Topics

LTSpice simulation of uBitx Power Amp

Jerry Gaffke
 

I've uploaded an LTSpice simulation of the uBitx Power Amplifier from Q90 on out through the IRF510's
but not including any of the transmit low pass filters, can be found here:
    https://groups.io/g/BITX20/files/KE7ER/ubitx_pa.asc
A pdf of the LTSpice schematic is here:
    https://groups.io/g/BITX20/files/KE7ER/ubitx_pa.pdf

I ran several simulations, all with a 0.05 volt pk-to-pk sine wave driven into C80 at the base of Q90.
Anything much higher and there was some definite gain compression going on.

With the IRF510 gates clipped out of the circuit, I saw 7.7v p-p of drive for each gate at 30mhz, 10.0v p-p at 7mhz.
Note that these signals were still loaded by the 47ohm resistors at R97 and R98

With the gates restored and the IRF510's powered from 12v, the gates had 3.0v p-p at 30mhz, 8.0v p-p at 7mhz.
Simulated power out (computed from the unfiltered p-p voltage into 50 ohms) was 5W at 7mhz and 0.9W at 30mhz. 
  
Will be interesting to see what this does if the gates are driven harder.
Also interesting to try a simulation of the WA2EBY amp using the same IRF510 models.

Jerry, KE7ER

Jerry Gaffke
 

I've now uploaded an LTSpice simulation schematic file that allows easy comparison of the uBitx and WA2EBY finals,
also a pdf version for those who have not installed LTSpice:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.pdf

I haven't played with it too much, haven't taken any notes.
But here's what I think I'm seeing:

The WA2EBY does perform better. 
Power in is about the same between the two, power out from the WA2EBY is generally a factor of 3 greater.
The gates on the WA2EBY have nice sine waves, the uBitx is quite distorted there by capacitive loading.
Input impedance on both is about 50 ohms at 7mhz, both go down to about 12 ohms at 30mhz,
as determined by monitoring the voltage after the 50 ohm source impedance resistor.
Both show a 4x power increase when switching from 30mhz to 7mhz, more than I would
have expected on the WA2EBY, perhaps I have some wrong values on the inductors?
The IRF510 model assumes worst case for Rds(on) and Qgate, though I doubt that's just too far off.

Jerry, KE7ER

John Backo
 

Very good, Jerry.

However, note that the output is NOT 50 ohms until AFTER the LPF.
The input to the LPF is somewhat lower. And the WA2EBY has a voltage advantage
over the 12V input to the drains, though the output of the drains is closer to 50 ohms.

So it is a bit of a misnomer to assume that the drain load is 50 ohms. This is true, incidentally,
of any AB linear amplifier (and is probably one of the sources of oscillation). I should think that the
true drain load in this configuration is closer to 10 ohms than 50 ohms. The LPF transforms it
to 50 ohms for the antenna.

But it is not unreasonable to assume an overall load of 50 ohms. But be aware though, that one of the functions of
the LPF is to produce this. It is not quite true at the drains. The actual impedance at the drains is a function
of the first capacitor/inductance stage of the LPF. Things are not quite as simple as they seem...

john
AD5YE

Jerry Gaffke
 


Hmmm,

I'm not assuming anything about drain loads, only that the load on the output of the transformer should be 50 ohms.
The LPF's on the uBitx and WA2EBY are symetrical, 50 ohms in and out, no impedance transformation there (at least for the fundamental).
Both amps are operating the IRF510's from a 24v supply, and the two drains of each amp
are bouncing about centered on 24vdc, out of phase.

The output should look very similar after we add in a low pass filter, 
though any harmonics will be knocked out to leave behind a clean sine wave.
 
Jerry


On Sun, Feb 4, 2018 at 10:53 pm, John Backo wrote:
However, note that the output is NOT 50 ohms until AFTER the LPF.
The input to the LPF is somewhat lower. And the WA2EBY has a voltage advantage
over the 12V input to the drains, though the output of the drains is closer to 50 ohms.

So it is a bit of a misnomer to assume that the drain load is 50 ohms. This is true, incidentally,
of any AB linear amplifier (and is probably one of the sources of oscillation). I should think that the
true drain load in this configuration is closer to 10 ohms than 50 ohms. The LPF transforms it
to 50 ohms for the antenna.

