Differential Cable Run

[Chaos_Klaus]

Hey,

I have another basic question regarding the pulstransformers.

My signals is 5V HIGH and 0V LOW. The longest possible time the signal is going to be HIGH is 11 cycles. 81,4s*11 makes about 0,895µs. At 5V that makes about 4,5Vµs.

If I choose a pulsetransformer that has more then 4,5Vµs, I'm fine right? Is 6Vµs enough or should I use a larger margin for error?

If the signal switches to LOW after 11 HIGH cycles, how fast does the transformer recover? The signal won't be continuously LOW for another 11 cycles.

What about actual inductance? I use 1:1 transformers. Does inductance matter at all? I mean it obviously forms a RL high-pass filter with the output impedance of the line driver but apart from that?

Thanks!

[rstofer]

In terms of the dual voltage power versus the number of pairs available after two pairs are used for Ethernet, have you considered one of the 'brick' type DC-DC converters?  Here is a unit that takes around 12V and provides +-15V for op amps:
http://www.digikey.com/product-search/en?keywords=ncs6d1215c

[Chaos_Klaus]

Using $4 Jacks with $20 dc converters is not a good solution.

I'd rather use a regular RJ45 connector for $0.5 and a pulsetransformer for less then $1.5.

I only use one pair for the data connection anyway. With PoE there is also no problem in using a data connection for power aswell.


By the way ... this is what the waveform looks when I push it through 1m of networkcable with two Ethernet jacks (incl. transformers and termination). Looks like the transformers are saturated, right? For some reason the bandwidth is also too low.

[Richard Crowley]

We don't know what transformer you are using?
We don't know how you have the transformer connected?
We don't know what you are driving the transformer with?
We don't know what the waveform looks like at the primary (input) side of the transformer?
We don't know how you have the scope probes connected or how the scope is set up?
So we don't really have any way of evaluating your scope photo.

This is my reference for connecting a microcontroller to 10BASE-T.
http://www.instructables.com/files/orig/FYK/AONS/HY4KUNRU/FYKAONSHY4KUNRU.pdf
It uses a standard transformer and a Microchip ENC28J60 to handle all the Ethernet protocol.
The transformer primary is center-tapped, and they have the center tap connected to +5V
Then they pull down on one or the other end of the primary winding to produce "positive" and "negative" pulses.
In the schematic PDF cited, the transformer and RJ45 connector are in the lower right corner,
and the output circuit is in the upper right hand corner.

Of course, you are not using a microcontroller like that ENC28J60.
But you should drive the transformer the same way, from two open-collector logic outputs operating differentially.

[vk6zgo]

The Cat5 cable you were intending to use will make a good "draw wire" to pull some coax through.

Then you can stay single ended,& get better suppression of radiated signals than from your twisted pair.
No magic gadgets which will degrade your signal needed!

[Chaos_Klaus]

@Richard Crowley

Sorry. The Picture was only really a bodged together example. I probably shouldn't have posted it. I don't have the line drivers, yet. So I've been driving the signal directly from the optical receiver jack through a series resistor (to limit current), but it's not really designed to do that. I'll post proper pictures when I get the drivers.

The pdf is helpful.

Why do I need to tie the center tap high?
When I use a differential driver, why do I actually need the transformer at all? I could just drive that differential signal into the cable with series termination, right?

My actual question was in the post above. When using a single transformer, how do I know what inductance to choose?

@vk6zgo

Coax does not cut it because I want to transfer three power rails aswell, as the rest of the thread indicates.

[Richard Crowley]

Why do I need to tie the center tap high?

So that your driving circuit can generate both positive and negative output levels by pulling one or the other end to ground.  Most digital outputs are much better at pulling things to ground vs. outputing the high-level supply voltage.  And besides, pulling to ground is the same no matter what voltage your logic operates at.

When I use a differential driver, why do I actually need the transformer at all? I could just drive that differential signal into the cable with series termination, right?

Yes and no.  Yes, if you got everything just right, you could reproduce both the dynamic and static source impedance. But you would never achieve the isolation you get from a very inexpensive transformer.  That is why that scheme is not used in the hundreds of millions of ethernet ports around the planet.

The example shown isn't strictly a differential driver. It is a pair of open-collector outputs that can only pull things to ground, and they are driven with opposite logic values. When one is open, the other is grounded, and vice-versa.

When using a single transformer, how do I know what inductance to choose?

If you use a transformer that is designed to drive 10BASE-T Ethernet, then it is designed to be the proper inductance, impedance, etc.
If you use some other random pulse transformer, then you are on your own with a complex design task.
That is why people use transformers that are pre-engineered for 10BASE-T twisted pair communication (as for Ethernet).

