The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> am-pm and spectral regrowth
https://designers-guide.org/forum/YaBB.pl?num=1301325888

Message started by aaron_do on Mar 28^th, 2011, 8:24am

Title: am-pm and spectral regrowth
Post by aaron_do on Mar 28^th, 2011, 8:24am

Hi all,

Can anyone give some comments on how am-pm and am-am relate to spectral regrowth? It seems to me that spectral regrowth is more directly related to iip3, although I can see how am-pm and am-am would cause spectral regrowth. Any good references would be much appreciated.

Thanks
Aaron

Title: Re: am-pm and spectral regrowth
Post by loose-electron on Mar 30^th, 2011, 7:23pm

Could you explain the question a bit better?

Title: Re: am-pm and spectral regrowth
Post by RFICDUDE on Mar 31^st, 2011, 6:24am

In principal AM-PM can contribute equally to spectral regrowth as AM-PM.
But practically it really depends on the circuit topology and technology (CMOS, HBT, etc.) as the combination of both that contribute to the spectral regrowth.

Typically AM-AM tends dominates in class A amplifiers, but I’m not so sure for class AB amplifiers. A primary source of AM-PM is nonlinear input capacitance and perhaps stray feedback. In either of these cases the frequency of operation should have a significant impact on the AM-PM contribution of an amplifier (AM-PM contribution should increase with frequency). I haven’t done a careful study of this, but certainly going from baseband to RF there is some frequency where nonlinear capacitance starts to distort the signal. Also, stray feedback increases with operation frequency.

I think there are some references on AM-PM, so I'll try to collect them.

Title: Re: am-pm and spectral regrowth
Post by loose-electron on Mar 31^st, 2011, 9:52am

Not too sure if spectral spreading issues are what hes asking about. That would have been my first guess, but hopefully a clarification is forthcoming.

Title: Re: am-pm and spectral regrowth
Post by albert john on Apr 1^st, 2011, 1:05am

The effects of amplitude and phase nonlinearities in a power amplifier, which lead to spectral regrowth, are examined in the context of a 3GPP W-CDMA forward link (base station) environment. Behavioral modeling and simulation techniques are applied first to independently examine the relationship between AM/AM and AM/PM conversion and spectral regrowth. This analysis suggests that compressing the peak envelope excursions by several decibels of gain compression still results in a tolerable level of spectral regrowth. Conversely, limited phase distortion appears as a key element in developing linear power amplifiers with linearity margin to the 3GPP specification. Experimental results, obtained by digitally predistorting a 3GPP W-CDMA compliant signal to correct only for the amplifier's AM/PM conversion, provide supporting evidence.

Title: Re: am-pm and spectral regrowth
Post by aaron_do on Apr 1^st, 2011, 3:03am

Hi all,

I thought my post had been given up on and haven't checked it recently. Anyway, to clarify, I am trying to understand which design parameter (AM-AM, AM-PM, IIP3 etc) has the most direct relationship to spectral regrowth. It seems that we have some spectral mask or ACPR specification, and based mainly on simulations/measurements we can get an idea what kind of AM-AM or AM-PM we need. Conversely, it seems the main reason we need AM-PM and AM-AM data is that it allows us to accurately estimate the spectral regrowth through simulation. However, it doesn't appear to provide any intuitive way to estimate ACPR or spectral regrowth. On the other hand IIP3 allows us to estimate spectral regrowth or ACPR, but cannot be used to simulate the actual spectral regrowth. Just looking for other peoples thoughts on this matter, and whether I have understood this correctly.

thanks,
Aaron

Title: Re: am-pm and spectral regrowth
Post by loose-electron on Apr 1^st, 2011, 2:30pm

Spectral spreading - spectral regrowth - in my interpretation anyway, are not the same as distortion products created by IP2, IP3 linearity issues.

One is the creation of spectral spurs due to the nonlinearity of the amplifier. The other is the broadening of the spectrum due to things like thermal noise.

One is important (IP2, IP3) due to the creation of distinct undesired products in the frequency planning of your receiver, or the creation of OOB spurs in a transmitter.

The other (spectral spreading due to inherent noise) is more of an issue in the design of a transmitter and its ability to maintain a "within channel" signal.

Not sure if the above is what you are looking for?

Title: Re: am-pm and spectral regrowth
Post by aaron_do on Apr 1^st, 2011, 8:20pm

Hi loose-electron,

I went over what you said and I think I understand it better now. My misunderstanding I think came from considering the signal as a whole bunch of discrete sinusoids existing at once. Thanks for the help,

Aaron

Title: Re: am-pm and spectral regrowth
Post by RFICDUDE on Apr 2^nd, 2011, 5:32am

Ah ok, I think I can help here.

We are talking about "nonlinear responses" when a particular signal is applied to a nonlinear circuit.

AM-AM and AM-PM is the first harmonic (fundamental) response when the amplitude of a single sinusoidal input signal is swept (large signal gain response). AM-AM AM-PM contains information about odd order nonlinearity (3rd order and higher) as the amplitude is swept such that the amplifier enters gain compression.

IIP3 is related to the 3rd order nonlinear response when a two-tone input signal is input to a nonlinear circuit. The intercept point is a practical abstraction making it easy to compare the linearity of circuits, but it is the measurement of a response to a specific stimulus signal. IM3 is particularly useful in circuits/systems where the amplitudes are such that the 3rd order distortion is the most significant even order contributor around the desired signal (i.e. receivers, LNA, downconverters, etc.).

