Psat calculation w/ load

Dear RakeshPRG,

Let me try to add a few comments to your questions. See if they make sense to you.

RakeshPRG said:

I am trying to find out Psat for an amplifier. The model uses an input current source of variable magnitude and the output power is plotted with values taken from direct plot->main form

As the magnitude of the input current is increased, the output power is supposed to reach a saturation level yielding the value of Psat. For this, I am using an AC sweep, since my intention is to plot Psat at various frequencies

The typical metric used to characterize the non-linear behavior of an amplifier is its 1 dB compression point. This occurs when the output power falls 1 dB below the power of an ideal amplifier with the same gain. Perhaps this is what you mean by "Psat". The non-linear behavior is not a hard limit, but the amount of compression increases as the input power (or output power) is increased. In your description of the simulation to assess Psat and from your pasted ADE-L analysis, it appears you are using an AC analysis. This is not appropriate since an AC analysis linearizes devices about their DC operating point. Hence, your simulation is not capturing the non-linear device behavior that causes compression. Typically, a number of transient analyses are performed at different frequencies and power levels to assess the 1 dB compression point. A transient simulation will capture non-linear device behavior.

RakeshPRG said:

So does it mean that there is an appropriate load resistor that I need to use for calculation? like a matching load but I suppose that would not work over wide frequencies. Can someone suggest me if this is a correct way for figuring out Psat /any other alternate ways?

The 1 dB compression point will be a function of both frequency and load impedance. For any given frequency, there will be some loads that produce higher 1 dB compression points than others. Similarly, for a given load impedance, the compression point will vary with the input frequency. For a given input frequency, part of the design process is to choose the load impedance that maximizes the 1 dB compression point. 

Do these comments help at all RakeshPRG?

Shawn

Dear RakeshPRG,

Let me try to add a few comments to your questions. See if they make sense to you.

RakeshPRG said:

I am trying to find out Psat for an amplifier. The model uses an input current source of variable magnitude and the output power is plotted with values taken from direct plot->main form

As the magnitude of the input current is increased, the output power is supposed to reach a saturation level yielding the value of Psat. For this, I am using an AC sweep, since my intention is to plot Psat at various frequencies

The typical metric used to characterize the non-linear behavior of an amplifier is its 1 dB compression point. This occurs when the output power falls 1 dB below the power of an ideal amplifier with the same gain. Perhaps this is what you mean by "Psat". The non-linear behavior is not a hard limit, but the amount of compression increases as the input power (or output power) is increased. In your description of the simulation to assess Psat and from your pasted ADE-L analysis, it appears you are using an AC analysis. This is not appropriate since an AC analysis linearizes devices about their DC operating point. Hence, your simulation is not capturing the non-linear device behavior that causes compression. Typically, a number of transient analyses are performed at different frequencies and power levels to assess the 1 dB compression point. A transient simulation will capture non-linear device behavior.

RakeshPRG said:

So does it mean that there is an appropriate load resistor that I need to use for calculation? like a matching load but I suppose that would not work over wide frequencies. Can someone suggest me if this is a correct way for figuring out Psat /any other alternate ways?

The 1 dB compression point will be a function of both frequency and load impedance. For any given frequency, there will be some loads that produce higher 1 dB compression points than others. Similarly, for a given load impedance, the compression point will vary with the input frequency. For a given input frequency, part of the design process is to choose the load impedance that maximizes the 1 dB compression point. 

Do these comments help at all RakeshPRG?

Shawn

In addition to Shawn's very useful points, there's a specific analysis (available for both pss shooting, and the hb anaysis) for finding the n-dB compression point efficiently (it's referred to as xdb analysis in Spectre RF). Rather than doing a simple power sweep (which can be difficult to ensure you have a suitable range over different levels), it does an adaptive search for the compression point. Yes, these are Spectre RF analyses, but they're not just for RF circuits...

Andrew.

Hi Andrew,

Andrew Beckett said:

there's a specific analysis (available for both pss shooting, and the hb anaysis) for finding the n-dB compression point efficiently (it's referred to as xdb analysis in Spectre RF). Rather than doing a simple power sweep (which can be difficult to ensure you have a suitable range over different levels), it does an adaptive search for the compression point.

Thank you for adding his important information and insights for RakeshPRG. I am sure that will be very useful to him as I did not include a path to finding the 1 dB compression point in a more efficient manner than a set fo transient simulations.

Thanks again,

Shawn

Hi Andrew and Shawn,

Thanks for those helpful tips. I tried the PSS analysis and it is my first time I worked on it. In this schematic, I used two ports and the input port has fin, Pin as variables.The goal of mine is to plot Psat Vs frequency. So I tried the PSS and used direct plot to extract P1dB (output referred) at different fundamental frequencies and plotting them separately in excel. I have few questions over the analysis.

Q1. Since the TIA is meant to use a current source at input, the usage of an input port (a voltage and series 50 ohm resistor) seems equivalent. But is this the right way?

Q2. I have to try the conservative/moderate/liberal settings to have convergence met. I also changed the Pin range and number of harmonics. Despite of that, when the fin is at 10G, the simulation doesnot converge.

Notices like "Zero diagonal found happens at the final sweeps although the initial ones run" , "Trapezoidal ringing", Why does this happen?

