After many readings that never explain, thanks to you, I finally understand how class E really works and most importantly, how to tune it. Thanks a lot 😁
❤ 👍 For the real amplifier I would recommend considering an LC matching network instead of a T network though. The main source of loss is the inductance (it typically has much lower Q than capacitors) and you ended up having two inductors in the matching network.
You are completely right! the inductors are a big issue when it comes to losses; I guess another thing I could have done was use a lower Q factor for the initial circuit - to have smaller inductor values; regarding T vs L matching - I guess this will highly depend on the application - if you need reduced harmonics for something like a HAM transmitter, even the single T network is insufficient... Anyway, I will not change the circuit since this way the practical build is better linked to the simulation
I just-just started building my very first transceiver and i really needed good info about power amps. And you nailed it, uploaded just right on time! Thank you!
Good presentation, but it might be helpful to emphasize that the class-E switches at the transmit frequency while the class-D has to switch at a large multiple of this frequency, generally at least five times higher and frequently ten times higher, resulting in much higher switching losses. There is also an argument to be made for having turn off occur with ZCS rather than ZVS. Finally, using a duty cycle less than 50% isn't illegal and has interesting effects. The major reason for emphasis on 50% duty cycle is that it makes the mathematics more tractable, but since you are a simulation guru, this shouldn't deter you! ;-)
The class E, inverse F, etc has remained mostly in the academic world, as the practical use is limited to CW, OOK, and FM etc. High order modulation techniques with high data rates require very linear amplifiers. Class E is like a more extreme class C, like exciting a resonant circuit, but uses better terminations for harmonics to restore a sinusoidal waveform. At higher power levels and frequencies getting a good driver becomes a challenge, which should be pretty much a squarewave, and output devices need very low internal capacitance, as it gets in the way of tuning and efficiency. Parts availability, and price, becomes an issue, so overall the concept remains of low use. Like the distributed amplifier, the concept goes back to the 1930's using tubes.
Interesting.... I've been designing a variant on this amplifier as an inverter driver for an isolated AC PSU (I want to make an octopus-style curve tracer that runs from 9VDC). With the inductor being the primary of a transformer.
Your timing is great. I'm trying to build a class E amp with a GAN device for the HF frequencies, and it works great in LTSpice. It's not so great in real life, so I'm anxiously waiting for your next video.
@@FesZElectronics Basically I'm not getting the voltage and current waveforms looking anything like they are supposed to. I'm using an EPC8010, which has a spice model, so I don't think it is non-ideal behavior of the transistor. The closest thing I can get to in LTspice to what I'm getting in the real world is putting stray inductance between source and GND, and between drain and C1, so I created some PCB, slavishly following the design layout recommendations from EPC, and they just came in, so I will see if things are better behaved.
In my simulations, I've seen that adding a gate driver (not just an ideal voltage source, but at least having a series resistor) will help with the realism; depending on the operating frequency, the inductor parallel capacitances will start to also be an issue, and also regarding the inductors, they all have quite a bit of series resistance - from skin effect...
I'm working on a CW-transceiver port backpack use and trying to optimize for everything regarding efficiency, features and size. I was working with class-C (netting only [in my mind] so-so performance) but after seeing the class-E topology/attributes have completely abandoned the class-C. Another consideration is SWR encountered with mis-tuned antennas; it's looking like class-E could be more tolerant of catastrophic mismatches (?). Thanks and looking forward to the build-video! 73...
I guess one of the big issues with class E, as with other switching amplifiers, is getting the transistor to switch fast enough at your frequency of interest. Regarding the connection to mismatched loads, I'm not sure what to say - as always I recommend a quick simulation - both to confirm the transistor you want to use can switch efficiently, and also to see the impact of the load mismatch..
hey, I know you have done videos on controlled sources. would you be willing to show a current controlled resistor? specifically I want to model a LDR like a NSL-32. thanks dan
Very informative series. Thank you very much. Could you please explain or give us example of how the diodes are used to give a binary number ? My second question is: why there is no impedance matching at the output of amplifier, is it 50 ohm?
Very nice video sieres. I've come across a problem and despite putting the same ltspice derivative I'm having the error "can't find the definition ... ". I also tried newer and older versions but it doesn't seem to work. I would love to hear a solution if you have any.
