picosecond test points

[John Larkin]

We're designing some laser driver boards, and I thought it would be
cool to add test points to some of the interesting circuit nodes.
Regular scope probes don't work at ps speeds, and probe grounding is
difficult. There are multi-GHz active probes, at roughly $1 per Hz.

So I was thinking that I could add a small, 0603 maybe, resistor to a
signal to be snooped, run a 50 ohm trace some small distance, and end
up at some structure that had a signal test point and a ground.
Something roughly like this:

http://db.tt/6rRcagTt

The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.

The "probe" would be a piece of hardline coax that runs to a sampling
scope input, with some provision for grounding. Something like a 3-5
GHz bandwidth should be feasible.

This looks promising if a little klunky. Any other ideas?


--

John Larkin, President 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

[John Larkin]

On Sat, 11 Feb 2012 12:57:53 -0800, John Larkin
<jjla...@highNOTlandTHIStechnologyPART.com> wrote:

>
>
>We're designing some laser driver boards, and I thought it would be
>cool to add test points to some of the interesting circuit nodes.
>Regular scope probes don't work at ps speeds, and probe grounding is
>difficult. There are multi-GHz active probes, at roughly $1 per Hz.

Per MHz, actually.

[Fred Bartoli]

John Larkin a écrit :

>
> We're designing some laser driver boards, and I thought it would be
> cool to add test points to some of the interesting circuit nodes.
> Regular scope probes don't work at ps speeds, and probe grounding is
> difficult. There are multi-GHz active probes, at roughly $1 per Hz.
>
> So I was thinking that I could add a small, 0603 maybe, resistor to a
> signal to be snooped, run a 50 ohm trace some small distance, and end
> up at some structure that had a signal test point and a ground.
> Something roughly like this:
>
> http://db.tt/6rRcagTt
>
> The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.
>
> The "probe" would be a piece of hardline coax that runs to a sampling
> scope input, with some provision for grounding. Something like a 3-5
> GHz bandwidth should be feasible.
>
> This looks promising if a little klunky. Any other ideas?
>
>

SD14?


--
Thanks,
Fred.

[Phil Hobbs]

John Larkin wrote:
>
> We're designing some laser driver boards, and I thought it would be
> cool to add test points to some of the interesting circuit nodes.
> Regular scope probes don't work at ps speeds, and probe grounding is
> difficult. There are multi-GHz active probes, at roughly $1 per Hz.
>
> So I was thinking that I could add a small, 0603 maybe, resistor to a
> signal to be snooped, run a 50 ohm trace some small distance, and end
> up at some structure that had a signal test point and a ground.
> Something roughly like this:
>
> http://db.tt/6rRcagTt
>
> The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.
>
> The "probe" would be a piece of hardline coax that runs to a sampling
> scope input, with some provision for grounding. Something like a 3-5
> GHz bandwidth should be feasible.
>
> This looks promising if a little klunky. Any other ideas?

Looks good. Another approach would be something like a BFP650 follower,
with the 50-ohm coax forming the emitter load when the probe is
connected. You'd need a very small base inductor to keep it from
oscillating.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

[lang...@fonz.dk]

On 11 Feb., 21:57, John Larkin

<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> We're designing some laser driver boards, and I thought it would be
> cool to add test points to some of the interesting circuit nodes.
> Regular scope probes don't work at ps speeds, and probe grounding is
> difficult. There are multi-GHz active probes, at roughly $1 per Hz.
>
> So I was thinking that I could add a small, 0603 maybe, resistor to a
> signal to be snooped, run a 50 ohm trace some small distance, and end
> up at some structure that had a signal test point and a ground.
> Something roughly like this:
>
> http://db.tt/6rRcagTt
>
> The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.
>
> The "probe" would be a piece of hardline coax that runs to a sampling
> scope input, with some provision for grounding. Something like a 3-5
> GHz bandwidth should be feasible.
>
> This looks promising if a little klunky. Any other ideas?
>

something like this: http://www.idinet.com/IdiNet/media/PDF/94.pdf?ext=.pdf

think mouser have them

-Lasse

[John Larkin]

I have a couple, but the tip is nasty and it's hard to ground.

John

[John Larkin]

OK, that leads to...

http://www.idinet.com/Test-Probes/Specialty-Probes/Coaxial-Probes/

What I don't understand is that all the inner probes seem to be flush
with the outer shell, which doesn't look like it would make a reliable
connection to the board.

The PCB layout for these would be an outer ground ring and an inner
via, with the signal trace on an inner or bottom layer. Or maybe an
outer broken/c-shaped ring for grounding, and the signal/pad on top.

[Fred Bloggs]

On Feb 11, 5:39 pm, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Sat, 11 Feb 2012 14:20:24 -0800 (PST), "langw...@fonz.dk"

It's a spring loaded plunger...

