Shielding boost converter for RF transceiver

[Buriedcode]

Hi,

I have several (ongoing, because I'm lazy) projects that are battery powered from a couple of AA cells, that use RF transceivers and/or recievers. Some with the CC1101, some with the RFM69HCW, and a few general basic 315/433MHz modules.

I've noticed that in many cases on my prototypes, the use of a boost converter stepping up the AA cells to 3/3 or 5V will either greatly redice the range of the RF link, or stop it working entirely.  I have used an LDO after the boost converter, as well as an inductor/cap filter to power these transceivers, but it is hard to know how well this works - things do seem to improve but nowhere near the performance of when it is powered by a higher voltage battery + linear reg.  I'm unsure how much is conducted noise through the power supply, and how much is radiated.

So, aside from looking for specialist low noise charge pumps - which aren't all that efficient - do you think it would be worth using a shielding can/RFI shield around the boost converter?  The bottom layer will be almost solid ground plane.  I'm even considering making up some PCB's to form modules - maybe 10x20mm with a shielding can to attempt to fully enclosure these, with only pins or cestellated holes for connecting VIN/VOUT. I haven't found existing modules that have the right Vin range (2-3V) and relatively low current (<100mA out).

I would just switch to a 9V and lienar reg, but I hate PP3, they end to be expensive and less capacity than AA's - which are available pretty much everywhere.  I could also go for a lithium pouch cell, but this doesn't really solve the problem since its voltage will still need a converter to provide power over its voltage range.

I guess my ain question is - is radiated EMI more of a problem for RF recievers than conducted?  We're talking single inductor sychronous boost converters Fsw range 300kHz - 1.2MHz, 2A Ipeak max. Typical single-cell-to-5v types.

[Phil1977]

Usually a small low power boost converter is just not big enough to radiate a lot of power by itself in the range <1GHz. But it can conduct a lot of EMI into its cabling and load, which again may be large enough to radiate.

That means: The first thing you should try is to place filters in the input and output of the converter. A good start would be a pi-filter in the input and output together with at least one common mode choke - usually at the input side, but if the input is only a small battery it could also be worth to try the common mode choke in the output line.

[Buriedcode]

Usually a small low power boost converter is just not big enough to radiate a lot of power by itself in the range <1GHz. But it can conduct a lot of EMI into its cabling and load, which again may be large enough to radiate.

As the batteries will often be connected to the PCB with wires - since the enclosure has provisions for AA battery clips, I guess these wires being anthing from 20 to 50mm could be part of the issue.  Honestly, it never really occured to be to look at the input even though thats where the high current switching happens.

That means: The first thing you should try is to place filters in the input and output of the converter. A good start would be a pi-filter in the input and output together with at least one common mode choke - usually at the input side, but if the input is only a small battery it could also be worth to try the common mode choke in the output line.

I think I'm gonna have to set up a small test bed for this.  I might not have the facility to measure EMI, but certain conducted noise, and I could do some basic range checks with different filters at different places. 

Cheers!

[Randy222]

Can you look at signal on something like tinySA ? "signal" as in actual signal in transmission line, and then clamp onto a piece of metal to see if EMI is there, or even place use a 314-433 antenna on a scope or tinySA placed near your device, see what shows up. Some time ago I notice my Rigol scope has a lot of noise around 1.5MHz. Turns out it's a local AM 1550 station.

[T3sl4co1l]

Perhaps more likely is differential noise, supply ripple, modulating the transceiver and mucking up its operation (voltage to frequency modulation i.e. it can't track a signal, or at least as well; amplitude modulation introducing sidebands and inter-channel images; etc.).  Higher frequencies may radiate from the device overall, but it's unlikely much harmonics are present in the several-100s-MHz range.  But who knows; varying amounts of those factors may all play a role in some configuration or another.

You can determine both, by putting the converter on a reference plane -- a solid piece of metal, usually a blank copper-clad PCB to make convenient soldered connections.  Put the converter there (whether module, or building a circuit directly on top e.g. Manhattan style), and mind where the connections go:
- Ground to plane, of course!
- Route input and output together, so they're co-located. If there is voltage drop along the plane, this minimizes how much of that voltage they pick up.  Note that, if connections come off opposite ends of a converter module, it's likely there's some (high frequency) voltage drop along it, which thus radiates into the wire pairs (when used as an inline module).
- Filter both input and output, with one or two stages of LC filtering.  Not much should be required, a couple uF and fractional-uH will suffice for high frequencies.  An electrolytic is a good idea as well, to provide ESR to dampen the filter (which otherwise doesn't have much termination resistance around it, which can lead to quite strong frequency peaks that cause problems for the module or connected circuitry).
- Note that a filter has a ground connection.  This is the reference against which the signals get filtered, i.e. for frequencies above cutoff frequency, they take on this potential instead.  You want input and output filters close together, to minimize the voltage between them, and you want them away from (not much straight-line distance is actually needed, just that it's not in the immediate vicinity of) the converter.
- Making the converter into a three-terminal module, common ground at a single point, eliminates common mode noise: by definition, there is only input and output (normal mode), there's no ground-to-ground voltage at all (or, it's trivially defined as zero, being the same point).

