How do change the variable range (Freq, Pout)? I would to examine lowers freqs and higher Pout. I managed to change the scale, but apparently there's variables defined somewhere? Apologies if needed- didn't see them on the equations page. Also - good work!
Hi Matthew, do you think it is possible to use a class E topoly for pulsed radar application ? I mean if it possible to pulse the class for examplo with 10% of duty cycle and 1KHz of prf.
Hi, in the case where the transistor capacitances are nonlinear, how would you go about finding the values of the parasitic capacitances for a specific transistor via ADS?
Dear Sir, I have a question. Can you differentiate Class-E and Class-J operations because I saw the required impedances of the two classes are similar!
Hi, Thank you for the nice presentation. It is very helpful. could you please tell me how can I replace the transistor GaN with PMZB380XN? or BSS138? and the frequency range to 13.56MHz? Thanks Regards Sam
Nabeel Fattah In the “ClassE_Design_Equations” data display, on the top left hand side of the “synthesis” page, there is a section called “User Inputs”. In that box, near the “Vknee” variable, there are more variables in small print which set the range for the sliders (you may need to zoom in to see these). To change the frequency range, set the Freq_Range variable as follows: [Min_Frequency::Frequency_Step::Max_Frequency]. Now the frequency slider should update to your desired frequency range. To set the exact frequency, just move the slider bar to that frequency. Regarding the device limits: in the “User Inputs” section, most of these can be found simply by looking at the device datasheet, or by looking at the DC IV curves. For example, Vmax is typically the Drain-Source breakdown voltage. For the intrinsic device capacitance, to a first order this is just the device output capacitance. To replace the GaN transistor with another transistor in any of the designs, just simply delete the GaN model and replace it with the model you are using, and rerun the simulation (click on the gear icon on top). You may need to adjust sweep limits to get meaningful information. For more detail on design procedure, there is also a webcast that you can watch for free at this link: www.keysight.com/main/eventDetail.jspx?cc=US&lc=eng&ckey=2527083&nid=-34333.1077793.08&id=2527083
Could you tell what are the equations you are designing with or from which book did you base yourself to design this PA amplifier ? When you set DC bias or output power, the synthesis tool automatically adjusts parameters, but, with which equations does it do it ? Thank you !
The equations used to generate the waveforms are based on a little bit of mathematical manipulation from Raab’s paper, “Idealized Operation of the Idealized Class E Power Amplifier”. These are found in the data display. The output power and subsequent waveform peak voltage value are ultimately what gets adjusted when you change the DC bias (and knee voltage).
First of all, thank you for this tutorial and the ADS tools. I'm trying to match the input and output networks with the CGH40006S GaN device, also from Cree libraries. At time 10':20'', the output network is being matched using the intrinsic device parameters but... how can I get the parameters of my device? I tried to use the S parameters provided by Cree (only the S22 for the output matching), but the results are not convincent... am I following the correct way? Also, at time 12':08'' it is sown the intrinsic and extrinsic waveforms but, how can I obtain the voltages at the intrinsic nodes from my device? Thanks in advance.
For this technique, it’s best to use a device model which has access to the intrinsic voltages and currents. Many Wolfspeed (formerly Cree) device models have this intrinsic access and then the nodes “vdsi” and “idi” will be pulled into your simulation dataset automatically when running harmonic balance. From there, you can use V/I to calculate the impedances at each harmonic frequency to then analyze intrinsic impedances directly on the device for those models when you sweep an external load. This technique is more explicitly described in the Class J video at 8:00 th-cam.com/video/N0dLVMoVQSI/w-d-xo.html. If you don’t have a device with intrinsic nodes available, get Cds and Cgs off the datasheet (for FETs) and then optimize your load through this passive network similar to what’s shown at 10:20. That will give you at least a good starting point from which you can then run more targeted loadpull simulations to arrive at the best load value.
