开云体育

ctrl + shift + ? for shortcuts
© 2025 开云体育
  1. QRPLabs
  2. Topics

Trying to understand linear PA 10W construction

 

Hello,
first of all, I apologize because I am not asking for advice about a QRP Labs product but rather an advice for my own homebrewing experiments. Recently I am into PA with IRF510 mosfets.
?
I looked at several linear amplifiers using IRF510 and most of them use a negative feedback network, usually a resistor between 470 and 1000 ohms connected between drain and gate, DC isolated by a large enough capacitor (1 nF, 10 nF, even 100 nF). QRP Labs is different because there is an additional inductance in series (1 uH).
?
I tried to estimate the effect of this inductance and it seems that in combination with the resistor used (220 ohms) it should produce a phase shift which is negligible on lower frequencies but rises to approximately 55 degrees at 28 MHz, effectively decreasing the negative feedback effect (by ~30% ?) and thus equalizing the difference between gain at lower and higher frequencies, which is usually the biggest problem of amplifiers with this mosfet.??
?
I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
Thank you if you can point me to some instructive source of information.
?
73 Jindra OK4RM??
 

开云体育

You pose a number of different questions regarding negative feedback and then bypassing considerations.

First bypassing: ?it is very common and good engineering practice to put an electrolytic capacitor, such as a 10uF 25V, in parallel with a 0.1uF (100nF) 50V ceramic, such as a 1206, and a 1nF (1000pF) 50V ceramic all in parallel so that each capacitor bypasses in the frequency range where they have low impedance. The smallest capacitor is positioned closest to the RF choke leading to the active device.

Second feedback: ?The additional small inductance can indeed reduce negative feedback and allow more gain at the higher frequencies where the device is dropping off. ?But this is as much an art as engineering, and my suggestion would be to try without at first, to see how your circuit performs. If it is needed then try it.

Dave

On Feb 18, 2025, at 11:12, ok4rm via groups.io <radio.miskovice@...> wrote:

?
Hello,
first of all, I apologize because I am not asking for advice about a QRP Labs product but rather an advice for my own homebrewing experiments. Recently I am into PA with IRF510 mosfets.
?
I looked at several linear amplifiers using IRF510 and most of them use a negative feedback network, usually a resistor between 470 and 1000 ohms connected between drain and gate, DC isolated by a large enough capacitor (1 nF, 10 nF, even 100 nF). QRP Labs is different because there is an additional inductance in series (1 uH).
?
I tried to estimate the effect of this inductance and it seems that in combination with the resistor used (220 ohms) it should produce a phase shift which is negligible on lower frequencies but rises to approximately 55 degrees at 28 MHz, effectively decreasing the negative feedback effect (by ~30% ?) and thus equalizing the difference between gain at lower and higher frequencies, which is usually the biggest problem of amplifiers with this mosfet.??
?
I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
Thank you if you can point me to some instructive source of information.
?
73 Jindra OK4RM??
 

Dave, regarding the (RF) bypass capacitors... I suppose we don't want an electrolytic capacitor in RF path, do we???
I should have mentioned which particular capacitors I had in mind:
In this document, there is schematic diagram at page 9, I had in mind capacitors C205, C206 (RF input) and C209, C210 (feedback).
As for the first two, I will certainly experiment what happens if I put different capacitors there. According to datasheets the 100 nF caps usually perform as real capacitors up to about 10 MHz, at higher frequencies the impedance rises again, suggesting there might be an inductive component in the complex impedance.?
There are methods like using two different capacitors in parallel but in such case they could theoretically form a parallel resonance circuit. Many years ago I read some account mentioning exactly this effect when combining capacitors of differing values and perhaps also different materials.
As for the feedback capacitors, if they exhibit some inductive component at frequencies above 10 MHz, this could be useful for further improving gain equalization.
?
In general, I believe that the actual components used in the 10W linear PA are result of Hans' lab experiments. But I was curious if what I assume is correct.
?
Thank you,?
Jindra
OK4RM
?
?
On Tue, Feb 18, 2025 at 05:28 PM, Dave VE3GSO wrote:

You pose a number of different questions regarding negative feedback and then bypassing considerations.
?
First bypassing: ?it is very common and good engineering practice to put an electrolytic capacitor, such as a 10uF 25V, in parallel with a 0.1uF (100nF) 50V ceramic, such as a 1206, and a 1nF (1000pF) 50V ceramic all in parallel so that each capacitor bypasses in the frequency range where they have low impedance. The smallest capacitor is positioned closest to the RF choke leading to the active device.
?
Second feedback: ?The additional small inductance can indeed reduce negative feedback and allow more gain at the higher frequencies where the device is dropping off. ?But this is as much an art as engineering, and my suggestion would be to try without at first, to see how your circuit performs. If it is needed then try it.