[lang...@fonz.dk]

On 11 Feb., 23:39, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Sat, 11 Feb 2012 14:20:24 -0800 (PST), "langw...@fonz.dk"

both outer and inner are spring loaded,
from the datasheet:

Spring Force:
Signal Conductor:
2.3 oz. @ .107 (2.72) travel
Shielding Plunger:
5.8 oz. @ .107 (2.72) travel

> The PCB layout for these would be an outer ground ring and an inner
> via, with the signal trace on an inner or bottom layer. Or maybe an
> outer broken/c-shaped ring for grounding, and the signal/pad on top.
>

yes something like that, maybe just a standard round test point with
a big enough void in the soldermask to get to the ground plane

btw. seen this?: http://koti.mbnet.fi/jahonen/Electronics/DIY%201k%20probe/

-Lasse

[Bill Sloman]

On Feb 11, 11:24 pm, Phil Hobbs

<pcdhSpamMeSensel...@electrooptical.net> wrote:
> John Larkin wrote:
>
> > We're designing some laser driver boards, and I thought it would be
> > cool to add test points to some of the interesting circuit nodes.
> > Regular scope probes don't work at ps speeds, and probe grounding is
> > difficult. There are multi-GHz active probes, at roughly $1 per Hz.
>
> > So I was thinking that I could add a small, 0603 maybe, resistor to a
> > signal to be snooped, run a 50 ohm trace some small distance, and end
> > up at some structure that had a signal test point and a ground.
> > Something roughly like this:
>
> >http://db.tt/6rRcagTt
>
> > The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.
>
> > The "probe" would be a piece of hardline coax that runs to a sampling
> > scope input, with some provision for grounding. Something like a 3-5
> > GHz bandwidth should be feasible.
>
> > This looks promising if a little klunky. Any other ideas?
>
> Looks good.  Another approach would be something like a BFP650 follower,
> with the 50-ohm coax forming the emitter load when the probe is
> connected.  You'd need a very small base inductor to keep it from
> oscillating.

I did something like that on the little up-grade board that put
ECLinPS into the critical bit of an old TTL-based pulse generator for
electron spin resonance - that would have been back around 1996.

The original designer wanted buffered test points so that he could
probe the critical signals without loading the trace, so I buffered
the relevant traces with BFR92A emitter-followers. We didn't bother
with a base-stopping inductor - something like an 0603 33R hard up
against the base did the job perfectly adequately.

--
Bill Sloman

[Okkim Atnarivik]

Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:
: Looks good. Another approach would be something like a BFP650 follower,

Phil, I was about to suggest that if you've ended up using the
BFP650 by following my example, you might be better off with the BFP640.
Or more modern ones with a higher hFE (I'm fond of the NESG3031 and
NESG4030).

Namely, my choice of BFP650 (and not the then more widely available
BFP640) was driven by the search of the lowest possible R_BB, because
I was deemed to encounter lower source resistances than the 50ohms for
which SiGe devices are generally optimized. I reasoned that the base
geometry of the BFP650, as a medium power device, is more likely yield
a low R_BB. In your applications BFP640 may work better with the "more
ordinary" impedance levels and ambient temperatures.

The "I was about.." part means that I'm somewhat puzzled now that I
looked up the recent Infineon data sheets. The BFP650 is now listed
as a SiGe:C part (the datasheet 2010-10-22) and its Gummel-Poon
model reads RB=6.376 . The preliminary datasheet I have, dated
Aug-16-2004, lists it as an ordinary SiGe part with RB=1.036 . So, I'm
wondering how much the device has changed from the versions I've been
using. I seem to recall that also in some other more recent data sheet
versions it was listed as a SiGe, not SiGe:C .

Given the fact that the most recent BFP640 datasheet (dated 2007-05-29)
reads RB=3.129, the BFP650 does not seem to have any noise advantage
any more, either.

In fact, there was a recent paper in the RSI where
the Jena group claim to have achieved 15 pV/rtHz with BFP640 in LHe,
as compared with my 75 pV/rtHz with BFP650. This is a puzzling result,
actually, because it is better than one would expect in the picture where
u_N originates from the collector shot noise acting on the r_E - provided
that the thermal voltage saturates at ~5..7 meV even when the ambient
temperature keeps going down. According to my data it does saturate in
BFP640 just like it does in all other SiGe's I've tried, when measured by
the transconductance. I'm wondering whether my transistors have always
been oscillating so high that I cannot see it (ie. >27 GHz) and the
V_T saturation is an artefact due to it...

Regards,
Mikko

[Okkim Atnarivik]

Okkim Atnarivik wrote:
>the Jena group claim to have achieved 15pV/rtHz with BFP640 in LHe

... a typo, sorry, they actually claim 35 pV/rtHz . And they seem
to have regenerated the emitter inductively, so the instability
hypothesis may indeed be true in my case.

Regards,
Mikko

[George Herold]

> -Lasse- Hide quoted text -
>
> - Show quoted text -

Yup, I was going to suggest a connector if you've got room.

Maybe just a ground ring on top for the hardline.. and some sort of
springie contact for the signal on the bottom.


gnd-----. .----gnd--
| |
|.-- |
-----*| *------
----+
^
+-springie contanct (phospur bronze?)

George H.

[Phil Hobbs]

I started by taking your advice, which was pretty valuable--thanks
again. I noticed the change in specs as well, when I ordered another
batch a couple of months ago. They changed the type number to BFP650H
instead of BFP650E when they changed the process. (The dogs.)

I'm using it as a cascode stage for a SKY65050 pHEMT, and it works
great. One of the best things about it is that it has effectively
infinite Early voltage, so you can get a lot of voltage gain out of a
single stage. Even for situations where you don't care so much about
ultralow noise, the combination of a gigantic f_T with a very high V_A
is unique in my collection. It does want to oscillate at 14 GHz if you
look at it crosswise, but a nice 5-ohm bead in series with the base
cleans that right up.

I'm going to be using it in a new front end design, as a
bootstrapped-bootstrap wrapped round a couple of parallelled BF862s. DC
coupling the bootstrap and using a current source load gets rid of the
thermal tails on the BF862s by keeping their dissipation constant.
(Running them just slightly above I_DSS gives them a zero tempco anyway,
but this one needs really good dc stability.)