But it is not unreasonable to assume an overall load of 50 ohms. But be aware though, that one of the functions of
the LPF is to produce this. It is not quite true at the drains. The actual impedance at the drains is a function
of the first capacitor/inductance stage of the LPF. Things are not quite as simple as they seem...

K9HZ <bill@...>
 

I have the .model for the RD series transistors if you care to run that too.  


Dr. William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ

 

Owner - Operator

Big Signal Ranch – K9ZC

Staunton, Illinois

 

Owner – Operator

Villa Grand Piton - J68HZ

Soufriere, St. Lucia W.I.

Rent it: www.VillaGrandPiton.com


email:  bill@...

 


On Feb 5, 2018, at 12:21 AM, Jerry Gaffke via Groups.Io <jgaffke@...> wrote:

I've now uploaded an LTSpice simulation schematic file that allows easy comparison of the uBitx and WA2EBY finals,
also a pdf version for those who have not installed LTSpice:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.pdf

I haven't played with it too much, haven't taken any notes.
But here's what I think I'm seeing:

The WA2EBY does perform better. 
Power in is about the same between the two, power out from the WA2EBY is generally a factor of 3 greater.
The gates on the WA2EBY have nice sine waves, the uBitx is quite distorted there by capacitive loading.
Input impedance on both is about 50 ohms at 7mhz, both go down to about 12 ohms at 30mhz,
as determined by monitoring the voltage after the 50 ohm source impedance resistor.
Both show a 4x power increase when switching from 30mhz to 7mhz, more than I would
have expected on the WA2EBY, perhaps I have some wrong values on the inductors?
The IRF510 model assumes worst case for Rds(on) and Qgate, though I doubt that's just too far off.

Jerry, KE7ER

K9HZ <bill@...>
 

Dunno about that.  I always design my LP filters at 50:50 ohms in and out... and my PA transformers are 50 ohms at the output.  I don't know why you would want to make the LP filter a transform too?


Dr. William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ

 

Owner - Operator

Big Signal Ranch – K9ZC

Staunton, Illinois

 

Owner – Operator

Villa Grand Piton - J68HZ

Soufriere, St. Lucia W.I.

Rent it: www.VillaGrandPiton.com


email:  bill@...

 


On Feb 5, 2018, at 12:53 AM, John Backo <jabac@...> wrote:

Very good, Jerry.

However, note that the output is NOT 50 ohms until AFTER the LPF.
The input to the LPF is somewhat lower. And the WA2EBY has a voltage advantage
over the 12V input to the drains, though the output of the drains is closer to 50 ohms.

So it is a bit of a misnomer to assume that the drain load is 50 ohms. This is true, incidentally,
of any AB linear amplifier (and is probably one of the sources of oscillation). I should think that the
true drain load in this configuration is closer to 10 ohms than 50 ohms. The LPF transforms it
to 50 ohms for the antenna.

But it is not unreasonable to assume an overall load of 50 ohms. But be aware though, that one of the functions of
the LPF is to produce this. It is not quite true at the drains. The actual impedance at the drains is a function
of the first capacitor/inductance stage of the LPF. Things are not quite as simple as they seem...

john
AD5YE



Jerry Gaffke
 

Have now uploaded version two of the simulation:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.pdf

Has the following changes:
Quiescent drain current reduced to 10ma on both.
Added a switching mechanism to show response at both 7 and 30 mhz with a single run.

Observations:

Power out on WA2EBY is still much lower at 30mhz than he reported in his QST article.
Input impedance at 30mhz falls to 12 ohms, which is the primary factor here.
If Vin source impedance is reduced from 50 ohms to 0 ohms, then 30mhz output exceeds 7mhz output.
 
Removing C3 increases output a bit at both 7 and 30mhz 

L3 helps make the exciter see a resistive load, but doesn't help much.
 
