3 dB, output only goes up by 2 dB. Eventually it gets worse; raise the input 3 dB, output goes up 1 dB. Finally, the output level will stop increasing because you've reached saturation on at least one of your gain stages.

Click to expand...

Thank very much you for the suggestion. It turns out I was saturating the cascaded amplifiers with the CW test signal input. Once I lowered the CW input signal down to -80 dBm the cascaded amplifiers are working as they should and I get the expected +10dBm output. However, when I use the noise source as the input, the output level drops to a -13dBm. The noise source has a 30dB ENR. When I measure it on the spectrum analyzer with a 100kHz Resolution bandwidth I get a 5 dB increase in the noise floor (from -85dBm to -80dBm). I assumed this was the input power from the noise source into the first amplifier, is this incorrect, is it really -174+30+10log(res BW)? The noise source has a 1GHz BW, so the power output after amplification is dependent on the bandwidth of the noise source and measurement instrument resolution bandwidth right? I am trying to figure out what my output power should be given a 90dB gain with an input noise source with 30dB ENR.

comsys_engineer said:

Thank very much you for the suggestion. It turns out I was saturating the cascaded amplifiers with the CW test signal input. Once I lowered the CW input signal down to -80 dBm the cascaded amplifiers are working as they should and I get the expected +10dBm output. However, when I use the noise source as the input, the output level drops to a -13dBm. The noise source has a 30dB ENR. When I measure it on the spectrum analyzer with a 100kHz Resolution bandwidth I get a 5 dB increase in the noise floor (from -85dBm to -80dBm). I assumed this was the input power from the noise source into the first amplifier, is this incorrect, is it really -174+30+10log(res BW)? The noise source has a 1GHz BW, so the power output after amplification is dependent on the bandwidth of the noise source and measurement instrument resolution bandwidth right? I am trying to figure out what my output power should be given a 90dB gain with an input noise source with 30dB ENR.

Click to expand...

In an ideal world, the output power read by the spec an should be just as you said, -174 + 30 + 10*log(RBW). However, there are a few things that will mess that up, because we aren't in an ideal environment.

First, the spectrum analyzer has it's own internal noise floor, so you won't see -174 dBm with Res BW set to 1 Hz. Put a 50 ohm load on the input of your Spec An, go to a super-narrow span and crank RBW down to 1 Hz, and see what your minimum noise floor is. In order to make good measurements, the rule of thumb is that you need to see the signal level/noise floor increase by at least 10 dB, otherwise your desired signal is getting swamped out by the noise inside the analyzer's receiver.

Second, when measuring noise figure using a noise source, you are using what is called the Y-Factor method. In a nutshell, the analyzer switches the noise diode on and off, measures the difference, and uses the gain with and without extra noise to compute NF. If the source has a 30 dB ENR (Excess Noise Ratio), that means the difference between noise-diode-ON and noise-diode-OFF is ~30 dB (look at the calibration chart typically stuck to the side of the source to see the variation). So, when the noise diode is turned off, the signal level coming out of it is more than zero. By measuring the difference, the Y-Factor method removes that offset.

I don't know that there is a direct way to measure the output power of a noise source without using a known amplifier of some kind. Noise sources are generally designed for the Y-factor method's on/off measurement, not used as an absolute reference. If you know the linear (uncompressed) gain of your amplifier chain, you can use that to compute the noise power generated by the noise source (in what ever bandwidth you choose, just use 10*log(BW) as needed). A linear amplifier will remain linear at all signal levels below it's compression point... that's fundamental.