Hi Charlie,
But we are still not done. As Bob pointed out, a RC-filter should be placed right at the RCA jack, that's is a good thing. But if the distance (i.e. leads) to the input stage proper is a bit long, we also need a shunt capacitor right there. Without such HF shunt, the behavior of the global NFB (if applied to the non-inverting input, of course) might be ill defined at HF.
Cheers,
Edmond.
But we are still not done. As Bob pointed out, a RC-filter should be placed right at the RCA jack, that's is a good thing. But if the distance (i.e. leads) to the input stage proper is a bit long, we also need a shunt capacitor right there. Without such HF shunt, the behavior of the global NFB (if applied to the non-inverting input, of course) might be ill defined at HF.
Cheers,
Edmond.
This RFI discussion has been very interesting an goes much deeper than any previous one I remember.
I am very bad att HF theory, but Bobs suggestion of a filter immediately att the input connector sounds (at least intuitively) wise. If I understand him right, he suggest basically a voltage divider type of RC link. Another possibility would be a Zobel filter, ie. a series RC link between signal and ground. In my limited understanding of HF theory those two variants would seem to serve the same purpose for HF, ie. an HF signal would see 50 Ohms to ground in both cases. Is that a correct understanding?
So far inputs has been discussed, but the output of an amp is also a potential RFI path into the amp. I remeber Rod Elliot writing many years ago that he suspected that might be a problem since RFI that sneaks in that way will have a fairly direct path to the input stage through the feedback network. His suggestion was to add a Zobel at the output connectors. For line level amps, Bobs solution would often be OK also at the output, if mirroring it, viewing the output as an input (for HF). Any thoughts on output RFI filters?
I am very bad att HF theory, but Bobs suggestion of a filter immediately att the input connector sounds (at least intuitively) wise. If I understand him right, he suggest basically a voltage divider type of RC link. Another possibility would be a Zobel filter, ie. a series RC link between signal and ground. In my limited understanding of HF theory those two variants would seem to serve the same purpose for HF, ie. an HF signal would see 50 Ohms to ground in both cases. Is that a correct understanding?
So far inputs has been discussed, but the output of an amp is also a potential RFI path into the amp. I remeber Rod Elliot writing many years ago that he suspected that might be a problem since RFI that sneaks in that way will have a fairly direct path to the input stage through the feedback network. His suggestion was to add a Zobel at the output connectors. For line level amps, Bobs solution would often be OK also at the output, if mirroring it, viewing the output as an input (for HF). Any thoughts on output RFI filters?
Charles Hansen said:
Well if they recorded in a high RFI environment much worse might be expected. I'm reviving an old experiment with transformerless ribbon mikes. I could not get a good answer on the noise limit posed by the air load and the resistive noise limit is absurdly low. I'm aiming at .1nV/rt-Hz and some fun bragging rights. A "hobby circuit" of course.
scott wurcer said:
Your right- the Quan-Tech has beads in it directly behind the socket, both on the older and the newer ones. However at microwave frequencies 2 pF is a substantial load. At 2 MHz or even 100 MHz its not a lot so additional capacitance may be necessary. I have built fixtures for the Quan-Tech without beads for Jfets without proiblems.
A microwave transistor can oscillate with just a socket in the way, and no external gain- that's why they come in stripline packages.
If lowest eN is what you are looking for at audio frequencies then any resistance in series is a problem. There are some ferrites that have significant loss at lower frequencies without getting too large. A Fair-Rite Fair-Rite 43 or 73 based bead might work but a lot depends on what is causing the interference.
Edmond Stuart said:
Hi Edmond,
You are right. As I pointed out in an earlier post, the RC network suggested at the RCA connector is definitely not the only one that should be in the path to the input stage. It is just the first line of defense for the very high frequencies. At the amplifier-proper, in close proximity to the input stage, there should be additional RC lowpass filtering, and this will well-define the impedance there as you suggest.
Cheers,
Bob
Charles Hansen said:
Hi Charles,
I'm not disagreeing with you, and I do not use ferrites or ferrite beads. However, let me ask some probing questions here.
Have you ever measured distortion or any other signal anomoly when a signal is passed through a circuit containing a ferrite bead that went away when the ferrite bead was removed?
What do you think the physical mechanism is for ferrite causing a sonic degradation? And why does that mechanism not result in something that we can measure, perhaps by non-conventional means?
Does the ferrite bead thing just fall into the category of perceived sonic differences that we just don't understand, what I call the X factor?
