PCB Design Guidelines To Minimize RF Transmissions | Hackaday

Some design guidelines for PCBs don't make much sense, and some practices seem to be too redundant. Usually, these are caused by the black magic of radio frequency transmission. This is either the unfortunate and unexpected consequence of the electronic circuit, or the magic and useful function of the electronic circuit, and a lot of design time is spent on reducing or eliminating these effects or adjusting them.

You want to know how important this is to your project and whether you should worry about accidental radiation. In terms of Baddeley importance level:

When the signal moves down the wire, an electric field is generated in the space around it. If it is a DC signal, the magnetic field will not change, so nothing exciting happens in the RF field, but all remain unchanged. Pure direct current is very rare. The battery can do this unless you perform any switching voltage regulation, but anything plugged into the wall power supply will generate a 50 or 60 Hz sine wave, which is then rectified, transformed, smoothed, poke and pushed into a DC voltage or the like s things. . In fact (depending on the quality of the power supply), the power supply will produce ripples and produce small changes in the DC voltage, thereby effectively generating a small varying electric field. Other things such as crystal oscillators, signal lines between chips, and memory buses all have constantly changing voltage signals that travel from one place to another along the wires. Therefore, electronic equipment is flooded with signals and changes the electric field. It is through the constantly changing electric field, through a large number of mathematical operations (mainly discovered by Maxwell, Faraday and Gauss), that the electric field becomes electromagnetic radiation.

The frequency of radiation is the frequency at which the electric field changes, and there are many factors that affect this. One is the shape of the wire through which the electric field propagates. If you have something called a differential pair, the electric fields falling along the wires cancel each other out, resulting in almost no transmission. If you do not have a wire connected to the other end, the signal will drop along the length and reflect back. If the length of the wire is adjusted so that it can amplify rather than cancel out when reflected, your antenna is good. Back to frequency, it is by no means a perfect sine wave. It is a combination of waves of different frequencies. The antenna receiver has an electronic device that will deconstruct those frequencies in a certain range to extract the signal. Modern things are mainly FM, so there is a main carrier frequency, and will change slightly with the change of the data signal.

The tracking antenna is an antenna made of a small copper wire on the PCB, which happens to resonate at certain frequencies. This may be intentional, such as the F antenna design used in a 2.4 GHz transceiver, or it may be accidental, such as causing a long ground wire to fall in. To avoid this, please carefully check the poured ground for any traces. Either eliminate them or put in a via so that the trace does not resonate. Keep the ground as old as possible. The more fingers you have, the more times you slice, stretch and separate them, and the more unintentional radiation you will suffer. Generally, unless an antenna is intentionally manufactured, do not have any wires connected to both ends. This can be applied to unconnected board IO. After all, if nothing is plugged into the IO connector, it is just a trace leading to nowhere. If your microcontroller is smart enough to detect when the cable is unplugged, please do not send any signals along the cable. Bind all unused IO to the end.

Traces near the edge of the circuit board and farther from the ground plane will radiate more electromagnetic interference. Through-hole stitching refers to placing the through-hole ring always (or as much as possible) around the PCB edge to the ground plane. You can also add through holes on the side of the signal trace to reduce the EMI of the trace. In addition, a certain dose of through holes should be used to connect the ground holes to the ground plane (if you have a separate ground plane on a 4-layer board, or your 2-layer board is mostly grounded at the bottom). This prevents accidental antennas and also ensures that the entire ground remains at the same potential.

Data sheets for microcontrollers and power conditioners have decoupling or bypass capacitors connected to the power supply pins. These chips do not often consume the same amount of power. They vary slightly with the work of the chip, sometimes requiring a brief increase in power. On the power pin, this looks like a rapidly changing signal. The purpose of the decoupling capacitor is to have a small power reserve next to the power pin, so that when the chip fluctuates violently and rapidly, the capacitor can eliminate those power requirements without spreading rapid changes across the power trace . Ferrite beads are usually used when connecting a switching power supply to a power supply board, because it can isolate noise from the power supply, so place it (along with a decoupling capacitor) next to the power supply output.

Why do you want to make wires that take longer than necessary? Sometimes you need to force tracking to take a fairly circuitous route (heehee) to get from one contact to another. The rule is more about prioritizing which routes will become shorter and which routes may be longer. Generally, the faster the signal travels on the wire, the higher the priority, and the shorter the ideal trace length. The crystal oscillator should be as close as possible to the microcontroller, and the wire should be directly between the two. With every increase of 1 millimeter, the greater the change in electric field and the greater the emissions. The UART can have longer wires because the signal does not change at such a fast speed and the positive voltage rail may meander all over the place. This is also a good habit, because faster signals mean you want shorter distances between components to minimize latency, but RF protection is also important.

Since your PCB will have some cable connections, board-to-board connections or chip-to-chip connections, each may have longer traces, so you can add some filtering to the traces to pass series resistance and bypass Capacitors to reduce its noise. The ground should be as close as possible to the noise source (usually a microcontroller).

These are super noisy things, the cheaper they are, the more shortcuts there are. Not only will they explode radio frequency radiation with 50/60 Hz harmonics, but also switching power supplies that usually work at hundreds of KHz will also generate a lot of noise. This output may be unstable, so a lot of noise will spread along the wire and radiate out until it reaches your project, and then the project can work under noisy power. Not to mention

.

One thing that has no effect is the angle of the line. It turns out that even if the speed exceeds 1 GHz, there is no measurable difference between the 90-degree angle in the trace and other angles in the radiated EMI.

If you want a more comprehensive application note

. Also, please see our

@Bob Badley

Thank you for explaining your limited knowledge of SAW, harmonics and FCC guidelines.​​​

The rest of this article is vague and accurate anecdotal advice and follows the advice of most application note publishers.

"Accurate and accurate anecdotal advice" = easy to understand, for more content, like most topics, we must read textbooks.

Suppose this is a way to eliminate "have" and "have". As if they were not enough to hinder the development of publications, it is now an IQ test.

Thank you! My goal is to please!

Can we let the FCC raid broadcast houses?

If they emit enough heat, maybe...

Depending on the situation, I have heard some stories (mainly around here, which is the basis for my comment),

They will knock on your door and ask if you know of any spread.

Then (if you show uncertainty or ignorance), they will (sometimes) help you investigate the source so that you can do some processing on the source, namely replacement and/or repair.

They are generally more useful than savage because they (most people) know that many causes of interference are unintentional (faults).

Even one of the links in the article shows this (neon one).

Usually in the case of maintenance, it is better (and cheaper) to find a HAM for testing interference (assuming they already know how to perform the test, and not a person who is only permitted just because of their own wishes) or contact your communication committee/ contact. Commission (FCC / OFCOM / ETC ..)

It's ham, not ham. Not an acronym.

At least these days, HAM is usually hard to be an asshole. It's like testing someone with a ham.

Indeed, 90-degree and 45-degree corners have little effect on EMI, but at 1 GHz, it may be important to maintain a constant impedance of the trace.

Indeed, I thought at the time that a sub-note should be added to the sentence. +1

However, for a constant impedance, 45 degrees is still bad. If you pay special attention to corner reflections, you need to model the miter more accurately than "use a 45 degree corner".

However, what I have seen about 90-degree and 45-degree angles is that the 90-degree angle in lower-quality manufacturing processes can cause traps and lead to acid corrosion. Nowadays, this problem can be alleviated even by photo-activated etchant.

I won’t say that it sucks-better than 90 degrees,-the corner can be modeled as a capacitor connected to the trace, and it is really small (1pF reactance at 1GHz is 160 ohms, I suspect that the corner is close to pF )

Nonsense, 160 ohms would be a disaster. You will see that 20 ohms is easy (it is like 1 dB), which is about 100 fF.

???? You are on the wrong path-at 1GHz, 100fF is 1600 ohm reactance. I understand that the upper limit of 1pF would be bad news at 1GHz, but some engineering decisions need to be made. Any impedance continuity on the transmission line is not good, but it is relative. At GHz frequency, if there is no reasonable size of the trace, the tolerance of the PCB material will not reach the limit, and other variables in the system may submerge the 45-degree angle.

Gah, I stupidly consider that the final impedance of the trace has changed, which is also completely wrong. In any case, if the circuit is tuned, even the extra capacitance of 100 fF may be a problem because you will shift the frequency. I regularly trim the ground plane under the pad to compensate for such small parasitic effects.

However, the "engineering decision" part is stupid: there is no reason to say that PCB design software can't do arcing instead of sharpening 45, and calculating the correct bevel to keep the trace width as constant as possible is just a matter of mathematics. It is not an engineering decision not to compensate. It is just a lack of knowledge or time, and both should be regarded as failures of the tools used.

Nice article, thanks for sharing.

This seemingly 90s website provides a lot of useful information on reducing EMI:

What a great domain name.

He is an authority on EMI, transmission lines, PCB layout rules, etc.

He happens to live in the middle of Jersey.

Let me add: Always, always think in a circular way. As we learned in Circuit Theory 101, a circuit is a closed loop, going from the source to the sink and back to the source. From direct current to daylight, the theory is applicable in all aspects. In order to minimize electromagnetic radiation, the length of the radio frequency circuit must be minimized. This means bypassing as soon as possible, but it is also important to consider how close the bypass capacitor is to the source of the problem (active device). Not only physically close, but also electrically. What does the whole cycle look like? Is there a clear ground loop, especially on two-layer boards? Is the capacitor physically close to the power supply pin? Is there enough activity like ferrite beads to ensure further reduction of EMI?

This is an interesting story in the conference hall. I have a client who failed because of spurious emissions at 150 MHz, which is very bad. All GPIOs sent by the microcontroller are "wet", so it may make sense to try to mitigate EMI on these lines. Do not. I asked where is the Ethernet PHY (50 MHz, harmonics are usually related to Ethernet). To be sure, the terminal connection on the 50 MHz clock line from the PHY to the transceiver is poor.

But the fact is: GPIO did not malfunction. There is a confusing factor here. Poor termination of the clock line leads to the enhancement of ringing and higher harmonics. These harmonics enter the IO ring of the microcontroller.

Bypass the poor.

In order to facilitate the rework of the microchip, the circuit board designer removes the bypass cap from the microchip. It definitely meets the intended purpose and makes the microcontroller easy to replace, but the added series inductance compromises the advantage of the 150 MHz bypass. As a result, the entire IO ring is allowed to have a higher impedance at 150 MHz, thereby introducing noise and transmitting through any IO connected to the power rail.

The mitigation method is twofold: add some serial terminations on the Ethernet clock line and make the microcontroller bypass closer to the IC. Fortunately, we were able to complete these changes within a four-hour observation window and saw substantial improvements. The changes made are folded into the next board rotation, and the product is now compliant.

Sorry, it should read "Bad terminal connection on the 50 MHz clock line from PHY to *microcontroller*."

This is a fascinating anecdote because it points out that a bypass that works on the basic principles may miss harmonics, which is not necessarily obvious to a novice!

Can you clarify the meaning of "IO ring"? Does it refer to the peripherals and connectors around the microcontroller on the board, or the structure inside the microcontroller die? Or something else? (This seems to be one of the terms, such as "power rail", which is so obvious to those who know they have never explained it, but it is confusing for novices who are not yet at this level. )

Sorry, too much time was spent with silicon designers. As you can guess, the IO ring is a structure on the die. The way the manufacturing process works is that all IO pads are on the edge of the die, so the circuits most directly related to IO actually form a ring around other processing circuits. Therefore, the IO ring.

Data sheets usually show simplified diagrams of IO pins, but I will briefly describe them here. There is usually a driver that includes a pair of "totem pole" FETs from a dedicated VDDIO power supply to VSS, an input buffer connected to the internal power supply, and some ESD management, usually as simple as a pair of diodes on the IO pins. , One to VDDIO and the other to VSS.

This is the beginning of the trouble. VDDIO must show low impedance to VSS in a wide frequency range, not just VSS VSS, which is most closely related to the IO circuit. If this does not happen, the noise will be coupled to the VDDIO power rail, and once there, it will propagate beyond any IO pins.

So how to get there? Multiple methods. The culprit is the IO driver itself. Whenever one of these drivers is turned on and off, it will draw current from the VDDIO power supply. The higher the impedance of the power supply, when used again* (please note that many microcontrollers have multiple VDDIO pins, each *must* must be independently bypassed), the power supply will cause more voltage ripples, And there will be more noise. Is coupled.

The auxiliary mechanism is a bit tricky and may be what happened in the above situation. Remember all these output drivers, input buffers and ESD structures? Well, each has a parasitic capacitance. It may be small, 1-2pF, but even so, what is the capacitive reactance at 150 MHz? 500-1000 ohms. This means you can effectively model the clock input as having a small resistance of 500-1k on the power rail. Similarly, once the noise enters the power rail, it can be propagated to any other IO pins with the help of active high-side drivers or simply through the same parasitic capacitance path as the power supply. The only way to avoid this situation is to carefully bypass the supply.

150MHz, right?

If bypassing is really critical to your design, you can try using embedded capacitor materials.

Please also refer to:

Thank you for citing these content, it is good to review it from time to time.

For most designs, bypassing is critical, but it doesn't need to be difficult, expensive or intrusive. In most cases, 1-2 0402 ceramic caps of 0.1-0.5 cents per supply pin are sufficient. The trick is to make sure you don't shoot in the foot. When I did the layout, the first step was to place the active components, followed by the bypass and any EMI or SI key components (ferrite, termination). When wiring is required, power and high-frequency signals are always the most important. Usually when this method is not adopted, I still call me, either a designer (unfortunately not recently) or an EMI consultant.

"I want to add: always, always think in a circular way."

Yes. absolute. What I want to say is "Remember that the current is circulating. You can see the signal path, but *you have to imagine the return path below it*."

In higher layer designs, many times you end up with multiple ground layers (not isolated grounds, but multiple layers), because if you stack signal layers on a single ground layer, crosstalk will occur. Also, sometimes you need stripline signals or CPWG, so there will be multiple ground planes there.

In these cases, it is often forgotten that when a signal passes from a layer where the ground loop is on *one* plane through a layer where the ground loop is on a *different* plane, the return current also requires a via to switch the layer. If you don't, that layer of switches can "completely" * mess up the impedance of the trace. Since the signal loop is now larger than you think, it will also cause an increase in EMI from this trace.

This is why the super fancy pants layout pack now has

Specifications, so they are automatically added during routing.

It is also important to remember that power planes also often return signals-therefore, when they are adjacent layers, it is best to avoid using high-switching wiring to span different power planes. For example, if layer 3 is a power layer and there are multiple planes there, try to avoid any traces on layer 4 going through these gaps.

If you have a signal to pass there, please place a coupling capacitor on the right side of the trace to couple the two planes (for example, 3.3V->5V in the simplest case), sometimes called "plane jumper" ". However, it would be better to avoid it completely.

"In this case, what is usually forgotten is that when the signal passes from a layer where the ground loop is on *one* plane to a layer where the ground loop is on a *different* plane, the return current also requires a via to switch the layer. If You don’t, that layer of switches can "completely" * mess up the impedance of the trace. Since the signal loop is now larger than you think, this will also cause the EMI of the trace to increase.”

I encountered this situation, routing the signal from the top layer to the bottom layer on a 4-layer PCB, and there were no bypass capacitors connected to the ground and power planes. What makes me more distressed is that I also did not consider the signal split across the power plane... well, the current loop of the signal might look like spaghetti:')

I'm already familiar with the current loop/current return path, but after making this mistake, the theory becomes "real" enough that I will no longer make this mistake in the near future.

Yes, the loop area should be the first, second and third points of this article, maybe to keep up with the unexpected stub antenna. Most things come from the loop area. Why use decoupling capacitors? Provide a low impedance path for the return signal to minimize the loop area. Why do we have a solid ground with almost no cuts? Provide a low impedance return path for the signal. and many more. Compared with the loop area, through-hole stitching is a very small thing.

Yes, something that emits 90 degrees is a myth, but I still make 45 pairs, just because I think it looks better.

Compared with 45s, 90-degree bending is more likely to become thinner when making PCBs, especially on household etching boards.

I have seen 90* elbows used in the Cisco WiFi card and they look like PCB etched "air core" inductors.

Although bending is more likely than the transmission capacity of the tracking track, it is more likely to define the track length that exists in the scene where the inductive coupling is performed (if any, so far, this is my guess).

Don't you do ESD / EFT testing?

Does IEC 61000-4-2 not apply to your product?

No, not at all.

Great article.

My master's degree has inspired me a lot, but PCB design has never involved any aspect. It turns out that this is the most critical aspect of my work since graduation. I have learned some of these things hard with failed prototype PCBs. I used to sew to solve thermal problems before, but I never considered RF.

>I have solved the thermal problem through sewing before, but I never thought about radio frequency.

OMG! Just search for "RF PCB" on the image, then go through the fence to see all the cute things.

Via stitching is usually performed to manufacture coplanar waveguides for stripline transmission lines.

The goal of PCB layout is to minimize the circuit loop area.

All traces not on the ground plane are antenna structures.

Antenna reciprocity. You generate a lot of emi at a certain frequency, and you like to receive it there.

The reason why radiation is radiated near the edge of the circuit board is that the transmission line generates a signal, so that the ground line width of the electric field line is 5 times that of the ground layer. Generally, you want the ground plane to be very close to the edge of the PCB and back to the signal plane when the signal plane contains high-speed or high-transition traces.

When you set STM32 to fast GPIO mode, ask yourself, does this really need to be fast? If the span is fast, so are its harmonics.

Find,

I see the working principle of high-quality PCBs, where there is no way to route any traces between layers for some reasons:

These traces (especially LVDS pairs) maintain very close spacing, and the PCB spacing between the traces and the ground plane is about 1/2 inch or 10mm.

Generally, the PCB is located in a metal housing, and all non-radio influence channels are located near the vents and connectors. (Even before, I have seen this structure in decent laptops)

>All traces not on the ground plane are antenna structures.

This is a problem because I usually only use two-layer PCBs.

Zero ohm resistors are your friend! You can get a cheap 1206 0 ohm resistor, and then whenever you need to wire, just put a resistor. Remember, despite this, you will cause some crosstalk between two crossing signals because their return current flows through the same part of the ground plane.

My guess is that it’s better to skip the signal overpower (or signal overpower) instead of skipping the signal over-signal, because in any case, the power supply will be high-frequency decoupled on the IC, so the actual physical connection back to the power supply is At low frequencies, however, the return current will be distributed across the board.

Another thing you can do is to put in more effort, but it can actually be better, it is a jumper (the smaller the size, the better, so you can make the pads as small as possible), And *install a jumper* at medium frequencies, the two sides of the ground plane are *different planes* (electrically connected everywhere), so the return current on one side of the plane cannot even see the other The side of the plane returning current.

The 0 ohm 1206 on the back can also be used, but unless you put the via on the pad, it will take up more space, which is usually difficult to construct. As long as you make sure that you don't run wires on top of the pads on the ground plane, the actual space on the ground plane is not a big issue. Just run the wire through the gap between the jumpers.

I often make 2-layer boards, and the 0 ohm jumper is completely my friend. The jumper on the back is a trick, one can say at once: "Hey, you are breaking the ground" and then point out no, no, you are just using the other side of it.

