gbyleveldt
AVForums Grandmaster
OK, so I've been tinkering with an amp of my own design for a while. Actually, it's incorrect saying my "own" design as it is pretty rare coming up with something unique these days. Let's rather say my own "interpretation" of various bits to be found on the net. See schematic attached.
The input stage is a basic vanilla LTP driving a bog standard, beta enhanced VAS. This drives a Diamond buffer with Thermaltrak transistors.
A few things are worth a mention:
1. While the LTP stage is pretty standard, this does have truly impressive capability. My personal preference is for a Mirror topology, but I've been battling my backside off now for a while to get the performance to the same level as the humble LTP. This is not a reflection on the Mirror Topology, just a refelction on my stupidity
I'm going to toy with using JFets at the input stage at a later time.
2. The VAS uses a small signal transistor, not the medium power unit usually found in this location. I can get away with it becuase of the diamond buffer, the benefit being that small signal, low Co, high speed transistors are very common. I'm toying with the idea of a differential VAS as well.
3. The Diamond buffer is just such a cool idea!! In simulation, this config has such impressive properties: Very High Bandwidth and distortion figures below 0.1% WITHOUT FEEDBACK
4. The current rage for output transistors is Thermaltraks and their use here is no exception. The Thermaltrak diodes form part of the Diamond buffer and resistor R4 & R8 set the bias of the system.
All the above, I believe, comes together to fulfill my views on what makes a good audio stage:
a) Make the unit have as much bandwidth as possible. While I have to meet the person that can hear above 20kHz, I believe it's important for the amp to go muuuuch higher. The reason is simple... Phase shift at the upper octaves ruins the stereo image, this I proved with my DAC.
b) Keep the higher order odd harmonics as low as possible, this in my opinion is more important than a low average THD.
Now, I've not built this yet. I did however spend quite a bit of time on it, tweaking it in the simulator to get the results I have. It should be understood that the figures obtained below aren't really possible in the real world, but that does not mean it has no significance. These figures I have are the best I could get out of various topologies and while their absolute level is unimportant, their relative values are, especially using the same transistor models from topology to topology.
Anyway, enough babbling. Here's the simulated performance at full power into 8R (around 110W RMS):
Harmonic Frequency Fourier Normalized
Number [Hz] Component Component
1 1.000e+03 3.190e+01 1.000e+00
2 2.000e+03 6.860e-05 2.151e-06
3 3.000e+03 9.327e-06 2.924e-07
4 4.000e+03 3.725e-06 1.168e-07
5 5.000e+03 8.571e-06 2.687e-07
6 6.000e+03 2.402e-06 7.528e-08
7 7.000e+03 3.206e-06 1.005e-07
8 8.000e+03 1.910e-06 5.989e-08
9 9.000e+03 1.871e-06 5.864e-08
Total Harmonic Distortion: 0.000220%
I'm putting this out there as a discussion piece and hope to do a prototype over the chrismas break. It should be noted that this is purely a DIY project and there aren't any commercial ambitions for this
The input stage is a basic vanilla LTP driving a bog standard, beta enhanced VAS. This drives a Diamond buffer with Thermaltrak transistors.
A few things are worth a mention:
1. While the LTP stage is pretty standard, this does have truly impressive capability. My personal preference is for a Mirror topology, but I've been battling my backside off now for a while to get the performance to the same level as the humble LTP. This is not a reflection on the Mirror Topology, just a refelction on my stupidity
I'm going to toy with using JFets at the input stage at a later time.2. The VAS uses a small signal transistor, not the medium power unit usually found in this location. I can get away with it becuase of the diamond buffer, the benefit being that small signal, low Co, high speed transistors are very common. I'm toying with the idea of a differential VAS as well.
3. The Diamond buffer is just such a cool idea!! In simulation, this config has such impressive properties: Very High Bandwidth and distortion figures below 0.1% WITHOUT FEEDBACK
4. The current rage for output transistors is Thermaltraks and their use here is no exception. The Thermaltrak diodes form part of the Diamond buffer and resistor R4 & R8 set the bias of the system.
All the above, I believe, comes together to fulfill my views on what makes a good audio stage:
a) Make the unit have as much bandwidth as possible. While I have to meet the person that can hear above 20kHz, I believe it's important for the amp to go muuuuch higher. The reason is simple... Phase shift at the upper octaves ruins the stereo image, this I proved with my DAC.
b) Keep the higher order odd harmonics as low as possible, this in my opinion is more important than a low average THD.
Now, I've not built this yet. I did however spend quite a bit of time on it, tweaking it in the simulator to get the results I have. It should be understood that the figures obtained below aren't really possible in the real world, but that does not mean it has no significance. These figures I have are the best I could get out of various topologies and while their absolute level is unimportant, their relative values are, especially using the same transistor models from topology to topology.
Anyway, enough babbling. Here's the simulated performance at full power into 8R (around 110W RMS):
Harmonic Frequency Fourier Normalized
Number [Hz] Component Component
1 1.000e+03 3.190e+01 1.000e+00
2 2.000e+03 6.860e-05 2.151e-06
3 3.000e+03 9.327e-06 2.924e-07
4 4.000e+03 3.725e-06 1.168e-07
5 5.000e+03 8.571e-06 2.687e-07
6 6.000e+03 2.402e-06 7.528e-08
7 7.000e+03 3.206e-06 1.005e-07
8 8.000e+03 1.910e-06 5.989e-08
9 9.000e+03 1.871e-06 5.864e-08
Total Harmonic Distortion: 0.000220%
I'm putting this out there as a discussion piece and hope to do a prototype over the chrismas break. It should be noted that this is purely a DIY project and there aren't any commercial ambitions for this