[John Larkin]

Funny you should say "cascode." I was just about to post the
following:

If I want a fast, low-capacitance, accurate, positive current source,
I could build a fairly slow active current source with a voltage
reference, a resistor, an opamp, and a p-fet. But it would have a lot
of capacitance and would be fairly slow. So I could cascode that with
a microwave-type PNP transistor, with a bit of base resistance or a
ferrite bead to keep it stable. But then I'd have the base current
error, and the good fast PNPs, what few there are, have mediocre
betas.

A PNP darlington is too slow.

So, how to correct for the base current error?

Two ideas so far: make a Darlington, but add a lowpass filter from the
higher-current transistor base, into the emitter of the second
transistor. The main transistor needed a base resistor anyhow.

Or sense the base current of the PNP, with opamps and such, and
increase the current of the precision/slow source to make up for it.

[Fred Bartoli]

John Larkin a écrit :

Try this :





Ic = 3mA

1K |
___ |/
5V>-----|___|----+------|
| |>
.-. |
| | |
1K | | |
'-' |
| |
|\ | ||-+
3V>--|+\ | ||<-
| >---------|-----||-+
.-|-/ | |
| |/ | ___ |
'--------------+--|___|-+
|
1K .-.
| |
| |1K
'-'
|
===
GND


Delta Ic < 3nA for 45<Beta<65 for ex.



--
Thanks,
Fred.

[Fred Bartoli]

Fred Bartoli a écrit :

That is 3nA for an opamp with 10K DC gain, 300pA for 100K,...

Maybe I should patent it and license it for one beer?

Oops, too late...

--
Thanks,
Fred.

[Jim Thompson]

Interesting cancellation of beta terms. But doesn't VAF of the NPN
bung thing up?

...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

[Phil Hobbs]

Cute!

[Phil Hobbs]

If you use a BFP650, VAF is huge (or even negative), even with a 40 GHz
f_T. See Figure 3 on the datasheet, http://tinyurl.com/6rkavak

[dagmarg...@yahoo.com]

On Feb 12, 2:15 pm, John Larkin

<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs
>
>
>
>
>
>
>
>
>
> <pcdhSpamMeSensel...@electrooptical.net> wrote:
> >Okkim Atnarivik wrote:
>

> Funny you should say "cascode." I was just about to post the
> following:
>
> If I want a fast, low-capacitance, accurate, positive current source,
> I could build a fairly slow active current source with a voltage
> reference, a resistor, an opamp, and a p-fet. But it would have a lot
> of capacitance and would be fairly slow. So I could cascode that with
> a microwave-type PNP transistor, with a bit of base resistance or a
> ferrite bead to keep it stable. But then I'd have the base current
> error, and the good fast PNPs, what few there are, have mediocre
> betas.
>
> A PNP darlington is too slow.
>
> So, how to correct for the base current error?
>
> Two ideas so far: make a Darlington, but add a lowpass filter from the
> higher-current transistor base, into the emitter of the second
> transistor. The main transistor needed a base resistor anyhow.
>
> Or sense the base current of the PNP, with opamps and such, and
> increase the current of the precision/slow source to make up for it.

It's sure a lot easier if you can flip the circuit into a sink. Then
you've got more choices than just a BFT92.

James

[Jim Thompson]

On Sun, 12 Feb 2012 18:39:45 -0500, Phil Hobbs

I only eyeballed the beta cancellation. It may well be that VAF drops
out as well... interesting!

[Phil Hobbs]

Jim Thompson wrote:
>
> On Sun, 12 Feb 2012 18:39:45 -0500, Phil Hobbs
> <pcdhSpamM...@electrooptical.net> wrote:
>
> >Jim Thompson wrote:
> >>

<snipped Fred's cute beta cancellation circuit>

> >>
> >> Interesting cancellation of beta terms. But doesn't VAF of the NPN
> >> bung thing up?
> >
> >If you use a BFP650, VAF is huge (or even negative), even with a 40 GHz
> >f_T. See Figure 3 on the datasheet, http://tinyurl.com/6rkavak
> >
> >Cheers
> >
> >Phil Hobbs
>
> I only eyeballed the beta cancellation. It may well be that VAF drops
> out as well... interesting!

One way of looking at the Early effect is a parasitic C-B conductance,
so I think Fred's circuit gets rid of that too, but only at low
frequency, since it requires the op amp and FET to do their thing.
Using a BFP650 ideally makes the output stiff at all frequencies from DC
to several gigahertz. I have _got_ to try that in a diode laser driver.

[Jim Thompson]

On Sun, 12 Feb 2012 19:27:03 -0500, Phil Hobbs

<pcdhSpamM...@electrooptical.net> wrote:

>Jim Thompson wrote:
>>
>> On Sun, 12 Feb 2012 18:39:45 -0500, Phil Hobbs
>> <pcdhSpamM...@electrooptical.net> wrote:
>>
>> >Jim Thompson wrote:
>> >>
>
><snipped Fred's cute beta cancellation circuit>
>
>> >>
>> >> Interesting cancellation of beta terms. But doesn't VAF of the NPN
>> >> bung thing up?
>> >
>> >If you use a BFP650, VAF is huge (or even negative), even with a 40 GHz
>> >f_T. See Figure 3 on the datasheet, http://tinyurl.com/6rkavak
>> >
>> >Cheers
>> >
>> >Phil Hobbs
>>
>> I only eyeballed the beta cancellation. It may well be that VAF drops
>> out as well... interesting!
>
>One way of looking at the Early effect is a parasitic C-B conductance,
>so I think Fred's circuit gets rid of that too, but only at low
>frequency, since it requires the op amp and FET to do their thing.
>Using a BFP650 ideally makes the output stiff at all frequencies from DC
>to several gigahertz. I have _got_ to try that in a diode laser driver.
>
>Cheers
>
>Phil Hobbs

Wonder if there's duality? I could certainly do with stiffer CMOS
mirrors.

