With a new Kenwood 5.1 receiver acquired from questionable sources, [PodeCoet] had no way to buy the necessary coax. He did have leftover Cat6 though. He knew that digital requires shielded cable, but figured hacking a solution was worth a try.
To give hacking credit where credit is due, [PodeCoet] spent over a decade enjoying home theatre courtesy of a car amp rigged to his bench supply. Not all that ghetto of a choice for an EE student, it at least worked. To hook up its replacement he pondered if Cat6 would suffice, “Something-something twisted pair, single-sideband standing wave black magic.” Clearly hovering at that most dangerous level of knowledge where one knows just enough to get further into trouble, he selected the “twistiest” (orange) pair of wires in the cables. Reasonable logic, one must select the strongest of available shoelaces for towing a car.
It worked perfectly unless anything electrical happened. Any electrical occurrence at all (plus, according to him, occasional Arduino voodoo hexes) caused the Kenwood’s audio relay to trip off and go into standby mode.
Probably gamma rays from the freemasons.
Umm, nope. Well, when one buys “second hand” products like this that is possible, but, Occam’s Razor gives you a more likely solution.
It turns out that replacing the cable with some shielded coax made the problem disappear. So that clears that up; unshielded Cat6 will not work for digital audio cable. That is something many of us did not definitively know until now.
To our smarmier readers who blew a gasket over [PodeCoet] trolling his Prof last month, allow yourselves a smug moment to bask in the light of his karma coming due. He owned up and submitted this fail. The rest of you should do the same so we can learn from your mistakes and enhance the greater knowledge of the community and well-being of mankind everywhere.
Fail of the Week is a Hackaday column which celebrates the learning potential of failed projects. Help keep the fun rolling by writing about your past failures and sending us a link to the story — or sending in links to fail write ups you find in your Internet travels.
95 thoughts on “Fail Of The Week: Cat6 != Coax”
Having something 100m of coax always comes handy.
Could’ve made it differential signalling(e.g. RS-485) and it would’ve worked with the CAT6 cable. Or use a simple current loop transmitter – added isolation is a bonus too.
(exposing my ignorance here) Is it the shielding that is important or the impedance? Did he actually need coax? or just shielded cable? (when I hear coax, I think tv cable coax and numbers like RG-58). Before bailing on the the cat-6, having got as far as noting the electrical noise, I would have tried attaching all the lines (or at least all the white/striped ones) to the common/shield to see if the collective twisting would have helped calm things down.
Impedance isn’t the problem.
Long cables have two main problems when it comes to transmitting fast-moving signals: induced noise from external EMF sources and the distributed inductance/capacitance of the cable itself.
Shielding reduces the noise from induced EMF by arranging the shield so it acts as a Faraday cage around the core wire. Connecting the shield to GND dumps any induced current into the low-impedance return path, and any voltage generated by the current gets included as part of the reference voltage. All voltages in a circuit are measured relative to the GND rail, so the calculation becomes (signal+(noise*C1))-(GND+(noise*C2)) where C1 and C2 are constants that represent the attenuation in both cases. At least some of the noise gets subtracted out.
Twisted-pair wire reduces noise by arranging the wires as two coaxial helices. The average path of both wires is exactly the same line, so in theory the induced current in both wires should be exactly the same. The same “subtract one from the other and the noise falls out” principle comes into play, but this time the two constants are practically the same. It works best when you use differential signals in the wires.
The parasitic inductance and capacitance of the wire create what’s called a ‘transmission line’ which introduces two problems: first, it takes a certain amount of time for the signal to travel along the line, and second, the inductance and capacitance form a low-loss resonant circuit. Signals passing through the line reflect back and forth, causing the ripples you see at the tops of the square wave pulses in the picture above (called ‘ringing’).
Getting rid of ringing requires proper termination and impedance matching, but usually involves increasing the resistance along a cable to dissipate energy. The trick is to get the amount of resistance necessary to get near ‘critical damping’, which gives the fastest possible response without ringing.
Sometimes the wires are pre-wired inside the wall, it is our best interest to get things connected without using additional wires. What does it take to use Cat6 to replace Coax? Would wiring 4+4 instead of 1+1 be better?
You need to covert from single ended to differential signal. How you do that depends whether you use analog or digital signal.
You need something called a “Balun” which stands for Balanced to Unbalanced. They are made for this exact purpose.
Exactly. For those curious about how this works with networking, those black epoxy blocks on network cards contain the baluns used to convert the single-ended digital signals on the card to & from the differential signals that travel down the twisted pairs. In theory, it might be possible to make up a hack, using the baluns from a pair of 1000bT NICs, into a general-purpose gizmo for sending digital signals like this down existing Cat-5E/6 structured-wiring systems.
I have actually tried to send a digital TV signal over 10 meters of Cat6 cable recently. I started by using one pair for the signal and ground. This gave me a couple of channels. I ended up using 2 pairs for signal and 2 pairs for ground without using the shielding. This gave me about 95% of all available channels, some of which were HD. I experienced some digital noise now and then however.
I once successfully ran composite video down a single pair on a short run – maybe 5m – of Cat-5 cable. I ran the power for the security camera down another pair. It worked very well.
Can’t you run composite through pretty much anything and get something intelligible out of it?
Yes, even if it is not 75 ohm impedance cable.
With coax, the core carries the signal and the shield carries the noise. The system pulls the shield signal out of the coax, and you have a purer signal.
With the twisted pair, they are sharing the noise AND the signal through induction. Remove the shield signal, and you remove the core signal. At the same time, your impedance is jacked because of induction, and it might as well be a short.
Ehm, I hope that was an attempt at a joke, mate!
“With coax, the core carries the signal and the shield carries the noise.”
well it is sort of true: the core carries the signal while the shield absorbs(and carries to ground) the noise. It’s not a complete description but it is partially true.
connecting all the other wires to ground could help
Agreed — help the other 6 wires do their best at mimicking a shielding jacket. I would also give a try bundling the orange pair together on the center pole.
The extra coupling from the grouns would also lower the cable impedance.
Agreed. When I first started making custom CAT5 cable runs for work, I figured why bother with unused wires? Keep it simple and crimp only the four actually used. It worked fine, at least until they fired up a TIG welder (an unintentional spark gap transmitter), then the whole network crapped out. Recrimping the ends properly completely solved the issue.
It should certainly help, at least a little bit.
Safer to say “from a flee-market”.
As in, a market one should flee from very quickly?
You’re missing two bits of context. He hints (jokingly) at it being acquired as part of a drug deal over the holidays weekend, hence all the stores being closed and him unable to buy coax until the next day.
Hence, this not really being that bad of a fail, more of a “make do with what you have” and “oh, that’s why we don’t do those things.”
There are multiple issues here:
1. twisted pair works for shielding when the current in both wires is the same and opposite and the readout is differential. If the source and receiver are grounded together through some other place(like mains) there is no current through the ground wire in the twisted pair, thus making the other wire act as a simple plain wire. Twisted pair is to be used only with differential signalling.
2. It’s the different cable impedance that can also screw things up. Twisted pair should be higher(100ohm) vs the 50-75ohm coaxial. If the terminating resistor is 50ohm, then you would get more attenuation
Simplest fix is probably to connect all other non connected wires to ground.
Unless digital audio switches at many 10s of MHz, or the cable is electrically long then the impedance of the line won’t make much difference. Rather the reason for shielded cable is that the signal is single ended, and the only noise immunity techniques available is to shield the signal conductor with the ground wire, or in other words coax. As an aside coax does not necessarily mean impedance controlled, guitar cables are coaxial, but do not specify a characteristic impedance.
The problem here is that twisted pairs offer no noise rejection/immunity for single ended signals. Furthermore, the close proximity of several unconnected wires within the cat6 cable might have made matters even worse as they will act as antennas to local EMI, and then couple this to the signal/ground twisted pair.
Guitar cabled operate at lower frequency and the impedance of the cable does not need to be matched to the one of the next stage, nobody cares about reflections here.
A 50ohm cable coupled with 50 ohm input of next stage transfers 50%. But a 100 ohm cable with a 50ohm next stage transfers way less, which might not be enough to get clear 1/0 levels.
50ohm source, 50ohm cable, and 50ohm load transfers 100% (assuming lossless). Either that or my network analyzer is faulty… S21 =0dB when source and load are matched.
Also for high frequency digital signals a mismatch results in reflections and ringing on transitions as the transmission line effects, and mismatch effects will be more severe for the higher harmonics.
If we’re talking in general it depends on the frequency and length of cable. If its electrically short (less than about 1/20 the wavelength) then the transmission line impedance makes very little difference (providing the cable doesn’t do something very strange so really its a lumped element), rather just the source and load impedance. For guitars actually the high impedance of the pickups means that a high impedance input is required for a preamplifier, otherwise signal is lost (particularly higher frequencies), in this instance its in the electrically short regime.
Given that the unit operated ok until there were electrical occurrences (guessing this means electrical things turning on and not orb lightning), I’d hazard a guess that impedance mismatch was not the culprit, but rather EMI. Sure the mismatch might mean its more susceptible to EMI due to signal loss already, but as the scope trace shows they’re pretty acceptable to begin with).
S21 is a ratio, so 0dB means some 2 things are equal, some thing gets split in two identical halves. I’ll let you read on the S parameters a bit to figure out why matched networks are not 100% efficient.
That amp is an analog amp, no digital audio.
we are all assuming the amplifier is using a digital input which is converted to analog and then amplified.
The manual for that amp says it has only line level audio inputs. No digital audio inputs.
Well, this “With a new Kenwood 5.1 receiver acquired from questionable sources, [PodeCoet] had no way to buy the necessary coax. He did have leftover Cat6 though. He knew that digital requires shielded cable, but figured hacking a solution was worth a try.” and this https://hackaday.com/wp-content/uploads/2015/01/had-coaxfail3.jpg?w=600&h=309 make me think the core problem is some digital transmission from a digital source to a digital receiver. whether that receives is in the amp itself or separate…. i don’t know.
Photo on the linked article shows optical and digital/coax inputs: http://ultrakeet.com.au/graphics/images/FC02/FC02_AmpCoaxInput.jpg
Just checked the manual – page 9 clearly shows three digital inputs, two of them are optical
Doh! I was looking at the car amp he was using. Talk about reading comprehension!
Truly digital amplifiers are very common nowadays. They work by using Delta Sigma to directly convert PCM to PWM/PDM.
A Balun on one end to do Unbalance to Balance (and impedance matching), then another on the far end to do the reverse. Unfortunately, the fix is way more expensive than getting the right cable.
Assuming that’s SPDIF or something digital – scope trace.
Actually you can make your own baluns from lengths of wire.
You have to make sure that the core in the Balun matches the frequency range of the signal.
“So that clears that up; unshielded Cat6 will not work for digital audio cable.”
Unless it’s carrying AES/EBU.
It would work if converted to differential first. Like using a RS485 converter.
Or a transformer (balun), like it’s usually done in AES/EBU XLR’s.
That’s because AES/EBU is balanced. It’s true that “digital audio” is a vague term but we can assume it means SPDIF.
Unshielded CAT6 works perfectly fine for AES/EBU over reasonable distances. Heck, a unshielded twisted pair of wires works for AES/EBU. The fact it’s consumer/single ended is the issue! Wrong cable for the signal.
AES/EBU in XLR(balanced) configuration is a 110 ohm transmision line spec, use a cheap microphone cable and you will likely get data coruption, I had many service calls for “bad” DAT recorders in studios where the fault was in the cable not the deck back in the ’80’s. CAT5/6 is 100 ohm and close enough for a short run – AES/EBU is specd to 100 metres with proper cable.
At a shop where I had worked there was a problem with a magnetic pickup sensor, and I fixed the problem with a phone cord (it was on hand and worth a shot) I wired the sensor up properly and wrapped the extra wires in a very similar fashion to the photo. On the other end I connected the wires, and all of the extras I stripped back and tied to ground. It was enough to keep the noise out of the line.
And it is 98% probable that phone cord is still in service.
I would think that what happened here is
a) impedance mismatch, the coax inputs/outputs for S/PDIF are 75 ohms, and the impedance of Cat-5 and 6 is around 100 ohms as I recall…
b) single-ended signal over balanced cable, without shielding.
The funny thing is that there is a balanced version of the interface that could probably work over Cat-5 and 6, but you would need the correct drivers and receivers to pull that off.
See Also: https://en.wikipedia.org/wiki/S/PDIF
SP/DIF *should* be transformer isolated, effectively putting a balun at each end. But cheapo consumer electronics mfrs sub the transformer for grounding the shield to chassis and using a resistor matching network.
Apparently EE students only take the micro-controller courses these days. He should have to transfer to liberal arts for a semester as punishment.
What, and make life *easier* ??? :p
A little google searching turns up a reasonably well done explanation of the difference between balanced and unbalanced transmission lines here:
And here’s a circuit he could have used to have actually made this work:
Whoops, my very own HaD fail. I seem to have missed that this was for digital audio, the link to the circuit I posted above will only work with analog audio obviously. I’m assuming this is S/PDIF, searching the Internet shows a number of websites that sell baluns to go from S/PDIF to cat5 which would be handy if you had a house that was prewired for Ethernet.
TL071s for transmitting SPDIF data? Seriously doubt that :-p
That circuit is ok for analog audio, e.g. to an old analog mixing console/multitrack recorder – those often have balanced inputs. It would be waaay too slow for SPDIF. Even for analog audio one uses rather NE5532/5534 or similar, because those TL071 are quite slow.
“Clearly hovering at that most dangerous level of knowledge where one knows just enough to get further into trouble”
…story of my life.
Yeah, but from here there is usually a clear cut to where you are going: either screw something up, get scared and stop or…. survive, learn something and move on.
we have used lot of cat5e cable at balanced lines. Actually we have color coded xlr-rj45 adapters what allows we to put 4 signal to same cable. No problem at all and works fine at least 100m. Naturally balanced is not digital audio and that my comment has nothing to do with this article (except audio) but I just want say that because if someone in here want organize big lan party and naturally you have lot of cat, it is just fine for balanced lines. But remember that dont try put DMX to same cable, it will leak to audio at least in utp cable.
Based on a standpoint of no knowledge of the subject, couldn’t you ground the unused cables, to act as a crude shield? I doubt it’d work properly, but it might reduce the problem. maybe.
You can use a part similar to a FIN1001M5X LVDS differential driver to convert SPDIF to differential SPDIF. That can be sent over a twisted pair to be received by a 100 ohm terminated receiver IC, such as a DS90LV018ATM. If you have a differential signal, then a compatible impedance twisted pair works very well.
LVDS parts do not have a large common mode range, so you don’t want to use them outside a PCB or a backplane. i.e. if there is a difference in ground voltages which can easily exceed that range. RS485 parts especially the old bipolar parts on the other hand can handle +/-10V common mode range would be more suitable. Use a couple of right type of Balun if you need even more common mode range.
Depends on the cable length. I’ve personally seen this solution work up to 18 feet, but haven’t tried it farther.
It is not the cable length that has the issue here. I am more concerned with the voltage different between ground at the two ends. If the difference in the ground levels is over the common mode range, then there is problem.
I mentioned PCB or backplane level as those shares the same grounds vs boxes that have different power supplies connected with external cables. e.g. I wouldn’t use LVDS in automotive because of shitty wiring harness
FYI: LVDS common mode range: +0.2V to +2.2V, so at most you can have +/- 1V difference between 2 grounds.
vs +/- 10V for a bipolar RS485 part.
The twists in the cable causes issues with anything over 5v and can generate RF interference. Use two different colors and this will help alleviate the issue.
Also be sure the pairs are twisted in the correct direction for your hemisphere.
If you’re not sure of the correct direction for your hemisphere, just flush a toilet and notice if it goes CW or CCW and twist the wires accordingly.