Spectral regrowth is the nonlinear response to an arbitrary stimulus signal exhibiting amplitude modulation. It is accurate to say that IM3 is the spectral regrowth when a two-tone signal is the stimulus and the amplitude is low enough that higher order nonlinearity (5th, 7th, etc) is not contributing significantly to the response.

Now AM-AM AM-PM and IM3 are characterizations of the nonlinearity because the stimulus signals are known and analytic, so the underlying nonlinearity can be extracted as power series coefficients by finding the values of coefficients that match the AM-AM AM-PM response or IM3.

Ok this is great info, but it doesn't answer your question about what to use during design.

For receivers, IM3 is good enough, but it can pay to do system level simulation with IM3 and then maybe envelope simulation with a modulated signal to determine any difference between the two. If everything stays 3rd order then the offset shouldn't change and IM3 can be used with good confidence. You can do something similar for IM2 situations too.

For transmitters, what I have seen work well is some designers correlate spectral regrowth against IM3 to find a crude offset between the two for modulated signals of interest. Then IM3 is use through the design phase since it is more straight forward to simulate and interpret the impact of design changes, but ultimately spectral regrowth needs to be checked using a AM-AM AM-PM model to pass signals through just to make sure to correlation is ok (the correlation is different for different signals). Also AM-AM AM-PM is only accurate for circuits that do not exhibit significant memory effects (IM3 dependent on tone-separation or tone-frequency). In this case AM-AM AM-PM will work for narrow tone-spacings but not so good for wider tone separation.

In the end you will likely need to use both IM3 and AM-AM AM-PM to gain confidence in the predictive spectral regrowth performance.

There are special cases such as OFDM where the signal can be modeled as a Gaussian or complex Gaussian amplitude modulation where the properties of the signal are known well enough to permit nonlinear calculations using the statistical properties of the assumed Gaussian signal. But this discussion gets a bit involved ...

The references below don't exactly answer your question, but they do contain some discussion about AM-AM AM-PM contributions to spectral regrowth.

J. S. Kenney and A. Leke, "Power amplifier spectral regrowth for digital cellular and PCS applications," Microwave J., vol. 38, no. 10, Oct., 1995, pp. 74.

Aikio, J.P.; Rahkonen, T.; , "A Comprehensive Analysis of AM–AM and AM–PM Conversion in an LDMOS RF Power Amplifier," Microwave Theory and Techniques, IEEE Transactions on , vol.57, no.2, pp.262-270, Feb. 2009

Sundstrom, L.; , "Fast and accurate ACLR estimation method," Radio and Wireless Conference, 2004 IEEE , vol., no., pp. 183- 186, 19-22 Sept. 2004

Title: Re: am-pm and spectral regrowth
Post by aaron_do on Apr 2^nd, 2011, 8:02am

Hi RFICDUDE,

thanks for the very comprehensive response. I think the point loose-electron was making (correct me if I'm wrong) is that spectral regrowth is not due to intermodulation. This sounds reasonable to me since we are only considering one signal. However, it doesn't mean that there is no correlation between IP3 and spectral regrowth since they originate from the same source (nonlinearity, although spectral regrowth can also originate from any other noise source).

I understand your point about how we check the correlation between spectral regrowth and IP3, and then use IP3 as a design parameter to shorten the design time, but I will have to read up a bit more about your point that AM-PM and AM-AM may not be accurate when memory effects come into play.

thanks for the references too,

cheers,
Aaron

Title: Re: am-pm and spectral regrowth
Post by RFICDUDE on Apr 2^nd, 2011, 11:01am

Hi Arron,

Actually spectral regrowth is due to intermodulation even with one signal. Whenever you pass an amplitude modulated signal through a nonlinear circuit you end up with intermodulation products. If the amplitude modulated signal is a two tone signal then you get distinct and measurable IM products. If the amplitude modulation is continuously varying then you end up with spectra that are from the cube (x^3) of the signal along with higher order spectra from (x^5, x^7 etc) and even products to from IM2.

Most AM modulated signals are bandlimited with some sort of baseband filter. So the intermodulation spectrum looks like the convolution of the bandlimited signal with itself. In the third order case you see the signal convolved three times which is why the spectral regrowth looks biggest right outside the desired signal and tapers off towards 3x the signal bandwidth (or until it hits the 5th order spectrum).

Now it is possible to have more than one carrier signal (either a multicarrier signal or an interfering blocker). In these cases you have both spectral regrowth to the intermodulation of the individual signals plus crossmodulation distortion between the carriers. So, some of the spectral regrowth comes from a convolution of the spectrum of each signal. Also, if you have spurious tones at the input of the amp then there will be mixing products around the tones that look like replicas of the input signal due to the reciprocal mixing.

I have only heard of spectral regrowth in reference to intermodulation distortion from AM modulated signals, so let me know if you are talking about something else altogether.

Cheers,

Title: Re: am-pm and spectral regrowth
Post by aaron_do on Apr 3^rd, 2011, 8:04am

Hi RFICDUDE,

your points were well made and appreciated. I will take some more time to try and understand the issue better.

thanks,
Aaron