@@FesZElectronics Yes of cource! m1: Can't find definition of model "si2304ds" and the derivative is the exact same with yours. I would appreciate any help. I also watced your video about models and subckt but i didn't managed to make it work. ***************************************************************** * NXP SI2304DS * * Polarity N-Channel * Ratings 30V/7.600000E-002OHMS/1.7A * * Date Created Thu Apr 19 06:31:03 2012 * ***************************************************************** * Model Generated by Transim Technology * Copyright(c) 2004 * All Rights Reserved * UNPUBLISHED LICENSED SOFTWARE * www.transim.com * * Contains Proprietary Information Which * is The Property of Transim or it's licensees. * * Commercial Use or Resale Restricted by * Transim License Agreement. * *****************************************************************
@@ΒασίλειοςΜ-β4ψ Did you set the prefix of the transistor in the schematic to be of type X? if you hold CTRL and right click on the transistor a "component attribute window" will appear, and here the prefix needs to be set to "X" to highlight its a subckt not a mosfet
I don't think there is a lower frequency limit; for the higher value, you will mostly be limited by being able to get the transistor to "switch" - you need to ensure rise and fall times smaller than the half period of the signal; there are practical designs going into the GHz range but usually does are custom made IC's, I'm not sure how easy it would be to make those with discrete components...
I wondered about this too, in the QEX paper (people.physics.anu.edu.au/~dxt103/160m/class_E_amplifier_design.pdf) and the calculator there use 1.0033 (for a Q of 5). I don't think it makes a particularly large difference in the result though. PS nice video, thank you :)
IF the input is a 50% PWM how do we transmit data? as in what type of modulation works in this case? do normal ones send also other duty cycles of waves or do they use OOK ing ?
Nice video! It happened just as I am experimenting with such amplifiers! Is there any workaround for when the Coss of the MOSFET is larger than the designed parallel capacitance?
Choose a different transistor :D In general you will want to have the Coss not just smaller, but significantly smaller since the Coss is not really a constant, so if that is the only resonant capacitance you are relying on it will create losses and distortion.
I guess you could try to design the circuit to run into a very small impedance - that would increase the needed parallel capacitance. Anyway, I think it will not be that easy to get the transistor to switch at 28MHz in the first place - you will need a very fast transistor. As always, I highly recommend that you try simulating the circuit to see how it might behave before building anything - of course if you have a simulation model available.
Both increasing the output power or decreasing Vdd can do the trick. Decreasing Vdd is not my favorite idea, as the MOSFET become less well behaved. Also, I'd like to constrain the power to .5~1W. This is too much for a 2N7000 to handle and the IRF510 has too much capacitance. Also the driving power of the Si5351 used as signal generator is really low (5mW), so I'm now considering using a class A or single sided class B to take it to 50~100mW and then pushing it through a class C to reach 1W.
By doing impedance matching my current gets very high compared to my output voltage it eliminates those reflections; however, I wish to not have a very high current since it’s for some coils and it makes the application more difficult
After many readings that never explain, thanks to you, I finally understand how class E really works and most importantly, how to tune it.
Thanks a lot 😁
You've been a major help here in the lab. Cheers!
❤ 👍 For the real amplifier I would recommend considering an LC matching network instead of a T network though. The main source of loss is the inductance (it typically has much lower Q than capacitors) and you ended up having two inductors in the matching network.
You are completely right! the inductors are a big issue when it comes to losses; I guess another thing I could have done was use a lower Q factor for the initial circuit - to have smaller inductor values; regarding T vs L matching - I guess this will highly depend on the application - if you need reduced harmonics for something like a HAM transmitter, even the single T network is insufficient... Anyway, I will not change the circuit since this way the practical build is better linked to the simulation
I just-just started building my very first transceiver and i really needed good info about power amps. And you nailed it, uploaded just right on time!
Thank you!
Good presentation, but it might be helpful to emphasize that the class-E switches at the transmit frequency while the class-D has to switch at a large multiple of this frequency, generally at least five times higher and frequently ten times higher, resulting in much higher switching losses. There is also an argument to be made for having turn off occur with ZCS rather than ZVS. Finally, using a duty cycle less than 50% isn't illegal and has interesting effects. The major reason for emphasis on 50% duty cycle is that it makes the mathematics more tractable, but since you are a simulation guru, this shouldn't deter you! ;-)
Really enjoy this amplifier series. Thank you FesZ!
Really good video, useful methods and practices shown. Thanks!
Great introduction to Class E amplifiers. Thankyou!
The class E, inverse F, etc has remained mostly in the academic world, as the practical use is limited to CW, OOK, and FM etc.
High order modulation techniques with high data rates require very linear amplifiers.
Class E is like a more extreme class C, like exciting a resonant circuit, but uses better terminations for harmonics to restore a sinusoidal waveform. At higher power levels and frequencies getting a good driver becomes a challenge, which should be pretty much a squarewave, and output devices need very low internal capacitance, as it gets in the way of tuning and efficiency.
Parts availability, and price, becomes an issue, so overall the concept remains of low use.
