I have been modifying (and hopefully improving) some basic amps for fun lately. One was a Yamaha RS300 and a Rotel RA 930-AX(for a friend). Main improvement on both was increasing bias to abt 30mV between emitters. Anyway, that is not the topic here.
In these cases I wanted to measure channel separation, witch I have not done much, since many of my DIY builds have been dual mono, and maybe I have not considered it to be a significant factor for SQ.
What I found was that separation was pretty bad on both amps, especially on higher frequencies (around 40dB from memory).
However, the most shocking to me was that the 'distortion separation' was abt 0!? What I mean is that the distortion components on the silent channel were about the same level as the distortion components on the driven channel, even if the fundamental was 40-60dB lower. This was almost the same on both of these amps.
I suspect it has to do with the common PS, and possibly that the front end is on the same rails as the output stage?
Could somebody explain this further?
In these cases I wanted to measure channel separation, witch I have not done much, since many of my DIY builds have been dual mono, and maybe I have not considered it to be a significant factor for SQ.
What I found was that separation was pretty bad on both amps, especially on higher frequencies (around 40dB from memory).
However, the most shocking to me was that the 'distortion separation' was abt 0!? What I mean is that the distortion components on the silent channel were about the same level as the distortion components on the driven channel, even if the fundamental was 40-60dB lower. This was almost the same on both of these amps.
I suspect it has to do with the common PS, and possibly that the front end is on the same rails as the output stage?
Could somebody explain this further?
In case I was unclear about the measurement, I had load resistors on both channels, feeding signal to one input, and shorting the other. I then looked at the FFT spectrum in ARTA for the channel with shorted input, while feeding the 10kHz test signal to the other channel. I could then see the distortion harmonics of the driven channel on the 'undriven' channel, almost at the same level as on the driven channel, even if the fundamental was way lower.
I did the same test on budget Denon surround receiver, and did not see the 'distortion crosstalk'. Channel separation was also slightly better than the two mentioned before. Again, bias from factory (as specified) gave a really ugly distortion spectrum with a lot of high order distortion. With proper bias, mainly 2nd harmonic was visible, and it's level was below -90dB in most cases. It is idling a bit warmer now, but not hot. Easy 'performance upgrade'.
I did the same test on budget Denon surround receiver, and did not see the 'distortion crosstalk'. Channel separation was also slightly better than the two mentioned before. Again, bias from factory (as specified) gave a really ugly distortion spectrum with a lot of high order distortion. With proper bias, mainly 2nd harmonic was visible, and it's level was below -90dB in most cases. It is idling a bit warmer now, but not hot. Easy 'performance upgrade'.
Last edited:
Crosstalk? The Rotel was a repair and 'tune-up' for a friend, but the Yamaha I still have. I can measure it again later and save some screenshots.
The "distortion harmonics" of a 10kHz signal are ultrasonic and probably not worth worrying about - crosstalk at higher frequencies is always going to be worse (due to stray capacitive coupling paths) but I am surprized its that high even in the ultrasonic area.
I'd double check all the grounding is solid from each channel back to the PS. Shared grounds and ground loops are a plausible reason for poor separation figures.
BTW stereo crosstalk of -30dB is roughly the threshold of audibility from what I've read, -40dB is something you can live with without worrying.
I'd double check all the grounding is solid from each channel back to the PS. Shared grounds and ground loops are a plausible reason for poor separation figures.
BTW stereo crosstalk of -30dB is roughly the threshold of audibility from what I've read, -40dB is something you can live with without worrying.
~ very little surprises me with budget gear ~
A few vintage HK330c receiver(s) I completely rebuilt... *the L and R speaker feeds shared a single ground from the star ground of the unit. I get it... would make for a less "cluttered" underbelly, cutting down on 2 more ground runs (including) second pair / speaker set B.
One of the first things I did, was correct that (while completely bypassing; speaker switching) running speaker lines direct, rerouting speaker B's single ground, "stealing it". I did not want switches in the speaker signal path.
*my point for explaining, is that it is so common for (even the good) manufacturers to make cheesy short cuts = that result in goofy stuff we see / measure / question. Multiple ways to make these bargain price toys [a little] better.
I am not knocking the HK330c, they are are actually ~very good sounding~ "when made right", correcting some of the manufacturing / cheap build nonsense.
A few vintage HK330c receiver(s) I completely rebuilt... *the L and R speaker feeds shared a single ground from the star ground of the unit. I get it... would make for a less "cluttered" underbelly, cutting down on 2 more ground runs (including) second pair / speaker set B.
One of the first things I did, was correct that (while completely bypassing; speaker switching) running speaker lines direct, rerouting speaker B's single ground, "stealing it". I did not want switches in the speaker signal path.
*my point for explaining, is that it is so common for (even the good) manufacturers to make cheesy short cuts = that result in goofy stuff we see / measure / question. Multiple ways to make these bargain price toys [a little] better.
I am not knocking the HK330c, they are are actually ~very good sounding~ "when made right", correcting some of the manufacturing / cheap build nonsense.