Also, be careful whether the module is common-ground or not.  A current-mode buck converter may have an output current sense resistor, and expects an isolated load.  You need to filter this separately as its own wire (same as input and output), rather than short it to GND plane.

Then you can put a shield over top the ground plane, which can be a few pieces of copper clad soldered together along the seams, to block most electric and magnetic field lines around the converter.

Example: a bunch of years ago I built this radio set,
https://seventransistorlabs.com/Radio_20m/Images/Radio_Wide.jpg
which uses a DC-DC converter (top-left) to take 12V and make 6.3V and 120V for the vacuum tube receiver.  The converter is contained within the five-sided copper-clad box.  The wires exit the open side, wrapping around to the lid where a little extra filtering is provided.  The receiver gets clean DC power and no noticeable conducted emissions couple into it; indeed, I can put an inductive loop probe on the antenna jack, and wave it around the converter and receive nothing, but sticking it inside the box picks up many... well, I don't have a measure of intensity, this receiver isn't calibrated, but needless to say the received power goes up by 20s of dB, and the intense crunch of white noise plays through the speaker.

Tim

[Buriedcode]

Can you look at signal on something like tinySA ? "signal" as in actual signal in transmission line, and then clamp onto a piece of metal to see if EMI is there, or even place use a 314-433 antenna on a scope or tinySA placed near your device, see what shows up. Some time ago I notice my Rigol scope has a lot of noise around 1.5MHz. Turns out it's a local AM 1550 station.

That antenna idea sounds interesting - I never really thought of what the reciever "see's" on its antenna, interms of SMPS noise.  Whilst not the same modules, I have noticed the fully shielded (as in RFI can) NRF240L01 modules I have are ok with SMPS, providing they have quite large bulk capacitance on their power inputs - where-as the cheaper unshielded ones barely reach 20m.  It's not a great comparison because a) 2.4GHz is a different band to the 434/315 and b) two different modules, by different manufacturers, with different layouts isn't exactly apples-to-apples.

Perhaps more likely is differential noise, supply ripple, modulating the transceiver and mucking up its operation (voltage to frequency modulation i.e. it can't track a signal, or at least as well; amplitude modulation introducing sidebands and inter-channel images; etc.).  Higher frequencies may radiate from the device overall, but it's unlikely much harmonics are present in the several-100s-MHz range.  But who knows; varying amounts of those factors may all play a role in some configuration or another.

Yes, this is the thing I'm trying to tease apart - how much is radiated, and how much is conducted through the power lines. I supply I could just put the battery + boost converter in a tin with just the power wires coming out, and see what that does.  It's a dirty approach, but combined with scoping the power lines to see just how much ripple - and any "fuzziness" there is (got a DS1054z so wont' see the HF stuff).   I suspect the layout of the protoype modules I've built are less than ideal - two of which are on single sided boards, albeit with a single sided 0.8mm copper clad board glued to the back and stitched to ground.  Reading up on large current loops between the input cap, L and switch shows that some of my layouts aren't great.

You can determine both, by putting the converter on a reference plane -- a solid piece of metal, usually a blank copper-clad PCB to make convenient soldered connections.  Put the converter there (whether module, or building a circuit directly on top e.g. Manhattan style), and mind where the connections go:
- Ground to plane, of course!
- Route input and output together, so they're co-located. If there is voltage drop along the plane, this minimizes how much of that voltage they pick up.  Note that, if connections come off opposite ends of a converter module, it's likely there's some (high frequency) voltage drop along it, which thus radiates into the wire pairs (when used as an inline module).
- Filter both input and output, with one or two stages of LC filtering.  Not much should be required, a couple uF and fractional-uH will suffice for high frequencies.  An electrolytic is a good idea as well, to provide ESR to dampen the filter (which otherwise doesn't have much termination resistance around it, which can lead to quite strong frequency peaks that cause problems for the module or connected circuitry).
- Note that a filter has a ground connection.  This is the reference against which the signals get filtered, i.e. for frequencies above cutoff frequency, they take on this potential instead.  You want input and output filters close together, to minimize the voltage between them, and you want them away from (not much straight-line distance is actually needed, just that it's not in the immediate vicinity of) the converter.
- Making the converter into a three-terminal module, common ground at a single point, eliminates common mode noise: by definition, there is only input and output (normal mode), there's no ground-to-ground voltage at all (or, it's trivially defined as zero, being the same point).
...
Then you can put a shield over top the ground plane, which can be a few pieces of copper clad soldered together along the seams, to block most electric and magnetic field lines around the converter.
..

Thats a lot to unpack, but my idea was indeed a "3 terminal device", albeit with extra connections for enable and a pin to select between two different output voltages (resistor to FB, to be grounded by host).  That should mean any IO filtering easier since, as you said, they share the same ground input/ouput, so the filtering components can sit right at those pins.

I know a shield is perhaps overkill, but given I was going to knock up some small converter modules anyway - adding the provision for a shielding can doesn't add to the cost of the PCB, and given the small size (~15 x 21mm?) the ones I've found aren't that expensive.

So it looks like I've got a few tests to perform to have some idea of how much is supply ripple, and how much radiated.. then the source of this noise, it's frequency compenents etc..  It's a lot of effort, and compounded by the fact I tend to use several different converters from several different manufacturers.  But perhaps I can pick just two for now.

I was surprised they don't exist as a product, but given the rather narrow application (single/double cell boost, <100mA, very low noise for radio applications etc..) and the fact that it seems half decent layout and filtering *should* be enough, I guess it just isn't worth manufacturering at a price point that would be practical.

At the risk of derailing the thread - anyone have any preferred boost convertes for such applications? We're talking low voltage input (1-3V), synchronous <100mA out.  I have quite a few MCP1640B's that work well, but really don't like anything over 50mA (@5V) from two cell input, and a few TPS61023.  But there are loads to choose from - too many to test.

[radiolistener]

do you think it would be worth using a shielding can/RFI shield around the boost converter?

Yes, it is crucial to use RF shielding around the DC/DC converter. The shielding helps minimize electromagnetic interference (EMI), which can affect other electronic components (such as RF receivers or sensitive ADC or PLL circuits). Also very important to use proper filters on both the input and output of the converter to reduce high-frequency noise that could propagate through the power lines.

Usually a small low power boost converter is just not big enough to radiate a lot of power by itself in the range <1GHz.

No, this is complete false claim. Even a small, low-power boost converter can radiate significant noise in the <200 MHz range, which can interfere with RF communications. This is especially true if the converter is not designed with EMI mitigation in mind and/or uses long wires without proper filtering close to the converter and/or lacks proper shielding.

The frequency band most affected by boost converters typically ranges from 20 kHz to 7 MHz. However, the impact can extend up to 150-200 MHz and beyond, as higher harmonics and switching frequencies can produce interference in these higher ranges. Boost converters can generate significant electromagnetic interference (EMI) within this broader spectrum, depending on their design and operational conditions.

It is practically impossible to completely eliminate EMI from a boost converter for radio receivers. However, it is possible to shift the boost converter's noise to unused parts of the spectrum and minimize it as much as possible using various techniques, such as filters and shielding. In this way, you can minimize the effect of the boost converter noise on your radio receiver.

[Phil1977]

Usually a small low power boost converter is just not big enough to radiate a lot of power by itself in the range <1GHz.

No, this is complete false claim. Even a small, low-power boost converter can radiate significant noise in the <200 MHz range, which can interfere with RF communications. This is especially true if the converter is not designed with EMI mitigation in mind and/or uses long wires without proper filtering close to the converter and/or lacks proper shielding.

I said it´s not emitting by itself, and physics strongly supports this statement. But no need to argue, we practically recommend the same thing. Filter the in- and output and -if necessary- shield the converter against near field couplings.

And it´s just nonsense that SMPS will always disturb transmissions. It´s only a question of how much engineering and components you invest into the filters. That´s why for many setups it´s just more reasonable to use linear converters, but with free choice of filter technologies and a solid metal backplane you can suppress all frequencies below the noise level of the most sensible receiver.

Edit: If you don't believe it, look at 6min30s clearly shows a more or less standard SMPS in a Rohde & Schwarz spectrum analyzer going from 5kHz to 3Ghz. And please don't tell me R&S is doing it wrong... 

[NiHaoMike]

I could also go for a lithium pouch cell, but this doesn't really solve the problem since its voltage will still need a converter to provide power over its voltage range.

Use LiFePO4 and you won't need a regulator to run 3.3V logic from it. The operating voltage range of a single cell just happens to be within what most 3.3V logic is specified to operate from.

[radiolistener]

I said it´s not emitting by itself, and physics strongly supports this statement.

No, this claim just demonstrates a lack of understanding of physics and practical RF experience. Even a battery-powered DC/DC converter with no external wires and galvanically isolated from any external connections still emits EMI noise. While this noise may be minimal and difficult to detect on a spectrum analyzer with a wide bandwidth span and a high noise floor, it can be easily picked up by a narrowband radio receiver, such as an AM broadcast receiver operating in the LW, MW, or SW bands. The fact that even isolated circuits can generate detectable EMI highlights the importance of proper design and shielding to mitigate these effects.

And it´s just nonsense that SMPS will always disturb transmissions. It´s only a question of how much engineering and components you invest into the filters.

The real nonsense is the belief that you can build a circuit that will never disturb a radio receiver. This is physically impossible. Even with an excellent design, including high-quality DC/DC converters, expensive filtering, and near-perfect shielding, EMI emissions can still be detected by a radio receiver with a narrower bandwidth and lower phase noise of it's local oscillator (LO). Furthermore, as technology advances, radio receivers become more sensitive and capable of detecting even lower levels of interference. This underscores the fact that while good engineering can significantly reduce EMI, it cannot entirely eliminate it.