I know you’re just joking but so many people have that misconception I thought this would be a good place to leave this:
Dead serious. That’s why GPSes have twice as many simultaneous channels as possible overhead satellites: half go through a right hand circularly polarized antenna and the others are for a left polarized antenna since satellites need to switch polarization as they cross the equator to deal with the change in electron spin.
Assuming it’s spdif he could have went full ghetto and go for the optical route with some fishing line or affixing a line of sight laser setup since the input is no different whether coax or optical just method of transfer :P
Raise your hands if you DON’T have a mess of coax just laying about? OMG how is this even possible? Hell, I have a 25 gallon tote dedicated to the stuff, and another 25 gallon tote dedicated to pc power supply cables. Seriously? if one thought about it just a little, they could make some!
I have a mess of cat5e laying around, as well as a tote of pc power supply cables, more stereo and/or video RCA patch cords than I’ll ever need, and some hidden device I haven’t found that turns microUSB-B cables into miniUSB-B cables when I’m not looking.
But I don’t have that much coax.
Unless coax is referring to something I’m not aware of. WHich could be the case. The connector in the article photo doesn’t look like any coax connector I’m familiar with.
Coax you’re used to are like how cable comes into your house, right? Ugly screw-on connectors.
The jack in the picture is RCA. Normally not shielded.
Shielding refers to a latticey-mesh that wraps around the cable (usually still underneath the final outer plastic jacket). It blocks stray signals from reaching the conductors and confusing the thing that reads the signals.
@PodeCoet – I don’t suppose you’re the very same but might you have participated in the h3c wargames way back when?
Nah man, sorry!
Assuming real S/PDIF, you have two problems. One is the it’s a low voltage signal, <1V, designed so that plugging an S/PDIF signal into an audio RCA input doesn't smoke your system. The second, as many have said, is the you need a differential driver. Put an RS-485 transceiver at either end (obviously, one in transmit, one in receive), shift the levels to work with the RS-485 chip inputs, and you've moved at least halfway toward AES/EBU, the original digital audio spec that Sony and Philips corrupted, er, modified to create their S/PDIF consumer version.
Whoa… Are you Dave ‘DiskSalv’ Haynie?
As a long time reader of Hack A Day I must applaud Matt Freund for his wonderful humour! (Yes that is how you are supposed to spell it)
Well, thanks but I don’t think any of my jokes made the cut. All the funny lines here are taken from [PodeCoet]’s hilarious failblog he submitted. That’s the man to applaud.
Wow. Digital / Analog brainfog.
This is going to hurt your brain!
Digital and analog signals go down cables. Cables can’t change voltage or transfer signal instantly because they have inductance and capacitance and they have to live in the real world where the laws of physics have greater voting power than one’s and zero’s.
Inductance and capacitance are *analog* properties so I am sorry to disappoint you by telling you that your digital cables are actually analog.
The output stage of a digital line (cable) driver is actually a modulator that translates a digital scheme into a real world analog signal to send down the cable. At the other end is a demodulator that translates the analog signal back into a digital scheme. So it really does not matter one iota weather the signal scheme is analog or digital because when it’s going down the cable it is analog anyway.
What does matter is shielding, impedance matching and in this case (most importantly) the signal type being balanced or unbalanced.
Cat 5 cable is rated up 100MHz. I would assume Cat 6 is higher so there is no issue about it being able to carry a 20KHz signal.
Balanced and unbalanced is *not* about how you *use* a cable it is about how it is designed to be used.
Balanced cables have twisted pairs so that *any* noise that is inducted into one wire of a pair is *also* inducted equally into the other.
If you have a signal at one end of 0 Volts and 5 Volts and you have a *noise* value of 20 Volts then at the other end the signals are 20 Volts and 25 Volts. *IF* this is going into a differential amplifier (which would be the case if the unit is designed for balanced cable) the the resulting signal is 5 Volts (25 – 20) – or the same as the input to the other end of the cable. If however the equipment receiving the signal was designed for unbalanced cable, it will earth one wire and measure the signal at 25 Volts ie signal plus noise (this is called an earth loop). If the receiving equipment has no earth then it will actually act as a balanced receiver because it’s ground signal will float in accordance with the noise plus signal. The earth loop goes all the way around through the house wiring from transmitter to the receiver and acts like a big antenna for noise as well. Then as I have mentioned, if it also goes through the signal cable, it completes the *loop*.
