Really? Suppose my signal has a maximum level of 2 volts and the noise spikes can go to 12 volts, would not a diode clipper set to just above 2 volts attenuate the noise spikes?
Philosophically this is not the same problem. Of course in any case the clipping also removes/degrades the desired signal. During clipping there is no information and everyone including Shannon is happy.
I could add some of the most secure communications are achieved by xoring cosmic noise with the data the two not being seperable by any reasonable process.
Last edited:
Hi Scott,
Sheesh!
-Chris 🙂
You just had to bring that up! Another avenue for John to make reference to.
Sheesh!
So, this has been developed to a high level in the form of record tick and pop suppression circuitry. Just bring back some old technology and you're golden!
-Chris 🙂
AVE...
I used once unshielded speaker cable to fix my headphones. I've heard every power transformer and every fan in my room. I tried ferrite beads and it didn't help too much - still there was more noise than signal. When I used shielded cable, there were no sound problems, even when I placed them over working power transformer or fan...
If you have noise spikes up to 12V, then start using shielded cables and stop creating stupid problems...
I used once unshielded speaker cable to fix my headphones. I've heard every power transformer and every fan in my room. I tried ferrite beads and it didn't help too much - still there was more noise than signal. When I used shielded cable, there were no sound problems, even when I placed them over working power transformer or fan...
How would you be sure there were no noise spikes below 2V? Noise tends to be gaussian in distribution. And when you attenuate the noise, its signal+noise that you attenuate so when a noise spike of 2V occurs on top of a signal at 1V you've still got 1V of noise.
I find myself in agreement with Scott, you're postulating Maxwell's demon here - the distinction between noise and signal is purely a mental one.
Have we fallen down the rabbit hole with Alice?
Noise is of course directional in the sense that it is a power flow. Cables etc. are not directional. To make a passive system directional you have to use fancy things like magnetically biassed ferrites. Then you are really exploiting electron spin, which is a quantum phenomenon. (Ooops - I just woke them!) Cables use electron charge, not spin.
As I said before, a device which can distinguish between signal and noise has to either exploit frequency difference or act as a Maxwell demon. I vote for the former.
Noise is of course directional in the sense that it is a power flow. Cables etc. are not directional. To make a passive system directional you have to use fancy things like magnetically biassed ferrites. Then you are really exploiting electron spin, which is a quantum phenomenon. (Ooops - I just woke them!) Cables use electron charge, not spin.
As I said before, a device which can distinguish between signal and noise has to either exploit frequency difference or act as a Maxwell demon. I vote for the former.
Bingo! (let me know if that doesn't translate well) Keep in mind both the source and the receiver produce noise! As the receiver often has a higher input impedance connecting it to the source can lower the output noise.
Reducing the noise not eliminating it doesn't violate the rules.
Noise often does have different characteristics that the signal.
Using noise as code is not the same issue. You have already processed it to have the same parameters as the signal.
Last edited:
No, the converse is true. As the source often has a lower output impedance connecting it to the receiver can reduce the input noise.
If the frequency range is different then a simple filter will do the trick. If you want to do cleverer things like exploiting autocorrelations then it is difficult to see how this can be done with a simple (i.e. non-intelligent) device. Maybe the 'purifier' is a quantum computer? (Joke!!)
No, We are saying the same thing just in a different order.
Depends on the noise type 1/f has increasing amplitude with decreasing frequency so a clamp will produce some results. Of course removing out of band noise with a filter is also useful. So if you can do that with a very very small DC resistance you just might have something.
No, input noise is noise generated in the input stage of a device. This can often be reduced by resistive loading from the previous stage. To reduce output noise, as you claimed, you would need a low input impedance in the following stage.
I don't think a clamp will help with 1/f noise. The reason is that although the noise spectral density goes up as frequency goes down, the available bandwidth across which the noise is developed also goes down. You could get an infinitely large signal, but you would have to wait an infinitely long time to see it. Even devices with significant 1/f noise (e.g. valves) don't usually have large jumps in their output which a clamp could limit, without severely limiting the signal too.
I don't think a clamp will help with 1/f noise. The reason is that although the noise spectral density goes up as frequency goes down, the available bandwidth across which the noise is developed also goes down. You could get an infinitely large signal, but you would have to wait an infinitely long time to see it. Even devices with significant 1/f noise (e.g. valves) don't usually have large jumps in their output which a clamp could limit, without severely limiting the signal too.
Sorry for the misunderstanding "Output" of the system not the preamp-amp junction.
As to the clamp this was very popular in automotive systems before the introduction of resistor wire or plugs. Very useful for mobile radio.
But having actually measured systems there are every so often spikes that can be clamped with good results.
It is time to ask about the "Let's Make a Deal" problem. Actually on topic! In the show the participant is ask to choose between three doors behind one of which is a valuable prize. After they choose the host shows what was behind one of the doors that was not chosen. It is always empty of a valuable prize. They are then allowed to change from the door of their first choice to the remaining door.
Those who do not understand the problem insist that although the chance of getting the prize at first was 1/3 after the reveal it has risen to 1/2 if they do not change doors. Those who get it know to always switch because the odds are 2/3 that the prize is behind the door they did not choose.
Why? And how does this affect designing a filter to separate signal from noise?
Those who do not understand the problem insist that although the chance of getting the prize at first was 1/3 after the reveal it has risen to 1/2 if they do not change doors. Those who get it know to always switch because the odds are 2/3 that the prize is behind the door they did not choose.
Why? And how does this affect designing a filter to separate signal from noise?
I agree. But that is not what the discussion is about.
When the clamp is active, all information is lost. For the duration of the clamping period, the information which is "lost" is the total energy of the spike plus some of the energy of the signal of interest.
The only information which is available during the clamping is the voltage level of the clamp, and the event duration. Nothing else gets through.
No law violation, lots of spike energy removed, and some of the signal as well.
Cheers, John
Ed:
It looks like the ringing has at least 10 cycles within a roughly 1 uSec timeframe. That would make the ringing approximately 10 Mhz. If you are saying that you have a slight lowering of this 10Mhz content, then you are indeed working in the RF range. Not an easy thing to do accurately.
What L would be required to give this result at 10Mhz?
Can you explain what the pic is showing? It says a peak to peak in the 11 volt range, with a 2 uSec period.. The time scale is 5 uSec and the vert scale is 5 mV, so I could use a little help with the explanation. Thanks.
Cheers, John
- Status
- Not open for further replies.