Forums

Regarding amplifier efficiency

Started by amal banerjee October 2, 2019
Could some electronics guru here please clarify this a bit ?

I have been experimenting SPICE RF amplifier designs in my
spare time,and I am getting confused by the differences between 
textbook expressions for amplifier efficiency etc., and my 
experimental findings.

Consider the textbook definition for efficiency as:
efficiency = Power delivered to load/DC power supplied

First, the power delivered to the load(e.g., standard 
50 Ohm) can be measured from the RMS values of current 
through the load and the RMS value of voltage across load.
On the other hand, the DC power value by definition is 
a steady fixed value.

However SPICE experiments give different answers. Take 
a single ended class B RF amplifier with BFR92A and a
parallel LC tuner for 500 MHz. Since the input, output
impedances need to be matched, the S parameters for 
Ic=10mA, Vce=10V are selected. A value for the collector
resistor is calculated. 

If the collector current is measured fro a trial transient 
analysis run, the collector current does not remain at 10 mA, 
In fact the Ic starts at around  8.5 mA and decreases to 
steady state value of 1 mA. So the DC power supplied by the
DC source decreases from the design value(10 mA) to 0.1x design 
value, and consequently only a RMS value for the supplied
DC power can be used in the calculations for rfficierncy. My
observations are based on real SPICE simulations, and differs
from the standard definition of DC power. 

Are there any flawws in this reasoning. All hints, suggestions 
are welcome. Thanks in advance.
    
   
 
amal banerjee wrote:

-----------------------


> > Consider the textbook definition for efficiency as: > efficiency = Power delivered to load/DC power supplied >
** Gotta get your definitions straight first. DC power = AVERAGE DC current times DC supply voltage. Load power = RMS current squared times load R. If load is not pure R, you have issues.
> > My > observations are based on real SPICE simulations, and differs > from the standard definition of DC power. >
** DC power is not up for re-interpretation. .... Phil
On 2.10.19 07:55, amal banerjee wrote:
> Could some electronics guru here please clarify this a bit ? > > I have been experimenting SPICE RF amplifier designs in my > spare time,and I am getting confused by the differences between > textbook expressions for amplifier efficiency etc., and my > experimental findings. > > Consider the textbook definition for efficiency as: > efficiency = Power delivered to load/DC power supplied > > First, the power delivered to the load(e.g., standard > 50 Ohm) can be measured from the RMS values of current > through the load and the RMS value of voltage across load. > On the other hand, the DC power value by definition is > a steady fixed value. > > However SPICE experiments give different answers. Take > a single ended class B RF amplifier with BFR92A and a > parallel LC tuner for 500 MHz. Since the input, output > impedances need to be matched, the S parameters for > Ic=10mA, Vce=10V are selected. A value for the collector > resistor is calculated. > > If the collector current is measured fro a trial transient > analysis run, the collector current does not remain at 10 mA, > In fact the Ic starts at around 8.5 mA and decreases to > steady state value of 1 mA. So the DC power supplied by the > DC source decreases from the design value(10 mA) to 0.1x design > value, and consequently only a RMS value for the supplied > DC power can be used in the calculations for rfficierncy. My > observations are based on real SPICE simulations, and differs > from the standard definition of DC power. > > Are there any flawws in this reasoning. All hints, suggestions > are welcome. Thanks in advance.
Only in Class A is the input power constant. In Class AB and B, the DC current varies according to the RF excitation level. The DC power implies an average input over the RF cycle. Your amplifier should have power supply filtering getting rid of the RF from the DC feed. Please post the .asc file, so we can see what you're doing. -- -TV
On Wednesday, October 2, 2019 at 4:46:20 AM UTC-4, Tauno Voipio wrote:
> On 2.10.19 07:55, amal banerjee wrote: > > Could some electronics guru here please clarify this a bit ? > > > > I have been experimenting SPICE RF amplifier designs in my > > spare time,and I am getting confused by the differences between > > textbook expressions for amplifier efficiency etc., and my > > experimental findings. > > > > Consider the textbook definition for efficiency as: > > efficiency = Power delivered to load/DC power supplied > > > > First, the power delivered to the load(e.g., standard > > 50 Ohm) can be measured from the RMS values of current > > through the load and the RMS value of voltage across load. > > On the other hand, the DC power value by definition is > > a steady fixed value. > > > > However SPICE experiments give different answers. Take > > a single ended class B RF amplifier with BFR92A and a > > parallel LC tuner for 500 MHz. Since the input, output > > impedances need to be matched, the S parameters for > > Ic=10mA, Vce=10V are selected. A value for the collector > > resistor is calculated. > > > > If the collector current is measured fro a trial transient > > analysis run, the collector current does not remain at 10 mA, > > In fact the Ic starts at around 8.5 mA and decreases to > > steady state value of 1 mA. So the DC power supplied by the > > DC source decreases from the design value(10 mA) to 0.1x design > > value, and consequently only a RMS value for the supplied > > DC power can be used in the calculations for rfficierncy. My > > observations are based on real SPICE simulations, and differs > > from the standard definition of DC power. > > > > Are there any flawws in this reasoning. All hints, suggestions > > are welcome. Thanks in advance. > > > Only in Class A is the input power constant. In Class AB and B, > the DC current varies according to the RF excitation level. > > The DC power implies an average input over the RF cycle. Your > amplifier should have power supply filtering getting rid of > the RF from the DC feed. > > Please post the .asc file, so we can see what you're doing. > > -- > > -TV
As I use HSpice and Ngspice, I am sorry I cannot supply the .asc file. I would be glad to provide the text netlist. I have power supply filtering. The DC power supply and TF BJT sub-circuit is as follows(I hope the diagram is not too complicated) VCC--/\/\--RFC inductor--BRF92A collector | Rc | | | --- ---- DC blocking cap --- Cap ---- | GND A current meter is inserted between Rc and RFC inductor to measure DC current from VCC.
On Tuesday, October 1, 2019 at 9:55:40 PM UTC-7, amal banerjee wrote:
> Could some electronics guru here please clarify this a bit ? > > I have been experimenting SPICE RF amplifier designs in my > spare time,and I am getting confused by the differences between > textbook expressions for amplifier efficiency etc., and my > experimental findings. > > Consider the textbook definition for efficiency as: > efficiency = Power delivered to load/DC power supplied > > First, the power delivered to the load(e.g., standard > 50 Ohm) can be measured from the RMS values of current > through the load and the RMS value of voltage across load. > On the other hand, the DC power value by definition is > a steady fixed value. > > However SPICE experiments give different answers. Take > a single ended class B RF amplifier with BFR92A and a > parallel LC tuner for 500 MHz. Since the input, output > impedances need to be matched, the S parameters for > Ic=10mA, Vce=10V are selected. A value for the collector > resistor is calculated. > > If the collector current is measured fro a trial transient > analysis run, the collector current does not remain at 10 mA, > In fact the Ic starts at around 8.5 mA and decreases to > steady state value of 1 mA. So the DC power supplied by the > DC source decreases from the design value(10 mA) to 0.1x design > value, and consequently only a RMS value for the supplied > DC power can be used in the calculations for rfficierncy. My > observations are based on real SPICE simulations, and differs > from the standard definition of DC power. > > Are there any flawws in this reasoning. All hints, suggestions > are welcome.
Yes, there are flaws. You said it was class B, implying "large signal." This is non-linear. It doesn't have "s-parameters" on the output because s-parameters are small signal parameters that presume *linearity*. You said it was class B, but then biased it as if it was class A, for "s parameters." This makes no sense, as evidenced by your results when you drove it with a signal. The quiescent bias current of an undriven class B amp approaches zero. "Impedance" is also a linear small signal assumption. The output of a non-linear amp does not have an impedance, strictly speaking. You should, as a first order approximation, be "load line matching," not "impedance matching.
On Thursday, October 3, 2019 at 3:59:52 PM UTC-4, Simon S Aysdie wrote:
> On Tuesday, October 1, 2019 at 9:55:40 PM UTC-7, amal banerjee wrote: > > Could some electronics guru here please clarify this a bit ? > > > > I have been experimenting SPICE RF amplifier designs in my > > spare time,and I am getting confused by the differences between > > textbook expressions for amplifier efficiency etc., and my > > experimental findings. > > > > Consider the textbook definition for efficiency as: > > efficiency = Power delivered to load/DC power supplied > > > > First, the power delivered to the load(e.g., standard > > 50 Ohm) can be measured from the RMS values of current > > through the load and the RMS value of voltage across load. > > On the other hand, the DC power value by definition is > > a steady fixed value. > > > > However SPICE experiments give different answers. Take > > a single ended class B RF amplifier with BFR92A and a > > parallel LC tuner for 500 MHz. Since the input, output > > impedances need to be matched, the S parameters for > > Ic=10mA, Vce=10V are selected. A value for the collector > > resistor is calculated. > > > > If the collector current is measured fro a trial transient > > analysis run, the collector current does not remain at 10 mA, > > In fact the Ic starts at around 8.5 mA and decreases to > > steady state value of 1 mA. So the DC power supplied by the > > DC source decreases from the design value(10 mA) to 0.1x design > > value, and consequently only a RMS value for the supplied > > DC power can be used in the calculations for rfficierncy. My > > observations are based on real SPICE simulations, and differs > > from the standard definition of DC power. > > > > Are there any flawws in this reasoning. All hints, suggestions > > are welcome. > > Yes, there are flaws. You said it was class B, implying "large signal." This is non-linear. It doesn't have "s-parameters" on the output because s-parameters are small signal parameters that presume *linearity*. > > You said it was class B, but then biased it as if it was class A, for "s parameters." This makes no sense, as evidenced by your results when you drove it with a signal. The quiescent bias current of an undriven class B amp approaches zero. > > "Impedance" is also a linear small signal assumption. The output of a non-linear amp does not have an impedance, strictly speaking. You should, as a first order approximation, be "load line matching," not "impedance matching.
Thank you very much for pointing out the errors in my reasoning.
amal banerjee wrote:

-------------------

> > Thank you very much for pointing out the errors in > my reasoning. >
** When doing computer sims - never forget the rule: " Garbage in = garbage out ". .... Phil
On a sunny day (Wed, 2 Oct 2019 20:53:37 -0700 (PDT)) it happened amal
banerjee <dakupoto@gmail.com> wrote in
<51b1b87f-35fe-43aa-8fb6-7a2edd5447d6@googlegroups.com>:

>As I use HSpice and Ngspice, I am sorry I cannot supply the .asc file. I would be glad to provide the text netlist. > >I have power supply filtering. The DC power supply >and TF BJT sub-circuit is as follows(I hope the >diagram is not too complicated) > >VCC--/\/\--RFC inductor--BRF92A collector >| Rc | >| | >--- ---- DC blocking cap >--- Cap ---- >| >GND > >A current meter is inserted between Rc and >RFC inductor to measure DC current from VCC.
From a practical circuit POV the current meter should be on a not-RF active point, if you are in the point betwwen Rc and RFC, then the current meter's capacitance to ground and test lead inductance will completely change the complex Rc RLC to something equal or worse than: L parasitic Rc-- L---L---RLC-- collector | === C parasitic | /// Freak! it may even oscillate. Not in spice, spice assumes a perfect 'meter' IE spice lies. So in practice it will suck due to internal meter impedance and lead impedances etc, making the whole test useless at 500 MHz. so that leaves this measuring method: V ---I meter---Rx --- RLC --- collector | | === --- --- --- | | added big RF decoupling capacitor (IE 22 nF LOL) /// /// I guess y'all know that but... Oscillating - or different impedance circuits may have a very different current draw....