typically stupid ED article...

[John Larkin]

On Tue, 8 Sep 2020 14:31:19 +0000 (UTC), antispam@math.uni.wroc.pl
wrote:

jlarkin@highlandsniptechnology.com wrote:

Even a 10M input, $20 DVM is a decent picoammeter.


Works nice with old meters. But digitally calibrated \"10M\" DVM
may have 11M input resistance, for such DVM to get any accuracy
one would have to calibrate it and multiply result by correcton
factor, not so nice. Also, significant part of \"10M\" may be
input leakage.

That\'s not a giant burden, to get a cheap picoammeter. My handheld
Flukes are really 10M. It\'s not hard to check.

I have nice cheap Chinse meter, lowest range is 10mV with
four digit resolution. Using it I can measure voltage of
a termocouple and see difference when termocouple is on table
and when it is on floor level. One gets funny results
measuring voltage on well discharged polyester capacitor.
There is substantial drift, indicating that most of
input \"resistance\" is in fact leakage. So, while quite
sensitive this DVM needs external resistor to measure
low currents.

I did a little experiment last week: charged a film cap to 5 volts and
connected it to my Fluke 8845, on its 10-volt HI-Z range. The DVM
slowly charged the cap, at about 15 pA.

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

https://www.dropbox.com/sh/5b2fs47b1qukefv/AAAQgOWmjbeYYsxglN23Py2ya?dl=0

I think there are even better opamps around now.


--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[whit3rd]

On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.

 

[Phil Hobbs]

On 2020-09-08 15:38, whit3rd wrote:

On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.

For a one-off, just using something like an LMC660CN (quad, 14-pin DIP)
mounted dead bug will make a pretty decent femtoammeter. You get 14
low-leakage standoffs for free, and you can use the other sections to
bootstrap the input protection circuitry.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[John Larkin]

On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.


For a one-off, just using something like an LMC660CN (quad, 14-pin DIP)
mounted dead bug will make a pretty decent femtoammeter. You get 14
low-leakage standoffs for free, and you can use the other sections to
bootstrap the input protection circuitry.

Cheers

Phil Hobbs

Insect abuse. I solder a few pins to the copperclad, or use bypass
caps as standoffs, and bend the signal pins out.

https://www.dropbox.com/s/oegve42oi34kt4e/Live_Bug.jpg?raw=1

(RCA gate. That pic must be pretty old.)

I can still see the part number and count in the usual direction.

Agree that air is the best insulator, and there\'s nothing wrong with
mid-air junctions. Plastic DIP packages seem to have fA surface
leakage.


--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[Phil Hobbs]

On 2020-09-09 13:23, John Larkin wrote:

On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.


For a one-off, just using something like an LMC660CN (quad, 14-pin DIP)
mounted dead bug will make a pretty decent femtoammeter. You get 14
low-leakage standoffs for free, and you can use the other sections to
bootstrap the input protection circuitry.


Insect abuse.

At least I put them out of their misery, unlike _some_ people.

I solder a few pins to the copperclad, or use bypass
caps as standoffs, and bend the signal pins out.

The pins get weakened a lot when you do that, though. In their natural
state they can handle a lot of torque from bending resistor leads and stuff.

https://www.dropbox.com/s/oegve42oi34kt4e/Live_Bug.jpg?raw=1

(RCA gate. That pic must be pretty old.)

I can still see the part number and count in the usual direction.

You probably open boiled eggs from the small end, too. That would figure.

I just put a gouge in the pin 1 end of the package using dikes, and
sometimes scribble the P/N on the copper with a fine-point Sharpie.

Agree that air is the best insulator, and there\'s nothing wrong with
mid-air junctions. Plastic DIP packages seem to have fA surface
leakage.

Yeah, they\'re surprisingly good. Of course you can\'t make a fA op amp
with a leaky package.

Cheers

Phil \"the only good bug is a dead bug\" Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[John Larkin]

On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.


For a one-off, just using something like an LMC660CN (quad, 14-pin DIP)
mounted dead bug will make a pretty decent femtoammeter. You get 14
low-leakage standoffs for free, and you can use the other sections to
bootstrap the input protection circuitry.


Insect abuse.

At least I put them out of their misery, unlike _some_ people.

I solder a few pins to the copperclad, or use bypass
caps as standoffs, and bend the signal pins out.

The pins get weakened a lot when you do that, though. In their natural
state they can handle a lot of torque from bending resistor leads and stuff.

_some_people just get everything right first time.

https://www.dropbox.com/s/oegve42oi34kt4e/Live_Bug.jpg?raw=1

(RCA gate. That pic must be pretty old.)

I can still see the part number and count in the usual direction.

You probably open boiled eggs from the small end, too. That would figure.

Boiling eggs is too much trouble, and I get burned handling them. I
have a new egg cooking technique that I will reveal if enough people
beg.