Dave

On Feb 18, 2025, at 11:12, ok4rm via groups.io <radio.miskovice@...> wrote:

Hello,
first of all, I apologize because I am not asking for advice about a QRP Labs product but rather an advice for my own homebrewing experiments. Recently I am into PA with IRF510 mosfets.
?
I looked at several linear amplifiers using IRF510 and most of them use a negative feedback network, usually a resistor between 470 and 1000 ohms connected between drain and gate, DC isolated by a large enough capacitor (1 nF, 10 nF, even 100 nF). QRP Labs is different because there is an additional inductance in series (1 uH).
?
I tried to estimate the effect of this inductance and it seems that in combination with the resistor used (220 ohms) it should produce a phase shift which is negligible on lower frequencies but rises to approximately 55 degrees at 28 MHz, effectively decreasing the negative feedback effect (by ~30% ?) and thus equalizing the difference between gain at lower and higher frequencies, which is usually the biggest problem of amplifiers with this mosfet.??
?
I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
Thank you if you can point me to some instructive source of information.
?
73 Jindra OK4RM??
 

Sadly there are almost no pure capacitors (or inductors or resistors) to be had
?
All normal components exhibit oddities (eg frequency or temperature or lead length related)?
?
Despite that the right component choice(s) and layout will make the device using them work over a good frequency range
?
i will just say that enraged IRF510s explode nastily so wear ballistic grade glasses when tinkering with feedback components or generally ‘close up’
 

Here's a huge trove of very readable information about building RF amps of all kinds;
?
Manfred has been posting this stuff since the 90's, his other material on electronics is well worth browsing.
The turtorials toward the end of that page are especially good.
?
Section 2.8 of EMRFD mentions how two different bypass caps in parallel can sometimes be worse than multiple caps of the same type.
The reason is that the inductance of the larger cap tied to the capacitance of the smaller cap can create a troublesome resonance.
This is the only place I've seen this mentioned, most designs seem to follow conventional wisdom and go with different sizes.
?
A nanovna is cheap, and very useful for evaluating how a cap (or multiple caps in parallel) will behave across various frequencies.
?
Electrolytics are designed for low frequencies, I doubt any of them would be much good at RF.
?
Ceramic caps are now commonly available for 20uF at 25v, so all but the largest capacitances can be handled without resorting to electrolytics.
However large ceramic caps can have unexpected behaviors, capacitance can vary by a factor of two with voltage.
Avoid the bleading edge combination of high capacitance and high voltage currently available in a given form factor.
?
Jerry, KE7ER
?
 

Of course you won't be using an electrolytic in the RF path.? That's why I prefaced my comments with "Bypass", where they do serve a purpose.

Yes, electrolytics can look like small value inductors at some high frequency, and yes they can form a low Q parallel circuit with one of the other low value bypass capacitors. The designer or builder needs to consider parasitic resonances in the design and fabrication of any amplifier.

A long time ago I was working on one of the point to point wired VHF tube amplifiers (burned out screen resistor) and noticed that the disk capacitors used for bypass had long leads, 5 to 6mm long, which I thought would negate their effectiveness.? I was told by someone much wiser than I that if the capacitors were mounted with short leads their bypass function would cease to be.? The leads and capacitor were purposefully series resonant at or near the frequency of operation. As you know a series resonant circuit looks like a short for RF, and a very effective bypass. There is always a way to make a component work!

