Thank you OS! Though in my eyes there are many inperfections. I had to switch from leaded to leadfree solder halfway resulting in many nonperfect solder joints.
For now in using through hole semi's because I expect to make mistakes and probably burn some transistors. Through hole ones I can replace myself. I cant assemble SMD at home. Also there would be no way to test and verify the circuit before ordering the smd assembled pcb's or would there?
I would love to design for SMD but i feel im not skilled enough to jump from spice simulation to SMD pcb yet. When the circuit and all values are finalized maybe ill do a smd version.
However, all components for the benchsupply test are soldered and no surprises appeared when firing up. If anything DC drift is much smaller with offset staying below 3mV.
My pleasure. I also made a spice version without pre-vas emitter followers like your Symasui but with cascoded vas. Yields about the sams linearity.
Yes though I wish I could have someone else do it more neatly and in less time😂
The board is now running from 50V supplies. After some mono listening tests I can happily conclude that the onboard PSU-section does not induce audible hum or buzz. Dead-silent which is to be expected with the savage PSRR of this topology.
Also hooked it up to arta.
Here is 13V7 RMS into 4 ohms:
Its getting rather tedious to discern the source distortion from DUT distortion and I can't really tell what the harmonics are like. Good results anyhow. PCB certainly performs better.
Also hooked it up to arta.
Here is 13V7 RMS into 4 ohms:
Its getting rather tedious to discern the source distortion from DUT distortion and I can't really tell what the harmonics are like. Good results anyhow. PCB certainly performs better.
H
HAYK
I find dissappointing the THD+N number, missing two zeros from sim.
What measures your sound card in closed loop?
What measures your sound card in closed loop?
True. Through its the best results ive gotten so far with my not so sophisticated setup. Im using a scarlett focusrite 2i4 as source and using its mic input for measurement. It does not disregard the distortion of the source. Source distortion is rather inconsistant so its hard to tell input distortion from output distortion. Theyre practically identical at this point. I wish I could somehow substract the source distortion from the amplifiers distortion.
H
HAYK
I have more doubt on your OPS to be imperfect. The perfect matching necessary, N to N and P to P might be an issue to get the OPS alone 0.03%. I don't think all the IPS+VAS is making nothing but 20db NFB.
Possibly. The fets are matched for Vgs at 300mA. They are within 5mV of eachother. We should test this in sim and see how much distortion that causes.
Also the N in THD+N is not caused by the OPS. Sim doesnt include PSU noise.
Also the N in THD+N is not caused by the OPS. Sim doesnt include PSU noise.
Last edited:
Currently facing a dilemma. LTspice shows lower inter-modulation distortion when more compensation is used. I was planning to try the quirky two-pole transitional miller compensation (2PTC). But now im not so sure anymore.
2PTC yields much lower higher harmonics but regular 2 pole yields 4 db less inter-modulation distortion. Reading from D Self and B Cordell indicates IMD is more important.
As simulated with current component values:
2 pole (as on pcb right now):
2PTC:
What do you think is the best approach. Playing it safe with compensation and having less inter-modulation. Or trying the 2PTC with a bit more inter-modulation?
Thanks and cheers,
Ruben
2PTC yields much lower higher harmonics but regular 2 pole yields 4 db less inter-modulation distortion. Reading from D Self and B Cordell indicates IMD is more important.
As simulated with current component values:
2 pole (as on pcb right now):
2PTC:
What do you think is the best approach. Playing it safe with compensation and having less inter-modulation. Or trying the 2PTC with a bit more inter-modulation?
Thanks and cheers,
Ruben
Hi All,
I need help verifying performance.
The compensation network remains untouched so far. I did increase the gain from 11X to 18.4X to yield 20VRMS output aka 100W into 4 ohms.
I tried arta's overlay function to find out what harmonics are added.
Here is the 1kHz signal fed to the catharsis:
Next I saved this as overlay and asked arta to show overlay difference on top when measuring the 100W output:
THD shows an increase of 0.0008% and THD+N shows an increase of 0.0002% input to output. Harmonics appear to have roughly the same relative proportions as on the input.
Meanwhile on top I can hardly make out the harmonics from the noise floor. Is all that 0.0008% THD hidden in the noise floor?
Can I conclude that indeed im mostly looking at the distortion fed to the input of the amplifier?
I basically need advice on how to verify the performance of this amp.
Much Cheers,
Ruben
I need help verifying performance.
The compensation network remains untouched so far. I did increase the gain from 11X to 18.4X to yield 20VRMS output aka 100W into 4 ohms.
I tried arta's overlay function to find out what harmonics are added.
Here is the 1kHz signal fed to the catharsis:
Next I saved this as overlay and asked arta to show overlay difference on top when measuring the 100W output:
THD shows an increase of 0.0008% and THD+N shows an increase of 0.0002% input to output. Harmonics appear to have roughly the same relative proportions as on the input.
Meanwhile on top I can hardly make out the harmonics from the noise floor. Is all that 0.0008% THD hidden in the noise floor?
Can I conclude that indeed im mostly looking at the distortion fed to the input of the amplifier?
I basically need advice on how to verify the performance of this amp.
Much Cheers,
Ruben
Last edited:
To see how changing the source affects THD readings I used my LG G7's quad dac as input signal. It's THD actually reads a bit lower but only gets me upto 19VRMS so 90W.