You could TRY using a pulse transformer that is designed for 110 ohm balanced line, and use the same transformer at the transmitting end as the receiving end and it might work. But I continue to question why you continue seeking a non-Ethernet solution for a simple 10BASE-T connection?  An ideal solution seems so easy, simple, and cheap, it is not clear why you don't want to use it?

[Chaos_Klaus]

I don't get what makes my signal equivalent to an ethernet signal. Ethernet uses Manchester-Coding and therefor can probably get away with smaller transformers. My signal on the other hand can be HIGH for as long as 0.9µs. So I need a transformer that can handle 4,5Vµs. The Ethernet transformers I see only have ET constants of about 2Vµs or 3Vµs.

And I keep looking for different solutions because I don't just want to get the job done. I want to learn how it works. I mean, knowing how to transformer couple a signal is not a bad thing to know right?

[Richard Crowley]

It might be useful to review the OSI seven-layer interconnection model: https://en.wikipedia.org/wiki/OSI_model
Because you are using Cat-5 UTP and 10BASE-T, you are using layer 1, the PHYSICAL layer.
That is not "Ethernet" at the LOGICAL layers (data-link or network or transport, etc.)

You can take advantage of the cable, connectors, and transformers pre-engineered for Ethernet.
That does NOT mean that you have to send "Ethernet-like" packets, datagrams, etc.
It only means that you are exploiting the already-available sending, transmission and receiving engineering designs.
And you are able to send a bit-stream into the transmitter driver, and get the bit-stream back out the other end.
That bit-stream can be the continuous ADAT data-stream. It does not have to be the "Ethernet" protocol of addressed and routed packets.

You seem to be making this a lot harder than it really is.
Yes, I agree that learning how to transformer-couple a signal is a good thing to know.
And winding your own coils and even making your own capacitors and resistors was instructive also.
But here in 2016 there are more useful things to learn and know beyond pulse-transformer engineering.

Pulse-transformer engineering is rather complex and arcane. In English we have an idiom: "bigger fish to fry".
Which means that you have more important details of your project to work out beyond 10BASE-T physical transport.
I am just afraid that you will get bogged down in working out optimum transmit and receive coupling for your Cat5, and that will delay working on the heart of the project.  That is why I suggested using pre-engineered "Ethernet" transformer solutions which are a very mature and very inexpensive solution.

[T3sl4co1l]

Pulse-transformer engineering is rather complex and arcane.

Well, no. I wouldn't use those words.  It's a well defined science, and quite simple to a designer with experience.  I suppose if ones' opinion is "frikken magnets, how do they work?", it would be a tautologically impenetrable topic.

(Not asserting you are one; just as an example. )

In English we have an idiom: "bigger fish to fry".

This is more the case.  (I don't feel that the other sentence goes together with this... but maybe I'm a boorish American.)

If you're going to use a pulse transformer, then you have to face the reality that:
1. You probably won't find an off-the-shelf part that fits exactly.  Work around it.  Or,
2. You'll be designing a custom manufactured part.  Costs time and $$$.  Or,
3. There's probably a better solution (the bigger fish) that doesn't need transformers at all.

Custom transformers aren't actually all that expensive.  Manufacturers have reasonable prices, on account of the common need for custom transformers.  You won't need to spend $10 or $20 -- but, you can easily spend $5 getting a perfectly suited custom component, versus a similar but suboptimal part that costs $0.50 off the shelf in quantity.

I suppose the "blinding dumb obvious", "big fish" answer here is: you're already using optical TOSLINK, so put an optical TOSLINK device on it and be done!!  Run the extra wires for power, run two off-the-shelf cables, or have a custom combo cable made (might cost the same as a custom transformer coupled signal path, overall!).

I guess we're beyond that point, though, so small fish it is!

Tim

[T3sl4co1l]

By the way ... this is what the waveform looks when I push it through 1m of networkcable with two Ethernet jacks (incl. transformers and termination). Looks like the transformers are saturated, right? For some reason the bandwidth is also too low.

http://s9.postimg.org/4852dmwfj/DS1_Z_Quick_Print14.png

Low bandwidth makes me suspect your drive impedance is too high...

The transformer's definitely not saturating.  The "droop" during long constant-level periods is due to inductance.  It would be sudden, after a delay, if it were saturation.  (Likely, your signal doesn't have enough strength to reach saturation anyway.)

That waveform looks pretty good, to me.  It's still making zero crossings.  The timing might not be quite right (code-dependent jitter).  Some DC restoration would help, or an equalizing filter, or a tracking-threshold detector.