Dave

On Thu, Feb 20, 2025 at 3:08?AM ok4rm via <radio.miskovice=[email protected]> wrote:
Dave, regarding the (RF) bypass capacitors... I suppose we don't want an electrolytic capacitor in RF path, do we???
I should have mentioned which particular capacitors I had in mind:
In this document, there is schematic diagram at page 9, I had in mind capacitors C205, C206 (RF input) and C209, C210 (feedback).
As for the first two, I will certainly experiment what happens if I put different capacitors there. According to datasheets the 100 nF caps usually perform as real capacitors up to about 10 MHz, at higher frequencies the impedance rises again, suggesting there might be an inductive component in the complex impedance.?
There are methods like using two different capacitors in parallel but in such case they could theoretically form a parallel resonance circuit. Many years ago I read some account mentioning exactly this effect when combining capacitors of differing values and perhaps also different materials.
As for the feedback capacitors, if they exhibit some inductive component at frequencies above 10 MHz, this could be useful for further improving gain equalization.
?
In general, I believe that the actual components used in the 10W linear PA are result of Hans' lab experiments. But I was curious if what I assume is correct.
?
Thank you,?
Jindra
OK4RM
?
?
On Tue, Feb 18, 2025 at 05:28 PM, Dave VE3GSO wrote:
You pose a number of different questions regarding negative feedback and then bypassing considerations.
?
First bypassing: ?it is very common and good engineering practice to put an electrolytic capacitor, such as a 10uF 25V, in parallel with a 0.1uF (100nF) 50V ceramic, such as a 1206, and a 1nF (1000pF) 50V ceramic all in parallel so that each capacitor bypasses in the frequency range where they have low impedance. The smallest capacitor is positioned closest to the RF choke leading to the active device.
?
Second feedback: ?The additional small inductance can indeed reduce negative feedback and allow more gain at the higher frequencies where the device is dropping off.? But this is as much an art as engineering, and my suggestion would be to try without at first, to see how your circuit performs. If it is needed then try it.

Dave

On Feb 18, 2025, at 11:12, ok4rm via <radio.miskovice=[email protected]> wrote:

Hello,
first of all, I apologize because I am not asking for advice about a QRP Labs product but rather an advice for my own homebrewing experiments. Recently I am into PA with IRF510 mosfets.
?
I looked at several linear amplifiers using IRF510 and most of them use a negative feedback network, usually a resistor between 470 and 1000 ohms connected between drain and gate, DC isolated by a large enough capacitor (1 nF, 10 nF, even 100 nF). QRP Labs is different because there is an additional inductance in series (1 uH).
?
I tried to estimate the effect of this inductance and it seems that in combination with the resistor used (220 ohms) it should produce a phase shift which is negligible on lower frequencies but rises to approximately 55 degrees at 28 MHz, effectively decreasing the negative feedback effect (by ~30% ?) and thus equalizing the difference between gain at lower and higher frequencies, which is usually the biggest problem of amplifiers with this mosfet.??
?
I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
Thank you if you can point me to some instructive source of information.
?
73 Jindra OK4RM??
 

I've never worried about the "insulation resistance" of a cap.?
In the case you give, the "insulation resistance" (not isolation) is 500 Ohms-Farads,
where the parallel leakage resistance across the cap in Ohms is multiplied by the capacity of that cap in Farads.
This leakage resistance varies linearly with the capacity in Farads for a given dielectric formulation
regardless of the all other factors, such as how physically thin the layer of dielectric is.
I don't understand why, and find this rather puzzling.? But again, it's not something I worry about.
A 1.0 uF cap is 0.000001 Farads, so 500 Ohms-Farads? means the insulation resistance is
actually 500/0.000001= 500,000,000 Ohms, or 500 megaOhms.
Since most of our RF circuits are working with impedances well under 1000 Ohms, the effect is microscopic.
?
Of somewhat more concern than the leakage resistance of many megaOhms is the series resistance
and series inductance.? Many datasheets fail to even mention this, and most of us just
assume those factors are low enough.? If it is of concern, they are easily measured with a nanoVNA
which will show nice graphs of resistance and reactance across an extremely wide range of frequencies.
In high power RF amplifiers, you will often see a number of caps in parallel used where high currents
are expected such a DC blocking cap at the RF output.? This reduces the total series resistance
and inductance by a factor equal to the number of individual caps.??
?
I do worry about leakage due to traces of water left under parts after washing and drying a board.
Especially around resistors used to divide down the feedback voltage in a regulator such as
R110,R112 or R111,R113 of the QMX+.? The datasheets for power supply parts seem to favor
extremely high values here so they can brag about nano-Amps of quiescent current.
I tend to keep the resistor to ground in that feedback network under 10 kOhms, if it were 100k
then pressing a finger on that part can be enough to cause the output voltage to increase.
?
If you care about a stable capacitance value over temperature and voltage, choose NP0 or equivalently C0P caps.?
That's a zero, not the letter 'O', as in zero temperature coefficient.
These get very large physically at capacitances of 0.01 uF or more, especially at the voltages found in a transmitter.
?
For bypass caps, DC blocking caps, and feedback caps such as found in an RF amp, we don't really care
if the capacitance varies.? These amps are often designed to work across a number of ham bands,
the impedance of a perfectly stable cap will vary drastically as you change the operating frequency from 3 MHz to 30 MHz.
X5R and X7R capacitors are well suited for this, the capacitance changes a little bit with temperature,
but can change by up to a factor of 2 with voltage in extreme cases where you insist on the maximum
capacitance and voltage rating in the smallest form factor possible.??
X5R capacitors can stuff more capacitance and a higher voltage rating into a given form factor than X7R capacitors.
X7R capacitors are rated for higher temperatures than X5R, that might be a consideration near the hot transistors of an RF amplifier.??
?
Jerry, KE7ER
?
?
On Thu, Feb 20, 2025 at 08:41 AM, Dave VE3GSO wrote:

I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
 


The following paper published at Designcon East back in 2005 by Steve Weir presents some interesting information on the effects caused by the interaction of multiple decoupling capacitors in power distribution networks.