LG G7 output as fed to catharsis:
Catharsis output at 19VRMS into 4 ohms:
0.0005% THD increase at 1kHz 90W. 0.0008% into 100W could be right. The 5k number should be very optimistic since I can only measure 3 harmonics.
LG G7 output as fed to catharsis:
Catharsis output at 19VRMS into 4 ohms:
0.0005% THD increase at 1kHz 90W. 0.0008% into 100W could be right. The 5k number should be very optimistic since I can only measure 3 harmonics.
The 2-Pole Transitional Compensation is now implemented on the PCB:
While on breadboard this gave me stability issues, the PCB with the much lower impedance PSU remains well behaved driving 4 ohms.
Tomorrow Ill try to see if there's any difference in distortion.
Much Cheers,
Ruben
While on breadboard this gave me stability issues, the PCB with the much lower impedance PSU remains well behaved driving 4 ohms.
Tomorrow Ill try to see if there's any difference in distortion.
Much Cheers,
Ruben
Last edited:
During testing I got the occasional blowout when firing up the amp.
Could be many reasons and stupidity but I fear random burst oscillation at startup as this did not happen with traditional compensation. When it happened it happened too fast to pick up any cues on DVM or oscilloscope. There was the sound of a short and she smell of silicone heaven. Perhaps the DC input went floating or the output got shorted. I wish I knew for sure.
Thankfully no components other than fuses and sense resistors were harmed.
Upon closer inspection 2TPC did give a little wiggle at clipping possibly confirming compromised stability:
2PC clips cleaner:
But there is a slew-rate penalty in going back to 2PC:
All of the above with 4 ohm load.
I'm now measuring THD from the -input terminal after the signal is attenuated by the feedback voltage divider. If my understanding is correct distortion should be equal before and after the feedback resistors. (Assuming resistor distortion through temperature coefficient is negligible) The input impedance of the probe is to high that attenuation shouldn't be affected much. This means no more external voltage divider, no more external ground wiring and no more readjusting the ADC pre-amp when comparing input against output. All of which seemed to affect measurements more than the amp itself. Now im getting much more consistent results. But are they correct? Is it fair to jugde the amps performance by looking at the attenuated signal at the -input? It is after all the signal that the amplifier is seeing.
Now im not seeing any difference between input and output at all. Not with 2PTC nor with 2PC. If can't confirm that 2PTC is indeed better im not risking the possibly maybe random instability. Though I'm not sure oscillation burnt the fuses and fancy resistors I don't just wanna have to worry about it. So I desoldered some and now am using 82p/12k 2PC again.
The following Arta shots are with plentiful 2PC. Like mentioned above the output is measured from the -input terminal.
Scarlett 2i4 as source is driving the catharsis to 20Vrms into 4 ohms.
At the +input:
Measured at -Input:
Here my LG G7 phone as source driving the amplifier to 19Vrms into 4 ohm.
At the +input:
At -Input:
Perhaps I should go back and forth between 2PC and 2PTC a bit more or build one each. As of now I have enough components and patience to build one more channel. Perhaps an Audio Precision could help me decide.
Much cheers,
Ruben
Could be many reasons and stupidity but I fear random burst oscillation at startup as this did not happen with traditional compensation. When it happened it happened too fast to pick up any cues on DVM or oscilloscope. There was the sound of a short and she smell of silicone heaven. Perhaps the DC input went floating or the output got shorted. I wish I knew for sure.
Thankfully no components other than fuses and sense resistors were harmed.
Upon closer inspection 2TPC did give a little wiggle at clipping possibly confirming compromised stability:
2PC clips cleaner:
But there is a slew-rate penalty in going back to 2PC:
All of the above with 4 ohm load.
I'm now measuring THD from the -input terminal after the signal is attenuated by the feedback voltage divider. If my understanding is correct distortion should be equal before and after the feedback resistors. (Assuming resistor distortion through temperature coefficient is negligible) The input impedance of the probe is to high that attenuation shouldn't be affected much. This means no more external voltage divider, no more external ground wiring and no more readjusting the ADC pre-amp when comparing input against output. All of which seemed to affect measurements more than the amp itself. Now im getting much more consistent results. But are they correct? Is it fair to jugde the amps performance by looking at the attenuated signal at the -input? It is after all the signal that the amplifier is seeing.
Now im not seeing any difference between input and output at all. Not with 2PTC nor with 2PC. If can't confirm that 2PTC is indeed better im not risking the possibly maybe random instability. Though I'm not sure oscillation burnt the fuses and fancy resistors I don't just wanna have to worry about it. So I desoldered some and now am using 82p/12k 2PC again.
The following Arta shots are with plentiful 2PC. Like mentioned above the output is measured from the -input terminal.
Scarlett 2i4 as source is driving the catharsis to 20Vrms into 4 ohms.
At the +input:
Measured at -Input:
Here my LG G7 phone as source driving the amplifier to 19Vrms into 4 ohm.
At the +input:
At -Input:
Perhaps I should go back and forth between 2PC and 2PTC a bit more or build one each. As of now I have enough components and patience to build one more channel. Perhaps an Audio Precision could help me decide.
Much cheers,
Ruben
Maybe bias surge?
My new frontend under test causes an intermittent "Z" noise during the first seconds, both from speakers and mosfets.
Could the mechanical stress disintegrate the chips?
Those sense resistors are 0.1 ohm 3W. (Ohmite 13FR100E) Could be that they just got too hot and failed after cooling down. If one fails the other side will over-current too. Good news is the mosfets survived.