Searching \ for '[OT] Raspberry Pi has bad mags' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: piclist.com/techref/index.htm?key=raspberry+has+bad
Search entire site for: 'Raspberry Pi has bad mags'.

Exact match. Not showing close matches.
PICList Thread
'[OT] Raspberry Pi has bad mags'
2012\03\09@061350 by RussellMc

face picon face
Can people please try to avoid a scattergunning of subject lines on
what is essentially the same topic. If a topic diverges enough that
the old subject line no longer serves well then by all means create a
new one, but on occasions there is a bad tendency to have several
threads running in parallel on the same topic so people cannot easily
follow what's being said.


Having subject lines that tried to be  usefully informative first and
to be cute only if there was searoom left over after being usefully
informative would be a bonus.



Russell

[The answer to question one is "They CAN, but ... " :-)

2012\03\09@134958 by YES NOPE9

flavicon
face
It is a new subject....  IMHO ..... I carefully consider whether to change the subject line
Sorry that you do not agree   .... do you think this subject line is cute ????
Gus


{Quote hidden}

> -

2012\03\09@182844 by V G

picon face
On Fri, Mar 9, 2012 at 5:22 PM, Herbert Graf <spam_OUThkgrafTakeThisOuTspamgmail.com> wrote:

>  Good point. I'm so used to working with parts that integrate the MAC and
> PHY. You're right, in most of the rest of the work, the PHY is often
> still an external component.
>

Can you give me an example of a part that integrate the MAC and PHY

2012\03\11@113103 by William Couture

face picon face
On Fri, Mar 9, 2012 at 12:58 PM, Herbert Graf <.....hkgrafKILLspamspam@spam@gmail.com> wrote:

> "Integrated Magnetics" has nothing to do primarily with noise
> suppression. Ethernet requires a transformer between the MAC and the
> actual jack.
>
> Traditionally this was always done by an external part. Look at pretty
> much any network card and chances are you'll see the transformer, it's
> easy to spot.
>
> To reduce costs, manus started putting the "magnetics" into the jack
> itself. This results in a slightly larger jack, but one less part to
> source, route for and mount.

My wife, also a computer "geek" but without an EE inclination was asking
me why anyone would design with a jack without built-in magnetics these
days.

Could anyone clue me in?

Thanks,
  Bill

-- Psst...  Hey, you... Buddy...  Want a kitten?  straycatblues.petfinder.org

2012\03\11@120403 by Spehro Pefhany

picon face
At 11:31 AM 3/11/2012, you wrote:
{Quote hidden}

To save a bit of money. The combination of an external transformer and a
passive jack is still a bit less costly (used to be a LOT less costly,
but the integrated ones have come down a lot in cost). Putting it in the
can costs more and probably involves as much _total_ labor but saves on
the footprint and PCB assembly costs.

Of course there are also plenty of applications for RJ45 jacks that don't
involve Ethernet.

--sp

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

2012\03\11@123843 by Isaac Marino Bavaresco

flavicon
face
Em 11/3/2012 13:03, Spehro Pefhany escreveu:
{Quote hidden}

Power Over Ethernet requires access to the center tap of the primary the
transformer. AFAIK the jacks with integrated transformers don't expose
the primary center taps.

Isaac

2012\03\11@125412 by Matt Bennett

flavicon
face
On Sun, March 11, 2012 11:03 am, Spehro Pefhany wrote:
> At 11:31 AM 3/11/2012, you wrote:

>>My wife, also a computer "geek" but without an EE inclination was asking
>>me why anyone would design with a jack without built-in magnetics these
>>days.
>>
>>Could anyone clue me in?
>>
>>Thanks,
>>    Bill
>
> To save a bit of money. The combination of an external transformer and a
> passive jack is still a bit less costly (used to be a LOT less costly,
> but the integrated ones have come down a lot in cost). Putting it in the
> can costs more and probably involves as much _total_ labor but saves on
> the footprint and PCB assembly costs.

In a previous life, I did Ethernet physical layer design/testing on
servers for a major computer manufacurer (rhymes with "Hell").  Without
fail, the first pass of Ethernet designs done would not pass the IEEE
802.3 physical layer specifications, *unless* they used an integrated
magnetics.  Return loss, signal integrity, or Hipot- doing the design
right (for interoperability) is quite hard.

It is pretty easy to get Ethernet to talk over a short cable, but if you
want to meet the BER ratings at maximum distance and minimum distance,
while also meeting safety specs (Ethernet (not PoE) is designed to be
electrically isolated) is far from a trivial exercise.

Using integrated magnetics saves a *huge* amount of time and vastly
improves your chances of design success.

Matt Bennett
Just outside of Austin, TX
30.51,-97.91

The views I express are my own, not that of my employer, a large
multinational corporation that you are familiar with

2012\03\11@130709 by Matt Bennett

flavicon
face
On Sun, March 11, 2012 11:37 am, Isaac Marino Bavaresco wrote:
> Power Over Ethernet requires access to the center tap of the primary the
> transformer. AFAIK the jacks with integrated transformers don't expose
> the primary center taps.

