Questions on damping resistors
william chan
Posts: 1,326
Hi all,
1. When are damping resistors required?
2. If the pcb tracks are short, are they still required?
3. What clocking frequency onwards will they be required?
4. How to calculate values of damping resistors?
5. Why some circuits use them and some circuits don't need them?
6. Can the PCB tracks be made thinner to avoid the need for damping resistors?
1. When are damping resistors required?
2. If the pcb tracks are short, are they still required?
3. What clocking frequency onwards will they be required?
4. How to calculate values of damping resistors?
5. Why some circuits use them and some circuits don't need them?
6. Can the PCB tracks be made thinner to avoid the need for damping resistors?
Comments
What does a resistor actually dampen? It is a completely passive component that essentially converts electrical energy to heat?
We generally use resistors to pull-up or pull-down a digital signal, or to divide voltage.
On the other hand, capacitors and coils are used to either create or damping oscillations. Thus, I am thinking you are wanting to know about cominations of capacitors and resistors that are used as a filter, not resistors alone.
Is that a better focus for you? If so, I am sure others can help you with this, and so can Google searches.
==========
Alternatively, are you mentioning resistors used in parallel with crystals to assure correct operation?
The Propeller doesn't seem to require any. Older microcontrollers needed these to assure that the Xtal would clock properly. Older microcontrollers also might require a couple of picofarad capacitors to perfect the circuit.
The problem with finding the right value is that you really need to bench test each and every crystal you install. And that requies a sophisticate piece of equipment that is too expensive for hobbyist and small manufacturers (costs thousand of US dollars). Most of us go away with a recommend approximation that worked, and never bothered with being so specific.
When I design board for SX chips, I just include pads for the resistor and caps ... just in case. Then I build and see what happens. If the device won't clock, I add values for caps and resistors that are recommended. If those don't work, I will sway until they do.
In sum, the Propeller is so much easier.
http://www.ti.com/lit/an/scba012a/scba012a.pdf
talks about damping resistors...
If you are not using ribbon cables or backplanes in your project, I think you really don't have to concern yourself with these.
What happens in long parallel wires is that you actually get a standing wave that will arrive at the end of the wire and reflect back toward the original device. This reflection corrupts the original data, so dampening resistors are used, at the far end.
+++++++
If you are using RS-422 and RS-485 transmitters and receivers, you do need these and they are generally 100-120 ohm values. Some engineers use a fancier configuration for very long or complex networks.
+++++++
The real dilemma here is that we tend to focus entirely on micro-controllers and digital circuits on small boards. Long wires come into play when you design over distances way beyond the circuit board as the wave lengths of signal are more corrupted as the distances increase.
Where do electronics engineers concern themselves with this.
A. Antenna to transmitters and antenna to recievers
B. Serial data connections that run hundreds of meters to kilometers
I am sure this is something I hadn't thought of... I hope others will comment.
+++++++
Historicall, the S-100 bus microcomputer had a backplane where damping resistors were extremely important. The card slots on your PC likely has dampening resistors to assure that the bus will behave well.
But if your are not creating a project with modular cardslots, you generally won't need these, unless you have a scheme for RS-422 or RS-485. Note that the TI document is all about transciever interfacing.. that implies the need to transmit and receive over long distances.
What are you up to?
Then why the LameStation has 12 damping resistors between the Propeller Chip and the graphic LCD when the transmission distance is only 3-4cm?
Consider: A capacitor and an inductor in parallel. The classic LC tank resonant circuit. When excited with some pulse that pulse energy will cause the tank to oscillate. With frequency f = 1 / (2 * PI * sqrt(LC)).
If that capacitor and inductor are perfect, i.e. no resistance, the oscillations will continue forever. The energy can go nowhere (Well, OK it will radiate EM into space, but let's ignore that for now)
If the there is some resistance in the circuit, say in the coil windings, the energy in the oscillations will be converted to heat and the oscillations fade to nothing exponentially. That is a resistor damping oscillations. More resistance more damping.
It's analogous to a swinging pendulum, friction will eventually stop the swinging. Add more friction and there is more damping.
Now, in a circuit one has inductance everywhere due to connection lengths and one has capacitance everywhere, between connections. So ringing is everywhere to.
I'll leave it to others to suggest when that ringing may become a problem and what to so about it. It's a black art to me. In the past I have relied on sensible circuit board and interconnect layout and normal decoupling capacitor usage. I don't recall ever having to apply damping to any of my troublesome circuits. They normally get fixed by trial and error
That would be limiting resistors in this case.
Correct this if the LCD it is using is 3.3 volt.
as a long transmission line! Think about the wavelength of a 1.5GHz pulse stream, you have
important harmonics upto perhaps the 9th, so 14GHz, wavelength in free space is 2cm, in
FR4 PCB about 1cm....
Basically everything in this regime is supposed to be a well terminated transmission
line to keep things purely resistive and preserve waveform shape (aka prevent ringing).
But the signals have to go into an IC package via pins or solder balls, where the impedance
is not maintained - you get reflections and resonances. Adding the series resistance (usually
a fraction of the characteristic impedance), such as 47 or 33 ohms will
damp down reflections (which have to go through the resistor and back again) a lot more
than the original signal (which will be attentuated a bit). The classic example of this is
USB where series resistors are used in D+/D- lines. Its less often needed with BGA devices
as the transmission line geometry is fairly well maintained going into the chip due to the
small physical size of the interface.
With unbalanced high speed signals the same principles apply, but the transmission line
is between the signal and ground plane. Sometimes you'll see a high speed bus routed with
ground (or Vcc) traces every third wire, so that each signal has a neighbouring ground trace.
The high speed 74LCX family have some examples of this, such as the 74LCX16374 and
74LCX16245. Also the pin out of DIMM modules follows a similar pattern, and every data
line has a small value series resistor on the module right next to the connector to help
control reflections from the inactive DIMMS during a transfer to/from another one.
I'd reckon from 50MHz upwards these issues start to become important, from 100MHz up
essential... The key insight is that a transmission line acts as a pure resistance and has
no low-pass characteristics until the wavelength of the signal is small enough to be within
an order of magnitude of the conductor spacing - so all the signal path design is to do with
propagation of waves along a transmission line and reflections and impedance mis-matches.
You don't think in terms of capacitance and inductance any more.
The signal may only be low frequency but if your logic is capable of fast edges those edges will generate the harmonics up to high frequencies as noted above.
So a single edge can cause multiple edges to be seen at the other end of the line.
Not usually a problem but it has caught people out.
Still, I'm surprised the LameStation has those resistors. Let's hope we hear from the developers about that.
Actually, in a bus or long wire configuration it is damping the relections of oscillations (at the far end), not the oscillations directly.
This is not a tuned oscillation context. That is yet another design topic and those resistors (if any) are likely not dampening.
Yes, there is paracitic resistance everywhere. But should we really become concerned with that in answering William's query?
And the main point is that bus trancievers are not generally useful in microcontroller projects. The TI document focuses on that topic. I am certainly not interested in building 8 to 32bit wide parallel buses and backplanes. SPI and I2C seem to avoid the problems with lesser chip counts. I actually had an IMSI S-100 computer at one time and never could get it working (no BIOS for the video/keyboard card, and very little to explain what was required).
Both the concept of Single Board Computer and System On Chip have removed the need to deal with these issues in most cases. When we need a long wire link, we use RS-422, RS-485, or USB. Even hard disks have switched to SATA, which is serial as those long parallel connections were troublesome. It is easier to chase down a bad wire in a serial bus than a 32 wire parallel bus.
How is the reflection of an oscillation different? In an LC tank the capacitor sees a reflection from the inductor and vice versa.
At the end of the day it's all down to Maxwell's equations and removing unwanted energy from the circuit.
The issue of switching from parallel to serial interfaces, like SATA, is interesting. Seems that when signals are fast keeping a whole bunch of them in sync in a parallel bus, which requires a clock, gets difficult. All the bits start arriving at different times. Better to just bang them out in serial as fast as possible.
I mean that the standing wave goes down the wire and hits the end and bounces back. The resistor is there to absorb the energy so that it does not bounce back and garble the information that is in the actual transmission.
Obviously, you never learn anything about antenna cable design, when the goal is to get the cleanest signal out to the antenna and the strongest, cleanest signal into to the radio receiver.
Oscillations are driven by power. Reflextions are driven by wrongful construction and termination of long wires, such as antenna lines, communication lines. Sloppiness causes problems.
You are the first that I have ever heard argue that there are reflections internal to an oscillator design that just might require a damping resistor. A bit absurd to me. But then you do love to debate so.......... hubris..... Such situation oscillation is desired and controlled to one and only one frequency in a sine wave.
You just a digital code monkey!
I don't like the use of words like "actually" to imply that what follows is known fact when it is not.
Quite so. I do no have an argument with that.
But, how is that different from my LC tank example. In terms of simple physics energy leaves the capacitor and is reflected back by the inductor. Then it's refelcted back by the capacitor. There is your oscillation.
No. Not exactly. You need to put some energy in to get the thing to oscillate in the first place. After that the oscillations drive themselves with no further power input.
Again, think of that pendulum that you push and let go. It swings happily all by itself. As does the LC tank. As does the transmission line. Really?
The name of the game with antenna cables, or any transmission line is impedance matching. That ensures the maximum transfer of energy from one end of the cable to the other. From transmitter to antenna. One does not usually want to waste power with resistors in an antenna circuit. Perhaps this is the first time you have heard it put that way. But physics is physics. Who said "require"?
However, as it happens, if you want to build a low distortion sine wave oscillator you will indeed be introducing damping so as to limit the amplitude of oscillation and ensure low distortion.
There you go with that "hubris" thing again. Seems most unfair.
It could be very well to tame the rising and falling edge ringing. There's a large high frequency component in sharp rise/fall edges that may require quelling.
Otherwise, the designers might have installed them "just because" they'd seen it done elsewhere. Having their input on this would be revealing.
Should I use poppycock or balderdash instead?
All these challenges to my writing style and its interpretation are difficult to anticipate. Sorry if I went overboard... must be the Lunar Eclipse/Full Moon/Blood Moon - Werewolf effect.
+++++++
Regarding the electronics and physics, we seem to have some fundamentally different points of view.
Oscillations without an input of energy with decay. That is why one winds or resets the weights a mechanical coocoo clock. With electricity, it this may less obvious. If my motorscooter begins to shake for some reason, I ease up on the throttle until oscillations abate (adding throttle will put you in the hospital).
What you consider to be included in oscillations, I consider to be 'Ringing' -- like ringing a bell. These naturally decay and are not part of long-wire reflections that I mention.
Cables are not merely about impedance matching, while RF amplifier stages are indeed about tight impedance matching.
I strongly suspect you are a bit weak about impedance matching. Audiophiles abuse the term, and so do many others. IN audio -- due to a wide audio bandwidth, the audio amp stages output impedance is often set lower, while the input impedance is set higher -- an intentional mismatch that works well. But the speakers have a relatively low impedance 4, 8, 16 ohm that is driven by a final power stage -- again, the right thing to do to make it all work, but not really a match.
In RF transmitter antennas, you have a particular carrier frequency that everything is tuned to the carrier frequency and impedances are tightly matched; and the cable to and from the antenna is merely standarized (50 ohm or 75 ohm coax) so that tuning is more predicatable. But the physical length of the cable might be set to optimize the carrier frequencies wave length. And at the end, there are design considerations to prevent a wave to relect back into the transmitter. Otherwise, the transmitter might actually burn out the final stage due to power reflected back into it.
RF oscillators and tank circuits tend to use there own modeling terminology, much based on their history and traditions of explaining. It is more like a fly-back effect than a reflection.
We simply disagree on what is the proper usage and context of 'reflection'. In my American university Physics education, we merely used 'reflections' in the topic of light -- photons bouncing off a surface. I learned the other usage from an ARRL Handbook (circa 1961 or earlier, in prepartion for a HAM radio license).
Of course, the Brits do say lift for elevator, bonnet for car hood, and much more. So carry on. I won't call you any more names.
I would suspect that the Lame Station wouldn't bother with insertion 'just because' they heard or saw this elsewhere. They likely had an actual problem and the problem went away with a few resistors added. In starting with remedies for wave shape, it is easiest to add a resistor first. Why? They are not as frequency dependent as capacitors and coils. There is always some inherent capacitance that might need tweeking.
Many large scale production boards (like the one inside your monitor) have many unpopulated pads that we provided 'just in case' a tweek was needed to get the right performance. It is easier to build a board with lots of extra alternative pads than to try to get the component count perfect.
My approach is to start by examining the following:
1) What is the characteristic impedance of the transmission line? The line could be single wire, PCB trace or traces, coax, twisted pair, flat cable or even air wired.
2) What is the rise time of the driving device from 20% to 80% in nanoseconds?
3) What is the length of the line to be driven in nanoseconds? 8 inches = 1 ns.
4) Is the line bi-directional or single ended? Single ended, single termination point is easier.
5) What is the source impedance of the driving device for both a high and low?
If the round trip delay (2T in the TI article) of the transmission line is less than the rise time of the driver then reflections are not an issue. In your TI article you may start with a circuit-signal like Fig 8 but you want to get to Fig 10. This is accomplished by matching the output impedance of the driver with the transmission line. If the line is 50 ohms and the driver is 8 ohms add a 33 47 ohm resistor in series at the source. The line (if open circuit at the far end) will reflect ~100% of the pulse but it will be absorbed by the series resistor( or dampened) at the driver/resistor output.
In Fig 12 you see how the reflection causes an undershoot of -1 volt. This is bad because the input diodes in the receiving device must absorb the excess energy or risk damaging the input transistors. Proper impedance matching corrects this as seen in the lower Fig 12 diagram.
This is only the tip of the iceberg and quickly gets complicated as you add multiple drop points, bidirectional lines, differential lines and as signals change layers in the PCB. Different PCB layers can have different impedance's that can only be moderately controlled by the board house. So I compromise impedance's and live with being off by 10% - 20% or more. Even with these imperfections the boards all function fine.
It might be easier to address your specific application since there are so many aspects to this subject.
Ummm... I take a different view.
You need to put some energy into a system to get something to ring (based on a resident resonance). After that you need to sustain the ringing with inputs of energy applied in the proper phase.
Otherwise, you have finally achieved perpetual motion. William seems to looking at ringing - not reflection, nor oscillations.
++++++++++
Aside from Heater and I, I see that others are giving answers to William's whole list of questions. I wasn't aware of the problems with 50Mhz and 100Mhz being barrier to board design. There are some very informative replies.. thanks to other's participation.
Start with the decision tree on page 11 of the TI app note and go from there. Keep in mind that the definition of a "long" line depends on the frequency/rise/fall time of the signal as mentioned in one of the posts. The signal line has distributed inductance and capacitance so adding a series resistor to the output will create an LRC circuit that acts like a low pass filter so the current and voltage rise time increases. The resistor also dissipates energy so any reflected signals are also reduced.
I think an analogy of sound echo could be used to explain the problem.
In a large hall you get a lot of echos, but in a small room you can't hear any echos. ( even though echos still happen )
This relates to long transmission lines having reflection problems not seen in short transmission line circuits.
In a large hall prone to echos, it is easier for the audience to hear you if you speak slowly compared when you speak at normal speed.
This relates to lower transmit frequencies having less problems than higher transmit frequencies.
So since the transmission line on the LameStation is only 3-4cm and the Propeller can never communicate with the LCD faster than 20Mhz, can we safely conclude that the 12 damping resistors are unnecessary?
I am trying to use a similar graphic LCD in my pcb design and seeing the damping resistors on the LameStation got me scared.
"You will be installing small 15Ω resistors into R4-R15. These are damping resistors. The LCD screen has to update very fast, so the microcontroller has to send a lot of data very fast. This can create interference on the line, so these small value resistors are put on each data line to reduce that noise."
Noise reduction (EMI?). The resistors apparently slow the prop edge rates to the LCD to reduce radiated noise or "interference on the line". Hmmm, this is a new one for me but an interesting idea. Makes sense. Just wondering which "line"? I've seen "FAST" parts switching bus pins at the same time and lifting the ground by a volt!! An 8 bit bus on a flat cable with one ground pin to handle all the switching currents coming back is pretty messy.
@William I'd say your options are 1) put in the 15 ohm resistors or 2) add more ground wires to the cable.
@Loopy: Nothing special happens at 50/100MHz, just a little bell goes off that says pay attention to high speed affects. It's the edge rates that carry the high frequency; a 100KHz square wave could have a 1GHz component.
-Phil
A second opinion? Or even one from the designer............
We can go around and around on theory and speculation, but nothing beats a test bench. Anyone that really is in a position to give a complete answer would have to run real tests for real data.
If you have a good scope and are willing to removed them, you can see exactly what happens before and after. (You can always put them back in place.)
The simple fact is the 'art' in electronics tends to be in tweeks that make devices reliable. And developers tend to be unwilling to share those particulars as they loose loyal customers to clones. It seems that Lame Station has given an honest and open explanation, but a scope would tell all.
@William Chan
You dilemma is simply the nature of copying a design. So provide pads for the resistors and start out without them. If operation is okay, don't add them. If problems occur, you don't have to reorder new boards, and you can use a scope to verify that your resistor values are optimal --- not a formula.
@Domanik
Thanks for keeping answers on the right track. I've learned a bit.
I think since the transmission tracks are short, the reflections would be much less than 1ns, meaning it can't be viewed on a scope. ( unless maybe a very expensive scope )
As the damping resistors are in series, I can't just leave the pads un-soldered. I would have to cut some tracks if I want to solder the shorted resistors.
My curiosity got the best of me so I simulated 3 possibilities: 1) no resistor 2) 15 ohms 3) 60 ohms
The PropOut 1, 2, 3 are: 40 ohm source impedance, 2ns rise/fall time, 23ns pulse width, 50ns cycle (20MHz), 0V to 3V.
The transmission line is 110 ohm, 0.8ns (6"), non-lossy.
Parasitics added: capacitance 5pf, .04uh inductance, .5 ohm series resistance.
LCD input pin of 6pf (estimated). Don't know if there's any termination resistance- 300 ohm kills most ringing & overshoot.
In #3, 60 ohm resistor, a close up look at node "A" shows the notch due to the reflection half way up the rising edge.
I'm guessing, in the original design, 15 ohms was added because it made it work.
Hey, what a great demonstration of what we are up against. I hadn't ever thought about transmission line models in Spice.
Is it possible the LCD is a 5v device? I'm guessing that if it is the logic thresholds are higher up than one would expect for a 3.3v device. Which I'm also guessing means one has less lee way for immunity to ringing.