I'm moving on to other things.
Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz
Here's my notes on how I got the WA2EBY inductor values, includes a python script for single layer coils:

# http://info.ee.surrey.ac.uk/Workshop/advice/coils/air_coils.html
#   Single layer coils, wheeler formula:  Henrys = 0.001 * N**2* R**2 / (228R + 254L)
#   where N is number of turns, R is radius in meters, L is length in meters (>0.8r)
def h(n, d, w): # turns, form diameter in inches, wire diameter in inches
   length = w*n/39.37 # length of coil in meters, assuming close wound
   radius = (d+w)/2/39.37 # radius of coil from middle of wire in meters
   henries = 0.001 * n**2* radius**2 / (228*radius + 254*length)
   return(henries)

L1, L2   9-1/2 turns #24 enameled wire, closely wound 0.25-in. ID       >>> h(9.5, 0.25,  0.02):  5.285e-07 = 0.5285uH
L3       3-1/2 turns #24 enameled wire, closely wound 0.190-in. ID      >>> h(3.5, 0.190, 0.02):  8.222e-08 = 0.0822uH
T1       10 bifilar turns #24 enameled wire on an FT-50-43 core  uH=(AL*Turns**2)/1000  AL=440     (440*100)/1000 = 44uH/winding
T2       10 bifilar turns #22 enameled wire on two stacked FT-50-43 cores. AL doubles (880*100)/1000 = 88uH/winding
T3       Pri 2 turns, sec 3 turns #20 Tefloncovered wire on BN-43-3312 balun core. AL=7000    (7000*4)/1000=28uH, (7000*9)/1000=63uH

Jerry,  KE7ER



On Sun, Feb 4, 2018 at 10:21 pm, Jerry Gaffke wrote:

I've now uploaded an LTSpice simulation schematic file that allows easy comparison of the uBitx and WA2EBY finals,
also a pdf version for those who have not installed LTSpice:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.pdf

I haven't played with it too much, haven't taken any notes.
But here's what I think I'm seeing:

The WA2EBY does perform better. 
Power in is about the same between the two, power out from the WA2EBY is generally a factor of 3 greater.
The gates on the WA2EBY have nice sine waves, the uBitx is quite distorted there by capacitive loading.
Input impedance on both is about 50 ohms at 7mhz, both go down to about 12 ohms at 30mhz,
as determined by monitoring the voltage after the 50 ohm source impedance resistor.
Both show a 4x power increase when switching from 30mhz to 7mhz, more than I would
have expected on the WA2EBY, perhaps I have some wrong values on the inductors?
The IRF510 model assumes worst case for Rds(on) and Qgate, though I doubt that's just too far off.

Jerry, KE7ER

Diver Martin <diver.martin@...>
 

> Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz

To quote George E.P. Box, "All models are wrong, but some are useful."

Spice models are just that, models.  They're rarely right.  But they can tell you useful things about some of the mechanics of what's going on inside the system.

On Mon, Feb 5, 2018 at 10:40 AM, Jerry Gaffke via Groups.Io <jgaffke@...> wrote:
Have now uploaded version two of the simulation:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.pdf

Has the following changes:
Quiescent drain current reduced to 10ma on both.
Added a switching mechanism to show response at both 7 and 30 mhz with a single run.

Observations:

Power out on WA2EBY is still much lower at 30mhz than he reported in his QST article.
Input impedance at 30mhz falls to 12 ohms, which is the primary factor here.
If Vin source impedance is reduced from 50 ohms to 0 ohms, then 30mhz output exceeds 7mhz output.
 
Removing C3 increases output a bit at both 7 and 30mhz 

L3 helps make the exciter see a resistive load, but doesn't help much.
 
I'm moving on to other things.
Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz
Here's my notes on how I got the WA2EBY inductor values, includes a python script for single layer coils:

# http://info.ee.surrey.ac.uk/Workshop/advice/coils/air_coils.html
#   Single layer coils, wheeler formula:  Henrys = 0.001 * N**2* R**2 / (228R + 254L)
#   where N is number of turns, R is radius in meters, L is length in meters (>0.8r)
def h(n, d, w): # turns, form diameter in inches, wire diameter in inches
   length = w*n/39.37 # length of coil in meters, assuming close wound
   radius = (d+w)/2/39.37 # radius of coil from middle of wire in meters
   henries = 0.001 * n**2* radius**2 / (228*radius + 254*length)
   return(henries)

L1, L2   9-1/2 turns #24 enameled wire, closely wound 0.25-in. ID       >>> h(9.5, 0.25,  0.02):  5.285e-07 = 0.5285uH
L3       3-1/2 turns #24 enameled wire, closely wound 0.190-in. ID      >>> h(3.5, 0.190, 0.02):  8.222e-08 = 0.0822uH
T1       10 bifilar turns #24 enameled wire on an FT-50-43 core  uH=(AL*Turns**2)/1000  AL=440     (440*100)/1000 = 44uH/winding
T2       10 bifilar turns #22 enameled wire on two stacked FT-50-43 cores. AL doubles (880*100)/1000 = 88uH/winding
T3       Pri 2 turns, sec 3 turns #20 Tefloncovered wire on BN-43-3312 balun core. AL=7000    (7000*4)/1000=28uH, (7000*9)/1000=63uH

