Your very own high-speed pattern generator

[Winfield Hill]

Can't afford a new Agilent series 81141A high-speed
pattern generator, at only $101,673 starting price?
My local rep writes to say that a refurbished N4901B,
the older model capable of 13.5GB/s, costs $32k, or
$54k for full BERT (not JBERT?), but only if ordered
by Feb 28th. Call ASAP to reserve a unit, he says.

 

[John Larkin]

Can't afford a new Agilent series 81141A high-speed
pattern generator, at only $101,673 starting price?
My local rep writes to say that a refurbished N4901B,
the older model capable of 13.5GB/s, costs $32k, or
$54k for full BERT (not JBERT?), but only if ordered
by Feb 28th. Call ASAP to reserve a unit, he says.

What sort of speed, patterns, and levels do you need?

The newish SRS clock box with the PRBS option ain't bad for $3K new.
We just got one, and it's interesting. It's a 2 GHz box, but they
limited the prbs to 1.55 GHz, presumably because the EclipsPlus shift
register dies somewhere below 2G. The layout is cute, seven SO-8
flipflops in a circle with an xor gate off to the side, like a frying
pan with a handle. It would go faster if they did the xor some other
way.

I've got it running over the weekend into an infinite-persistance
sampling scope to see if there are any outliers in the eye diagram.

I wonder if there's a market for a small pattern generator. It
shouldn't be really hard up to, say, 5 GHz.

John

 

[EE123]

 Can't afford a new Agilent series 81141A high-speed
 pattern generator, at only $101,673 starting price?
 My local rep writes to say that a refurbished N4901B,
 the older model capable of 13.5GB/s, costs $32k, or
 $54k for full BERT (not JBERT?), but only if ordered
 by Feb 28th.  Call ASAP to reserve a unit, he says.

What is JBERT?

 

[Fred Bartoli]

EE123 a écrit :

What is JBERT?

Being what J stands for for TI packages it's DILBERT.

 

[Hal Murray]

I wonder if there's a market for a small pattern generator. It
shouldn't be really hard up to, say, 5 GHz.

Some FPGAs come with high speed serial ports and an option
to bypass the 8b/10b or 64b/66b encoders. If your pattern
fits in the on chip ROMs, it should be straightforward to
emit an arbitrary pattern at 10 G bits/sec. (I haven't done
it, but it sounds like fun.)

 

[Rene Tschaggelar]

What sort of speed, patterns, and levels do you need?

The newish SRS clock box with the PRBS option ain't bad for $3K new.
We just got one, and it's interesting. It's a 2 GHz box, but they
limited the prbs to 1.55 GHz, presumably because the EclipsPlus shift
register dies somewhere below 2G. The layout is cute, seven SO-8
flipflops in a circle with an xor gate off to the side, like a frying
pan with a handle. It would go faster if they did the xor some other
way.

I've got it running over the weekend into an infinite-persistance
sampling scope to see if there are any outliers in the eye diagram.

I wonder if there's a market for a small pattern generator. It
shouldn't be really hard up to, say, 5 GHz.

Design your own SiGe chip ? Unfortunately the
chip fabs that do custom chips stick to CMOS.
Configurable CMOS that is. If you had to pay
for the photo masks yourself, one for each
layer, you'd spend a fortune just on them.

Rene

 

[John Larkin]

Design your own SiGe chip ? Unfortunately the
chip fabs that do custom chips stick to CMOS.
Configurable CMOS that is. If you had to pay
for the photo masks yourself, one for each
layer, you'd spend a fortune just on them.

Rene

No, just use a fast FPGA to make patterns at 400 MHz or so, and mux up
its outputs with some fast ECL, all stock parts. As someone noted,
many of the high-end fpga's already do gigabit serial streams, but I
don't know if they are general enough to be used directly as a pattern
generator.

Yesterday we were, independently, brainstorming a box that would
accept a customer trigger and generate some number of arbitrary
digital and analog outputs, with picosecond jitter and resolution on
all the output edges. We might do it if things slow down a little. We
occasionally get an inquiry for stuff like this.

John

 

[Phil Hobbs]

No, just use a fast FPGA to make patterns at 400 MHz or so, and mux up
its outputs with some fast ECL, all stock parts. As someone noted,
many of the high-end fpga's already do gigabit serial streams, but I
don't know if they are general enough to be used directly as a pattern
generator.

Yesterday we were, independently, brainstorming a box that would
accept a customer trigger and generate some number of arbitrary
digital and analog outputs, with picosecond jitter and resolution on
all the output edges. We might do it if things slow down a little. We
occasionally get an inquiry for stuff like this.

