This is a great thread.
I am waking it up again after 6 years of sleeping to see if people can continue to add value to what have previously been achieved.
Let me do a summary of various Output Stage circuits discussed so far.
The standard text book amplifier output is a Zobel (e.g. 8R + 100nF) from Vout to ground and a LR (e.g. 2uH || 8R) from Vout to speaker terminals. The main purpose of this output circuit is to provide unconditional stability to the amplifier output stage, ensuring that there is a low impedance path at RF and at the same time some impedance to prevent excessive RF current draw from capacitive load.
Regards,
Bill
I am waking it up again after 6 years of sleeping to see if people can continue to add value to what have previously been achieved.
Let me do a summary of various Output Stage circuits discussed so far.
The standard text book amplifier output is a Zobel (e.g. 8R + 100nF) from Vout to ground and a LR (e.g. 2uH || 8R) from Vout to speaker terminals. The main purpose of this output circuit is to provide unconditional stability to the amplifier output stage, ensuring that there is a low impedance path at RF and at the same time some impedance to prevent excessive RF current draw from capacitive load.
Regards,
Bill
Attachments
Last edited:
Next, This is the 2nd most popular implementation. Here the LR comes first then C to ground. According to Edmond Stuart, the C here prevents from RF ingress. It shunts RF to ground before it gets to the negative feedback of the amplifier and it does so much better than the standard implementation which does almost nothing beyond 2MHz.
Attachments
However, Eva indicated that the C could potentially cause LCR resonance because the C could interact with the speaker wire LCR and the speaker load. So Eva came up with a T circuit which does what the other circuits do while preventing from potential resonance with speaker wires and load (See post 304).
Attachments
Now, I am thinking that Eva’s circuit is very fine, except it has missed one point. The first RC must be placed very close to the amplifier’s output and the local decoupling ground and the speaker return ground, i.e. low inductance, probably on the output board, in order for the RC Zobel to be effective. So in Eva’s circuit the first L must be placed very close as well. The air core inductor can inject noise to the high impedance of the output transistor base or MOSFET gate, and the inductor can pick up noise from power supply noise. We really want to place the L close to the output speaker terminal, instead of the output board.
So I am proposing this circuit: a standard zobel right on the output board and a T circuit as Eva prescribed at the speaker terminal. The C of the T circuit can be connected directly on the nearest point on the chassis.
What do you think?
I am unable to model the correct values due to lack of skills and knowledge. If this circuit works, can somebody please help modelling it correctly and suggest the values of the components that should fit most amps?
So I am proposing this circuit: a standard zobel right on the output board and a T circuit as Eva prescribed at the speaker terminal. The C of the T circuit can be connected directly on the nearest point on the chassis.
What do you think?
I am unable to model the correct values due to lack of skills and knowledge. If this circuit works, can somebody please help modelling it correctly and suggest the values of the components that should fit most amps?
Attachments
C7 connected to speaker return terminal, not to Chassis.
If you want you can add an RF attenuating capacitor from Chassis to EVERY entry to the chassis. These are ADDITIONAL to the amplifier components.
I would use the 1nF suggested by H.Ott connected from Speaker Out to Chassis and another from Speaker Return to Chassis and another from Input RCA barrel to Chassis and another from RCA signal to Chassis. The IEC filter on the mains cable already has attenuation capacitance to Chassis.
The circuits with the R||L, that you are showing, are not an amplifier output Zobel. They are the Thiele Network (posts362&363), or modified versions of the Thiele Network (posts364&365).
The amplifier Output Zobel is the R+C from amp out to Main Audio Ground/Power Ground.
If you want you can add an RF attenuating capacitor from Chassis to EVERY entry to the chassis. These are ADDITIONAL to the amplifier components.
I would use the 1nF suggested by H.Ott connected from Speaker Out to Chassis and another from Speaker Return to Chassis and another from Input RCA barrel to Chassis and another from RCA signal to Chassis. The IEC filter on the mains cable already has attenuation capacitance to Chassis.
The circuits with the R||L, that you are showing, are not an amplifier output Zobel. They are the Thiele Network (posts362&363), or modified versions of the Thiele Network (posts364&365).
The amplifier Output Zobel is the R+C from amp out to Main Audio Ground/Power Ground.
Last edited:
All agreed!
I am hoping somebody can help to actually model the circuit to come up with LCR values that will work with common EF output and typical loudspeaker load.