While it may be possible (but very expensive) to design a DC/DC converter with good filters and shielding that minimally interferes with a broadcast radio receiver with audio output, creating such a buck converter for powering lasers used in laser spectroscopy is a very non trivial task. This is because it is much more sensitive to the noise. The strict ultra low noise requirements in such applications often necessitate the use of linear power supplies.

Edit: If you don't believe it, look at youtube 6min30s clearly shows a more or less standard SMPS in a Rohde & Schwarz spectrum analyzer going from 5kHz to 3Ghz. And please don't tell me R&S is doing it wrong... 

Your claim demonstrates a fundamental misunderstanding of physics and RF theory. Using a 3 GHz bandwidth results in such a high noise floor that you won't even detect the presence of relatively strong carriers from distant transmitters, even though these carriers would be clearly received by a radio receiver with a sufficiently high SNR.

Attempting to identify weak noise sources with such a wide 3 GHz bandwidth is impractical and scientifically unsound, which is why your example is flawed and frankly, laughable. 

[shabaz]

Attempting to identify weak noise sources with such a wide 3 GHz bandwidth is impractical and scientifically unsound,

Not read all your content (sorry, but you ramble a _lot_ and lots is often erroneous. I'm not trying to be rude, but it has to be said). That's not how a spectrum analyzer works. Take a look at the architecture of one, and compare and contrast with a radio receiver.

[radiolistener]

Not read all your content (sorry, but you ramble a _lot_ and lots is often erroneous. I'm not trying to be rude, but it has to be said). That's not how a spectrum analyzer works. Take a look at the architecture of one, and compare and contrast with a radio receiver.

Sorry, but if you believe that weak noise can be observed over a 3 GHz span, you are mistaken. This reflects a lack of understanding of how spectrum analyzers and radio receivers work, particularly how their bandwidth/span is directly linked to the noise floor. I recommend taking the time to carefully read and understand the principles behind these instruments to fully grasp how they operate.

Just try reading the theory carefully, without skipping sections of the text as you usually do. You’ll discover many new and interesting facts that you previously missed because you skim through and skip parts of the text. 

As a start point, I recommend considering why oscilloscopes with >=100 MHz bandwidth include a 20 MHz bandwidth mode and why exactly this narrow-band mode is specifically used for assessing power supplies noise. Then read carefully what is SNR, what is RBW, what is VBW, etc. And how selected bandwidth affects SNR.

[T3sl4co1l]

No, this claim just demonstrates a lack of understanding of physics and practical RF experience. Even a battery-powered DC/DC converter with no external wires and galvanically isolated from any external connections still emits EMI noise. While this noise may be minimal and difficult to detect on a spectrum analyzer with a wide bandwidth span and a high noise floor, it can be easily picked up by a narrowband radio receiver, such as an AM broadcast receiver operating in the LW, MW, or SW bands. The fact that even isolated circuits can generate detectable EMI highlights the importance of proper design and shielding to mitigate these effects.

At what distance?

The real nonsense is the belief that you can build a circuit that will never disturb a radio receiver. This is physically impossible. Even with an excellent design, including high-quality DC/DC converters, expensive filtering, and near-perfect shielding, EMI emissions can still be detected by a radio receiver with a narrower bandwidth and lower phase noise of it's local oscillator (LO). Furthermore, as technology advances, radio receivers become more sensitive and capable of detecting even lower levels of interference. This underscores the fact that while good engineering can significantly reduce EMI, it cannot entirely eliminate it.

While it may be possible (but very expensive) to design a DC/DC converter with good filters and shielding that minimally interferes with a broadcast radio receiver with audio output, creating such a buck converter for powering lasers used in laser spectroscopy is a very non trivial task. This is because it is much more sensitive to the noise. The strict ultra low noise requirements in such applications often necessitate the use of linear power supplies.

Curious, would you stake any money on that claim?

Serious; right here, right now. We can get a neutral 3rd party to escrow the money, and a small group together to judge the award.  That way there's no "well but I meant--" chicanery from either side, accept the result, done and done.  Waive right to lawsuit or arbitration, easy enough to handle the rules, in confidence, in public view here, make a little spectacle out of it, good times had by all, eh?

How much would you put on it? $10?  $100?

$1000? Might as well make it worth my time to build something, right?

The proposal would be something like... I build a module, and you have to detect near field or radiated emissions from it.  The module does something clearly nontrivial, like light a white LED from a AA cell, or output 5V from 12V with >40% efficiency.  Maybe the LED would be enclosed, maybe the cell as well; maybe input and output would both be wired to maximize chances of success.  Open to options.  Maybe near field would be too much of a "gimmie", and beyond the spirit of the above claims -- but just as well, I might allow any chance to succeed, short of tearing the thing apart.  Further stipulations might be that it has to use an inductive SMPS converter stage, perhaps Fsw(avg) above some minimum, so it's not simply a charge pump, or a dumb iron-core inverter, but reasonably representative of commercial practice, just cleaned up particularly well.

Alternately, a more usefully-didactic rendering -- maybe money doesn't get involved at all, rather the value is a case study for both participants and readers -- how about something like: you make a converter, as best as you can, but which you are still reading signal from.  Send it to me, and I'll fix it until you can't.  Then you figure out some other method maybe, to observe emissions from it.  Back and forth, so it goes.

I feel pretty confident in proposing and accepting such a bet; how about you?

Your claim demonstrates a fundamental misunderstanding of physics and RF theory. Using a 3 GHz bandwidth results in such a high noise floor that you won't even detect the presence of relatively strong carriers from distant transmitters, even though these carriers would be clearly received by a radio receiver with a sufficiently high SNR.

Attempting to identify weak noise sources with such a wide 3 GHz bandwidth is impractical and scientifically unsound, which is why your example is flawed and frankly, laughable. 

Ah, but you forget.  It works the other way, too.  How pray tell do you pick out e.g. a spread spectrum signal from below the noise floor?  If you don't know what the precise clock or message (spreading) signal is, if you don't have any coherent reference, how can you detect it?

People complain about spread spectrum being some kind of devil's bargain but it works for good reason.

Put another way: suppose you had access to the internal clock signal.  You could wire that to a lock-in amplifier, and pick up emissions practically across the world.  But how do you do that without the phase reference?  How do you do that incoherently?

It sounds like you have a method, which you should patent quickly, because you can make a hell of a lot more money on it than some stupid bet on the internet will.

Tim

[radiolistener]

At what distance?

about 6000 km

[T3sl4co1l]

I'm keying my flashlight to Morse code right now, what am I saying?

Tim

[radiolistener]

I feel pretty confident in proposing and accepting such a bet; how about you?

Unfortunately I will not be able to get your money due to limitations in my country.

But I can help you detect the noise from your circuit, and you can find it yourself. The issue is not whether it is possible (because it is possible) but rather how narrow the bandwidth will need to be. If your circuit is relatively low noise, you might need a radio receiver with an ultra-low noise LO, LNA with low NF at frequency of interest, and a sensitive ADC with DSP for processing. This could be expensive if you don't have a good SDR radio. However, we can try using your oscilloscope if it allows capturing long samples and processing them in MATLAB/Octave. I can assist you with the processing if you want to conduct that experiment."

How pray tell do you pick out e.g. a spread spectrum signal from below the noise floor?  If you don't know what the precise clock or message (spreading) signal is, if you don't have any coherent reference, how can you detect it?

You don't need to know the exact frequency to detect EMI from a DC/DC converter because it doesn't have a very stable carrier and has phase noise that creates a broader spectral hump around the main peak. All you need to do is detect the increase in the noise floor.

If you understand the math, you know that this can be achieved through simple averaging. Yes, it may take longer time to detect quieter noise due to the Shannon-Hartley theorem: C = B*log2(1+S/N).

Thus, a narrower bandwidth results in slower detection of the signal presence. But on the other hand, you are almost unlimited in your ability to narrow the bandwidth to detect a signal that is as weak as you need.

As you can see, there is no magic involved - just math.

People complain about spread spectrum being some kind of devil's bargain but it works for good reason.

As I mentioned before, a better way is to shift the noise into an unused part of the spectrum. This technique is used in SMPS power supplies for radios, which include a control knob to adjust the frequency and shift the noise away from critical frequencies.

Put another way: suppose you had access to the internal clock signal.  You could wire that to a lock-in amplifier, and pick up emissions practically across the world. 

But how do you do that without the phase reference?  How do you do that incoherently?

It looks obvious that you understand that it is possible, because you trying to limit the use of coherent LO. Isn't it? 

But how to build precise frequency standard is another story... We're talking here about the noise.

It sounds like you have a method, which you should patent quickly, because you can make a hell of a lot more money on it than some stupid bet on the internet will.

Yes, it’s possible, but it requires using a very precise and stable frequency standard, and the information transfer speed will be very low, which does not offer advantages over a classic high-power transmitter with an antenna.

This is how ELF transmitters work to send command orders to underwater submarines that carry nuclear weapons. These submarines may be at depths of 1 km or more, where conventional communications do not work. ELF transmitters use a 76 Hz (USA) or 82 Hz (Russian) carrier to transmit data, but the data transfer speed is low due to the narrow bandwidth (see Shannon-Hartley theorem). The message transfer speed is low, but it's enough to send a short encrypted message with an order to do nuclear strike or other command.

I'm keying my flashlight to Morse code right now, what am I saying?

First, I don't have the equipment to detect signals with such a low level. Second, I'm sure you're sending it too quickly, so it will be hard to distinguish from environmental noise. However, it's clear that by doing it you increase the environment noise level. Isn't it? 

[T3sl4co1l]

I feel pretty confident in proposing and accepting such a bet; how about you?

Unfortunately I will not be able to get your money due to limitations in my country.

Hah, fair enough. Under the circumstances, perhaps a donation would be in order? $1000 to the charity, relief fund, etc. of your choosing?

You don't need to know the exact frequency to detect EMI from a DC/DC converter because it doesn't have a very stable carrier and has phase noise that creates a broader spectral hump around the main peak. All you need to do is detect the increase in the noise floor.

If you understand the math, you know that this can be achieved through simple averaging. Yes, it may take longer time to detect quieter noise due to the Shannon-Hartley theorem: C = B*log2(1+S/N).

Sure, that works when the signal is coherent, just average the fuck out of the noise and out comes the signal.  But something is missing between precisely these two paragraphs:

You can't average that which is already noisy.

Both drop at precisely the same rate.

If I have a signal that averages as C = B*log2(1+S/N), added to a linear channel with noise that averages as C = B*log2(1+S/N), what margin of likelihood, or channel capacity per bit error rate, can you get through there?

Put another way: what confidence interval can you put on your receiver picking up -98.2736dB of atmospheric noise, versus -98.2661dB of atmospheric noise plus some potential source?  Can you identify the source -- if not unambiguously, then within some margin of likelihood?

At what point does the statistical variation of the background itself, so utterly overwhelm the source of interest, that it is indistinguishable under any and all possible analyses?

Yes, it’s possible, but it requires using a very precise and stable frequency standard, and the information transfer speed will be very low, which does not offer advantages over a classic high-power transmitter with an antenna.

This is how ELF transmitters work...

--It's not.

Two things:
1. You're completely missing the opposite side of the equation.  (I cannot ascertain whether this is accidental or intentional, but neither one is a good look?)

If *either of* the transmitter carrier, or the receiver LO, is noisy, the received signal is noisy -- potentially, lost in the noise floor.

Critically: the result is identical and indistinguishable whether we introduce phase noise, frequency modulation, or other spreading, into the transmitter's timing source, or the receiver's timing source.

Indeed we could have a referenced (perfectly coherent) system, where both transmitter and receiver are clocked from the same reference (pure or otherwise; it doesn't matter at this point), and over-the-air reception is good down to whatever averaging level works; that's a lock-in amplifier.  But suppose we add a PLL, modulated with incoherent FM or PM, inline to either transmitter or receiver: the whole thing crumbles apart, no matter how long the averaging is.  (Actually, maybe it still works for PM, if the phase error is less than 180°; it just takes longer to average out.  An undefined (i.e. evenly distributed) phase error however, isn't going to see anything. Likewise, an FM error, as a fixed frequency shift for example, completely eliminates what was otherwise a DC/baseband signal.  Maybe you know to retune the receiver, or pick out the tone in the audio output; but how do you know that's really the one anymore, or just some interfering signal?)

It is necessary and sufficient that *both* transmitter and receiver be pure, to whatever coding scheme the system uses (doesn't have to be single-tone, can be chirped, can be spread, whatever), in order to achieve communication between them.

Or, say in optical terms: suppose we have a two-arm interferometer.  If we add a random phase plate (say an electrically modulated one, white noise spectrum up to the frequency limit of the detector), to either arm, or both, the interference pattern is GONE, bupkis, nada.  Only when the signal and reference are coherent, do we detect a pattern -- receive a signal.

2. ELF transmitters work with very stable transmit frequencies, and very stable receiver tuning.  Both combined, means relative coherency is maintained, and a stable signal can be received, even if the signal levels are very low indeed.  (On top of which, the bitrate is quite low, but not by an extreme degree compared to the already extremely low carrier frequency.)  So it's an example which seems to prove my point and defeat yours?

First, I don't have the equipment to detect signals with such a low level. Second, I'm sure you're sending it too quickly, so it will be hard to distinguish from environmental noise. However, it's clear that by doing it you increase the environment noise level. Isn't it? 

Oh so you don't anyway... you're just blowing steam.

I mean, I guessed as much, two replies ago -- but you could've saved us all some effort (and yourself a lot of embarrassment*) by just saying so in the first place -- or not saying anything at all.

Look -- I don't know at what point you can still learn from this.  I've chosen to lean against this post more strongly than I might otherwise, and so don't expect much acknowledgement in return.  I still welcome a healthy and receptive conversation here.  But proclaiming half or un-truths as fact, without knowing your own gaps of knowledge about it?  Or worse still, if you are in fact aware of your own ignorance, or misrepresentation, and still choose to post it as fact?  Either way, that's not something I will allow to pass unchallenged.

Tim

*Or myself for taking the bait.

[radiolistener]

1. You're completely missing the opposite side of the equation.  (I cannot ascertain whether this is accidental or intentional, but neither one is a good look?)

If *either of* the transmitter carrier, or the receiver LO, is noisy, the received signal is noisy -- potentially, lost in the noise floor.

Critically: the result is identical and indistinguishable whether we introduce phase noise, frequency modulation, or other spreading, into the transmitter's timing source, or the receiver's timing source.

I didn't said that the DC/DC converter is a transmitter here. The DC/DC converter is not a transmitter, but an EMI source. Both the transmitter and receiver can use an ultra-low phase noise LO with a stable enough frequency to send the message. I assume that the LOs of the transmitter and receiver are synchronized with sufficient accuracy to consider them coherent.