Hi and thanks for the tool. I have a question. Is it possible to achieve 100% efficiency with a non-zero knee voltage. This tool suggests that it is but I am not quite sure. And is it possible to include the effect of a non-zero on-resistance when calculating the required load impedances? Thanks in advance. Best, Denizhan
To add in finite R-on, set the ‘rsat’ parameter in the ideal switch simulation, called waveform_verification (and concurrently set the knee voltage to zero since you’re kind of modeling that with R-on). The efficiency will drop. But probably your actual waveforms will still not match this for high frequency cases because the harmonic terminations impact how the voltage waveform hits the knee. In other words, if the circuit you build does not match the harmonic impedances out to infinity, the voltage trajectory around the knee will change, perhaps dramatically - which might render the R-on based efficiency estimate inaccurate.
At the end of the day, the objective of this tool is to provide ideal waveforms that can serve as a starting point for design, given a model with intrinsic access. The most important thing to get right is the peak voltage swing because of reliability. Adding a knee voltage simply reduces the effective supply value, enabling a better estimate of the peak swing. This is for mathematical convenience and is consistent with Sokal’s original work, but it’s obviously not completely accurate as it ignores the knee. To summarize, modeling the knee transition for Class E is pretty tricky - to the point where you are probably better off just making a guess and starting the design with the actual device to see what happens.
can i ask a question about class e pa?if i use tsmc mos model to design,can it be possible to operate at a frequency about 30GHz??And to select my output matching networks' value?
+Winnie Chang It will depend on your specs and your device: if the output power requirement is low enough (>>1W) and you can closely integrate the output match, it might be possible. You will need a very high frequency (>=120GHz) matching network, and a device large enough to deliver the power while also having very low parasitics. If the parasitics are too high, no matter what load is presented externally, the internal impedance will not change. As the video mentions, there are also reliability issues with Class E operation due to the large voltage swings involved which you will need to account for. Higher powers mean higher voltage swings. At low GHz frequencies, some have accomplished >1W Class E operation in standard CMOS technology using stacked configurations and power combining transformers, but as far as I know this hasn’t been scaled up to millimeter wave frequencies. So hopefully your Pout spec is well below 1W.
Hi First of all, thank you very much for the class-e workspace and synthesis tools. They are really helpful and I was able to create a class-e and simulate it in multisim. I have a question about the classe_network_preopt schematic shown at 10:14. In ADS2012, When I open this schematic, I notice that the EBOND4 and EBOND_shape components are replaced by empty boxes. I am guessing that these components are not available in this version of ADS since I get an error when I run the S-Parameters optimization simulation. My question is as follows: 1 where can I get/download the library with these components? 2 If the components are not available/compatible with ADS2012, can I use the BONDW components in the Passive-RF circuit library? Thanks and keep up the great videos.
chefboyclakie Thanks! The EBOND components are included with ADS2015, but they aren’t available in ADS2012. The EBONDS are nice because you can fine tune the wire profile to match what your lab bonder can do, but you can absolutely use the BONDW components in ADS2012 to do the same thing with the optimizer. In fact, you can probably use a lumped inductor to model the bonds if you’re looking to optimize for the first time, or even just short the component out entirely--since there are 4 wires in parallel, the series inductance is rather small anyway.
Keysight EEsof EDA If you are interested in upgrading or trying ADS2015, fill out this form and a Keysight engineer will contact you. www.keysight.com/find/eesof-sales-assistance
Hi, At the time 4':35" the presenter says there are some equations in graphical format to design a class E power amplifier with non 50% duty cycle. Could you please tell me where I can find those equations? (not in your work, the papers including the equations I mean)Thanks
+Jim Style “RF Power Amplifiers for Wireless Communications”, by Dr. Steve Cripps is a very good reference for PA Design. In the book, Dr. Cripps provides a graph which gives normalized Class E circuit values over a moderate range of conduction angles. Prior to this, there was also other work dealing with how the components values are impacted by finite resonator Q. As with conduction angle, this effect can also be described graphically or through a polynomial fit - a good reference is “Class-E High-Efficiency RF/Microwave Power Amplifiers: Principles of Operation, Design Procedures, and Experimental Verification” by Dr. Nathan Sokal -the inventor of the Class E topology.