[dagmarg...@yahoo.com]

On Feb 12, 4:49 pm, Fred Bartoli <" "> wrote:
> John Larkin a écrit :

> > Funny you should say "cascode." I was just about to post the

I saw a similar (but I think more complicated) current sink scheme
decades(?) ago in one of the magazines. It used an op amp diff amp,
bjt pass element, sensed base resistor drop to sense the base current,
and fed that back.

I don't think it was a cascode--just one pass element, IIRC.

I might have saved it somewhere, but I've got no idea where. Maybe
that's enough to jog someone's memory?

--
Cheers,
James Arthur

[John Larkin]

On Sun, 12 Feb 2012 18:39:45 -0500, Phil Hobbs

That is a very weird transistor.

But if I wanted a negative current source, I'd just cascode into a
PHEMT. I need positive current, and the people who make gaasfets
stubbornly refuse to make p-channel ones.

I could put a compound inductor directly in series with my mosfet
current source, I suppose.

[John Larkin]

On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmarg...@yahoo.com
wrote:

Sadly, I need a positive current source. My customer wants really
precise laser bias current, which I frankly think is unjustified, but
unfortunately he's the one with the money.

[Phil Hobbs]

It's hotter than a two-dollar pistol, too. Pity there aren't any PNPs.

With the beta cancellation trick, it might be interesting to try using a
BFT92 in inverted mode as the cascode device. Inverted RF transistors
have lower capacitance, and their f_T can even be higher than the same
device right-way-up.

[dagmarg...@yahoo.com]

Oh, this is it here:

http://www.edn.com/article/463825-Error_compensation_improves_bipolar_current_sinks.php


R5
.--47K---. .---<--< Iout
| | |
R4 | |\ | R2 47 |
Vref >---1K---+--|+\ | ___ |/
| >--+---|___|---+-| Q1
.-|-/ | | |>. 2n3020
| |/ --- Ccomp | |
| --- R6 | |
| | ___ 47K| |
| +---|___|---' |
| | ___ |
'-------+---|___|-------+
R3 1k |
.-.
| | R1
| | 1 ohm
'-'
|
===
GND


It says 2006, but it sure seems longer.

--------------
keywords (for future searches): EDN current source current sink base
current compensation error cancellation

"Error compensation improves bipolar-current sinks.(design ideas)
October 1, 2006 | Thompson, Brad; Granville, Fran

You can improve a current sink's accuracy by at least two orders of
magnitude by adding two standard 1%-tolerance resistors. As a bonus,
you also compensate for errors that a low-current-gain pass
transistor's base current introduces. To do so, you measure the
transistor's base current and add a proportionally scaled error term
to the source's reference voltage."
--------------


--
Cheers,
James Arthur

[dagmarg...@yahoo.com]

On Feb 12, 9:27 pm, Phil Hobbs

<pcdhSpamMeSensel...@electrooptical.net> wrote:
> John Larkin wrote:

> > But if I wanted a negative current source, I'd just cascode into a
> > PHEMT. I need positive current, and the people who make gaasfets
> > stubbornly refuse to make p-channel ones.
>
> > I could put a compound inductor directly in series with my mosfet
> > current source, I suppose.
>
> It's hotter than a two-dollar pistol, too.  Pity there aren't any PNPs.
>
> With the beta cancellation trick, it might be interesting to try using a
> BFT92 in inverted mode as the cascode device.  Inverted RF transistors
> have lower capacitance, and their f_T can even be higher than the same
> device right-way-up.

But for ordinary rail voltages it'll break down base-to-emitter,
that's the rub.

I was going to suggest John could just use a high rail voltage and a
bootstrapped resistor--then lots of current-source problems go away.

E.g., to be absurd, 1kV and a resistor gives 10ppm stability into a
10mV compliance range. Resistors are excellent, r.f.-wise. To be
less absurd, bootstrap the resistor.

--
Cheers,
James Arthur

[Okkim Atnarivik]

Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:
: I started by taking your advice, which was pretty valuable--thanks

Don't mention it. I have already gained more from your, Larkin's,
Win Hill's and many other's posts here in SED.

: again. I noticed the change in specs as well, when I ordered another

: batch a couple of months ago. They changed the type number to BFP650H
: instead of BFP650E when they changed the process. (The dogs.)

They did? The datasheet still list the part as BFP650 without suffixes
even though it is the SiGe:C version.

: I'm using it as a cascode stage for a SKY65050 pHEMT, and it works

Interesting hint, I should give it a try.

: great. One of the best things about it is that it has effectively

: infinite Early voltage, so you can get a lot of voltage gain out of a
: single stage. Even for situations where you don't care so much about
: ultralow noise, the combination of a gigantic f_T with a very high V_A
: is unique in my collection. It does want to oscillate at 14 GHz if you
: look at it crosswise, but a nice 5-ohm bead in series with the base
: cleans that right up.

The high Early seems to be characteristic to the SiGe's, it puzzled me
already in the thread http://tinyurl.com/8a44pzn . If you check the
page 4 of http://tinyurl.com/6qmnd4u you see it there too (and additionally
you get the large hFE = 400 ).

Regards,
Mikko

[Okkim Atnarivik]

John Larkin <jjla...@highnotlandthistechnologypart.com> wrote:
: But if I wanted a negative current source, I'd just cascode into a

: PHEMT. I need positive current, and the people who make gaasfets
: stubbornly refuse to make p-channel ones.