However, the DC/DC converter placed on the path will add some noise, which will reduce the SNR and make signal reception worse. This noise can be very small, may be so small that even not detectable with all existing hardware, but it still add noise. This is what I'm talking about.

And this is exactly what happens in topic-starter case. He don't use DC/DC as transmitter. It just works as EMI source which reduce SNR for communication channel between transmitter and receiver.

Oh so you don't anyway... you're just blowing steam.

I mean, I guessed as much, two replies ago -- but you could've saved us all some effort (and yourself a lot of embarrassment*) by just saying so in the first place -- or not saying anything at all.

This is an unfair manipulation on your part.

It’s like claiming that since you don't have the equipment to test if radio communication is possible at a distance of more than 100,000 km, it proves that electromagnetic waves cannot travel more than 100,000 km...

No, the lack of sufficiently sensitive equipment doesn't prove anything. If I don't have the equipment to detect the presence of a very small noise source, it doesn't mean that there is no noise at all from your keyed light.

PS: regarding to the bet I sent you private message

[radiolistener]

2. ELF transmitters work with very stable transmit frequencies, and very stable receiver tuning.  Both combined, means relative coherency is maintained, and a stable signal can be received, even if the signal levels are very low indeed.  (On top of which, the bitrate is quite low, but not by an extreme degree compared to the already extremely low carrier frequency.)  So it's an example which seems to prove my point and defeat yours?

No, it proves my point and refutes yours. The message can be received because its bandwidth is narrow enough to achieve a sufficient SNR to decode the message. If you introduce an EMI source with enough noise to degrade the SNR, it will be impossible to decode the message.

You can make the message bandwidth as narrow as you wish, reducing the noise power as much as needed, but this also requires a more precisely synchronized LO for both the transmitter and receiver. Therefore, your ability to listen for very weak signals is limited by the stability of your frequency standard.

And if you display the spectrum with a narrow enough bandwidth span on your receiver, you will see EMI noise from a nearby DC/DC converter, even if it has good shielding and very good filters.

To stop that discussion I can say from my experience that any kind of DC/DC converter, with any kind of shielding and filtering, produces high enough noise peaks (at least 20-30 dB above noise floor) to be visible on the spectrum waterfall of a medium-quality receiver. And there is no need to use very expensive ultra-low phase noise oscillator for LO to detect DC/DC module noise. 

The best result you can get with a 2 MHz DC/DC synchronized to the receiver LO is that it has fewer spurs, but its noise is still very visible. However, its noise is cleaner and more narrowband which is good for receiver.

I never seen DC/DC which don't have visible noise on the receiver spectrum waterfall. But even if you build it using modern cutting-edge technology, I’m sure it will still be easy to see its noise on a receiver. It might require a slightly less noisy LO, but I think it would still be visible even with a typical Chinese oscillator.

I can say this with confidence because I have tested DC/DC converters in practice and have seen their noise with my own eyes on the spectrum waterfall. I also know other people who have designed receivers using expensive DC/DC modules and extensive filtering measures, he has the same results. This is why I'm saying that it's practically impossible to have DC/DC which don't have visible noise on receiver.

[Andy Chee]

To stop that discussion I can say from my experience that any kind of DC/DC converter, with any kind of shielding and filtering, produces high enough noise peaks (at least 20-30 dB above noise floor) to be visible on the spectrum waterfall of a medium-quality receiver. And there is no need to use very expensive ultra-low phase noise oscillator for LO to detect DC/DC module noise. 

The power supply of my homebrew GPSDO has a DC-DC converter to maintain constant heater voltage for the OCXO as my battery voltage drains over the course of several hours of field portable operations.

I can see NO increase in the waterfall, and definitely not 20-30dB!  This is field operations away from urban electrical noise.

[Phil1977]

Edit: If you don't believe it, look at youtube 6min30s clearly shows a more or less standard SMPS in a Rohde & Schwarz spectrum analyzer going from 5kHz to 3Ghz. And please don't tell me R&S is doing it wrong... 

Your claim demonstrates a fundamental misunderstanding of physics and RF theory. Using a 3 GHz bandwidth results in such a high noise floor that you won't even detect the presence of relatively strong carriers from distant transmitters, even though these carriers would be clearly received by a radio receiver with a sufficiently high SNR.

Attempting to identify weak noise sources with such a wide 3 GHz bandwidth is impractical and scientifically unsound, which is why your example is flawed and frankly, laughable. 

Please read and understand some good introduction into spectrum analyzers. 3GHz is the total bandwidth of the device. You set the bandwidth for each scanned point with the RBW, which goes down to 1Hz at the shown model. The noise floor of the spectrum analyzer is specified to -165dbm typical - how does that compare with a decent receiver?

Radiolistener goes to my personal ignore list - he talks of personal experiences while other people talk about science.

[shabaz]

There's so much inaccurate information from radiolistener, I ignore it mostly. Sorry if that appears rude, but I'm not trying to be. I don't like saying it, but I feel I have to.

Here's a discrete DIY supply, measured before and after adding a cheap LDO IC (performance would have been even better with a more modern LDO with better PSRR). I'm no power supply engineer. I just took some effort to reduce harmonics and it could still be improved further. LDO (TLV740P) is not visible in the photo, because it's soldered on the underside.

By 1 MHz, it is already down to -140 dbV. That noise visible past that (around 50 MHz) is other equipment, not related to the supply (i.e., that noise is still present even when the supply is switched off).

This power supply would have no noticeable impact on a radio receiver at VHF. Nor HF.

[jonpaul]

Wrong converter topology.

use soft switching converters, resonant etc, set sw freq below band of interest.

j

[T3sl4co1l]

I didn't said that the DC/DC converter is a transmitter here. The DC/DC converter is not a transmitter, but an EMI source. Both the transmitter and receiver can use an ultra-low phase noise LO with a stable enough frequency to send the message. I assume that the LOs of the transmitter and receiver are synchronized with sufficient accuracy to consider them coherent.

Ah OK, you're still thinking of the OP situation, possible effects etc.

This conflicts with the usual understood meaning of "EMI source": what is a source but a[n unintentional] transmitter?  (Standards in English use this phrasing.)

Then, in terms of the lock-in example, you suppose that, more or less to say: powering one or the other unit with an SMPS, will introduce detectable aberration in link quality -- serving the role of the introduced PM/FM noise.  Would that be a reasonable statement of things?

To stop that discussion I can say from my experience that any kind of DC/DC converter, with any kind of shielding and filtering, produces high enough noise peaks (at least 20-30 dB above noise floor) to be visible on the spectrum waterfall of a medium-quality receiver. And there is no need to use very expensive ultra-low phase noise oscillator for LO to detect DC/DC module noise. 

Ok, this is much improved -- stating it as fact stands to mislead readers -- but stating it as opinion and experience leaves open the possibility that you just haven't seen every example, etc.  I apologize for the tone I took before.

Tim

[radiolistener]

Then, in terms of the lock-in example, you suppose that, more or less to say: powering one or the other unit with an SMPS, will introduce detectable aberration in link quality -- serving the role of the introduced PM/FM noise.  Would that be a reasonable statement of things?

What I wanted to say is that powering a receiver from an SMPS leads to some noise, which will be visible on the receiver. This doesn't mean it will break the link, because if the SNR is sufficient to maintain the link, you might not notice the noise. So, it depends on many factors—the signal level, the noise level within the receiver's bandwidth, etc. It may turn out that the main noise power of the SMPS is outside the receiver's bandwidth, and in this case, the degradation in link quality will not be noticeable within the receiver's dynamic range. In other cases, the noise power from the SMPS may overlap the receiver's bandwidth and significantly affect the link quality, but this still doesn't mean it will break the link. If the remote station's signal has enough SNR, the link will work, but the distance over which you can receive it will be affected.

However, if you don't notice the influence of the SMPS on the receiver at a certain frequency, it doesn't mean that it doesn't exist at other frequencies. That's what I meant when I said that any SMPS produces noise that can be detected by the receiver.

Please read and understand some good introduction into spectrum analyzers. 3GHz is the total bandwidth of the device. You set the bandwidth for each scanned point with the RBW, which goes down to 1Hz at the shown model.

I'm glad to see that now you understand the relationship between the selected bandwidth and the noise floor. This is great progress. Previously, you tried to find noise from an SMPS with a frequency span set from 5 kHz to 3 GHz.

Now, connect the output of your SMPS to the spectrum analyzer input through a DC block and try to check its noise with a 10 Hz span setting around the conversion frequency of your SMPS. Do you still not see the noise?

The noise floor of the spectrum analyzer is specified to -165dbm typical - how does that compare with a decent receiver?

What you still missing here is that sensitivity is linked with bandwidth and it's a big mistake to talk about sensitivity specified for 1 Hz bandwidth when you looking at 3 GHz bandwidth. Are you sure that the -165 dBm sensitivity is specified for a 3 GHz bandwidth? Or is it for 1 Hz? There's a huge difference.

For receivers typical bandwidth to talk about sensitivity is 500 Hz which is used for CW mode.
What is your spectrum analyzer MDS sensitivity at 500 Hz bandwidth?

This power supply would have no noticeable impact on a radio receiver at VHF. Nor HF.

Are you kidding? Your oscilloscope clearly shows SMPS noise from -40 dBV on VLF band to -115 dBV on SW band and -105 dBV on VHF band. -105 dBV is about -92 dBm power for 50 Ω. This is high enough to flood the weak signal with noise and make DX link impossible.

For example, the peak power of DX VHF station can have power level about -120 dBm or even less.

Typical MDS sensitivity of VHF receiver for 500 Hz bandwidth (CW mode) is about -140..-150 dBm, this is about -163 dBV.

And you talking that -40 dBV noise would have no noticeable impact on a radio receiver? Seriously? 

A VHF receiver that shows -105 dBV (-92 dBm) of noise from its own SMPS is just a piece of junk.