Thanks,

-Rakesh.

Dear Rakesh,

RakeshPRG said:

Q2. I have to try the conservative/moderate/liberal settings to have convergence met. I also changed the Pin range and number of harmonics. Despite of that, when the fin is at 10G, the simulation doesnot converge.

Notices like "Zero diagonal found happens at the final sweeps although the initial ones run" , "Trapezoidal ringing", Why does this happen?

As detailed in the Assembler and Explorer spectreRF product document for v19.1 at URL:

https://support.cadence.com/apex/techpubDocViewerPage?xmlName=spectrerfinexplorer.xml&title=Spectre%20Circuit%20Simulator%20and%20Accelerated%20Parallel%20Simulator%20RF%20Analysis%20in%20ADE%20Explorer%20User%20Guide%20--%20Single%20Input%20Large%20and%20Small-signal%20Analyses%20-%20Single%20Input%20Large%20and%20Small-signal%20Analyses&hash=pgfId-1065083&c_version=19.1&path=spectreRFinExplorer/spectreRFinExplorer19.1/Chap4.html#pgfId-1065083

using the trapezoidal integration method can lead to trapezoidal ringing. You might try changing the integration method to "gear2only" or simply changing your errpreset from "moderate" to "conservative". The "conservative" option, I believe, forces the use of a gear2 algorithm

I also noted in your pss GUI, you requested a transient analysis be performed. However, it appears you did not specify a time period to run the transient analysis. Unless you tell spectre to estimate how long to run the transient analysis, you should estimate a time based on how long it will take for your circuit to reach its steady-state solution. The pss analysis is trying to estimate the steady-state solution from the state of your circuit at the end of the transient simulation you specify. If you did not allow sufficient time for your circuit to settle to its steady-state solution (or near it) in your initial transient simulation, pss may not converge. It is a good idea to save your initial transient simulation (you do not have this option checked) to verify it does provide a good starting point for the subsequent pss analysis.

Shawn

Dear Shawn,

I appreciate your response. But my intention is to find out Psat, output power at saturation and not P1dB. I dont think theyre the same nor related somehow. I could find out P1dB output referred though but does that inform me something of Psat?

Dear RakeshPRG,

RakeshPRG said:

But my intention is to find out Psat, output power at saturation and not P1dB. I dont think theyre the same nor related somehow. I could find out P1dB output referred though but does that inform me something of Psat?

Perhaps you should define what you mean by "Psat". Saturation of an amplifier is not a "hard limit" to my knowledge. As a result,  the onset of gain compression (or equivalently amplifier saturation) is often characterized by a compression point at a given input frequency that relates the difference in gain from what a linear gain will provide. If you mean just the limiting value of output power after the input power has passed the 1 dB or 3 dB compression point, this value will be frequency dependent and might be more efficiently simulated by just choosing an input power level at a given input frequency where you know the amplifier will show non-linear effects. I believe the value of the limiting value of output power is not a "constant" either and will have some slope with input power.

Shawn

Shawn,

Youre right about choosing that input power that could lead to saturation. In my case, I have used input currents as low as 50uA to as high as 10mA, the results of which I added earlier (by mistake to previous threads/your reply 9days ago). This consists of both AC and transient simulations. I dont find the power output going to a maximum value in either of them.

The transient simulation however, produces output waveforms that are distorted due to non linearity as the input current exceeds 200mA. But I could not figure out how the waveforms relate to the output saturation powers.

RakeshPRG said:

For the AC analysis, the Power output goes up and never saturates even at the input current signal of magnitude 10mA which is awful lot coming from a Photodiode

This is hardly surprising, since the AC analysis is a linear analysis - it makes small-signal assumptions and linearises the circuit. That means that if you increase the input by a million, the output will also increase by a factor of a million too. This is commonly seen where people check the transfer function of a circuit using an AC analysis with an input AC magnitude of 1V - the output might then be 10,000V indicating a gain of 10,000 - clearly it will not really output 10,000V. So you will never see compression/saturation in a (standard) AC analysis.

Clearly you are seeing saturation/compression effects in your circuit when you simulate using transient, but since you've not described the meaning of "Psat" it's rather hard to tell you how to measure it. You could do this with a transient analysis run long enough for the circuit to settle, perform an FFT over the last cycle of your sine wave, and then plot the magnitude of the 1st harmonics versus your swept variable. That will increase linearly with the input amplitude until it starts to saturate, at which point the output power will deviate from the expected linear gain and hence will indicate saturation.

That's in essence what either the swept PSS or compression analysis (which I forgot, is only available for the hb - harmonic balance analysis) does. You can sweep the input power and the direct plot form allows you to plot the compression point - it will then plot the output power versus the swept input power (and PSS/hb ensures that everything is settled), and you can either choose to plot the output referred or input referred compression point. The output referred compression point is the output power at which the power is n dB (and you can pick n) lower the expected linear increase - maybe this is your Psat (hard to know, since you've not defined it, and it's not a universally recognised metric). The "compression" choice on the hb form allows this to be done more efficiently, because you can specify the goal up front (that you're measuring n dB compression at a particular node) and it will adaptively search for the result - but the results are plotted in a similar way.

Andrew.