Hi folks,
i wanted to see how the technical quite different Creek amplifiers Evolution 2 and Evolutuion 50A perform in comparison.
So i simluated the circuits. The attempt is not perfect, as i don't have models for the correct output transistors.
I took similar ones, so the tendency of the simulation should be valid...
The Simulation:
This is how they perform in THD:
Evo 2:
100W 4R @ 1KHz: 0.038%
25W 4R @ 1KHz: 0.013%
2.5W 4R @ 1KHz: 0.0024%
25W 4R @ 10KHz: 0.019%
2.5W 4R @ 10KHz: 0.017%
Evo 50A:
77W 4R @ 1KHz: 0.0057%
25W 4R @ 1KHz: 0.0032%
2.5W 4R @ 1KHz: 0.0029%
25W 4R @10KHz: 0.018%
2.5W 4R @10KHz: 0.027%
The Evo 2 shows distortions that are rising with higher levels. Even at 25W/4R the Evo 2 shows more k2 and k3 harmonics.
At 2.5W both perform quite similar, the 50A has just less k2 and a little more 100Hz hum:
At 10KHz the result is similar.
The circuit of the 50A is technically better, its much more energy efficient and keeps the the distortions low even with higher levels.
The 50A has a somewhat weak voltage supply filtering and can easily be improved.
Change 05/16:
Fixed wrong input signal scaling (peak value for signal), and used improved power supply model.
i wanted to see how the technical quite different Creek amplifiers Evolution 2 and Evolutuion 50A perform in comparison.
So i simluated the circuits. The attempt is not perfect, as i don't have models for the correct output transistors.
I took similar ones, so the tendency of the simulation should be valid...
The Simulation:
This is how they perform in THD:
Evo 2:
100W 4R @ 1KHz: 0.038%
25W 4R @ 1KHz: 0.013%
2.5W 4R @ 1KHz: 0.0024%
25W 4R @ 10KHz: 0.019%
2.5W 4R @ 10KHz: 0.017%
Evo 50A:
77W 4R @ 1KHz: 0.0057%
25W 4R @ 1KHz: 0.0032%
2.5W 4R @ 1KHz: 0.0029%
25W 4R @10KHz: 0.018%
2.5W 4R @10KHz: 0.027%
The Evo 2 shows distortions that are rising with higher levels. Even at 25W/4R the Evo 2 shows more k2 and k3 harmonics.
At 2.5W both perform quite similar, the 50A has just less k2 and a little more 100Hz hum:
At 10KHz the result is similar.
The circuit of the 50A is technically better, its much more energy efficient and keeps the the distortions low even with higher levels.
The 50A has a somewhat weak voltage supply filtering and can easily be improved.
Change 05/16:
Fixed wrong input signal scaling (peak value for signal), and used improved power supply model.
Last edited:
Dear Reina, Please post your LT spice files. It will help others to check without having to do lot of homework !
EVO 50A seems the most straight forward. Or whatever model without the dual differential.
Since it has all the typical known issues
For driving 4 ohm loads distortion would go slightly down with 2 pairs of output devices.
And the amplifier would have much better drive/gain/voltage swing with low impedance loads.
Far as slightly lower distortion for the amplifier design, the second gain stage / vas
Could be improved adding the typical " beta enhancement " type emitter follower driver.
KSC2690/KSA1220 are good choices to get the model to perform well.
unfortunately the PNP is hard to acquire but the NPN is readily available.
likely could use Toshiba TTA/TTC004.Q
1 more transistor could be added so the differential and 2nd gain stage both have dedicated
feedback type current sources, reference to ground with decoupling capacitor for least noise
Of course linear power supply is straight forward in real life with typical 10,000u to 20,000u / 10m or 20m capacitance
Since it has all the typical known issues
For driving 4 ohm loads distortion would go slightly down with 2 pairs of output devices.
And the amplifier would have much better drive/gain/voltage swing with low impedance loads.
Far as slightly lower distortion for the amplifier design, the second gain stage / vas
Could be improved adding the typical " beta enhancement " type emitter follower driver.