Attached please find the basic circuit.
I think that my proposed network should work. Let us say we come up with a set of LCR values for the components in the network which are the minimal values for unconditional stability. Starting from those values, increaseing the values should not upset stability, but only incur LOSS. I think the proposed network provides as much stability as others provide, except that it has higher loss. If the loss does not affect frequency response within audioband, we should be able to accept the loss. Correct me if I am wrong.
I have just modelled the proposed output network in SpeakerWorkshop. I assumed the amplifier output is a perfect voltage source.
When I used two ScanSpeak 8531 6.5 inch drivers in parallel as the load, there was about 0.2dB loss at 20kHz, and about 0.1dB LOSS at 5kHz.
When I used the Dynaudio Esotar T330D tweeter as the load, there was about 0.1dB loss at 20kHz, and basically no loss at 5kHz.
If I reduced the 1uH to 0.7uH for each inductor like Eva did, the loss was reduced but still there.
I think I prefer slightly higher inductor value for better RF attenuation and perhaps better stability. I think I can accept 0.1dB loss at 20kHz.
Can somebody please model it with LTSpice to look at it from the view point of the amplifier?
When I used two ScanSpeak 8531 6.5 inch drivers in parallel as the load, there was about 0.2dB loss at 20kHz, and about 0.1dB LOSS at 5kHz.
When I used the Dynaudio Esotar T330D tweeter as the load, there was about 0.1dB loss at 20kHz, and basically no loss at 5kHz.
If I reduced the 1uH to 0.7uH for each inductor like Eva did, the loss was reduced but still there.
I think I prefer slightly higher inductor value for better RF attenuation and perhaps better stability. I think I can accept 0.1dB loss at 20kHz.
Can somebody please model it with LTSpice to look at it from the view point of the amplifier?
Last edited:
How do you calculate 8R for both Zobel and LR? It seems 10R is commonly used.
The amplifier gain can change when the output load is not set to the nominal value.
This can become an unstable/less stable condition when the output load is disconnected.
The easy way is to add an HF load that prevents the amplifier seeing an infinite load impedance at HF. A resistor plus a capacitor draws near zero power in the audio band and yet provides the necessary load at HF where the stability may become a concern.
Often the resistor can be around the value of the nominal load resistance.
But it can vary a lot. Some of the National chipamps use very low values around 1r0 to 2r7 even though their nominal load impedance is 8ohms.
One needs to look at the stability margins of each individual amplifier to determine the optimum Zobel values.
Alternatively one can guess at a value that may work and this is where you often see 10r+100nF. This seems to work for many amplifiers, but probably indicates that the Designer has not found the optimum for his amplifier.
This can become an unstable/less stable condition when the output load is disconnected.
The easy way is to add an HF load that prevents the amplifier seeing an infinite load impedance at HF. A resistor plus a capacitor draws near zero power in the audio band and yet provides the necessary load at HF where the stability may become a concern.
Often the resistor can be around the value of the nominal load resistance.
But it can vary a lot. Some of the National chipamps use very low values around 1r0 to 2r7 even though their nominal load impedance is 8ohms.
One needs to look at the stability margins of each individual amplifier to determine the optimum Zobel values.
Alternatively one can guess at a value that may work and this is where you often see 10r+100nF. This seems to work for many amplifiers, but probably indicates that the Designer has not found the optimum for his amplifier.
Usually, the value of the Zobel Network is the same as nominal impedance of the load. Capacitor is such that it have a similary reactance (1/j.C.2pi.f) (f is the higher frequency at -3dB) Inductor is 1 to 2 µH, and the associated resistor 1/4 of the nominal load, perhaps 2.2 ohms. The final values are adjusted by "cut and try" whith rectangular signal and various loads for the best stability. It is essential to connect the oscillo probe at the final stages, to see the real output signal, inductor causes some ringing which can lead to misinterpretations.
