amplifier oscilates in LTspice

Also, my experience is that 2N3055’s, even modern ones can demand more than 200 mA of base current. 2N2222/MPS2907 solve that little problem nicely.
Agree
Just to state for the OP's situation I was under the impression he was using Tesla KD501/03 which have higher Hfe and Ft 2 MHz
which is why I choose the modern 2N3055 model. The Modspec models are much better for Hfe and thermal properties.
If I wanted to model " older" 2N3055 with lower gain and lower Ft with better thermal modeling. Would also use Modspec model
for MJE3055T. Gain is poor so yes would see more realistic Current needed to drive the Base. Rail sag at 4 ohms of course be realistic too.
Doubt 15 watts , Be closer to 12 watts. Of course in a " Model" I would change Ideal Voltage sources to have internal resistance.
So the would sag more like a linear power supply.

By the way great idea with the 2N2222, something I thought of as well. Or more specific going to metal can T0-18 was my thought, Not plastic 92 package.
If that is what you were hinting at. Aside that even plastic package still rated at higher currents.
1730678076158.png

Just to state for reference if anybody runs into this thread in the future. The PN or MPS packages are plastic T0-92
If wanting matching pair For metal can T0-18
2N2222A and 2N2907AL
Most manufactures here in the states like OnSemi have likely made those obsolete.
Seems like the only one making smaller metal cans = Central Semiconductor
 
what if i switched to an complementary transistor pair? i also have 4 "matched" trtansistors from an power supply
Yes complementary is possible and almost any topology can be used.
Just depends on topology and bias voltage to adjust the bias diodes needed.
What are the part numbers or transistor types. Are they BJT or Mosfet.
That changes a lot if we are going to bias Mosfets.
You have a working model now for the tesla BJTs
If im correct that was your stock for T0-3

As you can tell from my picture, I like metal can transistors and dont have issues with older Quasi.
 
You can buy them in the metal can or the TO-92. Your choice. I doubt Central Semi parts will blow up at 36 or so volts. Zetex (now part of Diodes Inc) makes plastic TO-92 compatibles that will take a whallop. Even 2 amp collector currents. Cost a buck apiece in onsety twoseys though. Worth it if you want current capacity.
 
Yes complementary is possible and almost any topology can be used.
Just depends on topology and bias voltage to adjust the bias diodes needed.
What are the part numbers or transistor types. Are they BJT or Mosfet.
That changes a lot if we are going to bias Mosfets.
You have a working model now for the tesla BJTs
If im correct that was your stock for T0-3

As you can tell from my picture, I like metal can transistors and dont have issues with older Quasi.
i have both teslas and 2n3055's

the power supply transistors are the mje13005
 
Agree
Just to state for the OP's situation I was under the impression he was using Tesla KD501/03 which have higher Hfe and Ft 2 MHz
which is why I choose the modern 2N3055 model. The Modspec models are much better for Hfe and thermal properties.
If I wanted to model " older" 2N3055 with lower gain and lower Ft with better thermal modeling. Would also use Modspec model
for MJE3055T. Gain is poor so yes would see more realistic Current needed to drive the Base. Rail sag at 4 ohms of course be realistic too.
Doubt 15 watts , Be closer to 12 watts. Of course in a " Model" I would change Ideal Voltage sources to have internal resistance.
So the would sag more like a linear power supply.

By the way great idea with the 2N2222, something I thought of as well. Or more specific going to metal can T0-18 was my thought, Not plastic 92 package.
If that is what you were hinting at. Aside that even plastic package still rated at higher currents.
View attachment 1376721
Just to state for reference if anybody runs into this thread in the future. The PN or MPS packages are plastic T0-92
If wanting matching pair For metal can T0-18
2N2222A and 2N2907AL
Most manufactures here in the states like OnSemi have likely made those obsolete.
Seems like the only one making smaller metal cans = Central Semiconductor
cdil makes metal 2n2222 and i can buy them from my "local" mouser for 1$ each, they may be a good idea for headphone amplifiers.
 
In this case we mean still using power transistors.
The drivers 3904/06 could be changed to metal package 2N2222/2N2907 instead of plastic.
As mentioned even in plastic 2222/2907 600ma were suggested than 100 to 200ma 3904/06
Metal can version close to 800ma

And there is Zetex or ZTX transistors in plastic that go up to 1000ma or higher in plastic.

I already blasted the model into clipping at 4 ohms.
AC current to drive 2N3055 base at these voltages and currents.
Was around 10ma for the CFP side and 25ma for the E2
plus 3 to 4 ma for DC bias

You de rate 60 to 70% for higher temperature Data sheets show cold or 25c temp.
Forget if 3904/06 is 100ma or 200ma at 25c

Power transistors were at 27c maybe ill crank them up to 80c and blast it 4 ohms
to see AC current.
 