Less common, but they do exist- I was able to find some at Molex pretty
easily, I suspect the other major players (Pulse, Bel, etc.) have them-
though they haven't really migrated to suppliers like Digikey.

Matt Bennett
Just outside of Austin, TX
30.51,-97.91

The views I express are my own, not that of my employer, a large
multinational corporation that you are familiar with

2012\03\11@131223 by Spehro Pefhany

picon face
At 12:37 PM 3/11/2012, you wrote:


>Power Over Ethernet requires access to the center tap of the primary the
>transformer. AFAIK the jacks with integrated transformers don't expose
>the primary center taps.
>
>Isaac

Quite a few of them do now, but they're more expensive, bigger, and have
more pins.

Eg.

http://www.belfuse.com/Data/Datasheets/0813-1X1T-57-F.pdf

2012\03\11@133235 by peter green

flavicon
face

> In a previous life, I did Ethernet physical layer design/testing on
> servers for a major computer manufacurer (rhymes with "Hell").  Without
> fail, the first pass of Ethernet designs done would not pass the IEEE
> 802.3 physical layer specifications, *unless* they used an integrated
> magnetics.  Return loss, signal integrity, or Hipot- doing the design
> right (for interoperability) is quite hard.
>   How many people are there who both care about it being "right" and are making
small enough volumes that the extra design effort is sigificant

2012\03\11@134108 by Spehro Pefhany

picon face
At 12:54 PM 3/11/2012, you wrote:

>Using integrated magnetics saves a *huge* amount of time and vastly
>improves your chances of design success.
>
>Matt Bennett
>Just outside of Austin, TX
>30.51,-97.91

Interesting. Every Ethernet switch that I've opened up has used
non-integrated magnetics, including the whack of Trendnet 8-port
gigabit switches that I upgraded to a year or so ago**.

I would have guessed that for a notebook, integrated magnetics
would be a no-brainer, and that the Ethernet layout would be
far from the most difficult part of the motherboard layout-
the high speed stuff between memory and processor etc. would
be where the challenge would be.

** They use four pieces of these parts for the eight 10/100/1000
ports, and a single 8-position shielded jack assembly.

http://www.bi-tek.com.tw/style/frame/templates4/product_detail.asp?lang=2&customer_id=828&name_id=40208&content_set=color_2&rid=0&id=119260

Wiring between the jacks and transformers, and wiring between the
chips and the transformers is pairs of conductors- probably
controlled impedance differential microstrip- I can see the
ground plane on an internal layer, but the stackup would have to
have a very thin layer of prepreg to get to 100R
differential Z given the dimensions (about 4 thou wide with around
10 thou spacing.

Best regards,

2012\03\11@134701 by Spehro Pefhany

picon face
At 01:32 PM 3/11/2012, you wrote:

> > In a previous life, I did Ethernet physical layer design/testing on
> > servers for a major computer manufacurer (rhymes with "Hell").  Without
> > fail, the first pass of Ethernet designs done would not pass the IEEE
> > 802.3 physical layer specifications, *unless* they used an integrated
> > magnetics.  Return loss, signal integrity, or Hipot- doing the design
> > right (for interoperability) is quite hard.
> >
>How many people are there who both care about it being "right" and are
>making
>small enough volumes that the extra design effort is sigificant?

As more than one Asian manager has said one way or another, "That costs
another 5 cents a unit. I ship 50,000 units a month, with that kind of money
I can hire another engineer" (and he probably still almost can..).

Best regards,

2012\03\11@141659 by Matt Bennett

flavicon
face
On Sun, March 11, 2012 12:41 pm, Spehro Pefhany wrote:
> At 12:54 PM 3/11/2012, you wrote:
>
>>Using integrated magnetics saves a *huge* amount of time and vastly
>>improves your chances of design success.
>>
>>Matt Bennett
>>Just outside of Austin, TX
>>30.51,-97.91
>
> Interesting. Every Ethernet switch that I've opened up has used
> non-integrated magnetics, including the whack of Trendnet 8-port
> gigabit switches that I upgraded to a year or so ago**.

I wouldn't call any Trendnet product high-end... :) At the level of the
stuff I buy at Fry's for my home (I have plenty of Trendnet stuff here)- a
standard non-managed switch is a commodity- price and operation are a
difficult trade-off, usually leaning towards price.  The IEEE standards
are not enforced- they are voluntary guidelines, but they give people
something to fall back on, and a direction to point the blame when things
fail.

A consumer setup tends to be a lot more tolerant of errors- TCP/IP will
hide a lot of issues until they become particularly egregious.  In a
server setup, where they are trying to get as much out of their network
equipment as they can, failures become more obvious, and the customers
will use that as leverage to push costs down in the next generation.