This is incorrect. The coil and its usual parallel resistor is to ensure that the load seen by the amplifier is essensially resistive at very high frequencies above the audio band. In general, most solid state amplifiers can become unstable if they must feed a capacitive load without the isolation provided by the coil at high frequencies. Certain speaker cables or speakers can be significantly capacitive and may destabilize some amplifiers that do not have a coil. The reason for this possible instability is two-fold. First, a heavy capacitive load can add an additional pole to the global feedback loop and reduce phase margin, sometimes to the point of oscillation. Secondly, even without global feedback, the output stage itself can become unstable if it has to drive a capacitive load. The most common example is an output stage that is essentially an emitter followers. Emitter followers do not like to drive capacitive loads, else they can become locally unstable. This often also applies to op amps. Virtually every properly designed preamplifier you will find has a resistor in series with the output of at least 50 ohms; this is to isolate the output stage from a capacitive load, like a long interconnect cable. This is essentially the same function as the series coil in a power amplifier. The coild is used in a power amplifier because its impedance is very low in the audio band, and thus the damping factor of the amplifier is preserved in the audio band and there is virtually no power loss in the network. inductance of the coil is usually in the range of 0.5 uH to 4.7 uH. A well-designed amplifier can be fine with 1-1.5 uH. A very well-designed amplifier can sometimes go as low as 0.5 uH. The parallel resistor is usually between 1.5 and 4.7 ohms, with 2.2 ohms being quite common.
An amplifier should be tested for stability with pure capacitive loads from 100 pf to 0.1 uf into no load, and 8 and 4 ohm loads at all power levels from 1 watt to full power. This can be a fairly stringent test. Sometimes even a 2-ohm load will be used. One wants to ensure that there will never be parasitic oscillations or oscillation bursts driving any combination of speaker cable, louspeaker and power level.
While parasitic oscillation bursts will tend to increase THD, do not assume that an acceptable THD reading will preclude parasitic oscillations with good confidence. I once had an amplifier whose THD was a generally-acceptable THD number of 0.01%, but I thought that was a little high for that particular design. Looking more carefully, I found a parasitic oscillation on the back porch of a sinewave at higher power levels. When I revised the design to rid it of the oscillation burst, THD under those conditions went down by a factor of somewhere between 2 and 5.
Cheers,
Bob
check out post #7 and #35 under
https://www.diyaudio.com/community/...uctor-on-my-class-ab-amp-output.346796/page-2
and post #14 under
https://www.diyaudio.com/community/...cell-for-solid-state-power-amplifiers.320639/
The question, where is the right position for connect the basic load (i. e. the Boucherot cell resp. Zobel network) is from my view dependent of the circuit topology from power amp (e. g. emitter-follower (EF) output or complementary feedback pair (CFP) output) and does not seem to have a universally valid answer. OTOH - neither in the well-known books regarding power amplifiers nor anywhere else can I find exact guidelines or rules in this case (e. g. in the book from D. Self "Audio Amplifier Design Handbook).
This question rises up in case of the power amplifier Horch 3.0s - schematic under
https://www.diyaudio.com/community/...d-3-0s-what-to-3-power-output-devices.379808/
This amp produce heavy oscillation around 3MHz and the Boucherot cell resp. zobel RC network comes after the serial inductor (on the speaker terminal - go to attachment 5+6 behind the schematic images).
https://www.diyaudio.com/community/...uctor-on-my-class-ab-amp-output.346796/page-2
and post #14 under
https://www.diyaudio.com/community/...cell-for-solid-state-power-amplifiers.320639/
The question, where is the right position for connect the basic load (i. e. the Boucherot cell resp. Zobel network) is from my view dependent of the circuit topology from power amp (e. g. emitter-follower (EF) output or complementary feedback pair (CFP) output) and does not seem to have a universally valid answer. OTOH - neither in the well-known books regarding power amplifiers nor anywhere else can I find exact guidelines or rules in this case (e. g. in the book from D. Self "Audio Amplifier Design Handbook).
This question rises up in case of the power amplifier Horch 3.0s - schematic under
https://www.diyaudio.com/community/...d-3-0s-what-to-3-power-output-devices.379808/
This amp produce heavy oscillation around 3MHz and the Boucherot cell resp. zobel RC network comes after the serial inductor (on the speaker terminal - go to attachment 5+6 behind the schematic images).
Last edited:
It decouples the amplifier' output from any capacitive loads that could cause instability. At high frequencies, a capacitive load looks like a short circuit, and that could damage an amplifier, while the inductor tends toward a high impedance at high frequencies, and that saves the amp from seeing what would be a dead short if the amp is connected to a capacitive load. Since those inductors are just coils with no ferrous cores in them, they don't saturate. A core is what would saturate. The resistor keeps the amp loaded at high frequencies where the inductor tends toward an open circuit. At audio frequencies, the coil is pretty much a dead short and has no significant effect on the audio.