80MHz single-ended receiver

I've recently been dealing with a plethora of high-speed serial data 
streams. LVDS is nice. The interface I've recently been given (80MHz 
serial, 3.3V 50R coax, data and clock, driven by a BUF602 video 
amplifier, single-ended and referred to ground) is not.

Right now I'm thinking of co-opting an LVDS receiver (with some biasing 
to bring the common-mode back up to around 1.2V) to receive the data 
stream. Short of using ECL, that's the best I can find in 
current/available components. Can't help feeling that there's a better 
way, but all of the single-ended receivers I've looked at fall far short 
of the 80MHz data rate.

Am I missing something? All sanity checks and 2"*4" cluesticks 
gratefully received.

Steve
http://www.fivetrees.com 

"Steve at fivetrees" <steve@NOSPAMTAfivetrees.com> wrote in message
news:GpSdnQ6ygtLbNvXbnZ2dnUVZ8seinZ2d@pipex.net...
> I've recently been dealing with a plethora of high-speed serial data
> streams. LVDS is nice. The interface I've recently been given (80MHz
> serial, 3.3V 50R coax, data and clock, driven by a BUF602 video
> amplifier, single-ended and referred to ground) is not.
>
> Right now I'm thinking of co-opting an LVDS receiver (with some biasing
> to bring the common-mode back up to around 1.2V) to receive the data
> stream. Short of using ECL, that's the best I can find in
> current/available components. Can't help feeling that there's a better
> way, but all of the single-ended receivers I've looked at fall far short
> of the 80MHz data rate.

What about using a BUF602 again at the receiving end? Look at the
application note in the datasheet, on the first page. Replace the 2k
resistor with 50 Ohm and use a 1:1 transformer parallel to the resistor to
couple the signal from the coax. If you need some gain, use a standard 1:4
balun and use 200 Ohm.

Meindert

"Meindert Sprang" <ms@NOJUNKcustomORSPAMware.nl> wrote in message 
news:4668fd7f$0$12180$e4fe514c@dreader17.news.xs4all.nl...
> "Steve at fivetrees" <steve@NOSPAMTAfivetrees.com> wrote in message
> news:GpSdnQ6ygtLbNvXbnZ2dnUVZ8seinZ2d@pipex.net...
>> I've recently been dealing with a plethora of high-speed serial data
>> streams. LVDS is nice. The interface I've recently been given (80MHz
>> serial, 3.3V 50R coax, data and clock, driven by a BUF602 video
>> amplifier, single-ended and referred to ground) is not.
>>
>> Right now I'm thinking of co-opting an LVDS receiver (with some 
>> biasing
>> to bring the common-mode back up to around 1.2V) to receive the data
>> stream. Short of using ECL, that's the best I can find in
>> current/available components. Can't help feeling that there's a 
>> better
>> way, but all of the single-ended receivers I've looked at fall far 
>> short
>> of the 80MHz data rate.
>
> What about using a BUF602 again at the receiving end? Look at the
> application note in the datasheet, on the first page. Replace the 2k
> resistor with 50 Ohm and use a 1:1 transformer parallel to the 
> resistor to
> couple the signal from the coax. If you need some gain, use a standard 
> 1:4
> balun and use 200 Ohm.

I probably should have mentioned that I need to stay DC-coupled. I'm 
provided with clock and data lines, and there is no guarantee of regular 
transitions on the data line.

Right now I'm pressing ahead with an LVDS-based design. It should work, 
but seems kludgey <shrug>.

Thanks for the suggestion, though.

Steve
http://www.fivetrees.com 

"Steve at fivetrees" <steve@NOSPAMTAfivetrees.com> wrote in message
news:F_adnf17PZ952PTbRVnygwA@pipex.net...
> I probably should have mentioned that I need to stay DC-coupled.

Ah, yes :-)

> I'm provided with clock and data lines, and there is no guarantee of
regular
> transitions on the data line.
>
> Right now I'm pressing ahead with an LVDS-based design. It should work,
> but seems kludgey <shrug>.

At what distance/attenuation does it have to work on? If you terminate the
coax at the receiving end, couldn't you still use a BUF602 as receiver, use
the same circuit but without the input cap?