> I would have guessed that for a notebook, integrated magnetics
> would be a no-brainer, and that the Ethernet layout would be
> far from the most difficult part of the motherboard layout-
> the high speed stuff between memory and processor etc. would
> be where the challenge would be.

Inside your own box, you have much more control- but when you have to
connect to something outside of your own control, the specs are the only
thing to fall back on. Build it to the spec, and it will work, as long as
the spec isn't broken. If you have 100% control of the environment, there
are plenty of special tricks that you can use (for example, adjusting
drive levels, pre-emphasis).



Matt Bennett
Just outside of Austin, TX
30.51,-97.91

The views I express are my own, not that of my employer, a large
multinational corporation that you are familiar with

2012\03\11@142414 by Herbert Graf

picon face

On 2012-03-09, at 6:23 PM, V G wrote:

> On Fri, Mar 9, 2012 at 5:22 PM, Herbert Graf <hkgrafspamspam_OUTgmail.com> wrote:
>
>> Good point. I'm so used to working with parts that integrate the MAC and
>> PHY. You're right, in most of the rest of the work, the PHY is often
>> still an external component.
>>
>
> Can you give me an example of a part that integrate the MAC and PHY?

The ENC MCHIP part does. The Wiznet parts as well.

There are I believe others, but those are the ones I'm most familiar with.

TTYL

2012\03\11@142830 by Matt Bennett

flavicon
face
On Sun, March 11, 2012 12:46 pm, Spehro Pefhany wrote:
>>How many people are there who both care about it being "right" and are
>>making
>>small enough volumes that the extra design effort is sigificant?
>
> As more than one Asian manager has said one way or another, "That costs
> another 5 cents a unit. I ship 50,000 units a month, with that kind of
> money
> I can hire another engineer" (and he probably still almost can..).

There is also the time aspect- at the beginning of the lifetime of a
product, the margins tend to be higher- if you can get your product to
market quicker, you will spend more time at higher margins and have higher
profit over the year. These get to be complex trade-offs. For public
companies, the goal is to maximize shareholder return.  Sometimes it makes
sense to hire the extra engineer, sometimes it makes sense to buy the more
expensive component, but the answer never seems to be cut-and-dried.



Matt Bennett
Just outside of Austin, TX
30.51,-97.91

The views I express are my own, not that of my employer, a large
multinational corporation that you are familiar with

2012\03\11@144427 by Matt Bennett

flavicon
face
On Sun, March 11, 2012 1:24 pm, Herbert Graf wrote:
>
> On 2012-03-09, at 6:23 PM, V G wrote:
>
>> On Fri, Mar 9, 2012 at 5:22 PM, Herbert Graf <@spam@hkgrafKILLspamspamgmail.com> wrote:
>>
>>> Good point. I'm so used to working with parts that integrate the MAC
>>> and
>>> PHY. You're right, in most of the rest of the work, the PHY is often
>>> still an external component.
>>>
>>
>> Can you give me an example of a part that integrate the MAC and PHY?
>
> The ENC MCHIP part does. The Wiznet parts as well.
>
> There are I believe others, but those are the ones I'm most familiar with..

A PHY and a MAC are two very different beasts- at its heart, a PHY is an
analog part, a MAC is a digital part. Getting them both to work right on
the same piece of silicon is a challenge, and leads to very costly
silicon. You will notice that the PIC32 is only available with an
integrated MAC, and a PIC18F97J60 (integrated MAC/PHY) is the most
expensive PIC18FxxJxx part (Digikey 100 piece pricing)

When it fits within the size constraints, I definitely prefer a separate
MAC/PHY - you can place the MAC where it is more convenient, and the PHY
very close to the mag/jack. If the choice was integrated MAC/PHY or
integrated mag/jack, lacking other constraints I will choose the
integrated mag/jack every time.

Matt Bennett
Just outside of Austin, TX
30.51,-97.91

The views I express are my own, not that of my employer, a large
multinational corporation that you are familiar with

2012\03\11@152612 by Isaac Marino Bavaresco

flavicon
face
Em 11/3/2012 15:24, Herbert Graf escreveu:
> On 2012-03-09, at 6:23 PM, V G wrote:
>
>> On Fri, Mar 9, 2012 at 5:22 PM, Herbert Graf <KILLspamhkgrafKILLspamspamgmail.com> wrote:
>>
>>> Good point. I'm so used to working with parts that integrate the MAC and
>>> PHY. You're right, in most of the rest of the work, the PHY is often
>>> still an external component.
>>>
>> Can you give me an example of a part that integrate the MAC and PHY?
> The ENC MCHIP part does. The Wiznet parts as well.
>
> There are I believe others, but those are the ones I'm most familiar with..
>
> TTYL


The first ENCs and PIC18Fx7Jxx were 10BaseT only.
Only the second generation ENCs are 100BaseT capable.


Isaac

More... (looser matching)
- Last day of these posts
- In 2012 , 2013 only
- Today
- New search...