110Mbs High speed digital isolators

On a sunny day (Mon, 07 Jan 2013 07:22:39 -0500) it happened JW wrote in :

Looks very interesting, but I do not see how 'recommended operating conditions' on page 2 can say 1us rise and fall time for the input signal, as that results in

On a sunny day (Mon, 07 Jan 2013 13:16:45 GMT) it happened Jan Panteltje wrote in :

Never mind, figured it out, is max rise and fall times.

They're neat devices. The only weird thing is the ambigous power-on state, but they share that with some other isolation schemes.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward" 
speff@interlog.com             Info for manufacturers: http://www.trexon.com 
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Interesting, GMR. GMR being static, they could've made it DC stable (if not necessarily appealing on current consumption), but they chose to do it by pulse instead. GMR may be sensitive to DC fields, though they don't specify.

The ADI magnetic isolators are about as fast (I think one is 50Mbps, but they may have up to 100, I haven't looked in detail). They're also stable at DC, in the presence of DC fields (basically until hall effect craps out the silicon itself), and have higher isolation ratings (5kV RMS short duration).

This device's susceptibility appears to be about independent of frequency, which may be useful.

Strange that they measure note 10: surely a PRBS with no more than 5 consecutive digits is only a 5 bit generator, not 66,535; and surely they mean 65535, and surely they mean 16 bit = 65535 state, not 65535 bits?

Tim

--
Deep Friar: a very philosophical monk. 
Website: http://seventransistorlabs.com 

"JW"  wrote in message  
news:gafle8pfktta78ui3rvvvl1j2502ncdbpq@4ax.com... 
>I came across these while repairing a Keithley 2001. Pretty neat device  
>as 
> it beats out opto-couplers by a factor of 10 or so. 
> 
> http://www.nve.com/Downloads/il71x.pdf 
> 
> Thought they might be of some interest.

Analog Devices has had isolators similar to that for some time (e.g. ADUM1100). I used some of their "ISO-Power" isolators a couple years ago. These include an isolated power supply (150mA) and go up to

150Mbps (e.g. ADUM5210). Quite pricey, though (they start at $5, IIRC).

--
I think what they're saying is that they're using a 16bit shift 
register to generate a maximal length (65535 states if you don't 
include the lock-up state) pseudo-random sequence, modified so that 
nowhere in the sequence will there be more than five consecutive ones 
or zeros.

Interesting - I hadn't realized that there were other players out there, this is the first time I've come across anything like these. The ADUM5210 does not seem to be available anywhere, though. On Analogs website it's listed as "Product Status:Pre-Release"

Could be. I was using the ARUM5200 (25kbps version), IIRC. I only needed 200Hz. ;-) There are several others on the product summary page, without the isolated supply, to 100Mbps. They're under a buck.

There's one thing about these devices, they do NOT work down to DC. You need to have a flip of the magnetic polarity at least every few us to reset the detector. Great for DC-balanced protocols like Manchester, but no good for stuff that shuts off the transitions when idle. I tried to use one for a PWM signal in a servo, and it didn't work at all. The manufacturer confirmed that it was not suited for the idle state.

Jon

There is a paragraph that says it needs a kick after power up, but I didn't see anything about a minimum transition rate. Did I miss something, or is that a crappy data sheet.

It's also fishy that note 10 says only 5 1s or 0s in a row.

[It also claims to operate at 110 Mbps over the full temperature and supply voltage range but the datails say 110 is typical and only guarantees 100.]

--
These are my opinions.  I hate spam.

The ADI parts work fine down to DC.

I recall trying those parts and not using them because of quirky behavior.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators

Ethernet over twisted pair guarantees this restriction, to control baseline drift.

I assume that they guarantee full rate over a defined portion of the shmoo plot (performance as a function of temperature and supply voltage).

Joe Gwinn

device as

Completely erroneous. IEEE 802.3 clause 7.3 clearly specifies Manchester encoding for 10 Mbit/s systems. Higher speed systems also use polarity balanced signaling. Other systems may vary.

voltage).

That is known as"bit-stuffing"

which is often used with synchronous protocols, such as HDLC (X.25) and CANbus. In order to reliably detect start/stop flags (long strings of "1"s), the actual message data must not contain long strings of "1"s. To prevent this, "0"s are added, if this rule would be violated within the message.

This has not been designed to remove the DC component, but I guess in an AC coupled system, a diode clamp could be used, similar to analog video sync level restoration in an AC-coupled video signal paths.

From the title, we are dealing with 110 mbit data, not 10 mbit data.

Joe Gwinn

device

DC.

transitions

miss

Manchester

polarity

So from a simple speed change you ASSUME that there must be a coding change that eliminates the DC balanced coding? Where are you going with this?

?-)

No assumption needed. I read the standards. And worked on the IEEE

802.3 (Ethernet) committee for a while, during the development of gigabit ethernet.

The 10 mbit waveform was 0 to -2 volts on coax, and could not be AC coupled. I do recall that it was manchester coded. I would diagnose such systems by taking the terminator off one end and plugging the coax cable into a scope input set to 50 ohms. One could see the whole conversation. And see if there was a problem.

The 100 mbit waveform was passively DC balanced, so it could be AC coupled. (It was the same signal design as for FDDI.)

The 1000 mbit waveform was actively DC balanced, so it could be AC coupled, but more importantly to prevent data-pattern-dependent bit errors due to baseline drift.

By the way, Ethernet standards are available gratis: .

If memory serves, the max run of five ones or five zeros originally came from the HDLC standard.

Joe Gwinn

Huh? 10Base-T and 100Base-T run on the same hardware, with the *SAME* transformers (i.e. AC coupled).