Meindert

"Meindert Sprang" <ms@NOJUNKcustomORSPAMware.nl> wrote in message 
news:466952db$0$14621$e4fe514c@dreader24.news.xs4all.nl...
> "Steve at fivetrees" <steve@NOSPAMTAfivetrees.com> wrote in message
> news:F_adnf17PZ952PTbRVnygwA@pipex.net...
>> I probably should have mentioned that I need to stay DC-coupled.
>
> Ah, yes :-)
>
>> I'm provided with clock and data lines, and there is no guarantee of
> regular
>> transitions on the data line.
>>
>> Right now I'm pressing ahead with an LVDS-based design. It should 
>> work,
>> but seems kludgey <shrug>.
>
> At what distance/attenuation does it have to work on? If you terminate 
> the
> coax at the receiving end, couldn't you still use a BUF602 as 
> receiver, use
> the same circuit but without the input cap?

The (incomplete) data I've been given is that the signal develops around 
1.7V across 50R at the receiver. This makes sense from a DC point of 
view (50R source, 3.3V power, 50R receiver), but implies that there is 
no significant attenuation along the cable. I'm not sure I can depend on 
this, so my design has to tolerate a smaller swing at the receiver.

Re using a BUF602 as a receiver: I did consider this to start with, but 
reading the datasheet I didn't get the impression it would do much for 
me. It seems to be characterised as a driver, rather than a receiver, 
and it's unity-gain, so I'd still need to restore the output levels. 
Also the data/clock nature of my signal means that I have to be careful 
about skew between the two signals. That's basically why I started 
looking at LVDS receivers... with appropriate biasing, and using e.g. a 
quad device with an external termination resistor, they're low-skew and 
relatively tolerant of the absolute amplitude of signal swing.

Steve
http://www.fivetrees.com 

Steve at fivetrees wrote:
> I probably should have mentioned that I need to stay DC-coupled. I'm 
> provided with clock and data lines, and there is no guarantee of regular 
> transitions on the data line.
> 
> Right now I'm pressing ahead with an LVDS-based design. It should work, 
> but seems kludgey <shrug>.

Why - LVDS are cheap and readily available ?

The only other option is some high speed comparator ?

Some LVDS devices have a watchdog system in them, so idle -> known 
state, so beware of that.

Often the Common mode range is good on LVDS.

-jg

Jim Granville wrote:

> Steve at fivetrees wrote:
> 
>> I probably should have mentioned that I need to stay DC-coupled. I'm 
>> provided with clock and data lines, and there is no guarantee of 
>> regular transitions on the data line.
>>
>> Right now I'm pressing ahead with an LVDS-based design. It should 
>> work, but seems kludgey <shrug>.
> 
> 
> Why - LVDS are cheap and readily available ?
> 
> The only other option is some high speed comparator ?
> 
> Some LVDS devices have a watchdog system in them, so idle -> known 
> state, so beware of that.
> 
> Often the Common mode range is good on LVDS.

This one, for example :  DS90LV028
has good CMR, and should suit 80Mbd well.
Dual Channel, in SO8, or smaller.

-jg

"Jim Granville" <no.spam@designtools.maps.co.nz> wrote in message 
news:4669c41c$1@clear.net.nz...
> Jim Granville wrote:
>
>> Steve at fivetrees wrote:
>>
>>> I probably should have mentioned that I need to stay DC-coupled. I'm 
>>> provided with clock and data lines, and there is no guarantee of 
>>> regular transitions on the data line.
>>>
>>> Right now I'm pressing ahead with an LVDS-based design. It should 
>>> work, but seems kludgey <shrug>.
>>
>> Why - LVDS are cheap and readily available ?
>>
>> The only other option is some high speed comparator ?

Yes, that's pretty much my thoughts too.

>> Some LVDS devices have a watchdog system in them, so idle -> known 
>> state, so beware of that.

Yes, noted. I'm biasing anyway to use it in single-ended mode.

>> Often the Common mode range is good on LVDS.
>
> This one, for example :  DS90LV028
> has good CMR, and should suit 80Mbd well.
> Dual Channel, in SO8, or smaller.

Heh - almost exactly the part I had in mind ;). I'm actually using its 
sibling, the DS90LV048, simply because a) I've just used it in another 
design and b) it's available from stock from Farnell and RS here in the 
UK.

Steve
http://www.fivetrees.com