A small signal transformer at one end would have fixed the problem because it removes the earth connection from the signal wires and breaks the earth loop.
Digital is a wonderful scheme that allows us to design things without a need to worry (much) about the analog world. It does *not* however take the analog features (bugs) way. They will always exist simply because the laws of physics are in fact analog.
Brainfog clearing? much?
In the past I have found an easy fix for ground loops is to disconnect the signal shield at one end.
My home stereo is two good quality PA amps with “earth lift” switches on the inputs.
I bet the guy would get a better signal if he utilised the remaining pairs as a crude shield (and possibly only grounded at one end)
It amazes me how one can take the mean of comments ranging from inquisitive to informative, amateur to PhD, and derive a solution. Moreover, one can infer the underlying physics, albeit the hard way. There are also other less technically useful but nonetheless enamouring metrics that percolate from hacker comments — Thank you all for contributing to this unique and priceless oracle founded in chaos, the pursuit of hacking coolness and the sharing of the knowledge behind it. Hackaday Rocks.
“So that clears that up; unshielded Cat6 will not work for digital audio cable”
Works just fine. His problem lies in not utilizing every conductor, the ringing shows this especially with light switching. Hello crosstalk. I’ve used unshielded and shielded ethernet for everything from S-VID to Digital Audio to wiring the 60W of LED that illuminates my 55 gallon aquarium. Use every conductor in thecable or remove that which is not going to be utilized, and make SURE you have a proper resistor in-line to ensure you are close to the required resistance level of the cable.
This was certainly a fail, but not due to the cable.
[PodeCoet] had no way to buy the necessary coax…
Why not just go to you local retailer, that sells audio accessories, or someother store.
And buy shielded audio patch cable:
— there is no functional or electrical difference between a
stereo RCA jack and a digital coax audio jack (assuming it is not optical).
for optical toslink:
Page says it was a long weekend and everything was closed
Ok, I did not see that on the other page…but, the post said:
[PodeCoet] had no way to buy the necessary coax…
One can only surmise from that, that either they didn’t know where to get the cable, or just couldn’t afford it.
You made a fair assumption. Heck, a dollarstore coax would have sufficed.
I thought it important to explain the context of someone who appears so knowledgeable failing so hard. The article I submitted made two mentions of the late holiday weekend/stores being closed. Both mentions were deleted before publishing.
using the term “ghetto” in this manner is racist a fuck
I think that you insisting the term “ghetto” is racist reveals your own prejudices and guilty conscience, not mine. I suspect you don’t even know what the word means.
“Ghetto” and “slum” aren’t necessarily synonyms, one could technically have a prosperous ghetto, but no one uses the word that way. A “ghetto” just means and area inhabited particular sub-grouping of people, not necessarily by race. And when it does refer to race, it does not exclude any in particular. There are absolutely white ghettos. It is a classist term, but one can be classist without being prejudiced there as well, simply using terminology to differentiate rich vs. poor.
The area around colleges with cheap housing, basement suites, etc, is almost universally referred to as a “student ghetto.” It’s where the affordable housing jam-packed with too many roommates is, all of which are fairly poor because they’re students. The grouping type that makes the ghetto being “student”.
In the slum connotation it’s equivalent to “redneck” or “trashy” but those words don’t fit this case. The way I’ve used it here is clearly to indicate a mild sense of poverty where new things are a frivolous luxury. I even used it in a flattering, badge-of-pride hacking sense.
That said, one can use all the proper language in the world and still be plenty racist. You avoiding the term ghetto while (seemingly) implying that any group of poor people are a non-white racial minority reveals *you* to be the one with the racist worldview.
In the end the only person who knows how I meant the word is me, and I’m proud of who I am and how I treat people, so, better luck next time.
When I worked in Hong Kong, I lived in a very expensive part of the island that was known as Gweilo Ghetto, being jam-packed with foreign white people.
That hack might’ve worked quite a lot better if they’d grounded all the unused wires at both ends.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)