+Jim Style In the workspace shown in this video, it is easy to change or adjust the conduction angle and resonator Q, with the component values automatically updating accordingly (there is a variable on the bottom right hand graph) - so both of these effects are accounted for.
how if i use branch line coupler on two stages class E power amplifier, is it better than push-pull or not? could you tell me how it works? thank you +Keysight EEsof EDA
Branch line is a 90 degree configuration. It can show strange behavior under mismatch conditions, if fed from a common voltage one or other PA may "load hog". Push-pull is 180 degree. Push-pull results in the cancellation of the second harmonic, which can be a great benefit. Under mismatch both amplifiers should load-pull equally.
Class E Amplifiers are used in wireless power transmission systems. To state the obvious: WPT systems are very similar to RF systems with the caveat being the transfer of power rather than information. In either system, there may be a limited amount of power to use for transmission OR there may be lots of power to transmit with, perhaps to deal with more spatial separation between Tx/Rx. For both cases, a high efficiency PA is going to be important: if there’s a limited power budget, you’ll want as much power as possible to go to charging as opposed to getting lost in the PA. In a high power system, thermal issues can also become problematic so a high efficiency PA is desirable to minimize heating. In most real systems, both of these requirements together drive the need for a good PA. As this video shows, Class E is a very good topology to achieve high efficiency. As far as wireless power transmission, there are some other benefits too. First, many WPT systems operate a relatively low frequencies (~7 MHz), so the initial design/synthesis method presented here may be more straightforward. For example, at these frequencies you might find that the impedance values from the Class E synthesis tool can be directly applied to the device without having to worry about significant parasitic impedance shifts. Additionally, the fundamental load of Class E is inherently inductive, which is good if you’re ultimately driving a big inductive charging coil. For more information, visit this website: www.keysight.com/find/eesof-ads-power-electronics-resources
+Cer Luigi Dela Cruz Theoretically speaking, you want to bias the amplifier input to obtain the proper conduction angle- for example, to achieve 180 degree conduction, you would set the input bias at the threshold voltage such that any positive signal will conduct while any negative signal will turn the device off (the “Class B” Bias condition). More information on biasing for basic classes of operation can be found in this video : How to Design an RF Power Amplifier: Class A, AB and B : th-cam.com/video/GhPqPVlDRPY/w-d-xo.html. Realistically though, the correct output waveforms occur with a combination of DC bias and RF input signal drive. So, there can be several combinations of bias + RF input drive (+input match too!) that produce very similar output waveforms. This is because if you decrease the bias and then increase the input drive level, you will end up at essentially the same place. At the end of the day, if you are biased relatively close to the point which gives the specified conduction angle, you can then fine tune the bias and drive level together to get the right Class E waveforms and also achieve good small signal gain and a favorable compression characteristic (AM/AM and AM/PM). For many PA designs this gain shape vs. input power is pretty crucial even though the true Class E mode only occurs at high powers- so that may ultimately be what drives the precise bias point selection.
THANKS FOR YOUR EFFORT MATT. PLEASE keep making more videos for RF APPLICATION USING GAN
How do change the variable range (Freq, Pout)? I would to examine lowers freqs and higher Pout. I managed to change the scale, but apparently there's variables defined somewhere? Apologies if needed- didn't see them on the equations page. Also - good work!
Hi Matthew, do you think it is possible to use a class E topoly for pulsed radar application ? I mean if it possible to pulse the class for examplo with 10% of duty cycle and 1KHz of prf.
Hi, in the case where the transistor capacitances are nonlinear, how would you go about finding the values of the parasitic capacitances for a specific transistor via ADS?
Dear Sir, I have a question. Can you differentiate Class-E and Class-J operations because I saw the required impedances of the two classes are similar!
How would you go about finding Imax value? Do you get this from a datasheet or from Id vs Vd graph?
Hi,
Thank you for the nice presentation. It is very helpful.
could you please tell me how can I replace the transistor GaN with PMZB380XN? or BSS138? and the frequency range to 13.56MHz?
Thanks
Regards
Sam
Nabeel Fattah
In the “ClassE_Design_Equations” data display, on the top
left hand side of the “synthesis” page, there is a section called “User
Inputs”. In that box, near the “Vknee” variable, there are more
variables in small print which set the range for the sliders (you may need to
zoom in to see these). To change the frequency range, set the Freq_Range
variable as follows: [Min_Frequency::Frequency_Step::Max_Frequency]. Now
the frequency slider should update to your desired frequency range. To
set the exact frequency, just move the slider bar to that frequency.
Regarding the device limits: in the “User Inputs” section,
most of these can be found simply by looking at the device datasheet, or by
looking at the DC IV curves. For example, Vmax is typically the
Drain-Source breakdown voltage. For the intrinsic device capacitance, to
a first order this is just the device output capacitance. To replace the
GaN transistor with another transistor in any of the designs, just simply
delete the GaN model and replace it with the model you are using, and rerun the
simulation (click on the gear icon on top). You may need to adjust
sweep limits to get meaningful information.
For more detail on design procedure, there is also a webcast
that you can watch for free at this link: www.keysight.com/main/eventDetail.jspx?cc=US&lc=eng&ckey=2527083&nid=-34333.1077793.08&id=2527083
Could you tell what are the equations you are designing with or from which book did you base yourself to design this PA amplifier ? When you set DC bias or output power, the synthesis tool automatically adjusts parameters, but, with which equations does it do it ? Thank you !
The equations used to generate the waveforms are based on a
little bit of mathematical manipulation from Raab’s paper, “Idealized Operation
of the Idealized Class E Power Amplifier”. These are found in the data
display. The output power and subsequent waveform peak voltage value are
ultimately what gets adjusted when you change the DC bias (and knee voltage).
Sorry, the paper title was incorrect. It should be "Idealized Operation of the Class E Tuned Power Amplifier".
No worries. I have realized that because there was only one match. Thank you !
excuse me please which software you are used to design class-e amplifier used in the video
+hussein ali Keysight Advanced Design System (ADS) Find out more info and apply for a free trial here: www.keysight.com/find/mytrial.rfmw.sm
Why doesn't any of the simulations run once I try to do it on my own?
How I can find the software on which you have been working. I'm really stuck. Thank you for your help
Hello benfadhel yosra,
Matt is using Keysight ADS. To apply for free trial of ADS visit: www.keysight.com/find/mytrial.rfmw.sm
One question, hoe does on open this file on ADS? When i open the file through ADS, theres a blank workspace.
First of all, thank you for this tutorial and the ADS tools.
I'm trying to match the input and output networks with the CGH40006S GaN device, also from Cree libraries. At time 10':20'', the output network is being matched using the intrinsic device parameters but... how can I get the parameters of my device? I tried to use the S parameters provided by Cree (only the S22 for the output matching), but the results are not convincent... am I following the correct way?
Also, at time 12':08'' it is sown the intrinsic and extrinsic waveforms but, how can I obtain the voltages at the intrinsic nodes from my device?
Thanks in advance.
For this technique, it’s best to use a device model which
has access to the intrinsic voltages and currents. Many Wolfspeed
(formerly Cree) device models have this intrinsic access and then the nodes
“vdsi” and “idi” will be pulled into your simulation dataset automatically when
running harmonic balance. From there, you can use V/I to calculate the
impedances at each harmonic frequency to then analyze intrinsic
impedances directly on the device for those models when you sweep an external
load.
This technique is more explicitly described in the Class J video at
8:00 th-cam.com/video/N0dLVMoVQSI/w-d-xo.html.
If you don’t have a device
with intrinsic nodes available, get Cds and Cgs off the datasheet (for FETs)
and then optimize your load through this passive network similar to what’s
shown at 10:20. That will give you at least a good starting point from
which you can then run more targeted loadpull simulations to arrive at the best
load value.
Could anyone please tell where can I get hardware for a specified frequency.
Hi and thanks for the tool. I have a question. Is it possible to achieve 100% efficiency with a non-zero knee voltage. This tool suggests that it is but I am not quite sure. And is it possible to include the effect of a non-zero on-resistance when calculating the required load impedances? Thanks in advance.
