VLF Transmission Using Soundcard

[sophiecentaur]

You are free to try, of course, but do you really have a tried and tested strategy? Just saying that you will use the "digital domain" could mean absolutely anything. Is there a suitable algorithm that you can use which will do all you require? I'm not sure.
Have you done anything towards working out the SNR you will need for this processing and the data rate you can achieve? That's the bottom line.

 

[sru2]

You are free to try, of course, but do you really have a tried and tested strategy? Just saying that you will use the "digital domain" could mean absolutely anything. Is there a suitable algorithm that you can use which will do all you require? I'm not sure.

Do you have experience with DSP? I am writing software that should be able to handle this.

Have you done anything towards working out the SNR you will need for this processing and the data rate you can achieve? That's the bottom line.

I started on this tonight. I hooked up a 500G roll of magnet wire to my sound card and took a look at the background noise of the system.

This first diagram shows the complete background noise for 0-20KHz.

To test the magnet wire, I brought an electric razor close to the wire (about 5cm) and slowly withdrew it (to about 30cm). The signal drop off was rapid as the distance increased.

I started to look for additional sources of noise and I found that I could create noise by refreshing Firefox.

Moving a laser mouse around also produces detectable noise.

Playing sounds through a Java application did not produce any noise, but playing an MP3 in Windows Media Player did. It would tend to suggest that the sound card was not the source of the interference, but I have yet to isolate the source. What is interesting is that this signal has a electronic signature revealing what song is being played and specifically what part of the song is being listened to. This should be of value to anyone interested in TEMPEST.

Finally the same song played through Youtube also reveals a unique signature that betrays information from your PC.

All of this is done unamplified in any way. The keyboard also produces unique signals, but it was too faint to be of much use in a small picture on this forum.

So noise at the receiver does not appear to be much of an issue. The next real question is how much energy does the receiver detect from a nearby transmitting source.

It looks like the MDS will be around -80dBm above 2Khz and that there is a 5dBm difference in noise between day and night. Given this, it may be best to amplify the transmitter, rather than the receiver.

 

[sophiecentaur]

Do you have experience with DSP? I am writing software that should be able to handle this.

I started on this tonight. I hooked up a 500G roll of magnet wire to my sound card and took a look at the background noise of the system.

So noise at the receiver does not appear to be much of an issue. The next real question is how much energy does the receiver detect from a nearby transmitting source.

It looks like the MDS will be around -80dBm above 2Khz and that there is a 5dBm difference in noise between day and night. Given this, it may be best to amplify the transmitter, rather than the receiver.

As you don't know the level of signal that you expect to receive, how can you say this? It's the ratio of signal to noise power that determines the performance. You do not know what level of signal to expect yet.

This is the basis of all communications engineering. Amplifying a signal that already has noise or interference added is very little use. You need to use as much TX power as you can afford.
I really do urge you to establish just how much signal you can get across the distance you need. Remember, your interfering sources may be right next to the receiver but your wanted source may be 100X as far away.

 

[sru2]

I moved the magnet wire roll out of range of the magnetic fields of my equipment. I am now receiving VLF and ELF stations, from the US, Germany, Britain and Russia. This includes extreme narrow-band Tacamo broadcasts.

If I am receiving this, some of which is 3000+ kilometers away, I should be able to hear a broadcast within 20m very clearly.

 

[sophiecentaur]

Do you know what Tx powers these foreign stations are using or what their field strength is at your receiver? Their field strength drops off at a much lower rate than your near field source will.
All I can say is that, without measuring signal level from the transmitting equipment, at the distance of interest, it could be wasting a lot of effort on other things. You can't lose by trying it out. If it's really as trivial a problem as you think then you may as well just do it early on. I cannot think of an example of radio communications design in which the actual signal levels were not considered very early on in the process. Perhaps you already know the carrier to noise ratio you need for a given error rate, using your proposed system - in which case, your interference measurements will tell you just what level you will need.
Your application may be a very specific one but how can you guarantee that there will be no local sources of interference when you use your system in an arbitrary location?

 

[sru2]

sophiecentaur, my signal generators arrive today, so I should be able to get some tests done sometime tomorrow.

I have one question about safety though and I was hoping you could clarify it for me.

I have been thinking about this near-field transmission. Am I right in saying that if I generate 3Khz sine and push it into an antenna two components will be generated? Firstly, an EM field localized to the antenna (near-field) and a free standing electromagnetic wave (radiated).

Let's leave how much is radiated to one side for the moment.

This time-varying magnetic field (near-field) will extend about 100Km at 3Khz. At close proximity the line of force are densely concentrated so induction is optimal. At 90Km the lines of force have diverged enough to make induction very weak.

Assuming a near ideal antenna exists within 100Km range (for 3Khz) and that it is directly connected to ground, at what distance from the transmitter does it become safe to begin broadcasting without this antenna causing a short? Could this antenna draw to much of a load? Would I need to make it high impedance or introduce a quick blow fuse?

Is there a formula for the divergence and the load it could draw?

 

[sophiecentaur]

The sig generators should give you some fun! Good luck with them; they should give you a really good feel for what's going on.

I'm not sure what your concern is about the effect of a distant structure. How could something so far away possibly couple with your transmit antenna enough to worry you? You will have a hard enough job getting usable signals across your small gap with low power sources so the effect of your hypothetical 'parasitic' cannot be a problem. Such structures can be a nuisance to receivers in their vicinity because they can modify the fields locally to them and cause a shadow / hole in the reception near them. Steel framed buildings / conurbations can play hell with the mf ground wave reception on rare occasions but that's another matter.

When working out the fields due to a radiator, you get several terms. One (/some) of them starts of with high values but are subject to high drop-off (1/r² and 1/r³) and there is on with a 1/r term which is the one that survives at great distance. The terms add together at all distances so there is a smooth transition between the regions. Where 1/r dominates is the one referred to as far field. Look at the Wiki pages about near and far field.

At 90Km the lines of force have diverged enough to make induction very weak.

is just stating the above in an arm waving 'geometrical' way.

 

[sru2]

I'm not sure what your concern is about the effect of a distant structure. How could something so far away possibly couple with your transmit antenna enough to worry you?

I was thinking the same thing. I thought it would be best to double check first.

 

[sru2]

If I have a transmitter with 5W, which produces an ERP of 50W, where does the additional power come from?

Is the gain supplied by a separate power source?

 

[vk6kro]

An transmitter plus an antenna would have an ERP of 50 watts with 5 watts into the antenna, not just a transmitter.

What it means is that in the same direction, you would need 50 watts to a dipole (used in its best direction) to give the same signal strength as the 5 watts produces with this antenna. So, the antenna has a gain of 10 dB.

There is still only 5 watts total, so you couldn't fry an egg with it or anything like that.

You could try reading this article:
http://www.w8ji.com/radiation_and_fields.htm
especially the big formula near the bottom quoted from Terman's book.

Look at the effect of frequency on radiation. Notice that frequency and wavelength are both used in the equation.

 

[sophiecentaur]

ERP (Equivalent Radiated Power) is a term used for transmitting installations and includes the power of the transmitter and the directivity of the antenna.
Antennae are never omnidirectional. They all have nulls in some direction and 'maxs' in others. A simple dipole will have a maximum radiated power around the 'equator' and zero at the poles.
Your 5W transmitting system will be putting most of its 5W in one general direction so that, on the main beam, you have the equivalent of a 50W transmitter (Equivalent Radiated Power). It is said to have 10dBi gain - that is to say 10 dB of gain relative to an isotropic antenna.

You won't be getting this sort of gain from a VLF antenna of any practical size.

[Edit: Rats, you got there before me. I seem to remember ERP being relative to an isotropic radiator and not a dipole? It's been a long time though. There's about 2dB of difference.]

 

[vk6kro]

I think it is EiRP for isotropic.

I checked with Wikipedia and they said that it was always for a dipole in the USA for FM transmitters. I had to look it up, though, because it isn't something you use every day.

Easier to explain than an isotropic antenna. I'm still looking for one of those.

 

[sru2]

Thanks for that guys, its been very helpful.

I have been focused on extending my visible bandwidth, lowering noise and removing mains hum for the last few days.

I figured out that the drivers with a sound card restrict the bandwidth to the audible range, up to 27Khz, before applying a high pass filter. By moving to ASIO to talk to the card directly on the line-in, rather than the microphone, I was able to get a visible bandwidth of 48Khz with a sampling rate of 96Ksps.

Applying a tracking filter I was able to strip most of the harmonics from the mains signals. That said, in doing so, I discovered secondary transmissions layered under the harmonics. Beginning at 150Hz, they are spaced around 300Hz apart. Given that I am running half a million samples through the FFT with a 96Ksps input, I am getting one pixel added to the spectrogram every six seconds. So, its difficult to tell if they are modulated, but distinct gaps are appearing so I assume so. I really need a good ADC and bandpass filters.

I am still running unamplified, but I am starting to see very weak signals emerge around 20Hz, 80Hz, etc. It looks like submarine comms. I'm also getting a lot of lightning strikes being detected. My noise floor is around -130dB at the minute, but that seems to be the card itself, as far as I can see I am dragging no noise in from the receiver whatsoever. This seems like a good approach, as I increase sensitivity I am able to eliminate noise as I go.

With respect to noise, have a look at the following diagram. This is around 26Khz, but I am seeing similar signals from around 1Khz. As I progress up the bands, I find random squiggles, but they appear to be confined to areas where the hum filter has reduced the dB. I think the automatic gain control is revealing them. As we approach the mid-20Khz, the signals start looking like the picture, rather than being confined to the edges of the purple bands that you can see (that's the hum filter).

Any idea what these signals are?

As a final question, I was thinking about noise reduction at the receiver. I know that one way to reduce thermal noise is to cool the antenna, cables and sampler. I was thinking of another way. If I suspend the antenna in a magnetic field, then it should apply a force to the electrons reducing their thermal movement. This should, in theory, reduce the noise floor and the intensity of that field will dictate by how much. The idea is that incoming signals of weaker intensities will then be detectable.

Has anyone tried this? What was your experience?

 

[the_emi_guy]

I was thinking about noise reduction at the receiver. I know that one way to reduce thermal noise is to cool the antenna, cables and sampler. I was thinking of another way. If I suspend the antenna in a magnetic field, then it should apply a force to the electrons reducing their thermal movement. This should, in theory, reduce the noise floor and the intensity of that field will dictate by how much. The idea is that incoming signals of weaker intensities will then be detectable.

Has anyone tried this? What was your experience?

Force on moving charge within magnetic field is perpendicular to both the field and the direction of motion, F = q(v x B). It will not slow it down or create drag, just change its direction (look up cyclotron frequency).
Using a preamp with low noise figure will insure that the electronic noise will be primarily thermal noise (which you really can't get rid of without serious expense).

 

[sru2]

Ok, I found this statement:

The cyclotron frequency or gyrofrequency is the frequency of a charged particle moving perpendicular to the direction of a uniform magnetic field B (constant magnitude and direction). Since that motion is always circular,[1] the cyclotron frequency is given by equality of centripetal force and magnetic Lorentz force

To me, that implies if I have a single bar magnet, the particles will flow along the lines of force in the direction opposite to their charge. Correct?

Now what happens if I have a second magnet, reversed, parallel to the first magnet?

That is, they are being accelerated in one direction by the first magnet, then accelerated in the opposite by the second.

If the forces are equal, then I should be able to "pinch" the particles and hold them still. That "pinch" will be proportional to the field strength.

Now as I understand it, this should drag down the signal strength right across the bandwidth, but will it also lower the thermal noise floor and, in doing so, increase sensitivity to weaker signals?

 

[sophiecentaur]

I think it is EiRP for isotropic.

I checked with Wikipedia and they said that it was always for a dipole in the USA for FM transmitters. I had to look it up, though, because it isn't something you use every day.

Easier to explain than an isotropic antenna. I'm still looking for one of those.

Ah yes, of course - EiRP!
Thanks for stirring my memory.

 

[sophiecentaur]

As a final question, I was thinking about noise reduction at the receiver. I know that one way to reduce thermal noise is to cool the antenna, cables and sampler. I was thinking of another way. If I suspend the antenna in a magnetic field, then it should apply a force to the electrons reducing their thermal movement. This should, in theory, reduce the noise floor and the intensity of that field will dictate by how much. The idea is that incoming signals of weaker intensities will then be detectable.

Has anyone tried this? What was your experience?

Did you stop and wonder why you have never seen 'noise reducing antennae' for sale, with permanent magnets attached? I think you are mixing your models up a bit. Adding a field will not produce damping - there would just be a possible curvature in the paths of electrons. If it would work, everyone would be doing it instead of needing high power transmitters and helium cooled parametric converters at the other end.

By and large, the main problem at low frequencies is interference and not thermal noise in the receiver. I can't think of any way that you could actually be able to distinguish what you seem to think is thermal noise from a floor of received interference and external noise. I am not aware of any use of cooled front ends in low frequency receivers. Read around the subject of man made noise at various frequencies. The audio bands are chock full of interfering EM, carried on mains wiring etc. BUT it is very low compared with the signal levels used in audio (no coincidence there, of course - just a pragmaatic engineering approach).

You seem to be identifying a lot of low level signals. How can you be sure they are not just artifacts of your system? I should also like to know what equivalent receiver bandwidth you are measuring with. How long (processor time) does it take you to identify these very low level signals?
Also, it would be interesting to know how much wanted signal power will be turning up in this bandwidth. Have you done any actual coupling EM measurements yet? You can be sure of nothing until you have found the Signal Level so that you can then work out the SNR, which is the bottom line.

 

[sru2]

Did you stop and wonder why you have never seen 'noise reducing antennae' for sale, with permanent magnets attached? I think you are mixing your models up a bit. Adding a field will not produce damping - there would just be a possible curvature in the paths of electrons. If it would work, everyone would be doing it instead of needing high power transmitters and helium cooled parametric converters at the other end.

In principle it would work, the issue is that it requires an intersection in the fields at every point along the length of the antenna. That's just not feasible without somehow integrating it into the antenna design. I'll play with the idea when I have time.

By an large, the main problem at low frequencies is interference and not thermal noise in the receiver. I can't think of any way that you could actually be able to distinguish what you seem to think is thermal noise from a floor of received interference and external noise. I am not aware of any use of cooled front ends in low frequency receivers. Read around the subject of man made noise at various frequencies. The audio bands are chock full of interfering EM, carried on mains wiring etc. BUT it is very low compared with the signal levels used in audio (no coincidence there, of course - just a pragmaatic engineering approach).

I seem to be able to control the interference quite well. I think the key is in the receiver design. If you build a high gain antenna, it pulls in all sorts of stuff from all over the world. So, the idea is to control the reception enough to reject distant weak signals, but not so much as to reject local weak signals. This reduces much of the interference and a tracking filter removes mains hum leaving the spectrum relatively clear right down to the noise floor. Then all you need to worry about is getting your transmitter above the noise floor for a given distance. Its a good approach for getting a good SNR at low frequency over a short distance.

You seem to be identifying a lot of low level signals. How can you be sure they are not just artifacts of your system?

My system is digital, the likely sources of interference would be oscillators and they tend to maintain a distinct frequency and do not erratically ramp up or down in frequency. The only other potential source is the mains itself, dragging in frequencies, but as far as I am aware its a filtered system and we receive a pure sine wave. I was able to eliminate all the noise I was detecting from my system using a filtered line-in. I no longer get interference from playing Youtube, music, etc., so there is no cross-talk on the card. Finally, I loaded my system to test the PSU and the result were negative, it had no impact of the signals.

It looks like they are radiated and they are hidden behind harmonics to hide them. I applied a tracking filter to remove mains hum and another mains system at 25Hz (I have no idea where this came from) and applied adaptive gain. The signals you see in this picture below are around -230dB (yes, -230dB on a sound card at room temperature).

I have checked the Spectrum chart and this band is allocated to the military where I am.

I should also like to know what equivalent receiver bandwidth you are measuring with.

Effect of FFT settings with fs= 96.0000 kHz:
Width of one FFT-bin: 183.105 mHz
Equiv. noise bandwidth: 274.658 mHz
Max freq range: 0.00000 Hz .. 48.0000 kHz
FFT window time: 5.461 s
Overlap from scroll interval: 75.0 %

How long (processor time) does it take you to identify these very low level signals?

3 secs per pixel, for around 1000 pixels, so the total time is 3000 seconds or 50 minutes to get a full screen picture.

Also, it would be interesting to know how much wanted signal power will be turning up in this bandwidth. Have you done any actual coupling EM measurements yet? You can be sure of nothing until you have found the Signal Level so that you can then work out the SNR, which is the bottom line.

Not yet, I have been bogged down with other projects, so I am working at this in my spare time. I am trying to perfect the computer, receiver and filter setup first. The system simply won't work without it being as close to perfect as possible. I think -230dB unamplified is amazing sensitivity and its probably now at its limits. So, the next tests will be related to transmission.

 

[sophiecentaur]

The system simply won't work either if there isn't enough signal, either. I don't see either why you are reluctant to do the experiment or how you can be so confident it's not a problem. It is usually very high up in any comms system list of things to specify and check.

Incidentally, your figures of "-230dB" don't mean much unless you specify what they are relative to and in what bandwidth you are measuring.

If your system is associated with a processor, then you have no idea what signals that processor could be generating as it is running its operating system and all those overheads.

Do you have a reference about this idea of permanent magnets? You say "in principle" it should work - but what principle?

 

[sru2]

The system simply won't work either if there isn't enough signal, either. I don't see either why you are reluctant to do the experiment or how you can be so confident it's not a problem. It is usually very high up in any comms system list of things to specify and check.

I had to build the receiver, test it, fix issues with my sound card, suppress hum, find suitable frequencies for tests and reduce noise.

You can't test without a receiver.

Incidentally, your figures of "-230dB" don't mean much unless you specify what they are relative to and in what bandwidth you are measuring.

Its an audio line-in, with a bandwidth of 48Khz. So, its dBm which is the standard for audio input. The signal strength is 1e-26 W.

If your system is associated with a processor, then you have no idea what signals that processor could be generating as it is running its operating system and all those overheads.

Its been ruled out through tests, its not local. Its most likely a series of military comms channels.

Do you have a reference about this idea of permanent magnets? You say "in principle" it should work - but what principle?

Search for magnetic constriction and magnetic confinement. Its a similar principle as the outer layer of a conductor is effectively a plasma with a frequency in the x-ray spectrum. I think its a field of research on it own. :)

 

[sophiecentaur]

Search for magnetic constriction and magnetic confinement. Its a similar principle as the outer layer of a conductor is effectively a plasma with a frequency in the x-ray spectrum. I think its a field of research on it own. :)

So why doesn't this turn up in every commercial antenna? How can you think that you can make it work when it's not available on the market? It would be worth £££ to sell to thousands of customers and, if you can do it, why aren't they all doing it? Just because someone writes a paper which suggests that it may be possible, you can't rely on using the technique.
Every link in the chain needs to be sorted - I agree - but you can't take it for granted that the 'small matter' of signal coupling will be no problem. Expending all your effort in one direction could prove a waste if the bottom line is that you just can't get enough signal to your receiver. Antenna technology is a huge part of comms and no one would dream of putting a satellite into orbit or a transmitter on a hill if they hadn't already a lot of confidence in the level of signal it can lay down in the proposed service area.
You have a problem with this project in that you are researcher, designer, technical advisor, procurer and works manager. You really don't want to spend months and months on a project that fails because one vital part was not considered. I am trying to ensure that at least one part has been sorted out - to save you disappointment. You can take nothing for granted, here. The receiver situation cannot be quantified until you know the actual signal levels you will be dealing with. You mentioned a required data rate at the end of all this. That implies (at the end of a lengthy link budget calculation) a go-no-go conclusion about the viability of your scheme with the given power levels and antenna design. You have asked a lot of basic questions that imply some holes in your knowledge of the whole scenario. I have suggested some areas in which you need to do further study before you can be sure of viability. The problem won't go away without some theory or measurements. Measurements would not be difficult and they would give you some ball-park figures about likely performance.

 

[sru2]

So why doesn't this turn up in every commercial antenna? How can you think that you can make it work when it's not available on the market? It would be worth £££ to sell to thousands of customers and, if you can do it, why aren't they all doing it? Just because someone writes a paper which suggests that it may be possible, you can't rely on using the technique.

Such a system would behave like a resistor in terms of current flow. It would cause signal strength to drop, heat to develop and perhaps electrical noise. Thus, it is of no use in most systems. If you work with very weak signals however, it has the potential to resist thermal noise by applying a force to the electrons. This should lower the noise floor and very weak incoming signals have the opportunity to create a detectable voltage.

The difficulty lies in engineering the lines of force to produce such an effect and making the production cost effective. If it proves to be cheaper to freeze a system and you get better results, then obviously that would be the route to choose.

As I said earlier, its a research project on its own.

Every link in the chain needs to be sorted - I agree - but you can't take it for granted that the 'small matter' of signal coupling will be no problem...

The idea is to amplify the transmission until a link is achieved, that's why I'm not too concerned with it and focused on the reception equipment.

 

[sophiecentaur]

I'm afraid that magnet thing just doesn't make sense. It flies in the face of basic communications theory. Thermal noise power is non- negociable. You can't reduce it without reducing the signal too.

I think I will have to leave you to find out the hard way about this. Received signal level is as important as any other factor in system performance. Why is a set-top antenna not adequate in fringe reception areas?
Good luck.

 

[vk6kro]

I recall seeing a design for an "audio loop" as used in libraries.

This is a purely magnetic system where the speaker output of an amplifier is just fed into a single wire loop around a room.

Then, within this loop, suitable coils of wire pick up enough signal to be amplified and fed to headphones.

This was an article in a fairly reputable magazine, so I guess it works.

 

[sru2]

Did a little testing today to see if the noise I was picking up yesterday was local. After much searching, I traced the source to a digital TV.

I've been able to reduce my noise floor further, its now at -245dBm. The interference from the TV, as can be seen in the picture below, is around -233.1dBm.

I'm afraid that magnet thing just doesn't make sense. It flies in the face of basic communications theory. Thermal noise power is non- negociable. You can't reduce it without reducing the signal too.

That's not entirely accurate, freezing will work, if you look up the equation on minimum detectable signal you will see that the driving force is temperature. In the case of magnetics, you are correct. It will drag down the signal right across the spectrum. That said, it has a similar effect to freezing at very low current levels. Rather than dissipating the heat through conduction, the electrons lose some of their thermal energy (which is motion) by being trapped between two or more fields. Very weak signals now have more energy than the thermal noise and become detectable.

So, its a good trade off for weak signal detection, drag down strong signals, lower the noise floor and pull in weaker signals.

The problem part is engineering an antenna that could create the necessary fields. I don't know if that is possible, or how effective it would be. As I said earlier, the principle is correct though.

I think I will have to leave you to find out the hard way about this. Received signal level is as important as any other factor in system performance. Why is a set-top antenna not adequate in fringe reception areas?

I have a balancing act to maintain. With a sound card, the weaker the signal the better the frequency separation must be in FFT to detect it and filter noise. With 96Ksps, using and FFT with an input length of 524288 samples, means it takes 5.5 seconds to get an output. At this rate, assuming that I use one low output as off and one high outputs as on, it means that per channel I get a bandwidth of around 12 bits per minute. In practice due to noise, this will drop to 6 or 3 as bit flipping would be common and different modulation methods would need to be used. That can be increased by increasing the sampling rate, applying digital filtering, etc. This is why I will be using a modified form of FDM, as it will provide a faster throughput.

The transmitter is the easy part which is why I am unconcerned with it. If I can't detect it, amplify it a bit.

I recall seeing a design for an "audio loop" as used in libraries.

This is a purely magnetic system where the speaker output of an amplifier is just fed into a single wire loop around a room.

Then, within this loop, suitable coils of wire pick up enough signal to be amplified and fed to headphones.

This was an article in a fairly reputable magazine, so I guess it works.

Its also used by deaf people to transmit signals to their hearing aids, it works with both the phone and TV.

http://deafness.about.com/od/assistivelisteningdevices/a/audioloops.htm

I had to fix one years ago for a friend. The loop ran around her living room, into a small box, which had a microphone in it. The microphone picked up sounds in the room, turned it into an alternating magnetic field which drove her hearing aid.

If that works well, this transmitter should be fine.

 

[sophiecentaur]

Unless you have a proper reference to your magnet idea, I'm afraid that I won't / can't just accept it as a possibility. I appreciate that the noise temperature can be reduced by cooling but only when the noise is straightforward 'front end' noise. How can a magnet alter that? (And bear in mind that the significant noise is not generated in the conducting part (copper) of the antenna but in the input resistance of the receiver.

You do not need a 'posh' receiver to measure the coupling between two coils at close distances. You can use whatever it was that you used to measure all those incredibly low level signals that you quote earlier on. If you are using a CW source, that should be trivial for your software.

 

[the_emi_guy]

I'm curious how your code is coming up with these very low noise floors. -245dBm is the thermal noise power in a 8E-8 Hz bandwidth. You would have to average for almost a year to hit this kind of number at room temperature. For some context, -245dBm is about 100 femto volts rms at 50 ohms.

 

[sophiecentaur]

I was wondering if we are having the 20log10 and 10log10 situation. It's easy done.

 

[sru2]

I'm curious how your code is coming up with these very low noise floors. -245dBm is the thermal noise power in a 8E-8 Hz bandwidth. You would have to average for almost a year to hit this kind of number at room temperature. For some context, -245dBm is about 100 femto volts rms at 50 ohms.

So was I, without cooling the receiver the noise floor should have been around -175dBm. The software that you see is actually Spectrum Lab, I'm using it as a reference point whilst I complete my own DSP library (I am currently writing the spectrogram control). I made sure I updated to the latest version and then re-ran the test.

I also trawled through Spectrum lab's help files in relation to the Automatic Gain Control and found this:

One of the downsides of the visual AGC is, you cannot tell the signal strength (voltage, power, or whatever) from the colour in the spectrogram display.

Spectrum/html/specdisp.htm#visual_AGC

So, I disabled it and trawled the output for the lowest value. As far as I can currently see, its around -185dBm which is pretty good for a sound card.

Even with the vAGC control on, the value doesn't change much, so I am assuming there was a bug in the older version I was using.

I was wondering if we are having the 20log10 and 10log10 situation. It's easy done.

I honestly don't know the source of the bug, but it does seem to be corrected in the latest version of Spectrum Lab.

 

[the_emi_guy]

So was I, without cooling the receiver the noise floor should have been around -175dBm.

-175dBm is 0.4nV at 50 ohms. Soundcard ADC stepsize is typically 40uV. Unless you have a lot of gain in front of soundcard you are probably encoding mostly zeros. Can you look at the raw ADC values?

 

[sru2]

-175dBm is 0.4nV at 50 ohms. Soundcard ADC stepsize is typically 40uV. Unless you have a lot of gain in front of soundcard you are probably encoding mostly zeros. Can you look at the raw ADC values?

I can't see the raw ADC values from the sound card itself. I do know that I can see a difference between -185dBm and -175dBm in terms of the output in Spectrum Lab.

Assuming 0.447V peak input on the line-in, with a 24 bit ADC the maximum voltage representation will be 50 nV. Pushing that into an FFT with half a million samples, I'm not surprised that we can get a resolution of -185dBm per bin.

On a side note, I discovered another source of interference around 45Khz comes from WIFI.

 

[sophiecentaur]

Earlier on in this thread, you talked of a target of 56K (bit/s?) but now we are talking in terms of a few bits per minute. Where are we exactly in this and what is your actual intended data rate?

 

[sru2]

Earlier on in this thread, you talked of a target of 56K (bit/s?) but now we are talking in terms of a few bits per minute. Where are we exactly in this and what is your actual intended data rate?

That was the dream. I am unconcerned with the final data rate, it was never going to be a competitor to WIFI. The limiting factor is the speed of the FFT output at a given resolution. This is dictated by the sampling rate of the sound card. Faster FFT output means less channels, slower means more channels. I need to work out what gives me the most throughput.

Of course I can always change the sound card for a high speed ADC and put the throughput through the roof, but that's another project. Sound card first.

 

[sophiecentaur]

That was the dream. I am unconcerned with the final data rate, it was never going to be a competitor to WIFI. The limiting factor is the speed of the FFT output at a given resolution. This is dictated by the sampling rate of the sound card. Faster FFT output means less channels, slower means more channels. I need to work out what gives me the most throughput.

Of course I can always change the sound card for a high speed ADC and put the throughput through the roof, but that's another project. Sound card first.

I think you'll find that the limiting rate is imposed by Mr Shannon. That imposes an upper limit to the information capacity of a channel - however 'fast' your signal processing happens to be.
And, as I have mentioned before, it depends upon the signal to noise ratio. If you have no idea what the wanted signal level is likely to be then you have absolutely no idea what the limit might be. If your uncertainty is in the region of 30dB (not all that pessimistic, imo) that could make a 10:1 difference in possible information rate. (log base 2 of 1000).

I would still like some information about this magnetic noise suppression system.

 

[sru2]

I think you'll find that the limiting rate is imposed by Mr Shannon. That imposes an upper limit to the information capacity of a channel - however 'fast' your signal processing happens to be.

I have a problem with Mr Shannon's work, or perhaps it is an interpretation of it. Read this:

The Shannon theorem states that given a noisy channel with channel capacity C and information transmitted at a rate R, then if R < C there exist codes that allow the probability of error at the receiver to be made arbitrarily small. This means that, theoretically, it is possible to transmit information nearly without error at any rate below a limiting rate, C.

The converse is also important. If R > C, an arbitrarily small probability of error is not achievable.

How can R exceed C?

Let me provide an example. Let's assume I open a channel on 20Khz. I define 1 bit to be exactly 1/20000 per second, in other words, a single period. The maximum I can send information (i.e. R) is 1/20000 per second, but the capacity (i.e. C) is 20000 bits per second.

Now this may seem strange until I mention this part. What size is a photon?

A hertz is defined as cycles per second. So a 1Hz photon is the same size as a 20Khz photon, that is, they are both 1 light second long. The wavelength of 14,990m must be packed into that 1 light second. This means, at 20Khz, each period is:

299 792.458m / 14990m = 19.99m

The question then becomes, is a photon 1 light second long, or is it 19.99m long and 20000 represent a 20Khz signal?

Now if this seems odd, consider the following. If the photon was 1 light second long, then every signal, regardless of its frequency, would take the same amount of time to detect (i.e. at least 1 second). Thus, a photon must be a single period.

This means that Mr Shannon is wrong, or his work has been interpreted incorrectly.

Did he understand quantum theory?