The same with SiGe PNP's - I would have wanted to make a push-pull
for LHe but there are no discretes available. An NPN-NPN push-pull
is a possibility, but awkward. Maybe the discretes will become available
soon, at least there are complementary IC processes producing such
devices as the OPA835/6 (which actually work in LHe but only marginally).

Regards,
Mikko

[dagmarg...@yahoo.com]

On Feb 12, 9:07 pm, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmargoodb...@yahoo.com

> Sadly, I need a positive current source. My customer wants really
> precise laser bias current, which I frankly think is unjustified,

If you sense and compensate base current perfectly, the current source
will be perfect. Maybe a flying cap sampling the base resistor would
do better than Fred's and the EDN ckts?

For speed beyond the BFT92, easiest are the hybrids you've already
done--outboard additions (L and R) that improve GHz response, the bjt
handles mid-range, and the op-amp handles d.c. accuracy.

Cascoding two BFT92s helps, eliminating Cfb feedback of the exposed
device.

To do better than that, bootstrap the exposed BFT92. I can think of a
few ways--pretty hairy.

> but
> unfortunately he's the one with the money.

No, it's actually good he's the one with the money. If you had it,
you wouldn't need him!

--
Cheers,
James Arthur

[John Larkin]

Unfortunately, most of the good fast parts are designed for RF use,
which means we only get n-types, and very badly characterized ones at
that. S-params and load pulls don't tell you much about DC or
switching behavior. You're lucky if they tell you Idss, very lucky if
they give you DC curves, and heaven is getting a Spice model.

I asked Mini-Circuits if one of their MMICs inverts the signal or not.
They didn't know.

Grrrrr.

[dagmarg...@yahoo.com]

On Feb 13, 9:28 am, dagmargoodb...@yahoo.com wrote:

> On Feb 12, 9:07 pm, John Larkin wrote:
> > On Sun, 12 Feb 2012 16:07:51 -0800 (PST), dagmargoodb...@yahoo.com
> > wrote:
>

What if you separate the a.c. and d.c. paths at the pass element too?
(Maybe this is what you meant by "lowpass filter at the base"?)

|
|<'
low z a.c. >----+---| BFT92
| |\
|-' |
d.c. bias >-<-| |
|---Z1--+
|
.-.
| | f.b.
| |
'-'
|

Z1 isolates FET capacitance @ rf.

James

[John Larkin]

On Mon, 13 Feb 2012 08:41:34 -0800 (PST), dagmarg...@yahoo.com
wrote:

One could just use a p-channel mosfet as the cascode. That could get
the source capacitance down to 20 pF, maybe better if one found the
right fet (I'm talking 50-100 mA maybe.) Then put a good wideband
inductor in its drain. "Good wideband inductor" is non-trivial. The
Piconics conical things are good, but expensive and hard to handle.
Making your own takes a bunch of parts... several different FBs and Ls
in series, with damping resistors.

[Phil Hobbs]

On 02/13/2012 03:12 AM, Okkim Atnarivik wrote:
> Phil Hobbs<pcdhSpamM...@electrooptical.net> wrote:
> : I started by taking your advice, which was pretty valuable--thanks
>
> Don't mention it. I have already gained more from your, Larkin's,
> Win Hill's and many other's posts here in SED.
>
> : again. I noticed the change in specs as well, when I ordered another
> : batch a couple of months ago. They changed the type number to BFP650H
> : instead of BFP650E when they changed the process. (The dogs.)
>
> They did? The datasheet still list the part as BFP650 without suffixes
> even though it is the SiGe:C version.

The distributors' stock of the E version all ran out a few months ago,
and all they have now is the H version. I haven't managed to get the
straight story as to just what they did.

>
> : I'm using it as a cascode stage for a SKY65050 pHEMT, and it works
>
> Interesting hint, I should give it a try.
>
> : great. One of the best things about it is that it has effectively
> : infinite Early voltage, so you can get a lot of voltage gain out of a
> : single stage. Even for situations where you don't care so much about
> : ultralow noise, the combination of a gigantic f_T with a very high V_A
> : is unique in my collection. It does want to oscillate at 14 GHz if you
> : look at it crosswise, but a nice 5-ohm bead in series with the base
> : cleans that right up.
>
> The high Early seems to be characteristic to the SiGe's, it puzzled me
> already in the thread http://tinyurl.com/8a44pzn . If you check the
> page 4 of http://tinyurl.com/6qmnd4u you see it there too (and additionally
> you get the large hFE = 400 ).

I've ordered some of those too, thanks. It's a bit worrying that there
are seem to be only about 100 of them in the world, according to FindChips.

Cheers

Phil
>
> Regards,
> Mikko

[George Herold]

On Feb 12, 4:49 pm, Fred Bartoli <" "> wrote:

> John Larkin a écrit :
>
>
>
>
>
> > On Sun, 12 Feb 2012 13:39:37 -0500, Phil Hobbs
> > <pcdhSpamMeSensel...@electrooptical.net> wrote:
>
> >> Okkim Atnarivik wrote:

> Fred.- Hide quoted text -

>
> - Show quoted text -

Would someone be kind enough to help me see how this compensates for
the base current? I've done the Vref-opamp-fet-resistor current
source.

Color me slightly confused.
(I'd be happy to pay the, one beer licensing fee,... to understand
it.)

George H.

[Phil Hobbs]

The summing junction is held still by feedback, and we assume the 5V
reference is also stable. Any base current thus divides 50:50 between
the two 1k resistors. Thus I_B /2 goes to the SJ, pulling it down by
(2k)(I_B/2) = 1000 ohms * I_B. To restore balance, the source of the
FET has to go up by the same amount, which requires an increase in drain
current of 1000*I_B/1000, i.e. exactly I_B.