John

I'd certainly like a 2**32-1 prbs generator that would go at least 5
Gb/s and would drive a telecom modulator (+24 dBm). Faster would be
better, but 5 Gb/s would definitely be worth buying if it were under,
say, $20k.

Cheers,

Phil Hobbs


PS: Now who would have suspected that Winfield Hill would stoop to FS
spamming SED?

 

[John Larkin]

I'd certainly like a 2**32-1 prbs generator that would go at least 5
Gb/s and would drive a telecom modulator (+24 dBm). Faster would be
better, but 5 Gb/s would definitely be worth buying if it were under,
say, $20k.

Cheers,

Phil Hobbs

That's do-able. Some GigaComm logic plus moderate dirty tricks could
make the sequence, and then a couple of the Hittite 20 GHz distributed
amps to drive your e-o modulator to 5-6 volts p-p. $1500 worth of
parts maybe.

PS: Now who would have suspected that Winfield Hill would stoop to FS
spamming SED?

Shocked, shocked.

John

 

[Winfield]

PS: Now who would have suspected that Winfield Hill
would stoop to FS spamming SED?

I was trying to make a joke - tongue in cheek,
hey, who who can afford $32 or 54k for a pulse
generator, let alone $101k? Sheesh! The email
I received from Agilent's rep was unsolicited,
and filled with excitement for this special deal
we'd jump at, but would disappear by the 28th.

 

[MooseFET]

That's do-able. Some GigaComm logic plus moderate dirty tricks could
make the sequence, and then a couple of the Hittite 20 GHz distributed
amps to drive your e-o modulator to 5-6 volts p-p. $1500 worth of
parts maybe.

Here's an idea I should save for a patent or perhaps it is too loony
for even that:

There are materials with non-linear optical properties that could be
used to cause a laser beam to deflect back and forth at a rate near
the GHz range. Rubidium vapor in the one I am thinking about right
now but there are some doped crystals that also would do this.

Two such devices at right angles could direct a Laser beam in a
circle. This circle would strike a light switching device such as an
LCD with a great many segments in a ring.

A bit of non-imaging optics could bring the light down to one location
or put it onto a fiber. A lot of light would be lots in the process
so the Laser will have to be modestly powerful.

At this point, if enough segments are in the light switch, you can
have light that is switching at a great many GHz.

 

[Phil Hobbs]

Here's an idea I should save for a patent or perhaps it is too loony
for even that:

There are materials with non-linear optical properties that could be
used to cause a laser beam to deflect back and forth at a rate near
the GHz range. Rubidium vapor in the one I am thinking about right
now but there are some doped crystals that also would do this.

Two such devices at right angles could direct a Laser beam in a
circle. This circle would strike a light switching device such as an
LCD with a great many segments in a ring.

A bit of non-imaging optics could bring the light down to one location
or put it onto a fiber. A lot of light would be lots in the process
so the Laser will have to be modestly powerful.

At this point, if enough segments are in the light switch, you can
have light that is switching at a great many GHz.

I've recently bought a couple of 40 Gb/s modulators (30 GHz 3 dB BW)
from Avanex for about $6k each. That's about as fast as I need to go
for the foreseeable future. They're about 70% efficient optically, but
need about +24 dBm worth of drive.

Cheers,

Phil Hobbs

 

[Phil Hobbs]

I was trying to make a joke - tongue in cheek,
hey, who who can afford $32 or 54k for a pulse
generator, let alone $101k? Sheesh! The email
I received from Agilent's rep was unsolicited,
and filled with excitement for this special deal
we'd jump at, but would disappear by the 28th.

Me too. Agilent FS spam was too funny to miss.

Cheers,

Phil Hobbs

 

[Michael A. Terrell]

I was trying to make a joke - tongue in cheek,
hey, who who can afford $32 or 54k for a pulse
generator, let alone $101k? Sheesh! The email
I received from Agilent's rep was unsolicited,
and filled with excitement for this special deal
we'd jump at, but would disappear by the 28th.

This is why high tech companies should never hire people who've sold
used cars.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

 

[Winfield]

I've recently bought a couple of 40 Gb/s modulators (30 GHz 3 dB BW)
from Avanex for about $6k each. That's about as fast as I need to
go for the foreseeable future. They're about 70% efficient optically,
but need about +24 dBm worth of drive.

At what frequency?

 

[krw]

This is why high tech companies should never hire people who've sold
used cars.

Why? Because they'd be too embarrassed to take the job?

 

[Phil Hobbs]

At what frequency?

They're only about 1 dB down at 20 GHz, iirc. At that sort of speed,
the phase nonlinearity is at least as important as the rolloff.

Cheers,

Phil Hobbs