AnsweredAssumed Answered

Measuring the Noise Figure of a Lossy Active Device

Question asked by KeithV on Dec 22, 2010
Latest reply on Dec 22, 2010 by andyowen

Like • Show 0 Likes0
Comment • 2

I am trying to measure the noise figure of an active DUT that should have a gain of around -30 dB and a noise figure of 30-40 dB.

I have trouble getting accurate results from the methods I have tried. I have an E4440A and have tried using the Noise Figure measurement personality with a 346C Noise Source. I have also tried the using a low noise figure amp after the DUT to boost the noise floor and calculated the noise figure using the noise floor of the combination and the gain of the low noise amp and DUT. From what I have read I have too much loss and/or too much noise to get a good repeatable measurement.

I also have read about cold-source measurements but this seems daunting to do manually on potentially several hundred units a month.

Anyone have suggestions on how I can get a good repeatable measurement without too much difficulty?

No one else has this question

Outcomes

tabbott

Dec 22, 2010 12:47 PM

Hi -

DUT match may be the issue. I recommend performing this measurement using isolators. That way if you do have a DUT match issue, the isolator will help and your NF measurements will become much more realistic and stable. If the isolators help, you can measure their insertion loss and compensate for that as before DUT loss or after DUT loss. When performing the noise source calibration make sure none of the devices are connected. You want to perform the noise source cal with the noise source connected directly to the RF input of the analyzer.

Regards -

Relevant Products;#E4440A,346C#

Actions

andyowen

Dec 22, 2010 4:53 PM

Hi,

When you that you are not getting accurate results, do you mean that the trace is excessively "jittery?" How does the gain measurement look? Similar? I would expect so, based on the numbers that you give here for you DUT. Can you please share the frequency of the measurement and perhaps post a screenshot of your measurement (if you can't share the frequency, you can still blank out the frequency axis on a screenshot)?

Since your DUT is very lossy, the difference between the output noise power when the noise source is "hot" and when it is "cold" (ie, the Y-factor) is VERY small. You pretty much lose the ENR of the noise source due to all of the loss in the DUT; the receiver is trying to measure the small difference between two small numbers so any added noise from the measurement system (your E4440A) would make this measurement jump up and down. The lower the noise figure of the measurement system, the less the Y-factor "jitters" and the less jitter you'll have on the NF and gain. So adding a preamp was a good first step but increasing the gain of the measurement system doesn't necessarily decrease the noise figure (unless you have a preamp with NF = 0 dB).

Some things to try:

1) Increase the measurement bandwidth on the instrument. This gives the receiver more noise power or more noise samples to work with (since your DUT doesn't really put out much noise). But if you have a narrowband DUT (say with a BPF on the input/output), keep the measurement bandwidth less than the DUT bandwidth.
2) Increase the number of noise averages. This has a similar effect to increasing the measurement bandwidth in that the receiver takes more noise samples but without increasing the actual instrument bandwidth.
3) Add a preamp as long as its gain and NF will actually improve the NF of the system. If the PSA already has an internal preamp and you are using it, it will be tough to improve it much more.

Now, all that said, you mentioned the direct-noise method. The direct-noise method measures the absolute output noise power for a given source temperature and measures the gain independently. Our PNA-X vector-network analyzer EXCELS at this measurement. The calibration is very easy with a 346C noise source and and our electronic calibration module. We can accurately measure the noise figure of a 40-dB pad, sweep a very high number of points in the fraction of that of a noise-figure analyzer, and correct for mismatch errors (which no other instrument on the market can do). Mismatch is a big error contributor to NF measurements, as tabbott mentioned.

Please refer to the PNA-X Noise Figure application note for more information on the direct-noise or "cold source" technique and the PNA-X brochure for general information on the PNA-X for S-parameters, IMD, gain compression, noise figure, and nonlinear device modeling.

Actions

2 Replies

tabbott Dec 22, 2010 12:47 PM
Mark Correct
Correct Answer
Hi -

DUT match may be the issue. I recommend performing this measurement using isolators. That way if you do have a DUT match issue, the isolator will help and your NF measurements will become much more realistic and stable. If the isolators help, you can measure their insertion loss and compensate for that as before DUT loss or after DUT loss. When performing the noise source calibration make sure none of the devices are connected. You want to perform the noise source cal with the noise source connected directly to the RF input of the analyzer.

Regards -

Relevant Products;#E4440A,346C#
Like • Show 0 Likes0
Actions
andyowen Dec 22, 2010 4:53 PM
Mark Correct
Correct Answer
Hi,

When you that you are not getting accurate results, do you mean that the trace is excessively "jittery?" How does the gain measurement look? Similar? I would expect so, based on the numbers that you give here for you DUT. Can you please share the frequency of the measurement and perhaps post a screenshot of your measurement (if you can't share the frequency, you can still blank out the frequency axis on a screenshot)?

Since your DUT is very lossy, the difference between the output noise power when the noise source is "hot" and when it is "cold" (ie, the Y-factor) is VERY small. You pretty much lose the ENR of the noise source due to all of the loss in the DUT; the receiver is trying to measure the small difference between two small numbers so any added noise from the measurement system (your E4440A) would make this measurement jump up and down. The lower the noise figure of the measurement system, the less the Y-factor "jitters" and the less jitter you'll have on the NF and gain. So adding a preamp was a good first step but increasing the gain of the measurement system doesn't necessarily decrease the noise figure (unless you have a preamp with NF = 0 dB).

Some things to try:

1) Increase the measurement bandwidth on the instrument. This gives the receiver more noise power or more noise samples to work with (since your DUT doesn't really put out much noise). But if you have a narrowband DUT (say with a BPF on the input/output), keep the measurement bandwidth less than the DUT bandwidth.
2) Increase the number of noise averages. This has a similar effect to increasing the measurement bandwidth in that the receiver takes more noise samples but without increasing the actual instrument bandwidth.
3) Add a preamp as long as its gain and NF will actually improve the NF of the system. If the PSA already has an internal preamp and you are using it, it will be tough to improve it much more.

Now, all that said, you mentioned the direct-noise method. The direct-noise method measures the absolute output noise power for a given source temperature and measures the gain independently. Our PNA-X vector-network analyzer EXCELS at this measurement. The calibration is very easy with a 346C noise source and and our electronic calibration module. We can accurately measure the noise figure of a 40-dB pad, sweep a very high number of points in the fraction of that of a noise-figure analyzer, and correct for mismatch errors (which no other instrument on the market can do). Mismatch is a big error contributor to NF measurements, as tabbott mentioned.

Please refer to the PNA-X Noise Figure application note for more information on the direct-noise or "cold source" technique and the PNA-X brochure for general information on the PNA-X for S-parameters, IMD, gain compression, noise figure, and nonlinear device modeling.
Like • Show 0 Likes0
Actions

Measuring the Noise Figure of a Lossy Active Device

Outcomes

Related Content

Recommended Content