Bear,
you might want to look here, where Nelson Pass raised the topic.
http://www.diyaudio.com/forums/showthread.php?postid=1343565#post1343565
Discussion continues over the few next pages, though not in a very linear manner
Simulations, based on reverse-engineered data, were also carried out by different people, which back up things pretty well, it seems.
- Klaus
you might want to look here, where Nelson Pass raised the topic.
http://www.diyaudio.com/forums/showthread.php?postid=1343565#post1343565
Discussion continues over the few next pages, though not in a very linear manner
Simulations, based on reverse-engineered data, were also carried out by different people, which back up things pretty well, it seems.
- Klaus
PMA said:
Not my intent, you could parallel any arrangement of TL UTP or whatever and get 8 Ohms characteristic impedance and it's hard simply due to the physics to get the C not to be problematic. I wonder about the old 300 Ohm twin lead with the glass rod insulators? 38 of those gives 8 Ohms.
Dunlavy's patent is a good example of what you can get away with at the USPTO.
bear said:
Right. What do you think the impedance of the loudspeaker is at, pick a number, 27 MHz? (The CB radio frequency) With all the long leads and all that? Probably pretty high.
If your speaker cable has an impedance of 40 Ohms or so (I think that's what zip cord is at several MHz - just an example...), then if RF is induced into the cable through pick-up as an antenna, then there will be associated voltage and current maxima and minima along the length of the cable. It's mostly a function of the length of the cable in wavelengths at the RF frequency. If the cable is terminated at both ends with something close to the characteristic impedance of the transmission line (again this the RF value we're talking about), then the maxima will be lower than would be the case of an unterminated line at that frequency. This is due to the termination mismatch and the resultant standing waves on the line.
You can look all this stuff up on the web. Loads of engineering books dating back from the 30's until today have information about this. This isn't some esoteric secret sauce invented in some cellar of a castle in the mountains.
So, if you want to terminate the line at RF, you need a suitable termination. Putting a resistor across the line would affect the audio range. So, instead add a small cap in series with the resistor. If you used a .01 uF cap in series with the 40 Ohm value mentioned above, the impedance of the termination is roughly 200 Ohms at 100 KHz. At least the magnitude of the impedance is. It gets higher as you go lower in frequency. That's not much in parallel with even a 16 Ohm horn. But, at 10 MHz, this same termination is about 41 Ohms, almost the same as the transmission line impedance. Do the same thing at the amplifier end, although a lot of amps already have Zobel networks there that would do pretty much the same thing.
The same thought process could be used with line level signals, although that's kind of tough when the input impedance of an amp or preamp is 10K or more. The same idea can be used on the ac/dc power inputs as well.
If you do something to reflect the energy, it just reflects back and forth, moving the voltage and current maxima and minima. Better to dissipate it.
Before people scream about the length of a speaker cable in terms of wavelengths not being relevant at audio, that's not the point. The length of a speaker cable *is* a significant portion of a wavelength at RF - maybe several wavelengths. The frequency band where we might get RFI problems from. The above is a possible way to minimize the effect of the cable at RF, not audio. I believe the original question was along the lines of how to reduce the affect of RF on a piece of audio gear...
scott wurcer said:
I saw a not well though through digital audio cable at CES that used 8 75 ohm coaxes in parallel. The muddy thinking was that 8 paralleled cables would reduce the cable errors by 1/8., like paralleled dacs. ??? I didn't want to open the discussion of "phase matching" the coaxes to the 1 pS level.
Paralleled coaxes might work, but to what end? Very few speakers are 8 ohms or even close. Do you put a matching network on the speaker end? And what happens to the actual response?
The referenced article from Audio Design Line was really interesting. The effects of RFI on IM was on a magnitude I would not have expected. It suggests that we should look at shielding even more carefully.
1audio said:
An amplifier speaker system is not a power matched system anyway. The amplifier output impedance at audio is close to being a voltage source (that damping factor thing...) and the loudspeaker is all over the place, although usually between 1 and 16 Ohms. Almost all speakers are designed with this in mind - the Pass current source amplifiers are one exception when applied to the right speakers.
More to the point is the RF situation.
Putting the stereo system in a well appointed screen room with an isolated power supply would solve this - no RF, no problem. You could probably get the acoustics pretty good in there, too, if you worked at it.
Then consider that low potential terminal block, connected to shield (if coax for speaker used), gets inside power amp untented. Audio components are not designed according to HF design rules, and there is a space for mysterious aids then (I would better not name directly). Consequences are assessed, not reasons.
CG said:
I think you are exactly right. Ideally, the speaker cable would like to see some kind of resistive termination at both ends at high frequencies. Whether this resistance should be on the order of the speaker impedance or line characteristic impedance can be debated. I suspect the characteristic impedance of most speaker cable is between 25 and 100 ohms.
The Zobel network in the amplifier is most helpful in this regard if it is on the speaker side of the coil (if used) and right at the speaker terminals.
I like to put Zobels on BOTH sides of the L-R coil, so that I can get the Zobel on the output stage physically very close to the output transistors for maximum effectiveness in suppressing output stage parasitic oscillations (emitter follower output stages like to see some resistive load out to very high frequencies). In some cases I like to use distributed Zobel networks on the output stage side.
Many loudspeakers may look somewhat resistive up to moderately high frequencies if the crossover ends up effectively placing a tweeter pad across the input at high frequencies, but this is certainly not always guaranteed.
These approaches help the differential RF pickup, but do nothing for the common mode pickup of the speaker cable acting as an antenna. In some cases that common mode RF current flowing in the speaker return ground might cause some trouble, depending on the amplifier's grounding topology.
Cheers,
Bob
Bob Cordell said:
*Very* good point about the common mode current (actually flowing on both the "hot" and "ground" - common mode).
What people call ground is really a misnomer at RF. That wire connected between the amplifier chassis, the electrical socket, and all the rest is hardly "ground." The impedance of that is sky high at RF. If attaching this wire affects the RF performance, it is only because it moved the resonant of the overall system to some frequency where there might not be any interference.
This will sound like absolute audio heresy, but perhaps the solution to the common mode problem is to coil some of the speaker cable - both conductors together - somewhere close to the amplifier. The common mode impedance will be pretty high. This is often done in the non-audio world to make broadband baluns. Since the "core" material is air, it's as linear as you can get. At audio, presumably the signal is entirely differential mode, so the effect at audio shouldn't be great.
Even at AF I came to dislike the term "ground" and I always try to not use "ground" in my thinking and drawings. In fact this avoidance of "ground" was the biggest step forward in my understanding of subtle details in electronics (audio/instrumentation).CG said:
I'd even go further and use iron-powder toroid cores, just on every cable (including mains). I had indeed success with these when mains problems occured (or with just plain 50m cable drums) at several occasions in homerecording studios, with the typical problem of someone hooking up his SMPS-powered laptop to the console, or problems with TFTs and the like.
Thanks to the ususal suspects like H.Ott, D.C.Smith, J.Brown...
- Klaus
PMA said:
PMA said:
Not particularly, I guess it suffices to say Dunlavy's patent text goes along unproductive lines IMO, such as the the 8 Ohm matching at audio frequencies.
On the second item, the new line of Neutrix XLR connectors were a measurable improvement on even EMI incursion from internal circuitry. The consruction is interesting, trying for a conplete ground shell with bypassing divided up in a ring around the passthruogh. Definately not the lowest cost construct.
EDIT - I have only tried these on battery powered equipment.
PMA said:
When I've either needed a balun or had RFI problems, yes.
But, I'd rather not spend any time debating. If you don't think my suggestion has merit or the explanation is full of el toro poo-poo, that's fine by me. I just offered a suggestion to a fellow DUY guy that I thought might be helpful.
Having fun? At least it got you thinking, but Mogami (I think it was them) made an 8 ohm coax cable, back in 1979 or so. Zobel networks were also used with the new, exotic cables at that time, especially the 'Polk' cable that was a real amplifier killer. It is the Q of the cable that is the problem, not the exact capacitance.
Have you noted that Mogami doesn't seem to offer cable, anymore?
Have you noted that Mogami doesn't seem to offer cable, anymore?
- Status
- Not open for further replies.