Originally Posted by DonH50
I did not read your post, sorry. Long day. I just responded to @m. zillch
<pause to read>
I can understand your viewpoint (yes, I have an EE degree; my current day job involves GHz interfaces, and most of my career focused on various data converter and RF circuits from hundreds of MHz to a few hundred GHz). I think we said pretty much the same thing but differ in how broadly we define "ground loop". While anything from DC to light modulating the ground could be properly considered a ground loop, for audio circuits by convention if not by standard (AES) it is limited to the 50/60 Hz current from the incoming (AC) power supply. The RF ground noise, while also a ground loop issue, is deemed RFI in the audio world and considered a separate noise component, at least IME. To an EE that is a distinction without a difference, but to an audio guy (who probably has limited RF experience) it is a big difference. Debugging and fixing a LF ground loop caused by coupled (E or M) field from the AC line is a different thing than dealing with RFI.
This is one of those things that is always gnarly. Technically the datastream on a digital link creates (injects) noise that couples into the receiver. That can occur on the signal (plus, hot) or return (ground) path. I tend to think of RFI as more a noise injector or aggressor rather than a (LF AC) ground loop that follows a much longer path and occurs at a much lower frequency. Thus I treat them differently (repeating above, sorry, end of a 60+ hour work week and as a reward get to work tomorrow (Saturday) since one of my tests is flaking out).
It is probably worth noting that differential (balanced) audio circuits do a pretty good job at handling ground loops (the LF AC type) by facilitating ground isolation since the AC shield is not (or should not be) the AC signal ground. But common-mode rejection falls off rapidly with frequency so even common-mode RF noise will generally be a problem to a differential circuit. We depend upon the isolated shield to "block" RFI from reaching the inner signal pair.
Bottom line is I do not disagree with your broader definition but in practice, at least that I have had in and out of audio, the two are treated differently in the audio world (by audiophiles and design engineers). I had the same problem when I first read about "passive bi-amping" in the context of AVRs: to me, passive bi-amping means a passive (RLC) line-level crossover vs. an active line-level crossover. This led to some rather confused posts before I understood what the HT folk meant by "passive bi-amping" (and a lingering question why anyone would do such a thing).
Or we can agree to disagree, I need to get some sleep.
Great to read posts from a fellow engineer, BTW! - Don
Edit: I am familiar with Murata's (and other's) CM chokes and their benefits; we use them extensively for decoupling and filtering of CM RF noise. Again, a difference in how broad the definition of "ground loop" is in the audio (not RF) world.
Nice to talk to an EE so we can actually get deeper into theory. I looked at your profile, you are an analog IC designer. I was an analog IC designer in 1980s for two years working for a company called Exar Corp. But I was working mainly on bi-polar IC design and the speed was very slow. I know now is CMOS or even more advanced stuffs. My main experience is in systems both analog and digital, mainly high speed circuits ( for my days that is up to 2.5GHz stuffs). I worked a lot on signal integrity for pcb design and big systems.
I designed USB interface before, but that was USB 2 that is only 480MHz stuffs. This is my analysis. We should all know the current loop is the main cause of EMI radiation. EMI flux is proportion to the current through the loop and the area enclosed by the current loop. Ground loop is one form of current loop.
The ground loop current might be RF, but digital signals are random, it can be bursts of signal that can down mix into audible frequency. Also, RF signal can down mix into low frequency and it audible.
Look at my drawing:
1) I drew the transmitting side on left side and I called the ground as GND1. I have a differential driver Tx driving the USB +D and -D through the cable to the receiver Rx on the right side.
2) I put 2 different sources of noise V1 and V2. V2 is the noise on the Tx amp that induce a common mode noise voltage onto +D and -D at the same time. But V2 is not on the ground shield as shown.
3) I put V1 between GND1 and GND2 of the Tx side and Rx side resp. This noise will drive between GND1 and GND2.
4) On the receiver side, there is a Rx amp that receive the differential +D and -D and convert back to single end. The Rx amp should have hysteresis and good common mode rejection ( very common for differential LTP input stage).
5) There are two ways of termination. For USB is 80ohm differential. I have two possibility of terminations, I labeled (A) and (B). For (A), it's just an 80ohm across +D and -D, it's differential termination. This is the most likely case. The other case I labeled (B). It's 40ohm on both +D and -D to GND2. This is not the likely case, 80ohm across +D and -D is the most likely case.
Now, lets look at how V1 and V2 produce current on the ground and on the +D and -D.
1) Let's look at current I
2 that caused by V
2. In case (A) where 80ohm across +D and -D, the common mode input impedance of the Rx is high, I
2 is literally zero. In the case (B) where there is two 40ohm resistor to ground from +D and -D. the noise current produced by V
2 will be I
2/20ohm= 50mA.( 40ohm//40ohm = 20ohm)
2) Lets look at ground current I
1 caused by V
1. it will be very large even though V
1 is small. I gave an example of V
, and resistance between GND1 and GND2 is 0.1ohm, I1 = 1A
You can see V
1 which is the ground noise cause a lot more current flowing through GND1 and GND2. I drew the ground as a big loop as it goes from the Tx through the power cord to Rx. You know EMI produced is proportion to I1 and the area of the loop. That's where all the EMI is generated and affect the circuit.
If you put a ferrite, it can give as high as 50ohm impedance separating GND1 and GND2, so even with V
. Just a little impedance will go a long way in reducing EMI noise. This EMI noise is completely due to ground loop.
this is my analysis. A lot of noise is cause by random digital noise, that actually can generate audio noise by itself.
Nothing is 100%, I can be wrong, and also, this likely not covering all cases of noise problem. But I think this is very likely the noise problem and it can be verified by just a simple ferrite clip and a minute.
Of cause there's also another possibility that the pcb layout of the transmitter side and the receiver side are so bad that there is ground loop in the transmitter that radiate EMI into the air and the receiver has trace that forms a loop and let the EMI flux cut through the loop and induce current noise and cause the noise in the audio. But the chance of this is smaller. The symptom will be if you move and turn either the transmitter or the receiver to a different position, the noise will change. That would be a good indication the noise is through the air, not the ground loop formed by the USB cable.