Pulse transformer utilization and design, and electromagnetics in general (transmission lines, antennas, radio theory, high speed theory..) depends upon impedance matching.  Get that right, and most problems disappear.  (You still have to worry about reflections, but that's part of the point: having everything well terminated at each end.)

Tim

[Richard Crowley]

It's a well defined science, and quite simple to a designer with experience. 

Certainly, no question.  But @Chaos_Klaus is not "a designer with experience", and neither am I, for that matter. At least experience designing inductive components.  I could probably study it and become a journeyman at calculating and winding my own pulse transformers.* But that seems pretty prosaic and uninteresting when you can get the solution already worked out for you, practically for free.  And the project was presented as a way to extend ADAT-style bitstreams and power over a distance of several meters. If the project is now learning how to design and build pulse transformers, then it has become a different question.

I am trying to make the point that @Chaos_Klaus has a bit-stream that he wants to send over 110-ohm balanced UTP.  And there are perhaps 100s of millions of examples of pre-enginered solutions for EXACTLY that situation inside everything on the planet that has an RJ45 ethernet jack.  The transformers, reference circuits, connectors, etc. are available pre-engineered, off the shelf at prices so low, the shipping will probably cost more than the goods.

I suppose the "blinding dumb obvious", "big fish" answer here is: you're already using optical TOSLINK, so put an optical TOSLINK device on it and be done!! 

Just because the bitstream is ADAT (which uses TOSLINK optical physical layer) doesn't mean that the source is optical.  And even if it were, TOSLINK/ADAT uses those low-quality, dirt-cheap, extruded "light pipe" which has a very limited useful distance. 1m is pushing it for reliability because it was designed as an ultra-cheap optical solution for adjacent gear. It was never designed for distances much over 1m.  To be sure, there are 3rd party gadgets that interface proper, drawn-glass-fiber optical cable to the cheap rectangular plastic TOSLINK connectors. But they are mostly over-priced, audiophool snake-oil products.

Run the extra wires for power, run two off-the-shelf cables, or have a custom combo cable made (might cost the same as a custom transformer coupled signal path, overall!).

I would sincerely love to know where you can get custom optical/copper cable made in small lots (300-1000m) at reasonable prices.  No cable manufacturer I've ever heard of would react to such a request without laughing.  And optical fiber is way too fragile for "tactical" uses unless you use the heavy duty "tactical grade" fiber which sells for corresponding "heavy-duty" prices.

UTP Cat5 10BASE-T is quite a logical solution for sending an ADAT bitstream and three power buses. And using transformers made for implementing Ethernet just seems to me a couple of orders of magnitude easier, cheaper, and faster than any alternative.

* Actually, I may have to delve into the art as I have an application that could use that technology in an unconventional way.  I will open a new thread rather than hijacking this one.

[Richard Crowley]

If you don't need isolation, an alternative for driving/receiving a balanced data cable would be SN75176 or MAX485.
Many people use those for driving/receiving over UTP, even for multi-drop, not just for point-to-point.
Since this application is simplex (one direction only), you can hard-wire the chip for transmit at one end, and for receive at the other end.
Assuming they are capable of handling the ADAT bit-rate.

[T3sl4co1l]

I am trying to make the point that @Chaos_Klaus has a bit-stream that he wants to send over 110-ohm balanced UTP.  And there are perhaps 100s of millions of examples of pre-enginered solutions for EXACTLY that situation inside everything on the planet that has an RJ45 ethernet jack.  The transformers, reference circuits, connectors, etc. are available pre-engineered, off the shelf at prices so low, the shipping will probably cost more than the goods.

For sure, it's not hard to find!

I would sincerely love to know where you can get custom optical/copper cable made in small lots (300-1000m) at reasonable prices.  No cable manufacturer I've ever heard of would react to such a request without laughing.  And optical fiber is way too fragile for "tactical" uses unless you use the heavy duty "tactical grade" fiber which sells for corresponding "heavy-duty" prices.

Well, that kind of example would probably be done by a cable assembler -- it might be slightly less dirty than wire ties.  Example, everything inside spaghetti sleeving so it looks more or less intended.

If you don't need isolation, an alternative for driving/receiving a balanced data cable would be SN75176 or MAX485.
Many people use those for driving/receiving over UTP, even for multi-drop, not just for point-to-point.
Since this application is simplex (one direction only), you can hard-wire the chip for transmit at one end, and for receive at the other end.
Assuming they are capable of handling the ADAT bit-rate.

Yeah, the bitrate might be a small challenge; RS422/485 translators top out in the 10-30Mb range, so you just need to make sure you don't get a slow kind.

USB (Full Speed) is also around that bitrate, but the PHYs are almost exclusively integrated into devices so that's not much help.

Tim