Best,
Denizhan
It’s never possible
to achieve 100% efficiency for Class E because no real transistor is a perfect
switch.
To add in finite R-on, set the ‘rsat’ parameter in the ideal switch simulation, called waveform_verification (and concurrently set the knee voltage to zero since you’re kind of modeling that with R-on). The efficiency will drop. But probably your actual waveforms will still not match this for high frequency cases because the harmonic terminations impact how the voltage waveform hits the knee. In other words, if the circuit you build does not match the harmonic impedances out to infinity, the voltage trajectory around the knee will change, perhaps dramatically - which might render the R-on based efficiency estimate inaccurate.
At the end of the day, the objective of this tool is to provide ideal waveforms that can serve as a starting point for design, given a model with intrinsic access. The most important thing to get right is the peak voltage swing because of reliability. Adding a knee voltage simply reduces the effective supply value, enabling a better estimate of the peak swing. This is for mathematical convenience and is consistent with Sokal’s original work, but it’s obviously not completely accurate as it ignores the knee. To summarize, modeling the knee transition for Class E is pretty tricky - to the point where you are probably better off just making a guess and starting the design with the actual device to see what happens.
can i ask a question about class e pa?if i use tsmc mos model to design,can it be possible to operate at a frequency about 30GHz??And to select my output matching networks' value?
+Winnie Chang
It will depend on your specs and your device: if the output power requirement is low enough (>>1W) and you can closely integrate the output match, it might be possible. You will need a very high frequency (>=120GHz) matching network, and a device large enough to deliver the power while also having very low parasitics. If the parasitics are too high, no matter what load is presented externally, the internal impedance will not change.
As the video mentions, there are also reliability issues with Class E operation due to the large voltage swings involved which you will need to account for. Higher powers mean higher voltage swings. At low GHz frequencies, some have accomplished >1W Class E operation in standard CMOS technology using stacked configurations and power combining transformers, but as far as I know this hasn’t been scaled up to millimeter wave frequencies. So hopefully your Pout spec is well below 1W.
Hi
First of all, thank you very much for the class-e workspace and synthesis tools. They are really helpful and I was able to create a class-e and simulate it in multisim.
I have a question about the classe_network_preopt schematic shown at 10:14.
In ADS2012, When I open this schematic, I notice that the EBOND4 and EBOND_shape components are replaced by empty boxes. I am guessing that these components are not available in this version of ADS since I get an error when I run the S-Parameters optimization simulation.
My question is as follows:
1 where can I get/download the library with these components?
2 If the components are not available/compatible with ADS2012, can I use the BONDW components in the Passive-RF circuit library?
Thanks and keep up the great videos.
chefboyclakie Thanks! The EBOND components are included with ADS2015, but they aren’t available in ADS2012. The EBONDS are nice because you can fine tune the wire profile to match what your lab bonder can do, but you can absolutely use the BONDW components in ADS2012 to do the same thing with the optimizer. In fact, you can probably use a lumped inductor to model the bonds if you’re looking to optimize for the first time, or even just short the component out entirely--since there are 4 wires in parallel, the series inductance is rather small anyway.
Keysight EEsof EDA If you are interested in upgrading or trying ADS2015, fill out this form and a Keysight engineer will contact you. www.keysight.com/find/eesof-sales-assistance
Thank you very much
Hi, At the time 4':35" the presenter says there are some equations in graphical format to design a class E power amplifier with non 50% duty cycle. Could you please tell me where I can find those equations? (not in your work, the papers including the equations I mean)Thanks
+Jim Style
“RF Power Amplifiers for Wireless Communications”, by Dr.
Steve Cripps is a very good reference for PA Design. In the book,
Dr. Cripps provides a graph which gives normalized Class E circuit values over
a moderate range of conduction angles. Prior to this, there was also
other work dealing with how the components values are impacted by finite
resonator Q. As with conduction angle, this effect can also be described
graphically or through a polynomial fit - a good reference is “Class-E
High-Efficiency RF/Microwave Power Amplifiers: Principles of Operation, Design
Procedures, and Experimental Verification” by Dr. Nathan Sokal -the inventor of
the Class E topology.
