This, on the assumption the kink will move someplace recognizable, right? Tom, what kind of tests have you done highlighting crossover distortion? Does the precision amp wash away all the A/B switching or is it just suppressed below the noise floor?
I think they're still hashing out how it will work. From what I've understood, it relies upon a work in progress model (see Taguchi) and he/she? doesn't have the wherewithal to do it alone. He does want to direct it though and that is where I see the impasse. I think the concept of strictly defined audio postulates that have a good possibility of an electronic correlation is brilliant. It makes it harder for both golden ears and lab rats to lie to each other and has the potential to fix a lot of what's wrong in the audio business today. I check in on them every once in a while to see where they're at. I kind of feel bad for the OP, I think he's got a good idea but he probably underestimated the ego paranoia of his audience.
Tom,
check out the (rather long) Precision Oscillator and Notch Filter threads over in the Equipment&Tools forum for distortion measurements that exceed, and complement, the AP27xx built-in stuff -- non-adjustable, of course, but very useful for fixed frequency THD and IMD measurements at the "esoteric" levels. Notably Victor's Oscillator, together with using sync'ed averaging on the AP, can give more insight than plain THD+N vs Power or Frequency.
Those can be wrapped in a composite as well, with probably only a few minor components changes vs the 3886 as they are from the same general design family. To me, the 3886 makes the most sense as it is the strongest chip in the family.
It's not the instrument per se but the presentation of the data. You can use the AP and present averaged long FFT's that will resolve the harmonics from the noise. As I said that THD+noise plot that everyone wants on their datasheet averages all the noise in an 80kHz BW at every point obscuring the harmonic content except at very high amplitudes and contains virtually no information.
I agree. But that's what the builders want, so that's what they'll get...
I plan to camp out in the lab tomorrow night and get some measurements done. I will include an averaged FFT showing the fundamental and whatever harmonics may be visible above the noise floor. As you say, I doubt the harmonics will emerge from the noise floor until the amp is approaching the peak output power.
~Tom
If you watch the instrument as the power is swept, it's pretty obvious that the kink is from a range switch. The AP goes 'click' and the graph moves up a tad. If you used a voltage divider to divide the MOD86's output voltage by, say, 2 before entering the AP, the kink would move to a higher power. Assuming the resistive divider wouldn't degrade the SNR of the measurement, that could work to remove the kink in the plot. It's largely of academic interest, though.
I don't have any way to separate the cross-over distortion from the other distortion mechanisms present. The cross-over distortion is low enough that I can't see it on an oscilloscope (even when zooming in on the zero crossing). If cross-over distortion was an issue in this design, the THD+N wouldn't be this good. Also, as you mention, the LME49710 would perform the error correction needed to drive the cross-over distortion to zero anyway.
~Tom
I decided to camp out in the lab tonight rather than tomorrow. I have SMPTE data as well as updated measurements of the THD+N vs power and at higher output powers. I now get 35+ W into 8 Ω. I took some measurements of the harmonic spectrum to get the THD rather than the THD+N.
The THD at 1 W output power comes in at 0.0002 % (-115 dB).
A measurement of the noise floor is attached. This is measured without any kind of shielding and using a real power supply (toroid + rectifier + 2x22000 µF).
Expect more measurements to trickle in tomorrow and over the weekend.
~Tom
The THD at 1 W output power comes in at 0.0002 % (-115 dB).
A measurement of the noise floor is attached. This is measured without any kind of shielding and using a real power supply (toroid + rectifier + 2x22000 µF).
Expect more measurements to trickle in tomorrow and over the weekend.
~Tom
Attachments
P2P composite... Good luck with that. That's a howler in the making right there.
Anybody considering a point-to-point build of any precision circuit aught to read my LM3886 P2P vs PCB thread. I present plenty of data there. Conclusion: A fully optimized P2P layout performs nearly as well as a fully optimized PCB. To ensure that a P2P build has been assembled optimally, a THD measurement is needed. A fully optimized PCB, on the other hand, will result in an optimized build every time.