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

I just put a gouge in the pin 1 end of the package using dikes, and
sometimes scribble the P/N on the copper with a fine-point Sharpie.

Call the SPCI.


--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[Phil Hobbs]

On 2020-09-09 18:08, John Larkin wrote:

On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

You can get cheap fA-bias current opamps and make your own very
sensitive current meter easily.

yeah, but... lots of current meter applications have HV sources, and
require protection on the inputs. There\'s also the little problem of PC
board leakage (some teflon standoffs and discrete FETs are the old
solutions for this).

The real problems with build-your-own are more packaging than
semiconductor performance.


For a one-off, just using something like an LMC660CN (quad, 14-pin DIP)
mounted dead bug will make a pretty decent femtoammeter. You get 14
low-leakage standoffs for free, and you can use the other sections to
bootstrap the input protection circuitry.


Insect abuse.

At least I put them out of their misery, unlike _some_ people.

I solder a few pins to the copperclad, or use bypass
caps as standoffs, and bend the signal pins out.

The pins get weakened a lot when you do that, though. In their natural
state they can handle a lot of torque from bending resistor leads and stuff.

_some_people just get everything right first time.



https://www.dropbox.com/s/oegve42oi34kt4e/Live_Bug.jpg?raw=1

(RCA gate. That pic must be pretty old.)

I can still see the part number and count in the usual direction.

You probably open boiled eggs from the small end, too. That would figure.

Boiling eggs is too much trouble, and I get burned handling them. I
have a new egg cooking technique that I will reveal if enough people
beg.

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

Plus the shoot-through and asymmetry teaches you a lot about bypassing.

I just put a gouge in the pin 1 end of the package using dikes, and
sometimes scribble the P/N on the copper with a fine-point Sharpie.

Call the SPCI.

Society for the Promotion of Cool Inventions? Bring \'em on.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Bill Sloman]

On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:

On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm.

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.

<snip>

--
Bill Sloman, Sydney

 

[Lasse Langwadt Christensen]

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:

On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm.

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.

what are you on about? John saw that an ATC04 was just as fast with a bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard and the swing is only 450mV or 4.5mA

 

[John Larkin]

On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
<langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm.

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.



--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[Phil Hobbs]

On 2020-09-10 12:07, John Larkin wrote:

On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm.

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

Hopefully with fewer tweaks. Some of those distributed-deflection tubes
were monsters.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Chris]

On 09/10/20 17:16, Phil Hobbs wrote:

On 2020-09-10 12:07, John Larkin wrote:
On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin
wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin
wrote:
I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver
warm.

National Semiconductors LVDS uses the same voltage swing for the
same reason. Most of us have known about this for some forty years
now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a
bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard
and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

Hopefully with fewer tweaks. Some of those distributed-deflection tubes
were monsters.

Cheers

Phil Hobbs

I remember chasing the ripples down the line in Tek 551 distributed
amplifiers. Pain in the whatsit to get right and very time consuming.
Varied with temperature and tube (12BH7 ?) life as well...

Chris

 

[John Larkin]

On Thu, 10 Sep 2020 18:16:43 +0100, Chris <xxx.syseng.yyy@gfsys.co.uk>
wrote:

On 09/10/20 17:16, Phil Hobbs wrote:
On 2020-09-10 12:07, John Larkin wrote:
On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin
wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin
wrote:
I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver
warm.

National Semiconductors LVDS uses the same voltage swing for the
same reason. Most of us have known about this for some forty years
now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a
bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard
and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

Hopefully with fewer tweaks. Some of those distributed-deflection tubes
were monsters.

Cheers

Phil Hobbs



I remember chasing the ripples down the line in Tek 551 distributed
amplifiers. Pain in the whatsit to get right and very time consuming.
Varied with temperature and tube (12BH7 ?) life as well...

Chris

The 545 also had a differential signal delay line that was made of
tunable t-coils. The tuning procedure was iterative. Must have taken
hours. I think the 585 had two cascaded banks of distributed amps.

The 519 managed 1 GHz, with no amps at all. The deflection plates were
distributed, vaguely the same idea.

The early Tek stuff was hand-wired and had selected parts and lots of
tuning. Labor intensive.

There are lots of IC distributed amps around now, mostly to drive
telecom fiber modulators. They are fairly expensive ($200 range) and
have wimpy voltage swings.


--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[Chris]

On 09/10/20 20:44, John Larkin wrote:

On Thu, 10 Sep 2020 18:16:43 +0100, Chris<xxx.syseng.yyy@gfsys.co.uk
wrote:

On 09/10/20 17:16, Phil Hobbs wrote:
On 2020-09-10 12:07, John Larkin wrote:
On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin
wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin
wrote:
I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver
warm.