KSC2690/KSA1220 are good choices to get the model to perform well.
unfortunately the PNP is hard to acquire but the NPN is readily available.
likely could use Toshiba TTA/TTC004.Q
1 more transistor could be added so the differential and 2nd gain stage both have dedicated
feedback type current sources, reference to ground with decoupling capacitor for least noise
Of course linear power supply is straight forward in real life with typical 10,000u to 20,000u / 10m or 20m capacitance
Last edited:
These amplifiers are slim, minimalistic amplifiers, with a careful use of materials, optimized for audiophile requirements.The biggest weakness lies in their design for 8-ohm loads. 4-ohm speakers overload the transformer and overheat the output transistors. 🙁 The Evo 50A will then shut down quite quickly.
The 50A's good measured values result from the linearization of the VAS stage with the 100pF MKP film local feedback capacitor. My improvement is based on removing the miller capacitance of the VAS stage and reducing the crossover distortion of the Class A/B output transistors by including the low-frequency output signal in the local VAS feedback.
The 50A's good measured values result from the linearization of the VAS stage with the 100pF MKP film local feedback capacitor. My improvement is based on removing the miller capacitance of the VAS stage and reducing the crossover distortion of the Class A/B output transistors by including the low-frequency output signal in the local VAS feedback.
The mod is a buffered VAS with Transitional Miller Compensation (TMC).
You could reduce R22 further down to 100 Ohm.
No surprise. It is only one output pair
Which can technically do 70 watts. With large enough heatsink.
Regardless not shutting down or not.
Better drive and less distortion comes from more output devices.
Larger transformer fixes that.
80 watt amp wants 160 watts power x2 330VA transformer is minimum, for few dollars more 500VA
and 2 output pairs. Makes the amp wake up and not be typical off the shelf.
Then decreasing the load of the output stage on the Vas.
Or if a simple 2 stage output Darlington is only used, then Vas current needs to go up with proper transistor.
Only way to fix regardless of compensation. Is to Darlington the vas / buffer / beta enhancement
Then the typical next step is further reduction of the load on the Vas with better gain output stage.
As we know with Darlington, same old triple.
Then Vas can remain typical TO-92 transistor.
And it is Darlington output stage , so will be same old 100ma bias current per device and correct size heatsink.
Which can technically do 70 watts. With large enough heatsink.
Regardless not shutting down or not.
Better drive and less distortion comes from more output devices.
Larger transformer fixes that.
80 watt amp wants 160 watts power x2 330VA transformer is minimum, for few dollars more 500VA
and 2 output pairs. Makes the amp wake up and not be typical off the shelf.
Then decreasing the load of the output stage on the Vas.
Or if a simple 2 stage output Darlington is only used, then Vas current needs to go up with proper transistor.
Only way to fix regardless of compensation. Is to Darlington the vas / buffer / beta enhancement
Exactly
Then the typical next step is further reduction of the load on the Vas with better gain output stage.
As we know with Darlington, same old triple.
Then Vas can remain typical TO-92 transistor.
And it is Darlington output stage , so will be same old 100ma bias current per device and correct size heatsink.
As noted the output stage still loads the Vas the same, and the Vas still loads Differential the same.
improving is reducing load on the differential, most improvement is reducing load on the Vas.
Easily done with 1 transistor for large change in total THD, further improved with just 2x output pairs.
Thanks for your remarks. I don't plan on building a different/new amplifier. I just look for easy mods. The simple circuitry of the Evo 50A also offers the advantage of shutting off working currents in the event of an overload, thus preserving the output relay.
@jxdking said
Why do you think this resistor value should be decreased?
I was looking for the largest stable value, as i don't want to load the VAS stage more than necessary or reduce the slew rate.
@jxdking said
Why do you think this resistor value should be decreased?
I was looking for the largest stable value, as i don't want to load the VAS stage more than necessary or reduce the slew rate.
Last edited:
The corner frequency of 1n and 1K is about 160KHz. It is too low for me. 1.6MHz is about right value when I implement TMC.
The resistor here is bootstrapped by the output. I don’t think it will have noticeable effect on the slew rate.
Hi @jxdking ,
no, simulation shows that the R22 value can be increased, R22 may be removed also. But with going below 1k the circuit is not stable with capacitve loads anymore.