?

Rsy Anderson WB6TPU

?
 

Jerry, thank you for explanation. Being real "amateur" in RF I assumed this was a real resistance :) I can learn something new every day :)
?
73 Jindra
?
On Thu, Feb 20, 2025 at 07:17 PM, Jerry Gaffke wrote:

I've never worried about the "insulation resistance" of a cap.?
In the case you give, the "insulation resistance" (not isolation) is 500 Ohms-Farads,
where the parallel leakage resistance across the cap in Ohms is multiplied by the capacity of that cap in Farads.
This leakage resistance varies linearly with the capacity in Farads for a given dielectric formulation
regardless of the all other factors, such as how physically thin the layer of dielectric is.
I don't understand why, and find this rather puzzling.? But again, it's not something I worry about.
A 1.0 uF cap is 0.000001 Farads, so 500 Ohms-Farads? means the insulation resistance is
actually 500/0.000001= 500,000,000 Ohms, or 500 megaOhms.
Since most of our RF circuits are working with impedances well under 1000 Ohms, the effect is microscopic.
?
Of somewhat more concern than the leakage resistance of many megaOhms is the series resistance
and series inductance.? Many datasheets fail to even mention this, and most of us just
assume those factors are low enough.? If it is of concern, they are easily measured with a nanoVNA
which will show nice graphs of resistance and reactance across an extremely wide range of frequencies.
In high power RF amplifiers, you will often see a number of caps in parallel used where high currents
are expected such a DC blocking cap at the RF output.? This reduces the total series resistance
and inductance by a factor equal to the number of individual caps.??
?
I do worry about leakage due to traces of water left under parts after washing and drying a board.
Especially around resistors used to divide down the feedback voltage in a regulator such as
R110,R112 or R111,R113 of the QMX+.? The datasheets for power supply parts seem to favor
extremely high values here so they can brag about nano-Amps of quiescent current.
I tend to keep the resistor to ground in that feedback network under 10 kOhms, if it were 100k
then pressing a finger on that part can be enough to cause the output voltage to increase.
?
If you care about a stable capacitance value over temperature and voltage, choose NP0 or equivalently C0P caps.?
That's a zero, not the letter 'O', as in zero temperature coefficient.
These get very large physically at capacitances of 0.01 uF or more, especially at the voltages found in a transmitter.
?
For bypass caps, DC blocking caps, and feedback caps such as found in an RF amp, we don't really care
if the capacitance varies.? These amps are often designed to work across a number of ham bands,
the impedance of a perfectly stable cap will vary drastically as you change the operating frequency from 3 MHz to 30 MHz.
X5R and X7R capacitors are well suited for this, the capacitance changes a little bit with temperature,
but can change by up to a factor of 2 with voltage in extreme cases where you insist on the maximum
capacitance and voltage rating in the smallest form factor possible.??
X5R capacitors can stuff more capacitance and a higher voltage rating into a given form factor than X7R capacitors.
X7R capacitors are rated for higher temperatures than X5R, that might be a consideration near the hot transistors of an RF amplifier.??
?
Jerry, KE7ER
?
?
On Thu, Feb 20, 2025 at 08:41 AM, Dave VE3GSO wrote:
I have one more question regarding the 0.1 uF and 1 uF capacitors used in the RF path. I was interested if these are X7R material, so I tried to find a datasheet () where I could see some impedance plots. It seems that X7R capacitors can have isolation resistance as low as 500 ohms at temperatures around 60°C. What is also interesting is that higher capacitance values, such as 0.1 uF, exhibit increasing impedance at ferquencies above 10 MHz (X7R, 100 nF, in the datasheet mentioned above).
?
My question(s):
- is it "safe" to use "big" X7R capacitors as (de)coupling? In QRP Labs PA's there are some 100 nF and 1 uF capacitors where I suspect their impedance might increase quite a lot above 10 MHz.
- or should I look for some other type of material? Maybe using several smaller 1 to 10 nF SMD capacitors in the input path instead of one 0.1 uF?
?
?
?
?

1 - 9 of 9