Jerry,  KE7ER



On Sun, Feb 4, 2018 at 10:21 pm, Jerry Gaffke wrote:

I've now uploaded an LTSpice simulation schematic file that allows easy comparison of the uBitx and WA2EBY finals,
also a pdf version for those who have not installed LTSpice:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.pdf

I haven't played with it too much, haven't taken any notes.
But here's what I think I'm seeing:

The WA2EBY does perform better. 
Power in is about the same between the two, power out from the WA2EBY is generally a factor of 3 greater.
The gates on the WA2EBY have nice sine waves, the uBitx is quite distorted there by capacitive loading.
Input impedance on both is about 50 ohms at 7mhz, both go down to about 12 ohms at 30mhz,
as determined by monitoring the voltage after the 50 ohm source impedance resistor.
Both show a 4x power increase when switching from 30mhz to 7mhz, more than I would
have expected on the WA2EBY, perhaps I have some wrong values on the inductors?
The IRF510 model assumes worst case for Rds(on) and Qgate, though I doubt that's just too far off.

Jerry, KE7ER




--
Martin Held - AE7EU
http://ae7eu.com/
-------------------------------------------------------------------------------
If there aren't any questions, then what is there to learn?

K9HZ <bill@...>
 

Mod ls are Starting points without spending $. 


Dr. William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ

 

Owner - Operator

Big Signal Ranch – K9ZC

Staunton, Illinois

 

Owner – Operator

Villa Grand Piton - J68HZ

Soufriere, St. Lucia W.I.

Rent it: www.VillaGrandPiton.com


email:  bill@...

 


On Feb 5, 2018, at 12:47 PM, Diver Martin <diver.martin@...> wrote:

> Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz

To quote George E.P. Box, "All models are wrong, but some are useful."

Spice models are just that, models.  They're rarely right.  But they can tell you useful things about some of the mechanics of what's going on inside the system.

On Mon, Feb 5, 2018 at 10:40 AM, Jerry Gaffke via Groups.Io <jgaffke@...> wrote:
Have now uploaded version two of the simulation:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX_2.pdf

Has the following changes:
Quiescent drain current reduced to 10ma on both.
Added a switching mechanism to show response at both 7 and 30 mhz with a single run.

Observations:

Power out on WA2EBY is still much lower at 30mhz than he reported in his QST article.
Input impedance at 30mhz falls to 12 ohms, which is the primary factor here.
If Vin source impedance is reduced from 50 ohms to 0 ohms, then 30mhz output exceeds 7mhz output.
 
Removing C3 increases output a bit at both 7 and 30mhz 

L3 helps make the exciter see a resistive load, but doesn't help much.
 
I'm moving on to other things.
Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz
Here's my notes on how I got the WA2EBY inductor values, includes a python script for single layer coils:

# http://info.ee.surrey.ac.uk/Workshop/advice/coils/air_coils.html
#   Single layer coils, wheeler formula:  Henrys = 0.001 * N**2* R**2 / (228R + 254L)
#   where N is number of turns, R is radius in meters, L is length in meters (>0.8r)
def h(n, d, w): # turns, form diameter in inches, wire diameter in inches
   length = w*n/39.37 # length of coil in meters, assuming close wound
   radius = (d+w)/2/39.37 # radius of coil from middle of wire in meters
   henries = 0.001 * n**2* radius**2 / (228*radius + 254*length)
   return(henries)

L1, L2   9-1/2 turns #24 enameled wire, closely wound 0.25-in. ID       >>> h(9.5, 0.25,  0.02):  5.285e-07 = 0.5285uH
L3       3-1/2 turns #24 enameled wire, closely wound 0.190-in. ID      >>> h(3.5, 0.190, 0.02):  8.222e-08 = 0.0822uH
T1       10 bifilar turns #24 enameled wire on an FT-50-43 core  uH=(AL*Turns**2)/1000  AL=440     (440*100)/1000 = 44uH/winding
T2       10 bifilar turns #22 enameled wire on two stacked FT-50-43 cores. AL doubles (880*100)/1000 = 88uH/winding
T3       Pri 2 turns, sec 3 turns #20 Tefloncovered wire on BN-43-3312 balun core. AL=7000    (7000*4)/1000=28uH, (7000*9)/1000=63uH