That'll work great at DC, but any fast voltage swing at the collector
will make a mess, due to the slow loop dynamics and high impedance at
the base. The simple way to fix it is to use a transistor with infinite
Early voltage and bypass the daylights out of the base, which is how the
BFP650 came up.

[Jim Thompson]

Just run the math. Assume Ic is output, base current is Ic/beta,
emitter current is Ic*(beta+1)/beta

You'll find a need for 2*VR-VH > 0 (VR=3, VH=5 in this example)

And Ic can't be 3mA as shown in the example ;-)

[John Larkin]

But that part has negative (positive?) Early voltage, namely negative
output impedance. I wonder if that's thermal or wideband.

So, it would oscillate with an LC in its collector and no base drive!

[George Herold]

On Feb 13, 12:20 pm, Phil Hobbs

> hobbs at electrooptical dot nethttp://electrooptical.net- Hide quoted text -

>
> - Show quoted text -

Great, Thanks... I was missing that the current flowing from the
base-1k-SJ then had to go somewhere.... (adding it's bit to the sense
resistor.)

George H.

[George Herold]

On Feb 13, 12:36 pm, Jim Thompson <To-Email-Use-The-Envelope-I...@On-

> | E-mail Icon athttp://www.analog-innovations.com|    1962     |
>
> I love to cook with wine.     Sometimes I even put it in the food.- Hide quoted text -

>
> - Show quoted text -

Thanks Jim... Not so much doing the math, but I was missing where the
currents were flowing.

George H.

[dagmarg...@yahoo.com]

On Feb 13, 11:56 am, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Mon, 13 Feb 2012 08:41:34 -0800 (PST), dagmargoodb...@yahoo.com

Using the BFT92 as the cascode gets you to ~1.5pf all by itself.

It all depends on the particulars (a ramp generator has different
needs than a laser diode, etc.).


James

[John S]

Right! It's 1mA for this example.

John S

[Phil Hobbs]

It looks like it's thermal, because it only happens at high collector
currents. Down where you'd actually want to use it, it's flat like a
ruler.

I'm seriously considering resurrecting that constant-dissipation trick
for thermal tracking in noise cancellers, i.e. jack up V_CE as I_C goes
down so as to avoid thermal tails and offsets. It was worthless with
VAF=12V Si RF devices, but with VAF >= 1 kV, it might be competitive.

[BobW]

On 2/11/2012 12:57 PM, John Larkin wrote:
>
>
> We're designing some laser driver boards, and I thought it would be
> cool to add test points to some of the interesting circuit nodes.
> Regular scope probes don't work at ps speeds, and probe grounding is
> difficult. There are multi-GHz active probes, at roughly $1 per Hz.
>
> So I was thinking that I could add a small, 0603 maybe, resistor to a
> signal to be snooped, run a 50 ohm trace some small distance, and end
> up at some structure that had a signal test point and a ground.
> Something roughly like this:
>
> http://db.tt/6rRcagTt
>
> The resistor could be 450 or 950 ohms, for a 10:1 or 20:1 ratio.
>
> The "probe" would be a piece of hardline coax that runs to a sampling
> scope input, with some provision for grounding. Something like a 3-5
> GHz bandwidth should be feasible.
>
> This looks promising if a little klunky. Any other ideas?
>
>

I would add a high speed connector. The best I've seen (so far) is the
Huber & Suhner MMPX line. They are tiny but rated to about 68GHz.
However, with coax and end connectors, your bandwidth will be reduced
from that.

We're looking at the latest LeCroy scopes to look at 37GHz signals.
Their new 10 series has up to 60GHz of bandwidth and 160GSamples/s (real
time). It's only about $400K :-|. They use 2.4mm connectors. I'm not
sure if you can get a direct MMPX<->2.4mm without adapters.

Bob

[Robert Baer]

Oh..i like that non-functional circuit on page two..

[Robert Baer]

Yeah; that floating gate is a bit puzzling..

[dagmarg...@yahoo.com]

On Feb 13, 11:05 pm, Robert Baer <robertb...@localnet.com> wrote:
> George Herold wrote:
> > On Feb 12, 4:49 pm, Fred Bartoli<"">  wrote:
> >> John Larkin a écrit :

> >>> So, how to correct for the base current error?
>
> >> Try this :
>
> >>                       R1            Ic = 3mA
> >>                     1K               |
> >>                       ___            |/
> >>              5V>-----|___|----+------| Q1
> >>                               |      |>.
> >>                              .-.       |
> >>                           R2 | |       |
> >>                           1K | |       |
> >>                            '-'       |
> >>                               |        |
> >>                  |\           |     ||-'
> >>             3V>--|+\          |     ||<- Q2

> >>                  | >---------|-----||-+
> >>                .-|-/          |        |
> >>                | |/           |   ___  |
> >>                '--------------+--|___|-+

> >>                                   R3   |
> >>                                   1K  .-.
> >>                                       | |R4

> >>                                       | |1K
> >>                                       '-'
> >>                                        |
> >>                                       ===
> >>                                       GND
>
> >> Delta Ic<  3nA for 45<Beta<65 for ex.
>
> >> --
> >> Thanks,
> >> Fred.
>

> > Would someone be kind enough to help me see how this compensates for
> > the base current?  I've done the Vref-opamp-fet-resistor current
> > source.
>
> > Color me slightly confused.
> > (I'd be happy to pay the, one beer licensing fee,... to understand
> > it.)
>
>

>    Yeah; that floating gate is a bit puzzling..

It's not floating, some union guy just bashed it into the op amp with
his forklift. (fixed)

All the base current comes from +5v through R1. Suppose ib(Q2) falls
by 1uA. Where does it go? What does that do?