National Semiconductors LVDS uses the same voltage swing for the
same reason. Most of us have known about this for some forty years
now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a
bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard
and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

Hopefully with fewer tweaks. Some of those distributed-deflection tubes
were monsters.

Cheers

Phil Hobbs



I remember chasing the ripples down the line in Tek 551 distributed
amplifiers. Pain in the whatsit to get right and very time consuming.
Varied with temperature and tube (12BH7 ?) life as well...

Chris




The 545 also had a differential signal delay line that was made of
tunable t-coils. The tuning procedure was iterative. Must have taken
hours. I think the 585 had two cascaded banks of distributed amps.

The 519 managed 1 GHz, with no amps at all. The deflection plates were
distributed, vaguely the same idea.

I remember seeing a 519 around 1966-67. It was being used for laser
research fwir. 1 Ghz real time bandwidth, but as you say, not that
sensitive with y input direct to the plates. The bandwidth was
revolutionary for it\'s time though. Scope camera on it as well. Yes,
am that old but try to stay awake .

The early Tek stuff was hand-wired and had selected parts and lots of
tuning. Labor intensive.

I think some of HP kit was the same and can be very difficult to restore
because of the hand selected parts. Can\'t just replace with
a stock 2n part number and get the same results.

There are lots of IC distributed amps around now, mostly to drive
telecom fiber modulators. They are fairly expensive ($200 range) and
have wimpy voltage swings.

Still not trivial to design and build a very high speed, high voltage
amplifier...

Chris

 

[John Larkin]

On Thu, 10 Sep 2020 23:05:42 +0100, Chris <xxx.syseng.yyy@gfsys.co.uk>
wrote:

On 09/10/20 20:44, John Larkin wrote:
On Thu, 10 Sep 2020 18:16:43 +0100, Chris<xxx.syseng.yyy@gfsys.co.uk
wrote:

On 09/10/20 17:16, Phil Hobbs wrote:
On 2020-09-10 12:07, John Larkin wrote:
On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin
wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin
wrote:
I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver
warm.

National Semiconductors LVDS uses the same voltage swing for the
same reason. Most of us have known about this for some forty years
now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a
bigger swing

and LVDS isn\'t some National Semiconductors special it is a standard
and the swing is only 450mV or 4.5mA

And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

Hopefully with fewer tweaks. Some of those distributed-deflection tubes
were monsters.

Cheers

Phil Hobbs



I remember chasing the ripples down the line in Tek 551 distributed
amplifiers. Pain in the whatsit to get right and very time consuming.
Varied with temperature and tube (12BH7 ?) life as well...

Chris




The 545 also had a differential signal delay line that was made of
tunable t-coils. The tuning procedure was iterative. Must have taken
hours. I think the 585 had two cascaded banks of distributed amps.

The 519 managed 1 GHz, with no amps at all. The deflection plates were
distributed, vaguely the same idea.

I remember seeing a 519 around 1966-67. It was being used for laser
research fwir. 1 Ghz real time bandwidth, but as you say, not that
sensitive with y input direct to the plates. The bandwidth was
revolutionary for it\'s time though. Scope camera on it as well. Yes,
am that old but try to stay awake .

Los Alamos used many 519s with cameras to record certain one-shot
events. They also had a home-made scope that recorded longish
transients on eight stacked scan lines of a pretty big CRT. Recording
one DHART shot was a massive project.

Los Alamos Sales had a bunch of 519s out in the rain in the parking
lot. I didn\'t want to lug one home, so they let me have a CRT for $20
or something.

https://www.dropbox.com/s/r6c3zkwlqrayt53/519_CRT.JPG?raw=1

It\'s single-ended distributed-deflection on the vertical axis. The
screen was tiny.

http://w140.com/tekwiki/wiki/519

I like the tube but the scope itself was really ugly. It used some big
ole transmitting tube in the horizontal sweep.

The early Tek stuff was hand-wired and had selected parts and lots of
tuning. Labor intensive.

I think some of HP kit was the same and can be very difficult to restore
because of the hand selected parts. Can\'t just replace with
a stock 2n part number and get the same results.


There are lots of IC distributed amps around now, mostly to drive
telecom fiber modulators. They are fairly expensive ($200 range) and
have wimpy voltage swings.


Still not trivial to design and build a very high speed, high voltage
amplifier...

We recently released this:

http://www.highlandtechnology.com/DSS/J270DS.shtml

The output stage baby board took me lots of simulation and three
official PCB iterations. Tiny things start to matter below 1 ns. Spice
stops being believable down there.




--

John Larkin Highland Technology, Inc trk

The cork popped merrily, and Lord Peter rose to his feet.
\"Bunter\", he said, \"I give you a toast. The triumph of Instinct over Reason\"

 

[John Miles, KE5FX]

On Thursday, September 10, 2020 at 12:45:00 PM UTC-7, John Larkin wrote:

The 545 also had a differential signal delay line that was made of
tunable t-coils. The tuning procedure was iterative. Must have taken
hours. I think the 585 had two cascaded banks of distributed amps.