Last edited:
Remember Munch from Law and Order: “What did you expect??? What did you reasonably expect???”

When you run drivers that hard you can also get thermal runaway, when Tj of the driver is 50-100C hotter than the output - and there is no way for your bias scheme to know that. The innards of a TO-92 will heat up/cool down almost on a cycle by cycle basis, but not quite. It can be enough to trigger serious cross conduction if you’re crossing through zero and driver Tj cool-down is lagging by 20 ms.
 
The input filter removes RF or Radio Frequency interference.
So depends if you have issues or not , and what band.

AM here in the states 540 kHz to 1800 kHz
FM 88 to 108 MHz
4G cell 40 to 100 MHz
lower band 5G cell 600 to 1 GHz

So the 1k resistor I used combined with cap to ground is simple low pass filter.
1730770609479.png

1k and 100p = 1.6 MHz
Basically kills all cell and FM

If you have AM interference it has to go lower to 540 kHz
There is only so many common values so 330p capacitor
1k and 330p = 482 kHz
That pretty much kills everything

Funny enough usually 47p to 100p removes most problems.
Some Churches here broadcast AM right in their parking lots.
So if your really close to a broadcast antenna
Usually have to use 330p

In your schematic R2 and C3 also form lowpass filter for stability.
Depends were your phase margin is.
Since your gain is high with 22k resistor you likely only need 1.2p to 3.3p
to get the amp stable. Too high of capacitor value will reduce the slew rate or could make the phase margin
go positive. Causing stability issues. Depends. Just adjust to min value needed to make the amp stable.
That can vary greatly depending on transistor types and opamp bandwidth.
One of the many points to use a model

1.2p being the lowest value with 22k puts you at 6 MHz
So if that causes issues with phase margin. Often you see amps with lower values
Like 10k to 18k so with 1.2p you can get higher frequency filters.
Such as 10k with 1.2p will get you up to 13.2 MHz if needed.

Most audio opamps shouldnt be over 15 MHz
then again when people use 100 MHz opamps and wrong values they blame the " opamp"
for stability
 
Last edited:
You have nothing to set the bias, You will be way up at 170 to 200ma without something.
Use at least a set resistor or something friendly is adjustable.
Look at post #51 R11 sets bias. I added the 10k parallel to be adjustable.
By doing so if the pot is damaged over time. bias wont fly up to the moon
since the set resistor is still there.


If your trying to PCB this the opamp does need power rail decoupling.
Also shown in post #51 this is udder basic opamp technique.
The small 100n have to be absolutely as close to the power pins as possible.
The larger 10u also close. Due to size will be slightly farther.

Also the power transistors should be decoupled on the board.
Usually 470u is good enough also 100n parallel . Also to be a close to the power transistor
pins as possible.

Real world things oscillate with out power rail decoupling.
Otherwise standard for all opamps and power amps.

The diodes thermal track, they need to be close to power transistor heatsink.
Normal stuff as well for audio amps
 
The input filter removes RF or Radio Frequency interference.
So depends if you have issues or not , and what band.

AM here in the states 540 kHz to 1800 kHz
FM 88 to 108 MHz
4G cell 40 to 100 MHz
lower band 5G cell 600 to 1 GHz

So the 1k resistor I used combined with cap to ground is simple low pass filter.
View attachment 1377242
1k and 100p = 1.6 MHz
Basically kills all cell and FM

If you have AM interference it has to go lower to 540 kHz
There is only so many common values so 330p capacitor
1k and 330p = 482 kHz
That pretty much kills everything

Funny enough usually 47p to 100p removes most problems.
Some Churches here broadcast AM right in their parking lots.
So if your really close to a broadcast antenna
Usually have to use 330p

In your schematic R2 and C3 also form lowpass filter for stability.
Depends were your phase margin is.
Since your gain is high with 22k resistor you likely only need 1.2p to 3.3p
to get the amp stable. Too high of capacitor value will reduce the slew rate or could make the phase margin
go positive. Causing stability issues. Depends. Just adjust to min value needed to make the amp stable.
That can vary greatly depending on transistor types and opamp bandwidth.
One of the many points to use a model

1.2p being the lowest value with 22k puts you at 6 MHz
So if that causes issues with phase margin. Often you see amps with lower values
Like 10k to 18k so with 1.2p you can get higher frequency filters.
Such as 10k with 1.2p will get you up to 13.2 MHz if needed.

Most audio opamps shouldnt be over 15 MHz
then again when people use 100 MHz opamps and wrong values they blame the " opamp"
for stability
the only am station where i live is at 220khz and some romanian ones that i can barely hear at night, am is dead where i live