Jerry,  KE7ER



On Sun, Feb 4, 2018 at 10:21 pm, Jerry Gaffke wrote:

I've now uploaded an LTSpice simulation schematic file that allows easy comparison of the uBitx and WA2EBY finals,
also a pdf version for those who have not installed LTSpice:
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.asc
    https://groups.io/g/BITX20/files/KE7ER/WA2EBYandUBITX.pdf

I haven't played with it too much, haven't taken any notes.
But here's what I think I'm seeing:

The WA2EBY does perform better. 
Power in is about the same between the two, power out from the WA2EBY is generally a factor of 3 greater.
The gates on the WA2EBY have nice sine waves, the uBitx is quite distorted there by capacitive loading.
Input impedance on both is about 50 ohms at 7mhz, both go down to about 12 ohms at 30mhz,
as determined by monitoring the voltage after the 50 ohm source impedance resistor.
Both show a 4x power increase when switching from 30mhz to 7mhz, more than I would
have expected on the WA2EBY, perhaps I have some wrong values on the inductors?
The IRF510 model assumes worst case for Rds(on) and Qgate, though I doubt that's just too far off.

Jerry, KE7ER




--
Martin Held - AE7EU
http://ae7eu.com/
-------------------------------------------------------------------------------
If there aren't any questions, then what is there to learn?

Jerry Gaffke
 

I suppose.
The LTSpice model of the IRF510 does assume worst case, perhaps that is part of it.
But the WA2EBY model is off from the performance reported in the QST article by more than I would expect when operating at 30mhz.
And the uBitx models pretty close to the performance we actually get.

Regardless, looks like we'd do well to hack the uBitx to use something closer to the WA2EBY circuit.
And if that doesn't give the required performance at 30mhz, then drive the IRF510's from a lower impedance source.
LTSpice is whispering to me that if driven hard enough, the IRF510's will do 30W at 30mhz.

Jerry, KE7ER


On Mon, Feb 5, 2018 at 10:47 am, Diver Martin wrote:
> Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz
 
To quote George E.P. Box, "All models are wrong, but some are useful."
 
Spice models are just that, models.  They're rarely right.  But they can tell you useful things about some of the mechanics of what's going on inside the system.
 

Diver Martin <diver.martin@...>
 

Looking at the schematic, has anyone tried changing the turns ratio on T10, or monkeying with the preamp?  It would lower the drive level, but also lower the drive impedance.  One might also increase the resistance slightly of R90/93 (increasing the drive bias of the preamp stage, dangerous) or lowering the emitter degeneration resistors (R941, 911, 942 and 96) from 22 ohms to something like 10 ohms (Degeneration is useful here though, it provides linearity to the amplifier).

I use IRF530's in a different amplifier design that puts out 50W with about 3W in at 30MHz.  Not sure what the preamp stage is rated for, I'd assume ~0.5 to 1W tops.

On Mon, Feb 5, 2018 at 11:29 AM, Jerry Gaffke via Groups.Io <jgaffke@...> wrote:
I suppose.
The LTSpice model of the IRF510 does assume worst case, perhaps that is part of it.
But the WA2EBY model is off from the performance reported in the QST article by more than I would expect when operating at 30mhz.
And the uBitx models pretty close to the performance we actually get.

Regardless, looks like we'd do well to hack the uBitx to use something closer to the WA2EBY circuit.
And if that doesn't give the required performance at 30mhz, then drive the IRF510's from a lower impedance source.
LTSpice is whispering to me that if driven hard enough, the IRF510's will do 30W at 30mhz.

Jerry, KE7ER


On Mon, Feb 5, 2018 at 10:47 am, Diver Martin wrote:
> Let me know if anybody figures out why my model of the WA2EBY amp shows the output to be so much lower at 30mhz
 
To quote George E.P. Box, "All models are wrong, but some are useful."
 
Spice models are just that, models.  They're rarely right.  But they can tell you useful things about some of the mechanics of what's going on inside the system.
 




--
Martin Held - AE7EU
http://ae7eu.com/
-------------------------------------------------------------------------------
If there aren't any questions, then what is there to learn?