NF problem of (Tx/Rx switch + LNA + down mix1) + down mix2

[wccheng]

mix rf tx rx

Dear all,

I have successfully simulated the NF of combination Tx/Rx switch + LNA + down mix1. It gives me the NF = 4.5 dB. As I combined this with down mix2, the NF is very high. It gives NF = 8 dB. However, the NF of stand alone down mix2 is 4 dB only. Moreover, I have checked the gain of Tx/Rx switch + LNA + down mix1. It gives Gain = 20 dB. What is the possible problem of this high NF value? How could I fix it?

Thanks

 

[madengr]

qpss qnoise

Where are your image filters? The LNA is probably acting as a filter for mix1, but what do you have between mix1 and mix2?

 

[biff44]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

Yep--what he said!

 

[wccheng]

I don't insert any filter in between. Is it need to insert for better NF? After the mix1, the frequency is 1.4GHz around. Does any filter available? Could make it on chip?

Thanks

 

[madengr]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

You need to attenuate the noise at you image frequency since that is also down-converted to 2nd IF. You either use a BPF around your 1st IF or a notch at your image frequency. It really depends on if your IF and LO is fixed or variable. You could also use an image reject mixer. This is covered in receiver text books so read up on that. Yes, you can do a low-Q 1.4 GHz filter on chip. If your IF1 image is far enough away you can get away with it.

I had an RX with a very low level, noisy oscillation in the LNA, just enough to raise the system NF by 2dB. Too low to be seen by spectrum analyzer, but a 15pF bypass cap directly on the IC lead-frame (or a finger) killed it.

There cam be lot's of reasons by NF is high, but start with the obvious, especially in a simulation.

 

[wccheng]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

However, my mix2 will mix down the input frequency=1.4GHz to the DC. Does the filter useful? It seems that it doesn't have any image frequency problem if I down convert to DC.

You need to attenuate the noise at you image frequency since that is also down-converted to 2nd IF. You either use a BPF around your 1st IF or a notch at your image frequency. It really depends on if your IF and LO is fixed or variable. You could also use an image reject mixer. This is covered in receiver text books so read up on that. Yes, you can do a low-Q 1.4 GHz filter on chip. If your IF1 image is far enough away you can get away with it.

I had an RX with a very low level, noisy oscillation in the LNA, just enough to raise the system NF by 2dB. Too low to be seen by spectrum analyzer, but a 15pF bypass cap directly on the IC lead-frame (or a finger) killed it.

There cam be lot's of reasons by NF is high, but start with the obvious, especially in a simulation.

 

[biff44]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

I do not know the particulars of your system, but lets say you were trying to receive a 2 GHz signal.

So let me postulate a way to do that. If you had a 1.9 GHz local oscillator, mix the desired signal down to 100 MHz, input that to an analog-to-digital converter, then you would have a receiver system.

If there was broadband noise in the vicinity of 2.0 GHz (noise of the world, noise figure of your input LNA or mixer, etc), then that noise would be downconverted to 100 MHz, and would set your signal to noise ratio. However, what about the noise at 1.8 GHz? Isn't there noise there too? So that noise would mix with the 1.9 GHz LO, and also come out at an IF frequency of 100 MHz. Since both bandwidths of noise are incoherent, you would read a system noise figure that is 3 dB higher than expected.

To get rid of it, if after the LNA you had a bandpass filter that passed 2.0 GHz, but rejected 1.8 GHz by at least 15 dB, then you would be all set. Conversely, if your mixer was an image reject mixer, you would only downconvert the 2 GHz, but not the 1.8 GHz inputs, and again you would be fine.

IF you downconvert to baseband in the first mixer, yes you do not have to worry about the image noise. But you have to be careful with additional mixers, that what you are downconverting to baseband has only one input-bandwidth of noise, and not a 2nd or third one due to image noise.

 

[wccheng]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

In my system, the mix1 will down convert the received 3G-5GHz frequency band signal to the 1.9GHz. Then the mix2 will mix with LO=1.9GHz with this input RF=1.9GHz to the DC. Are you mean that I need to insert one filter in-between of these two mixers? The question is that mix2 is mixing the input RF=1.9GHz to the DC directly. It should not mix the image frequency too. Am I right?

wccheng

I do not know the particulars of your system, but lets say you were trying to receive a 2 GHz signal.

So let me postulate a way to do that. If you had a 1.9 GHz local oscillator, mix the desired signal down to 100 MHz, input that to an analog-to-digital converter, then you would have a receiver system.

If there was broadband noise in the vicinity of 2.0 GHz (noise of the world, noise figure of your input LNA or mixer, etc), then that noise would be downconverted to 100 MHz, and would set your signal to noise ratio. However, what about the noise at 1.8 GHz? Isn't there noise there too? So that noise would mix with the 1.9 GHz LO, and also come out at an IF frequency of 100 MHz. Since both bandwidths of noise are incoherent, you would read a system noise figure that is 3 dB higher than expected.

To get rid of it, if after the LNA you had a bandpass filter that passed 2.0 GHz, but rejected 1.8 GHz by at least 15 dB, then you would be all set. Conversely, if your mixer was an image reject mixer, you would only downconvert the 2 GHz, but not the 1.8 GHz inputs, and again you would be fine.

IF you downconvert to baseband in the first mixer, yes you do not have to worry about the image noise. But you have to be careful with additional mixers, that what you are downconverting to baseband has only one input-bandwidth of noise, and not a 2nd or third one due to image noise.

 

[madengr]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

It seems that it doesn't have any image frequency problem if I down convert to DC.

If your signal is DSB (double side band) than it has better be mathematically symmetric about the carrier since the USB (upper side band) and LSB (lower side band) are folded on top one another when mixing down to DC. Again, that's why an image reject mixer is used since it cancels one of the sidebands. I'm not sure what you are trying to receive but direct to DC downconversion is usually not done with just one mixer, usually an IQ demodulator with an added poly-phase filter or Hilbert transformer on one of the outputs, both of which are then summed. You may be OK down converting BPSK with this method but I don't know about any kind of quadrature modulation. Analog SSB modulation and you will double your noise figure. FSK would be pointless unless it's coherent with a true phase reversal.

 

[fanshuo]

I think the IQ demodulator can also be implemented in baseband

 

[rillyxue]

a very good topics.

Added after 53 seconds:

i think it may be some hamonics.

 

[noiseless]

Please check ur simulation setup. There are two LOs, how do u simu noise of total system? qpnoise?

 

[wccheng]

Re: NF problem of (Tx/Rx switch + LNA + down mix1) + down mi

I use QPSS + QNOISE

Please check ur simulation setup. There are two LOs, how do u simu noise of total system? qpnoise?