Kurt Rosenfeld of TekWiki found a whole box of vintage internal documents
a while back, including factory cal procedures. The original procedures
are probably buried in there somewhere:

http://w140.com/tekwiki/wiki/545
http://w140.com/tekwiki/wiki/585
http://w140.com/tekwiki/wiki/Distributed_amplifier

-- john, KE5FX

 

[Bill Sloman]

On Friday, September 11, 2020 at 1:50:42 AM UTC+10, lang...@fonz.dk wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm..

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.

what are you on about? John saw that an ATC04 was just as fast with a bigger swing

But you can\'t send it anywhere and expect the signal to retain the fast edges without pushing it into a transmission line.

> and LVDS isn\'t some National Semiconductors special it is a standard and the swing is only 450mV or 4.5mA

It\'s a standard now, but it started off as a range of National Semiconductor parts. At the time there were several back-plane standards that didn\'t work well for rack-length back planes and fast signals. I\'d been using ECL for that kind of job, but it meant a -4.5V negative rail, and a -2V rail for the terminating resistors. When LVDS became widely available life got quite a bit easier.

--
Bill Sloman, Sydney

 

[Bill Sloman]

On Friday, September 11, 2020 at 2:07:19 AM UTC+10, John Larkin wrote:

On Thu, 10 Sep 2020 08:50:35 -0700 (PDT), Lasse Langwadt Christensen
lang...@fonz.dk> wrote:

torsdag den 10. september 2020 kl. 09.35.07 UTC+2 skrev Bill Sloman:
On Thursday, September 10, 2020 at 8:08:30 AM UTC+10, John Larkin wrote:
On Wed, 9 Sep 2020 15:29:56 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-09 13:23, John Larkin wrote:
On Wed, 9 Sep 2020 12:24:50 -0400, Phil Hobbs
pcdhSpamM...@electrooptical.net> wrote:

On 2020-09-08 15:38, whit3rd wrote:
On Tuesday, September 8, 2020 at 7:59:34 AM UTC-7, John Larkin wrote:

I was shocked when I tested that ACT04. I was used to CMOS being slow,
and its edges were as fast as 10K ECL. And a lot bigger.

ECL was always designed to drive 50R transmission lines.

A 1.2V swing into 50R is 24mA, and that\'s enough to get the driver warm.

National Semiconductors LVDS uses the same voltage swing for the same reason. Most of us have known about this for some forty years now, but John gets shocked when he runs into it.


what are you on about? John saw that an ATC04 was just as fast with a bigger swing

But didn\'t seem to be aware that terminating that larger swing at the other end of a transmission line was going to be tricky.

and LVDS isn\'t some National Semiconductors special it is a standard and the swing is only 450mV or 4.5mA
And standard ECL swings about 0.8, not 1.2. The voltage across a
pulldown resistor could be anything.

If the resistor is being used to terminate a transmission line there\'s a big temptation to return it to a -2V rail.

There are all sorts of ways of skinning that particular cat

> A few ECL parts, like EL89, swing close to 2 volts. That can be handy.

It\'s a very specialised part designed to drive lines that are driven from a resistance equal the cables\' characteristic impedance, and terminated at the receiving end another such resistance, halving the signal swing at the receiving end

https://www.onsemi.jp/pub/Collateral/MC10EL89-D.PDF

\"The device is especially useful in Digital VideoBroadcasting applications; \" which tend to involve rather long cables and don\'t tolerate much in the way of reflection.

CML usually swings about 0.4 at the termination. I haven\'t tried them
unterminated; I should.

It might be educational. You should have got that bit of education at Tulane, but it was probably covered in one the lectures you skipped.

We\'re playing with the idea building a distributed amplifier with
discrete fets; I have an EE prof doing the math. The arbitrary goal is
to make 50 volt pulses with 50 ps edges. If the output stage works,
the next problem will be to come up with a fast gate-line driver with
sufficient swing. The problem keeps moving left.

Bill Percival invented - and patented - the distributed amplifier in 1936. I met him once when I was working at EMI Central Research.

https://en.wikipedia.org/wiki/Distributed_amplifier

The Tek 545 used a distributed amp made from tubes. We should be able
to do that with fets.

I did it with broad-band transistors in 1983. It sort of worked - we got two 800psec wide complementary pulses swinging through 7V out of a string of three long-tail pairs of cheap 5GHz broad band transistors, but two bigger - and much more expensive - HP rf transistors eventually got us to 500psec with a much simpler circuit.

You can now buy cheap 110GHz bipolar broad-band transistors, which might get you down to 50psec but they won\'t do 50V, unless you stack up a few transmission line transformers. Exotic FETs are a different ball game.

--
Bill Sloman,Sydney