Okay, now read Phil's post.

HTH,
James Arthur

[Gerhard Hoffmann]

On 02/13/2012 04:21 PM, John Larkin wrote:
> On Mon, 13 Feb 2012 10:30:41 +0200 (EET), Okkim Atnarivik
> <Okkim.A...@twentyfour.fi.invalid> wrote:
>
>> John Larkin<jjla...@highnotlandthistechnologypart.com> wrote:
>> : But if I wanted a negative current source, I'd just cascode into a
>> : PHEMT. I need positive current, and the people who make gaasfets
>> : stubbornly refuse to make p-channel ones.
>>
>> The same with SiGe PNP's - I would have wanted to make a push-pull
>> for LHe but there are no discretes available. An NPN-NPN push-pull
>> is a possibility, but awkward. Maybe the discretes will become available
>> soon, at least there are complementary IC processes producing such
>> devices as the OPA835/6 (which actually work in LHe but only marginally).
>>
>> Regards,
>> Mikko
>
> Unfortunately, most of the good fast parts are designed for RF use,
> which means we only get n-types, and very badly characterized ones at
> that. S-params and load pulls don't tell you much about DC or
> switching behavior. You're lucky if they tell you Idss, very lucky if
> they give you DC curves, and heaven is getting a Spice model.
>

It's even worse if you want a fast PNP that is space-proof.
Say, a BFT93 in a metal can. Nothing. If there was one, you can be
sure that its qualification has timed out 10 years ago...

Yesterday, I changed a Wilson current mirror from BFT93 to
Intersil HFA3096. Used to work nicely with BFT93, but oscillates
like hell with the HFA3096 if it gets more than 1 mA :-((

I wonder what will happen with these extra-wide-body hermetic
flatpacks in place of the SO-16s. Probably nothing good.
But at least there is a rad-hardened version.

Is there a sure way to calm down a Wilson current mirror without
too much sandbagging or introducing noise? It MUST work into
a capacitive load.
A simple mirror seems to be stable.

> I asked Mini-Circuits if one of their MMICs inverts the signal or not.
> They didn't know.
>

I think they just relabel semiconductors they buy somewhere else, and
they may even switch manufacturers. I think that was the case for ERAs.

And the phase noise of old SRA-1 used to be better than it is now.
But it was never specified, so we cannot complain. Still have NOS.

> Grrrrr.
>

Rrrrrrrrrright!

Gerhard

[Gerhard Hoffmann]

On 02/14/2012 01:22 AM, BobW wrote:

> We're looking at the latest LeCroy scopes to look at 37GHz signals.
> Their new 10 series has up to 60GHz of bandwidth and 160GSamples/s (real
> time). It's only about $400K :-|. They use 2.4mm connectors. I'm not
> sure if you can get a direct MMPX<->2.4mm without adapters.

In december, we had one of these 80 GSPS real time thingies from Agilent
for a week to test, made me quite greedy.

Gerhard

[Phil Hobbs]

Put a 10-100 ohm ferrite bead in series with each of the bases.
(including the diode-connected ones).

[John Larkin]

On Tue, 14 Feb 2012 08:48:15 +0100, Gerhard Hoffmann <dk...@arcor.de>
wrote:

Right. They trashed the ERA5 beyond recognition. Caused us grrrrrief.

>
>And the phase noise of old SRA-1 used to be better than it is now.
>But it was never specified, so we cannot complain. Still have NOS.
>
>> Grrrrr.
>>
>
>Rrrrrrrrrright!
>
>Gerhard

[John Larkin]

On Tue, 14 Feb 2012 08:56:40 +0100, Gerhard Hoffmann <dk...@arcor.de>
wrote:

I have a Tek 11801 sampler with a 40 GHz head. Costs about $2K on
ebay.

[Gerhard Hoffmann]

On 02/14/2012 04:16 PM, John Larkin wrote:
> On Tue, 14 Feb 2012 08:56:40 +0100, Gerhard Hoffmann<dk...@arcor.de>
> wrote:
>
>> On 02/14/2012 01:22 AM, BobW wrote:
>>
>>> We're looking at the latest LeCroy scopes to look at 37GHz signals.
>>> Their new 10 series has up to 60GHz of bandwidth and 160GSamples/s (real
>>> time). It's only about $400K :-|. They use 2.4mm connectors. I'm not
>>> sure if you can get a direct MMPX<->2.4mm without adapters.
>>
>> In december, we had one of these 80 GSPS real time thingies from Agilent
>> for a week to test, made me quite greedy.
>>
>> Gerhard
>
> I have a Tek 11801 sampler with a 40 GHz head. Costs about $2K on
> ebay.

That was at my customer's.

I myself have a hp54750a with 20 and 50 GHz plug-ins, 9 ps risetime.
But it is not real-time. One cannot trigger on the one bit error in
a 10 GB/s fiber optic link that may or may not happen this night.

5 years ago, we could have been faster by 6 weeks during the
development of a XFP transceiver.

[John Larkin]

Wow, that's pretty bad.

Are they suggesting that, since it's beta bracketed, it should be beta
biased? Certain parties Would Not Approve. Makes sense for RF amps.