Like the distributed amplifier, the concept goes back to the 1930's using tubes.
Interesting.... I've been designing a variant on this amplifier as an inverter driver for an isolated AC PSU (I want to make an octopus-style curve tracer that runs from 9VDC). With the inductor being the primary of a transformer.
Your timing is great. I'm trying to build a class E amp with a GAN device for the HF frequencies, and it works great in LTSpice. It's not so great in real life, so I'm anxiously waiting for your next video.
What sort of issues do you have? I mean what exactly is different in the real circuit compared to the simulation
@@FesZElectronics Basically I'm not getting the voltage and current waveforms looking anything like they are supposed to. I'm using an EPC8010, which has a spice model, so I don't think it is non-ideal behavior of the transistor. The closest thing I can get to in LTspice to what I'm getting in the real world is putting stray inductance between source and GND, and between drain and C1, so I created some PCB, slavishly following the design layout recommendations from EPC, and they just came in, so I will see if things are better behaved.
In my simulations, I've seen that adding a gate driver (not just an ideal voltage source, but at least having a series resistor) will help with the realism; depending on the operating frequency, the inductor parallel capacitances will start to also be an issue, and also regarding the inductors, they all have quite a bit of series resistance - from skin effect...
Excellent knowledge! Thanks Fesz!
Much enjoyed this vid.
Thanx for great content
Yay! I’ve been waiting for you to do this for ages! Thank you!
Ah, awesome, I've been waiting for your take on Class E
I'm working on a CW-transceiver port backpack use and trying to optimize for everything regarding efficiency, features and size.
I was working with class-C (netting only [in my mind] so-so performance) but after seeing the class-E topology/attributes have completely abandoned the class-C.
Another consideration is SWR encountered with mis-tuned antennas; it's looking like class-E could be more tolerant of catastrophic mismatches (?).
Thanks and looking forward to the build-video! 73...
I guess one of the big issues with class E, as with other switching amplifiers, is getting the transistor to switch fast enough at your frequency of interest. Regarding the connection to mismatched loads, I'm not sure what to say - as always I recommend a quick simulation - both to confirm the transistor you want to use can switch efficiently, and also to see the impact of the load mismatch..
Thank you! I'd asked for this, and appreciate you doing it.
It takes the time it takes, but I eventually do get to the various subjects. Anyway, I also found this one quite interesting and fun.
hey, I know you have done videos on controlled sources. would you be willing to show a current controlled resistor? specifically I want to model a LDR like a NSL-32. thanks dan
Excellent video!
Very informative series. Thank you very much.
Could you please explain or give us example of how the diodes are used to give a binary number ?
My second question is: why there is no impedance matching at the output of amplifier, is it 50 ohm?
Very nice video sieres. I've come across a problem and despite putting the same ltspice derivative I'm having the error "can't find the definition ... ". I also tried newer and older versions but it doesn't seem to work. I would love to hear a solution if you have any.
Maybe you can share some details - what is the exact derivative and the full error code
@@FesZElectronics Yes of cource! m1: Can't find definition of model "si2304ds" and the derivative is the exact same with yours. I would appreciate any help. I also watced your video about models and subckt but i didn't managed to make it work.
*****************************************************************
* NXP SI2304DS
*
* Polarity N-Channel
* Ratings 30V/7.600000E-002OHMS/1.7A
*
* Date Created Thu Apr 19 06:31:03 2012
*
*****************************************************************
* Model Generated by Transim Technology
* Copyright(c) 2004
* All Rights Reserved
* UNPUBLISHED LICENSED SOFTWARE
* www.transim.com
*
* Contains Proprietary Information Which
* is The Property of Transim or it's licensees.
*
* Commercial Use or Resale Restricted by
* Transim License Agreement.