I never thought of using 0 ohm resistors as jumpers. Thank you for your suggestion!

Don't panic, just keep the loop area small. Imagine that the return signal should be as close as possible to the trace, but on the ground plane. For each cut, it must move outward. You can reduce the distance according to the cross stitch stitches I think. Pop your far trajectory up to the top, and then swipe down. This creates an additional path for the ground plane. If you trace the loop area, you can find many areas in which you can manipulate the ground plane without much impact, especially under the IC. Creating a two-layer board with a good low-inductance ground* and* power plane is a bit of an art form, especially when you have a lot of signals.

A few examples:

Thanks for your suggestions and examples.

My most recent job is to obtain medical equipment through radiation and emission tests.

Regarding conduction emissions from power supplies: We tried many power supplies from large (not so major) manufacturers, all of which claimed to comply with conducted emission limits. Guess what? Without external line filters, none of them will pass. Plan to install one.

Another area of ​​difficulty is the appearance of a broad peak around 150 MHz. This is due to the color display being fixed in a metal mount (but not screwed or otherwise fixed, as this is a high-quality display designed to be packaged in consumers) ) Plastic clamshell shell). It does have a metal case, but that is not enough, we have to use copper tape to glue the display case to the metal bracket.

The RFI test is fun! ! ! (Because you learned a lot, and when you got the "pass", the boss thought you were smart)

We paid a lot of money to a guy equipped with all the equipment and antennas for our CE test...I want his job. ;)

He just pointed out the problems, and didn't have to solve them...

Although knowing the location of the impersonation in a 100pf capacitor can "eliminate the edges" and squeak within the limits, it does make you look like a practitioner of black RF technology. ;)

"Regarding conducted emissions from power supplies: We tried many power supplies from large (not so major) manufacturers, all of which claimed to comply with conducted emission limits. Guess what? Without external line filters, none of them would pass . Plan to install one."

Check your PCB layout/grounding. If the device passes CFR Title 47 Part 15B, you may have done it.

We have similar experience in purchasing power modules from well-known manufacturers. Finally, we got a report from the lab telling them that the device failed the emission test, they went and redesigned their power supply to actually meet the specifications announced in the data sheet. It's not that we are a customer who buys 100,000 products every year, they bend back to continue doing business with us. If we buy 1,000 dollars a year, we have already had a bumper year.

When I re-read the beautiful picture on their data sheet, it said "When installed in a compatible device"

Now, both you and I know that there is almost nothing except the power supply itself which causes conducted emission, but I think that sentence is their "swinging room".

Oh, and all consumables are "medical grade", which means that in addition to emissions, they should also meet the leakage requirements of IEC 60601-1. That way, they did it. Silly me I think that if they can solve the problem correctly, then emissions will be a breeze.

Possible explanation: first look at the size of the power supply filter, and then look at the size of the switching power supply. There is not enough space to install these filter components in a switching power supply... and competitors are constantly applying pressure to reduce the size of the power supply.

IEC60601 means that because the products are connected by cables, the requirements for leakage current are very strict.

Leakage current is easily coupled through the large-capacity Y capacitor on the input AC-DC EMI filter. Other sources are grounded heat sinks, which capacitively couple noise into the grounded conductor.

Note that many power supplies have a Y capacitor between the secondary DC output ground and the primary side common. This is to reduce the loop area as much as possible to avoid radiating too much EMI.

For hard switching (fixed frequency, non-resonant) power supplies, it is difficult to work without or without Y capacitors. These types of supply can be said to be more stable, but this may be because they are easier to understand and model.

Y caps are used as filters for common mode noise and common mode chokes

If the method of grounding the power supply of the product enclosure is bad, be prepared for filtering.

For traces that deliberately carry RF or HF signals, circular traces are better than slanted traces, because this can reduce signal impedance and reduce ringing. Looking at things like WiFi or even ZigBee recently used, you may find that the trajectory of the antenna feeder is circular and has a flowing curve instead of a sharp 90 degree bend. You want to send as much signal as possible to the antenna and radiate in that direction in the desired direction. Looking at the impedance matching traces on the board with USB, PCIe or other buses with differential pairs, you will find that the corners are rounded instead of 90 degrees.

The reason (according to Johnson, high-speed digital design) is that the width of the transmission line increases slightly at right-angle corners. Therefore, keep the trace width constant and minimize the impedance discontinuity, because the impedance discontinuity will cause reflection.

I believe the reason for this is that it is easier to calculate impedance, not better.

In fact, you can still use right-angle traces, you only need to cut off the outer corners to keep the impedance roughly the same. 2.4GHz is definitely the range where the signal begins to appear such impedance discontinuities. Another trick is to cut holes in the ground plane below. This is usually used under filter covers or series resistors because the width of the pad is significantly larger than the trace.

If your project is not a deliberate radiator, but puts it in a metal box, there may be any debris you want inside, right? Nothing works out of the box.

As long as there is no opening in your box, or the connectors and cables connected to the next box are 100% shielded, it's fine. ;)

Radio frequency is like water, it will leak out from any joints or non-metallic connectors.

If it's that simple.

The noise is coupled with the wiring through the outside. The plastic switch "leaks" and the LED is plastic.

Your metal box can also be a radiator.

> Everything is unlimited.

What does it do? If there is no I/O and UI.... I can imagine that this will work for temperature and acceleration data loggers, nothing more!

Once you have power input, any type of I/O, switches, indicators, displays, your box is no longer sealed. Keyboard matrix scanning can be incredibly noisy, especially if you use GPIO to scan without limiting the slew rate of the signal. (Slew rate = speed of voltage change = sharpness of square wave edge = rich harmonic content of generated EMI.)

Similarly, LED matrix scanning. MAX7219 is a popular chip without slew rate limitation. It is the same as MAX7221, but adds this function. If you are worried about EMI, please guess which one to use!

This is not just for commercial products. If you use radio signals yourself, the noise from your own equipment may be the biggest problem. Because it is so close, it can solve the problem without being particularly loud.

If you have field wiring, it will couple the noise you generate in the box and radiate/conduct out of the box.

FWIW, in the presence of millivolt (sometimes smaller) signals, I will deal with very large RF interference, no, you can’t make the box that tight. You just can't. The exact number makes me incomprehensible, but the performance of about 60dB is good (washer, not an accidental slot antenna, no corrosion, perfect conductor, braided coaxial cable soldered to the copper box, passing through the decoupling ferrite/capacitor, Then it flows continuously through –

When I use the various parasitic capacitances of 50kv (from coaxial cable to ordinary air) on the HV line, the Q value tends to be very high, even if the power supply has an internal "assumed resistive" ballast, the peak current will be unexpected The arc can easily reach kiloamps-kilovolts. For example, megawatts. Now, it is a few feet away from us, working in microwatts, and trying to tell the accurate gamma spectrum or the accurately detected neutron count....

Therefore, megawatts/microwatts – 10^12 – exceed 60dB!

The result is that you have to shield both ends, carefully perform power decoupling and everything you can think of, and then make some modifications until the actual effect is good, so as not to fool your own measurement results. By the way, I have been doing RF/EMI since the early 60s. Knowing the theory is not all, although it will help-at least when you find a bug, you know the reason for the fix; ~}

Since I am doing "research grade" work-in a sense, it is more lenient than "production grade" because it only takes long enough to get answers... and the output of my test bench often explodes A few meters of computers even if they are not connected to anything-I do use arduinos and raspies-as Eben said, it is a low-consequence environment, and it is cheap and easy to replace, just in case the EMI incident is greater Beyond plan.

Please note that in real life, a fusion reactor like the one I am using does not need this sensitive instrument near the reactor-this is research! Some things you only need to find out once, and then you can make them bigger.

The sewing example seems to be similar to the old wincor (Nixdorf/Siemens) product we used (1994 to 1999) in structure, PCB grade, soldering quality and power supply layout.

In addition, some of the above products also use VCC (5 and/or 12v) as their PCB shielding LOL

Have enough bypass Vcc to ground, until AC current

Why not use a chip with a socket for an external oscillator? Use SMD tapered clamps and have enough space to accommodate the oscillator.

Board thickness....

In order to connect the two pins to the top, nonetheless, there is still no way to stop them. The unused pins have 10-14pF harmonic/over-oscillation suppression capacitors.

The next problem is that the chip will be used in a decent metal tablet: unless insulated, short-circuit the clock source.

Capton tapes can only be used for so long (like any design failure year?), and then the PC can only "die".

Generally Joe doesn't know how to solve it, but we will at least guess that it is just to replace the Kapton tape.

The thickness of the board depends on the laminate design, and it is entirely up to the designer to decide which laminate to use.

Because there is more than one type of external oscillator, the size (including size) of each oscillator is different.

Because the oscillator in the socket will be bulky, take up a lot of circuit board area, relatively expensive, low reliability under high vibration or acceleration, and will increase inductance, parasitic capacitance, etc.

You neglected to block...good for you, good for them. Good for everyone.

The title is wrong! The best headline is: How to not write an article.

I think you inadvertently are the whole article.

You *must* be careful with ferrites! Steve Weir published a great article at DesignCon 2011, but unfortunately, it may seem to be offline.

However, the "Analog" article also introduced this.

Ferrites, together with decoupling capacitors, form a good LC filter, which is why they are used. But why not use ordinary inductors? Because the resonance LC peak *amplifies* the signal, but does not attenuate the signal, and the ferrite has a loss (resistive) at its self-resonant frequency.

Except for your hat* move the LC resonance back to the sensing area! *This means that if you are not careful, the ferrite beads + LC filter can *amplify* the switching noise without attenuating it!

The solution is to simulate it! You may need to add a dominant pole (a capacitor + a small resistor in series, or a cr-pin electrolytic cell with high ESR) to suppress the LC resonance of the ferrite.

For this reason, Wurth Elektronik's ferrites are great because they provide a complete set of LT Spice simulations for all ferrites. Therefore, you can determine the best solution very, very quickly.

I found this article and it looks great;

Ah, Wayback Machine! His introduction to the ferrite bead part is

, This is a simple quick reading and instructions on how to deal with them.

I like it and i will join

This is a very timely article for me, because I am currently taking a course on radio receivers. The last class discussed pcb design. I seem to think of something that takes about 1/10 of the frequency wavelength distance. A wire Become a feeder? In any case, don't listen to those who hate the article, thank you for your article.

According to experience, we say that any trace longer than 1/10 of its transmission frequency distance should be regarded as a transmission line.

Although it is usually only used for sine waves, for square waves we tend to look at the rise and fall times.

There is a white paper (MT-097) for analog devices, they said to multiply the rise time (in nanoseconds) by 2 and use the result as the maximum trace length (in inches), and then suggest to treat it as a transmission line .

If you don't think of it as a transmission line, all signals will degrade, but by applying a rule of thumb, you will be fine in most cases.

"...This is by no means a perfect sine wave; it is a combination of waves of different frequencies."

I think you mean this is by no means a perfect sine wave. It is a combination of multiple sine waves of different frequencies. "

It is also worth noting that inputs (such as ADC pins) can also be noise sources. TI microcontrollers are notorious for this behavior. Regardless of the MUX selection, the ADC clock will be output through the pin.

It sounds wrong.

Are there no separate AVDD and AVDSS or VREF + and VREF- on the chip?

We have sorted out the pages about RF PCB functions, and briefly outlined some commonly recommended RF materials and bonding materials that can be used in various applications and industries, including consumer electronics, military/aerospace and medical.

Check it out here:

Please be kind and respectful to help make the comment section great. (

)

The site uses Akismet to reduce spam.

By using our website and services, you expressly agree to our placement of performance, functionality and advertising cookies.