+Jim Style
In the workspace shown in this video, it is easy to change
or adjust the conduction angle and resonator Q, with the component values
automatically updating accordingly (there is a variable on the bottom right
hand graph) - so both of these effects are accounted for.
how if i use branch line coupler on two stages class E power amplifier, is it better than push-pull or not? could you tell me how it works? thank you +Keysight EEsof EDA
Branch line is a 90 degree configuration. It can show strange behavior under mismatch conditions, if fed from a common voltage one or other PA may "load hog". Push-pull is 180 degree. Push-pull results in the cancellation of the second harmonic, which can be a great benefit. Under mismatch both amplifiers should load-pull equally.
Hi Sir, I cannot able to download the said files you have provided in your link.
Krisha, Thanks for pointing this out. Something is wrong. We will get it fixed ASAP.
Keysight EEsof EDA Small technical glitch, but it's fixed now. Go ahead and try the download link again. www.keysight.com/find/eesof-how-to-classe
Keysight EEsof EDA Thank you Sir.
i heard class E amplifier is also use in wirelss power trnsmission..is it true? if yes how?
Class E Amplifiers are used in wireless power transmission systems. To state the obvious: WPT systems are very similar to RF systems with the caveat being the transfer of power rather than information. In either system, there may be a limited amount of power to use for transmission OR there may be lots of power to transmit with, perhaps to deal with more spatial separation between Tx/Rx. For both cases, a high efficiency PA is going to be important: if there’s a limited power budget, you’ll want as much power as possible to go to charging as opposed to getting lost in the PA. In a high power system, thermal issues can also become problematic so a high efficiency PA is desirable to minimize heating. In most real systems, both of these requirements together drive the need for a good PA. As this video shows, Class E is a very good topology to achieve high efficiency. As far as wireless power transmission, there are some other benefits too. First, many WPT systems operate a relatively low frequencies (~7 MHz), so the initial design/synthesis method presented here may be more straightforward. For example, at these frequencies you might find that the impedance values from the Class E synthesis tool can be directly applied to the device without having to worry about significant parasitic impedance shifts. Additionally, the fundamental load of Class E is inherently inductive, which is good if you’re ultimately driving a big inductive charging coil. For more information, visit this website: www.keysight.com/find/eesof-ads-power-electronics-resources
i tried a royer oscillator for WPT...but i want to produce more voltage at receiver side to charge multiple devices..any suggestion?
Hello, Why no videos for class D ? :(
How do you bias a class e amplifier?
+Cer Luigi Dela Cruz
Theoretically speaking, you want to bias the amplifier input
to obtain the proper conduction angle- for example, to achieve 180 degree
conduction, you would set the input bias at the threshold voltage such that any
positive signal will conduct while any negative signal will turn the device off
(the “Class B” Bias condition). More information on biasing for
basic classes of operation can be found in this video : How to Design an RF Power Amplifier: Class A, AB and B : th-cam.com/video/GhPqPVlDRPY/w-d-xo.html.
Realistically though, the correct output waveforms occur
with a combination of DC bias and RF input signal drive. So, there
can be several combinations of bias + RF input drive (+input match too!) that
produce very similar output waveforms. This is because if you decrease
the bias and then increase the input drive level, you will end up at
essentially the same place.
At the end of the day, if you are biased
relatively close to the point which gives the specified conduction angle, you
can then fine tune the bias and drive level together to get the right Class E
waveforms and also achieve good small signal gain and a favorable compression
characteristic (AM/AM and AM/PM).
For many PA designs this gain shape vs.
input power is pretty crucial even though the true Class E mode only occurs at
high powers- so that may ultimately be what drives the precise bias point
selection.
+Keysight EEsof EDA Thanks for that information. :)
Hi,I'm students I want to help me to design class E operation at 1.8 G ,I want help for you to learn me steps design, thanks for this video
Why
Русские есть?)