*
*****************************************************************
.SUBCKT SI2304DS DRAIN GATE SOURCE
LD DRAIN 5 4e-10
RLD2 DRAIN 5 1.00530964914873
RLD1 5 4 0.0023
LG GATE 1 7.97740315722523e-10
RLG GATE 1 2.00494009227702
LS SOURCE 8 6.94325122020549e-10
RLS2 SOURCE 8 1.74502936203407
RLS1 8 7 0.00952211428140778
RDS 7 4 3000000000 TC=-0.005
RS 6 7 0.0001
RD 3 4 0.0408969766276831 TC=0.00464827672211182,2.47227840906323E-05
RBD 9 4 0.0279781624816791 TC=0.00464827672211182,2.47227840906323E-05
DBD 7 9 DBD
M1 3 2 6 6 MINT
RGS 2 6 1000000000
CGS 2 6 1.157e-10
RG 1 2 1.97713365715578
* CGD
C11 11 12 1E-12
V11 11 0 0Vdc
G11 3 2 VALUE { V(13, 0)*I(V11) }
E11 12 0 3 2 1
E12 13 0 TABLE {V(12)}
+ -20 172.8
+ -12 171.5
+ -10 171
+ -5 169.1
+ -4 169.3
+ -3 170.5
+ -2 275.8
+ -1 247.7
+ -0.5 229.5
+ -0.2 219
+ -0.1 213.4
+ 0 204.15
+ 0.1 195.9
+ 0.2 185.2
+ 0.5 165.8
+ 1 133.1
+ 2 104
+ 5 73
+ 10 55.3
+ 12 51.3
+ 20 41.5
+ 25 37.9
+ 30 35.3
.MODEL MINT NMOS(Vto=2.55 Kp=5.815550e+000 Nfs=1160000000000 Eta=100
+ Level=3 L=1e-4 W=1e-4 Gamma=0 Phi=0.6 Is=1e-24
+ Js=0 Pb=0.8 Cj=0 Cjsw=0 Cgso=0 Cgdo=0 Cgbo=0
+ Tox=1e-07 Xj=0
+ U0=600 Vmax=2000)
.MODEL DBD D(Bv=38.6 Ibv=1.000000E-005 Rs=1E-6 Is=1.53489332570028e-13
+ N=1 M=0.44 VJ=0.5 Fc=0.5 Cjo=1.097e-10 Tt=1.85e-08)
.ENDS
@@ΒασίλειοςΜ-β4ψ Did you set the prefix of the transistor in the schematic to be of type X? if you hold CTRL and right click on the transistor a "component attribute window" will appear, and here the prefix needs to be set to "X" to highlight its a subckt not a mosfet
What are the typical upper power and frequency for the E class amplifier?
I don't think there is a lower frequency limit; for the higher value, you will mostly be limited by being able to get the transistor to "switch" - you need to ensure rise and fall times smaller than the half period of the signal; there are practical designs going into the GHz range but usually does are custom made IC's, I'm not sure how easy it would be to make those with discrete components...
Thank you!
Hi @FesZ Electronics> What does it mean 1,154 in the expression ?
You mean at 4:38 for calculating Pout? its just a number "1.154". I am not sure of the exact way in which it was determined.
@@FesZElectronics Is it possible that this is the way to calculate the output voltage peak?
I wondered about this too, in the QEX paper (people.physics.anu.edu.au/~dxt103/160m/class_E_amplifier_design.pdf) and the calculator there use 1.0033 (for a Q of 5). I don't think it makes a particularly large difference in the result though.
PS nice video, thank you :)
IF the input is a 50% PWM how do we transmit data? as in what type of modulation works in this case? do normal ones send also other duty cycles of waves or do they use OOK ing ?
My guess: narrow band FM.
Nice video! It happened just as I am experimenting with such amplifiers! Is there any workaround for when the Coss of the MOSFET is larger than the designed parallel capacitance?
Choose a different transistor :D In general you will want to have the Coss not just smaller, but significantly smaller since the Coss is not really a constant, so if that is the only resonant capacitance you are relying on it will create losses and distortion.
@@FesZElectronics you just crushed my hopes of having a high efficiency PA for my 28.1MHz beacon. :-D
I'll try to make it class C then ...
I guess you could try to design the circuit to run into a very small impedance - that would increase the needed parallel capacitance. Anyway, I think it will not be that easy to get the transistor to switch at 28MHz in the first place - you will need a very fast transistor. As always, I highly recommend that you try simulating the circuit to see how it might behave before building anything - of course if you have a simulation model available.
Both increasing the output power or decreasing Vdd can do the trick. Decreasing Vdd is not my favorite idea, as the MOSFET become less well behaved. Also, I'd like to constrain the power to .5~1W. This is too much for a 2N7000 to handle and the IRF510 has too much capacitance.
Also the driving power of the Si5351 used as signal generator is really low (5mW), so I'm now considering using a class A or single sided class B to take it to 50~100mW and then pushing it through a class C to reach 1W.
Why does the output power of the class-E amplifier have different pk-pk without adding the compensating network.
By doing impedance matching my current gets very high compared to my output voltage it eliminates those reflections; however, I wish to not have a very high current since it’s for some coils and it makes the application more difficult
I wanna see a Class E audio amplifier, it must be possible somehow.
Perfect.
So it suit uses in an IF tuner, isn't it.
Is it possible to add a control loop to this clasa of amplifier?
I'm not sure - are you thinking of controlling the duty cycle based on the output amplitude?
@@FesZElectronics yes