Every now and then I put my ear up to the Acoustats to see if a tiny buzz has been created by recent changes. Sometimes I think I hear a sound but most often it's the air conditioning compressor (either mine or my neighbor's which is just outside the living room window), the refrigerator, or something else. I pride myself nowadays on having a very quiet system. Despite it's complexity. This is partly because of design, and partly hard work like I just did today and yesterday.
It wasn't always that way. For 30 years I just took a small amount of hum and noise for granted. Most often I could hear a quiet hum distinctly from the listening position with nothing playing. Then around 2000--my 31st year of audio foolery--I finally started taking hum and noise seriously, and I strangely found that the biggest source of hum in my system then was not my MC225 tube amplifier, which I had long suspected of needing a refurb, but my pristine looking solid state 4 way analog crossover, the highly esteemed Pioneer "Series Twenty" D23. Apparently because of a failing power supply, it was injecting hum directly into the audio signal. It was not, that time, a ground loop. I never fixed the D23, instead I tried many other approaches to doing speaker crossovers, and ended up using DSP around 2006 and ever since. No matter how much DSP you do, and no matter how many DSP processors, there should be no hum and noise added. Though you might get ground loops.
A few days ago I was thinking this time I was hearing a very tiny buzz, just barely audible. It didn't sound like the air conditioning compressor or other noises happening right then, and it was clearly at it's loudest right in front of the Acoustats. It wasn't easy to be sure, for awhile I thought it was just my hair rubbing against the Acoustat "socks". I went back and forth between the speakers hearing something similar in both. I was thinking to myself this could be the Acoustat power supply failing. So, I shut off the Hafler 9300 amplifier. It was pretty clear this made the tiny buzz go away.
I then though of doing something I hadn't done in awhile. I hooked my Fluke 8060 A DVM to the Acoustat terminal with nothing playing. For a lot of reasons, the Fluke is not necessarily an optimal way to measure audio signals. But it's very convenient.
I was shocked to see 2.23 mV on the meter, way more than I would have expected.
So then I put shorting plugs into the Hafler 9300, and measured 0.00 mV. The Hafler not only was not at fault, it is one quiet amplifier.
So then I tried muting the Emotiva Stealth DC Dac driving the Hafler. The buzz measured barely different. While it was still muted, I pulled out the XLR connector for the AES/EBU digital signal. And then again the noise measured 0.00 mV.
Clearly this noise is a ground loop from the source of the digital signal to the Emotiva. I was immediately thinking I knew the Emotiva wasn't perfect wrt AES input. I was thinking AES should not have a hum and noise problem because it's balanced, but the Emotiva implementation isn't very good. I'm not sure if it uses a transformer. I remember maybe reading a review which complained about something like that. (But the Emotiva's AES transformer or lack of transformer is not actually at fault I concluded later, as I will explain.)
I started thinking I should buy a better DAC. (And perhaps I should.) But meanwhile I could switch to one of the DACs I have in storage, including an Audio GD Dac 17, and an available Denon DVD 9000. As far as the Audio GD, I need to do some retesting on it, after a wrongly blamed it for causing the Krell FPB 300 to shut down frequently. I want to be sure I know how good or bad the Audio GD is before putting back online. Meanwhile if I used the Denon or most other DACs I have in storage I'd need to convert the AES signal to Coax. I have several converters for that, but do I know if THEY are isolated or will just propagate the ground loop via the shield of the Coax? With any of these converters, or even just the Emotiva itself, I might be tempted to use Toslink, which would certainly eliminate the ground loop problem. So, if I'm going to use Toslink with some other DAC, why don't I just use Toslink with the Emotiva DAC that's already set up?
I tried Toslink and indeed it reduced the noise down to 0.00 mV with the Emotiva muted.
But I don't like Toslink. I think it's a weak connection very much more subject to jitter than coax or AES. So after reading online about ground loop issues with AES connections, I decided to do something radical that several people suggested, except perhaps not the way they would do it. I took the 3 foot AES cable I had been using, and cut out the shield for about 1/2 at the end using wire strippers and cutters. Along with the shield, there was a drain wire I cut too. I patched it up with lots and lots of white electrical tape.
Sure enough, this also reduced the noise down to 0.00 mV with the Emotiva muted. And it worked just fine apparently playing music.
Ground loop fixed! It rarely gets this good, noise going down from 2.3mV (horrible) to 0.00mV (perfect, so good it's unreal) with one little change costing nothing.
Then with nothing playing, and in fact the Tact preamp itself was muted, I unmuted the Emotiva. I was shocked to see the noise rise up to around 1 mV (sometimes as high as 1.5mV). It was still silent at the speakers, but it almost seemed like 1/2 of the measured noise had come back.
Perhaps this was because I hadn't cut the #1 wire also, I wondered. (Actually, the shield and drain wire carry the #1 pin wire, so in fact I had cut those as well, though possibly not a connection between the Emotiva chassis and the shell of the XLR connector.)
I tried the Toslink unmuted and sure enough, it had slightly more of this noise voltage, over 1.6mV. That a Toslink connection also had this noise, with no XLR connected to the Emotiva, proves that the noise is not being "introduced" by my having cut the shield on the cable. The noise appears to be related to the digital signal itself interacting with the Emotiva.
I now figured this was probably very high frequency noise generated by the DAC, perhaps caused by the ASRC, even for the supposedly muted signal. Perhaps the Toslink connection had even more because it added jitter on top of jitter.
I was still uncomfortable about this new measured noise the next day and I decided to do some more tests. I bypassed the ASRC containing miniDSP units, and the Crystal 8620 containing Behringers (I'm not sure if they enable ASRC or not) and plugged the DAC straight into the muted (and with no signal playing on anything either) Tact 2.0 RCS preamp, which uses fully synchronous digital receiving and transmission. The spurious noise was about the same. So it was not being "caused" by the ASRC's.
But I still really wanted to be sure this was not hum or buzz or any audible noise like that, even if it might not be audible simply because it was much lower than before (and it was barely audible then).
So I got out my Meguro Noise Meter, which has an A Weighting filter enabled.
I ran these tests with the shield-broken XLR cable. With the Emotiva muted, I measured 0.076mV at the output of the Hafler. With the Emotiva unmuted, I measured 0.13mV on the Meguro.
So the A weighted noise does not show the dramatic increase to 1.5mV shown in the Fluke's semi-wideband measurement. This proves in fact the noise remaining, after fixing the ground loop issue, is essentially inaudible. It is likely to be supersonic noise.
I can see some strange noise on the spectrum display of the Behringer DEQ that does midrange EQ after a miniDSP does the crossover. Even when the Tact is muted, there is noise around -140dB or so shown on the Behringer display. THAT may be the effect of dither and/or the ASRC in the miniDSP's. But the ultrasonic noise we are measuring is much higher than -140dB, compared to the full output of the amplifier (around 34 volts) it is -106dB down.
So it's probably mostly above 20kHz, the cutoff on the Behringer spectrum display.
Not much to worry about. The big thing this time was the ground loop which was easily fixed. I grabbed another XLR cable, without the broken shield, and measured noise at 1.8mV on the Meguro.
I got the same results with Toslink and AES disconnected, so once again it's not my broken shield, though for that I repeated the Toslink only connection with the Fluke, and consistently measured just over 1.6mV, about the same as my Coax connection.
BUT, wait...I tried having no input connected to the Emotiva at all, and still measured the same everything...
Apparently this is just the wideband noise of the Emotiva DAC when it is operating, and it is slightly quieter in audible noise when it is muted, but way quieter in ultrasonic noise. Here I should add that I have the Emotiva gain control set to +2.5dB. The Gain of the Hafler 9300 is 29. So the noise appears to be about 0.0016V / (29 * 1.4), which would only be 94dB below 2V. This seems rather high, though I've never paid much attention to the "Unweighted" noise specs of DAC's if I've ever seen them at all. Virtually all of the specifications we read are "A weighted," which makes most everything sound better than it is. The weighted measurement on the Meguro is just over 20dB better, which would be a bit above 114dB, which is a bit below spec.
It's strange, very strange to me, that the DAC has way more wideband noise than my amplifier. Way more audible frequency noise too, though so low in level it's hardly audible in practice. I would have never expected this. I wonder if this DAC is failing. I already had the Emotiva DAC for one living room amplifier fail (so I can't run ABX amplifier tests right now, I have to disconnect input cables to switch amplifiers). I'm going to have to look at other DACs. But the audible portion of this noise problem appears miniscule, so it's not a high priority as was fixing the ground loop. Back when I didn't care about hum and noise as much, I had the opinion that a little bit of hum and noise might actually make things sound better. That's still a possibility I haven't fairly tested. But now I presume that noise should be reduced as much as easily possible (without doing something that might increase distortion, etc).
It occurs to me that whenever you have a DAC with an AES connection, and outputs to coax audio, it is certain to propagate a ground loop. The ground of the coax output is going to be referenced to the shield ground of the incoming AES. It therefore makes sense to me that Pin 1 should be completely disconnected on AES devices having coaxial output. AND the XLR shell should be isolated as well.
Or, maybe it's just that the Emotiva isn't grounded to it's AC input. Perhaps gear with AES input MUST BE grounded, so that there is a separate chassis ground to which the AES shield is ultimately referenced, which is barely in contact with the active circuitry except perhaps through a small capacitor and/or resistor.
The other possible solution is to have a cable (like mine now) with the shield and ground connections completely broken. Rather than crudely strip off the shielding as I did, it would be better to make the cable this way from the beginning. But so far as I can see, nobody sells a shield-broken AES cable, only a pin 1 lifter (which might not work if the shells are connected). The shield should be broken on the male XLR pin side (outgoing signal).
Meanwhile, the importance of the tiny buzz I fixed, which could be faintly heard with ear right up to the Acoustat socks, is seriously questioned by the following line of reasoning. This buzz problem that I just noticed and fixed must always have existed to a greater or lesser degree when I used an AES connection to the Emotiva to a single ended amplifier connection. That means when I was comparing the Krell FPB 300 and the Hafler 9300 and in blind level matched ABX tests, and found no audible differences in music, I was comparing an amp which had this problem (the 9300) and an amp which didn't, because of its balanced input connections (the Krell). If little buzzes like this make a difference at all, at least THAT should have been audible in the blind testing. But it wasn't.
So, this would seem to show that faintly audible noises that you can just barely hear with an ear up against the speaker, are not important in ordinary listening. The buzz measured approximately 2mV at the speaker terminals. The rated output of the amplifier is about 34V. Thats about 85dB down from peak level, or about 55dB down from soft passages.
What does that say about noises that are -110dB or -140dB ? Perhaps not even worth thinking about ever.