--

John Larkin, President
Highland Technology, Inc

jlarkin at highlandtechnology dot com

http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators

Custom laser controllers

Photonics and fiberoptic TTL data links

VME thermocouple, LVDT, synchro acquisition and simulation

[Phil Hobbs]

On 02/14/2012 10:16 AM, John Larkin wrote:
> On Tue, 14 Feb 2012 08:56:40 +0100, Gerhard Hoffmann<dk...@arcor.de>
> wrote:
>
>> On 02/14/2012 01:22 AM, BobW wrote:
>>
>>> We're looking at the latest LeCroy scopes to look at 37GHz signals.
>>> Their new 10 series has up to 60GHz of bandwidth and 160GSamples/s (real
>>> time). It's only about $400K :-|. They use 2.4mm connectors. I'm not
>>> sure if you can get a direct MMPX<->2.4mm without adapters.
>>
>> In december, we had one of these 80 GSPS real time thingies from Agilent
>> for a week to test, made me quite greedy.
>>
>> Gerhard
>
> I have a Tek 11801 sampler with a 40 GHz head. Costs about $2K on
> ebay.
>
>

Plus it doesn't run Windows!

[Phil Hobbs]

Well, you can trigger off it--delay lines are easy in fibre--but you
only get one data point per trigger. Just choose that point well. ;)

[John Larkin]

On Tue, 14 Feb 2012 18:46:18 +0100, Gerhard Hoffmann <dk...@arcor.de>

A realtime scope would be handy around here for snooping PCI Express
signals. Maybe pricing for a, say, 4 GHz scope (with probes!) will get
reasonable some day.


--

John Larkin, President
Highland Technology, Inc

jlarkin at highlandtechnology dot com

http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators

Custom laser controllers

Photonics and fiberoptic TTL data links

[Phil Hobbs]

You can get a TDS694 for $5k-ish. 3 GHz, 10 Gs/s.

[Gerhard Hoffmann]

On 02/14/2012 02:34 PM, Phil Hobbs wrote:

> Gerhard Hoffmann wrote:

>> Yesterday, I changed a Wilson current mirror from BFT93 to
>> Intersil HFA3096. Used to work nicely with BFT93, but oscillates
>> like hell with the HFA3096 if it gets more than 1 mA :-((
>>
>> I wonder what will happen with these extra-wide-body hermetic
>> flatpacks in place of the SO-16s. Probably nothing good.
>> But at least there is a rad-hardened version.
>>
>> Is there a sure way to calm down a Wilson current mirror without
>> too much sandbagging or introducing noise? It MUST work into
>> a capacitive load.
>> A simple mirror seems to be stable.

No, the simple npn mirror wasn't stable, too :-(

>
> Put a 10-100 ohm ferrite bead in series with each of the bases.
> (including the diode-connected ones).

That's what I meant with sandbagging & noise. 50R makes it stable,
but it adds to Rbb. I won't get ferrites for fear of noise pickup.
As usual, no silver bullet. But many thanks, anyway.
And I like your book!

sigh, Gerhard

[George Herold]

Ouh, ferrites can add to magnetic pickup. I never thought of that.
Thanks,
George H.
(a mini toroid?)

[Bill Sloman]

Probably because it doesn't happen. A chunk of ferrite - aka a ferrite
bead - can provide a low permeability short-cut for any magnetic field
in the immediate vicinity, but ferrite beads are small and don't have
much of an immediate vicinity. The flux concentrations they produce
are going to be really hard to detect.

--
Bill Sloman, Nijmegen

[George Herold]

After posting I realized that a ferrite 'bead' is just like a one turn
toroid. So it is pretty hard to couple in any field.
It might be fun to try and measure. A one loop 'pick-up' coil versus
a coil with a ferrite bead on it.

George H.
>
> --
> Bill Sloman, Nijmegen- Hide quoted text -

[John Larkin]

Loops in the PCB trace layout will be better magnetic pickups than a
ferrite bead.


--

John Larkin, President 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

[Nico Coesel]

Thats way too expensive for such an old scope. I strongly doubt you
can actually use this scope at 3GHz. It has BNC inputs (where most
other equipment uses N or SMB for anything over 500MHz). There several
offered on Ebay but they are not getting sold for 5k.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

[Phil Hobbs]

(a) You can buy four of them for the price of anything comparable new;
(b) They work great.
(c) Nobody spending his own money buys for what the dealers are asking.

I've got one on my shopping list for 2012--we'll see!

[Fred Bartoli]

John Larkin a écrit :

And the comments above the wonderful circuit is pretty bad as well...

> Are they suggesting that, since it's beta bracketed, it should be beta
> biased? Certain parties Would Not Approve. Makes sense for RF amps.
>
>


--

Thanks,
Fred.

[John Larkin]

On Wed, 15 Feb 2012 17:38:38 GMT, ni...@puntnl.niks (Nico Coesel)
wrote:

BNCs work fine at 3 GHz. But you'd need active probes in a lot of
situations, like snooping serial busses, and they can get expensive.

[Gerhard Hoffmann]

On 02/15/2012 03:55 PM, John Larkin wrote:

> Loops in the PCB trace layout will be better magnetic pickups than a
> ferrite bead.
>

Especially if there are unrelated pieces of ferrite sitting inside
the loop, yelling "All you fluxes, come here, here is a well-paved
road".

[John Larkin]

On Wed, 15 Feb 2012 20:48:02 +0100, Gerhard Hoffmann <dk...@arcor.de>
wrote:

A surface-mount ferrite bead is not much magnetic volume. I don't know
how they are constructed, but I don't think they are coils... maybe
just serpentines. Beads typically have a few uH of base indictance.

http://www.vishay.com/docs/ilb_ilbb_enote.pdf

This claims that beads are "closed magnetic circuits" and "inherently
shielded."


--

John Larkin, President
Highland Technology, Inc

jlarkin at highlandtechnology dot com

http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators

Custom laser controllers

Photonics and fiberoptic TTL data links

[John Larkin]

Yeah, could you please translate that into English for us?


--

John Larkin, President 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