Thanks,
Bob
Re: FET/BJT
I'm going for FET but I have enough experience to know that creating that extremely low transconductance and operating at a net >100mA drain current is asking for 100MHz+ oscillations especially with everything spread around like it has to be.
Basicly I don't think you can build this thing without some gate stoppers and in this case I can't stand even a few Ohms of resistance so I want to try the SMD beads. The ones I ordered are .3 Ohms DC so I should be OK.
PHEONIX said:
I'm going for FET but I have enough experience to know that creating that extremely low transconductance and operating at a net >100mA drain current is asking for 100MHz+ oscillations especially with everything spread around like it has to be.
Basicly I don't think you can build this thing without some gate stoppers and in this case I can't stand even a few Ohms of resistance so I want to try the SMD beads. The ones I ordered are .3 Ohms DC so I should be OK.
scott wurcer said:
.1 nV/rt-Hz is about 20 dB better than almost any single transistor.
You could calculate back somewhat if you know the turns ratio of the transformer. My Lundahl catalog lists a ribbon mike transformer as having a 1:1:1:1:37 ratio and a .2 Ohm primary (X4) resistance. The intrinsic noise of the ribbon is much lower than the ambient noise floor. (The same issue is present in moving coil cartridges.) (The obsessive way would be to put the mike in a vacuum to see what the electrical noise floor is.)
With a ribbon the input bias current may be enough to push the ribbon out of the gap if you don't take it into account. And every passing breeze may slam the system to the rails.
I'm interested in what you learn since a ribbon mike preamp is on my to-do list as well.
Bob,Bob Cordell said:
Dimitri has, although this might seem to be more obvious case (speaker line):
http://www.diyaudio.com/forums/showthread.php?postid=1410537#post1410537
- Klaus
Bob Cordell said:
No, I gave up trying to measure those kinds of things many years ago. I would spend days trying to find some measurement that would correlate to something that was blatantly obvious in a listening test and come up empty handed. I realized I would be a lot better off spending that time just doing listening tests.
Bob Cordell said:
It's clearly something to do with magnetism. Here's how the listening tests went:
a) We took a ferrite ring and made a common-mode RFI filter for the AC mains by winding all three leads around the ring. Probably about 3 turns or so. We were keeping all the leads well spaced to minimize capacitive coupling. The ring was probably about 3/8" cross-section and about 2-1/2" diameter.
b) When the filter was installed in the power amp, things sounded better. There was less grain, a "quieter" perceived noise floor, and a more musical overall presentation. Exactly the kinds of things that you find by reducing RFI in any piece of audio equipmentn.
c) After about a year we revisited these listening tests. Then we found that taking the ferrite out resulted in *better* overall sound. Removing the ferrite brought back some of the problems caused by RFI, but it also got rid of a glassy hardness that made the amp sound "shouty" and annoying.
d) The sound with the ferrite in reminded me of listening to film capacitors that had steel (copperweld) leads. So I got a length of galvanized cold water pipe and laid it next to the speaker cable. The glassy hardness and shouty quality returned.
e) At this point it was pretty clear that hysteresis due to magnetic materials was probably the culprit. So we got a bulk tape eraser and demagnetized the ferrite. This solved the problem -- the hard and shouty quality was gone, and the good sonic qualities of reducing the RFI were retained.
f) The only problem was that within a day you could hear the ferrite becoming magnetized again. After two or three days, the hardness and shoutiness was back in full force.
I thought about building in an automatic degausser (like on a TV set), but decided a different approach would be better. So I thought about it for a while and invented a device that does the exact same thing as a ferrite (absorbs RF energy and turns it into heat) but was made completely from non-magnetic materials. This gave us the best of both worlds -- the quietness and lack of grain resulting from eliminating the RFI, yet without the glassy hardness that ferrites eventually induce.
For a while a lot of people used ferrites in many different applications. It was especially common to see them on power cables and interconnects, usually under a heat-shrink jacket. Once I was tuned into the sound that ferrites produced, I could never stand to listen to a product that used them and have avoided them ever since.
Could you measure the effect? Probably, given enough time and enough equipment. I'm sure that it would be very small and difficult to measure. Then even if you found something people would argue that the effect was too small to be audible. I don't play those games any more. I can learn a lot more by performing carefully controlled listening tests. Sometimes I even learn something by reading these forums...
Charles,Charles Hansen said:
do you think it is possible that the choke windings weren't arranged 100% perfect to achieve 100% common mode function which then with any DC on the mains line could have been responsible for the re-magnetizing? I cannot think of any other mechanism big enough in magnitude to achieve this. "Ground" currents to other devices I would expect to be too low in magnitude and those should be AC anyway.
