MJL21193 said:
What happens when you try to make an adder with caps? That does a lot of harm to a signal instead of just a little bit. You want DC right?
Last night looking at the old amp revealed PSR of about 70 dB.
If you filter the power line with a simple RC network the PSR would improve by theoretically but you are loading the filter so the improvement will be less than theoretical.
Adding a pass transistor and filtering the base, the power supply remains low impedance. Okay you do get a volt drop across the transistor, but you could achieve quite high PSR this way.
The picture shows a filter 36 dB down at 100 Hz and this means the amp PSR would be 106 dB (add amp PSR of 70 dB).
If you filter the power line with a simple RC network the PSR would improve by theoretically but you are loading the filter so the improvement will be less than theoretical.
Adding a pass transistor and filtering the base, the power supply remains low impedance. Okay you do get a volt drop across the transistor, but you could achieve quite high PSR this way.
The picture shows a filter 36 dB down at 100 Hz and this means the amp PSR would be 106 dB (add amp PSR of 70 dB).
Attachments
Nico Ras said:
Hi Nico,
I employed the simple RC low pass filter to improve the mid/high range PSRR. As shown in my plot, this gives good simulated results. This also gives good real world results without too much complication. If I want to improve this further, I would split the frontend of the amp from the main supply and use a separate, regulated supply for the front end. For my tastes, my ears and my expectations, this is not necessary.
Over in my Abomination thread, I was shown how to do a simple test that demonstrates the audibility of the power supply ripple. In the test, detailed here, I could not hear anything from the speaker.
It is good to shoot for the absolute lowest, but in most cases it is not necessary. Numbers are misleading and real listening is what counts.
I have taken another look at the "bootstrap" resistor, prompted by Nico. I have found that it isn't making the difference that I first experienced. Granted, a lot has changed in the design since I first employed it, and the optimizations that have been done may have made up for the deficiency it was covering.
I have decided to delete it, rather than leave it for "old time sake" It doesn't do anything bad, but it also doesn't do anything good.
I have updated the plan:
🙂
I have decided to delete it, rather than leave it for "old time sake" It doesn't do anything bad, but it also doesn't do anything good.
I have updated the plan:
🙂
Nico Ras said:No you haven't. Bootstrap resistor is there in all its glory
Hi Nico,
Not the VAS bootstrap. The frontend resistor to the feedback loop - R28 on the old schematic. Remember, the one that you said made no difference?
MJL21193 said:
Oh Johnny Boy,
can't be, your ears haven't ripened enough to distinguish between head and tail of a wire.
jacco vermeulen said:
Ah, but I do know the female from the male end 😀
I like to keep an open mind to these things but sometimes it gets a trifle silly.
I have been playing with RMAA and decided to do some tests on the Patchwork amps that are being used in my 6 channel amp.
Tests were at about 20 watts output into 8 ohms. Signal was attenuated by a resistor divider. Ignore the stereo crosstalk figures, as the attenuator used a common ground.
The scores:
THD:
Noise:
Frequency response:
IMD:
I just posted results for the same tests on the K10A amp for comparison.
Tests were at about 20 watts output into 8 ohms. Signal was attenuated by a resistor divider. Ignore the stereo crosstalk figures, as the attenuator used a common ground.
The scores:
THD:
Noise:
Frequency response:
IMD:
I just posted results for the same tests on the K10A amp for comparison.
I have completed this project and I will be running some final sims and posting the details. There has been some interest in the design and there are a few builders out there, so I have some final construction details to post.
First, a complete and final schematic:
First, a complete and final schematic:
Final board layout:
Board is single sided 3" x 5" and all outputs, drivers and Vbe are mounted underneath.
Bill of material:
Board is single sided 3" x 5" and all outputs, drivers and Vbe are mounted underneath.
Bill of material:
My simulation of the amp puts the THD at ~0.001% at about 50 watts output. I ran some more tests with RMAA, this time getting all of the settings right (checked the "do not monitor" box in the recording properties - I didn't do this with the results above and it makes a big difference) and the results are below. Actual output from the amp is about 20Vrms, so ~50 watts into 8 ohms.
First the numbers. First 2 tests are for the soundcard itself. The second 2 are with the amp in the loop. I ran these tests several times to verify consistency.
Soundcard THD = 0.0019, therefore the amps THD = 0.0002% (0.0021 - 0.0019 = 0.0002). Very good!
Frequency response comparison between the soundcard only and the amp looped in:
Noise spectra, soundcard only compared to amp looped in:
IM distortion, again a comparison between the soundcards native IMD and with the amp looped in:
Finally, the THD spectra. Soundcard alone is the white trace, amp in the loop is the green trace. We see 3k and 6K (these are present in the same magnitude for the soundcard test) but they are below -100db:
First the numbers. First 2 tests are for the soundcard itself. The second 2 are with the amp in the loop. I ran these tests several times to verify consistency.
Soundcard THD = 0.0019, therefore the amps THD = 0.0002% (0.0021 - 0.0019 = 0.0002). Very good!
Frequency response comparison between the soundcard only and the amp looped in:
Noise spectra, soundcard only compared to amp looped in:
IM distortion, again a comparison between the soundcards native IMD and with the amp looped in:
Finally, the THD spectra. Soundcard alone is the white trace, amp in the loop is the green trace. We see 3k and 6K (these are present in the same magnitude for the soundcard test) but they are below -100db:
- Status
- Not open for further replies.