Sunday, October 17, 2010
The total noise reduction was achieved through various means NOT
Update and Edit: Additional testing has cast doubt on the original claims of this post below. It now does appear that 6dB of noise reduction in the supertweeter amplifier output was achieved by using a 6dB attenuator. I now measure 21mV, 0.64mV (A weighted) with no attenuator and 9.7mV, 0.33mV (A) with attenuator. Specifically re-measuring the change of putting most equipment (plugged into power conditioner/UPS) on dedicated amplifier circuit instead of regular (crazy!) household circuit, I have found that change makes zero measureable difference in noise level. Surprise Surprise. I have now switched from using a 3dB attenuator on midrange amplifier Parasound HCA-1500A to using its built-in level control to achieve 6dB attenuation. The Harrison Labs 3dB attenuator I was using seemed to make the sound slightly funny sounding in the midrange probably because it presents 3K ohm load to the Behringer. But the Harrison Labs 6dB attenuator on the supertweeter amplifier, which has no level controls, does not seem to have that effect, probably because ears can't hear it at such high frequencies. I watched the voltage output on the tweeter amplifier for awhile, it seems it stays below 1 volt (!) which is much less than a watt output (on a 250 watt amplifier). Based on that test, there should be no problem with 12dB attenuator which will be ordered today. [End edit, now back to original erroneous post.]
Yesterday's noise reduction exercise (the supertweeter which used to make a rough fizz now only makes a barely audible background hiss as close as you can get to it) mainly focused on use of interstage attenuation. Actually, only about half of the noise reduction came from attenuation. The other half came from switching to short shielded equipment cords on 3 main pieces (UPS, Tact, DCX), which definitely made a measureable but small difference, switching the power conditioner to the dedicated amplifier circuit, and using some better shielded cables. Unfortunately, my measurements of all this are incomplete, but the total noise reduction was from 21mV at the output of the tweeter amp (pretty hefty noise level) to 9mV at the amp, a greater than 6dB reduction. The use of a 6dB attenuator didn't quite reduce noise 6dB, more like 4dB according to my measurements, likely due to other noise sources not yet fixed that are not being attenuated.
These noise measurements are wideband, made with my nice Meguro noise meter shown above. I didn't do A weighted measurements consistently unfortunately. The audible noise reduction seemed much greater than 6dB actually, which is possible because it's possible the audible noise has been reduced more than the inaudible noise. In fact, that often happens. The A weighted noise levels are much much lower. I figured out that is how Behringer can meet their noise specifications for the DCX 2496. The "A" weighting ignores a huge amount of high frequency whine from the Behringer's switching power supply. A weighting is not a completely accurate weighting according to audibility, either. I understand that the B weighting is better but nobody uses it. I think noise is bad whether it's considered "audible" or not. Even the "inaudible" noise modulates audibly with lower frequencies. And it may well be that our hearing system is affected by tones we don't actually perceive; there is at least one peer reviewed research paper which suggests that.
One funny thing in all this was that I didn't intend to go this way. I had long figured the supertweeter noise was ground loop noise with all the lower harmonics filtered by the supertweeter's build-in crossover. However, when I tested for ground loop noise by temporarily ground lifting the tweeter amp (using a grounding adapter) I discovered that made no difference. To be sure I was hearing everything, I put a full range speaker on the amp instead of the supertweeter.
So then I got involved switching the line cords to the nice new ones (SJT shielded 14 gauge, much nicer than freebie SVT), etc., which wasn't making much difference, so I asked myself "Just where is this noise coming from."
I then put my meter on the Behringer itself, with no actual digital input, and the wideband noise level was an astoundingly high 0.7 mV. So that's where the noise is coming from! Then I immediately realized an attenuator would be a quick fix.
Now I'm a little worried about the 3dB attenuator on my midrange amp. The voltage levels there are pretty high, and the Behringer may be getting distorted into the 10K load or less of the attenuator. Now I'm thinking I may just use those volume controls on the back of my Parasound HCA-1500 after all. They maintain a relatively high load impedance for the Behringer, and may sound better for that reason if not others. In the end, I'd like to replace those volume controls with fixed resistor networks of the lowest noise resistors.
Posted by Audio Investigator at 8:11 PM