And on that non-magnetic solution, is there a patent # / other info you are at liberty to share or is this proprietrary?
- Klaus
Re: Re: FET/BJT
As noted in my other posts, I would strongly recommend avoiding the ferrite beads. There are a few things to check:
a) Try building the circuit and seeing if you really do have an oscillation problem before trying to cure something that may not exist. I've had very good luck with JFET's and stability.
b) Gate stoppers may not really present a problem in the real world. Sure the noise figure may go up compared to a few hundred milliohms of thermal noise from the ribbon, but the real question is what is the voltage sensitivity of the ribbon element? In other words, if you can get 10 mV of AC with an 80 dB SPL source, you can pretty easily make a preamp that has 1 uV of input referred A-weighted noise using a single pair of FET's and gate stoppers. That would give an equivalent noise SPL of 0 dB, which would be fine.
A similar situation occurs with phono stages. It's not necessary to approach the thermal noise of the source impedance, it's enough to make sure that the noise is well below the signal voltage. (Of course, phono stages have an advantage here, as the RIAA equalization can be used to reduce the noise level in at least some topologies.)
scott wurcer said:
As noted in my other posts, I would strongly recommend avoiding the ferrite beads. There are a few things to check:
a) Try building the circuit and seeing if you really do have an oscillation problem before trying to cure something that may not exist. I've had very good luck with JFET's and stability.
b) Gate stoppers may not really present a problem in the real world. Sure the noise figure may go up compared to a few hundred milliohms of thermal noise from the ribbon, but the real question is what is the voltage sensitivity of the ribbon element? In other words, if you can get 10 mV of AC with an 80 dB SPL source, you can pretty easily make a preamp that has 1 uV of input referred A-weighted noise using a single pair of FET's and gate stoppers. That would give an equivalent noise SPL of 0 dB, which would be fine.
A similar situation occurs with phono stages. It's not necessary to approach the thermal noise of the source impedance, it's enough to make sure that the noise is well below the signal voltage. (Of course, phono stages have an advantage here, as the RIAA equalization can be used to reduce the noise level in at least some topologies.)
KSTR said:
I don't really know. Anything is possible. But if that is what is causing the problem (and I don't think it is, because you can hear the sonic problems with ferrites even when using a tubular "bead" with zero turns), it would be pretty impractical to make sure that each choke winding was wound "100% perfect".
KSTR said:
No, unfortunately we invented that device around 1998 when the currency crisis in the Far East caused many high-end companies to go out of business. We didn't have enough money then to patent the idea. It's too bad because I think the device could be mass produced and be cheaper than ferrites. Then the computer industry would want to buy them. Just think how rich we would be if we made even $0.01 on every computer made!
I'm sorry... but I am often times confused!
I thought that the stuff in AC mains is AC, and that is what demagnetizes things that have become magnetized? Where is the source of the DC that magnetized/biased the ferrite core? Or is there some other effect going on here??
And, are there non ferrous things that absorb RF other than a coil of wire acting like an inductor?? Did I read that right? Seems like unless this got invented already and I am missing something that this is an invention that is worth a huge amount of $money$??
I'm sure it all makes sense, just that I don't know about it.
_-_-bear
I thought that the stuff in AC mains is AC, and that is what demagnetizes things that have become magnetized? Where is the source of the DC that magnetized/biased the ferrite core? Or is there some other effect going on here??
And, are there non ferrous things that absorb RF other than a coil of wire acting like an inductor?? Did I read that right? Seems like unless this got invented already and I am missing something that this is an invention that is worth a huge amount of $money$??
I'm sure it all makes sense, just that I don't know about it.
_-_-bear
Charles Hansen said:
Hi Charles,
Thanks for this thoughtful answer. There is a lot of interesting stuff to ponder here.
Two things in particular jump out, if I read you right:
First, the fact that this was on the AC side of the power supply, which is not directly in the signal path. This then raises the question of whether what the ferrite was doing to the sound was as a result of it somehow distortiong the AC, failing to filter something coming in from the outside (or modifying in some way that garbage), or, finally, whether the ferrite was causing some kind of magnetic field of junk to be coupled into the signal path of the amplifier.
Second, the suggestion that the sonic degradation was not from the ferrite, per se, but rather an accumulated magnetization of the ferrite.
Cheers,
Bob
- Status
- Not open for further replies.