@compensation:
The 50A is designed to work without an inductor at the output. So i need to compensate in a way, that i have enough phase margin for all capacitive loads.
From simulation experiements with a square wave input signal i know that most criticical are Cload values like 0,5µF and above.
How did i get the values?
At first R22 is removed. Starting point is C2. In the orginal Amp 100pF is used here. This value determines the maximum slew rate. So i don't want to increase it too much. With 120pF i can use a C7 value as low as 680nF to get the same series capacitance. Simulation shows that a little larger value improves stability.
Now i can simulate with R22. The simulation shows, that i get better distortion values with a lower value but it needs to be at least 1K to be stable under all conditions.
The idea is that i want to keep the local VAS miller feedback compensation for high frequencies, but for lower frequencies we can use the output voltage as feedback to linearize the output stage.
In the end I have chosen a little higher value for C7 (1nF) and a lower value for R22 (1K).
no, simulation shows that the R22 value can be increased, R22 may be removed also. But with going below 1k the circuit is not stable with capacitve loads anymore.
@compensation:
The 50A is designed to work without an inductor at the output. So i need to compensate in a way, that i have enough phase margin for all capacitive loads.
From simulation experiements with a square wave input signal i know that most criticical are Cload values like 0,5µF and above.
How did i get the values?
At first R22 is removed. Starting point is C2. In the orginal Amp 100pF is used here. This value determines the maximum slew rate. So i don't want to increase it too much. With 120pF i can use a C7 value as low as 680nF to get the same series capacitance. Simulation shows that a little larger value improves stability.
Now i can simulate with R22. The simulation shows, that i get better distortion values with a lower value but it needs to be at least 1K to be stable under all conditions.
The idea is that i want to keep the local VAS miller feedback compensation for high frequencies, but for lower frequencies we can use the output voltage as feedback to linearize the output stage.
In the end I have chosen a little higher value for C7 (1nF) and a lower value for R22 (1K).
Last edited:
capacitive loads create issues at high frequency.
Resistive loading is typical solution, or increase differential Degen.
Which it needs anyways since 22 ohms is close to non existent degen.
The emitter resistors of the power devices can be increased.
From .22 ohms to .27 even .33 ohms.
Most will scream that this will ruining " damping"
When adding another pair of output transistors and even using .47 ohms completely solves the damping issue.
the model is simplified with a ideal voltage source for bias.
A actual VBE multiplier in a amplifier for Darlington would have base resistors to the drivers.
100 ohms is common but can actually be lower. Then 100n is tied across VBE
Which is a big help for high frequency issues being created by the output section.
A triple or E3 will add power supply decoupling for the power transistor drivers. otherwise mandatory.
With E2 often left out, but is still beneficial regardless.
It is good that some sort of non ideal power supply is used in the model.
Since capacitance and power rail decoupling actually does something.
Is also easy to use ideal voltage source and add internal resistance.
1 to 2 ohms is enough to make the model perform more realistic.
And any capacitive decoupling actually does something.
I like that your model does show 50Hz AC mains noise at 100Hz
Be interesting to know how much basic rail decoupling capacitors lower it.
Resistive loading is typical solution, or increase differential Degen.
Which it needs anyways since 22 ohms is close to non existent degen.
The emitter resistors of the power devices can be increased.
From .22 ohms to .27 even .33 ohms.
Most will scream that this will ruining " damping"
When adding another pair of output transistors and even using .47 ohms completely solves the damping issue.
the model is simplified with a ideal voltage source for bias.
A actual VBE multiplier in a amplifier for Darlington would have base resistors to the drivers.
100 ohms is common but can actually be lower. Then 100n is tied across VBE
Which is a big help for high frequency issues being created by the output section.
A triple or E3 will add power supply decoupling for the power transistor drivers. otherwise mandatory.
With E2 often left out, but is still beneficial regardless.
It is good that some sort of non ideal power supply is used in the model.
Since capacitance and power rail decoupling actually does something.
Is also easy to use ideal voltage source and add internal resistance.
1 to 2 ohms is enough to make the model perform more realistic.
And any capacitive decoupling actually does something.
I like that your model does show 50Hz AC mains noise at 100Hz
Be interesting to know how much basic rail decoupling capacitors lower it.