~Tom
Anybody considering a point-to-point build of any precision circuit aught to read my LM3886 P2P vs PCB thread. I present plenty of data there. Conclusion: A fully optimized P2P layout performs nearly as well as a fully optimized PCB. To ensure that a P2P build has been assembled optimally, a THD measurement is needed. A fully optimized PCB, on the other hand, will result in an optimized build every time.
~Tom
Last edited:
hand building high frequency prototypes is possible - dead bug, gnd plane perf board, add "surf boad" for smt...
(the sip adaptors do add inductance but you can glue the flat 9000 series onto the gnd plane perf) http://www.capitaladvanced.com/9000ser.htm
with sub MHz feedback gain intercept of the power chip amps you really aren't pushing hand wiring limits - its the 50 MHz op amp that needs very local feedback, bypassing
(the sip adaptors do add inductance but you can glue the flat 9000 series onto the gnd plane perf) http://www.capitaladvanced.com/9000ser.htm
with sub MHz feedback gain intercept of the power chip amps you really aren't pushing hand wiring limits - its the 50 MHz op amp that needs very local feedback, bypassing
Last edited:
its easy to get ridiculous resolution of nonlinearity distortion measurement with 2 tone IMD stimulus and filters passing/amplifying the difference product/blocking the test tones - don't even need that clean of source as long as you can generate the 2 tones without IMD, only harmonics, then sum linearly
do need the filter to knock down the stimulus so the analyzer front end/ADC nonlinearities don't limit measurements either
I have managed ~160 dB floor with these techniques and long averaging
saw "bad" caps giving ~ -110 dB product, fixed with polystyrene
do need the filter to knock down the stimulus so the analyzer front end/ADC nonlinearities don't limit measurements either
I have managed ~160 dB floor with these techniques and long averaging
saw "bad" caps giving ~ -110 dB product, fixed with polystyrene
Last edited:
Tom, I have been reading your PTP vs. PCB thread and as jcx points out, PTP doesn't have to be bare wire in air. I have 1/16" teflon sheet and foil to help me simulate a proper pcb. A scowling laugh isn't what I expected from someone who vowed in an earlier post to 'help any diyer who is willing to put in the work'. No one is taking anything away from what you have accomplished. At first glance this design looks like a winner but all the numbers aren't in yet and too few people have heard it to conclude its sonic excellence. I'm rooting for you and that's why I'm trying to corroborate what you've done with my PTP attempt. As a courtesy, I will not use an LM3886 but an LM3875. I will be more than happy to send it to you when complete for you to test it and highlight the superiority of your PCB. Are we still friends?
I wasn't laughing. Nor was I trying to start a peeing contest. In Post #170, I merely pointed out that a well executed P2P will perform nearly as well as a well-executed PCB, but that it is much, much easier to get good performance, repeatedly using a PCB. I have posted the data to back up this claim in the LM3886 P2P vs PCB thread.
Friends? Absolutely! No worries.
~Tom
And the 5th harmonic is the highest order that pops out of the noise floor. 4th order is long gone. Not too shabby!
Pretty darn good, if you ask me. But, of course, I may be "slightly" biased...
The all low-order (mostly 2nd and 3rd harmonic) distortion reminds me a lot of the distortion profile of a typical single-ended triode amp, except, the THD is much, much lower in case of the Modulus-86.
~Tom
Will OPA1611 be any better for this project?
Tom,
It seems that the OPA1611 is better than the LME49710. Its too bad that the OPA1611 does not come in DIP8 package that you have used in your project. However, I wonder if its worth using the SOIC/DIP adapter to replace the LME49710 with OPA1611 to check if it brings any further improvements as we are crunching numbers anyway.
You can report of any sonic improvements also.
Tom,
It seems that the OPA1611 is better than the LME49710. Its too bad that the OPA1611 does not come in DIP8 package that you have used in your project. However, I wonder if its worth using the SOIC/DIP adapter to replace the LME49710 with OPA1611 to check if it brings any further improvements as we are crunching numbers anyway.
You can report of any sonic improvements also.
Last edited: