signal indicator

[Cursitor Doom]

On Sun, 05 May 2019 07:10:00 -0700, Joerg wrote:

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.

....and the following week he came back and said, "Could you do it for
50c?"




--
This message may be freely reproduced without limit or charge only via
the Usenet protocol. Reproduction in whole or part through other
protocols, whether for profit or not, is conditional upon a charge of
GBP10.00 per reproduction. Publication in this manner via non-Usenet
protocols constitutes acceptance of this condition.

 

[Joerg]

On 2019-05-04 13:58, John Larkin wrote:

On Sat, 04 May 2019 12:13:26 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 08:31, John Larkin wrote:
On Sat, 04 May 2019 07:03:47 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-03 08:18, John Larkin wrote:
On Fri, 3 May 2019 06:41:33 -0000 (UTC), Jasen Betts
jasen@xnet.co.nz> wrote:

On 2019-05-03, John Larkin <jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

maybe use transistors instead of diodes?

Yes. BFT25 maybe, 5 GHz Ft.


Probably an ordinary BFS17 would also do. RC series in front of the
base, base at near zero DC via another resistive path. The BE junction
rectifies some and then an LED in the collector path is lit. Very little
base current will do because modern LED in the collector path at 1mA can
generate light that is uncomfortably bright. It'll pulse but very fast.

It is not very elegant but less parts.

BFS17 is great precisely because it's fast but not too fast. But
probably a tad slow to detect 2 GHz. Ft is just about 2 GHz.


That's ok, you aren't relying on it being a perfect amplifier. Only to
have the BE junction rectify a little and cause enough CE current to
keep an LED lit. The LED won't see much of the RF. In essence the same
that happens in EMI cases but where this isn't desired, where cell phone
RF >1GHz rectifies itself at the first BE junction of even slow opamps
and causes a demodulated signal in the output.


The dual doubler thing looks good, from a few hundred KHz to over 2
GHz, with 5 pF loading on each side of my diff ECL logic.

https://www.dropbox.com/s/axonmf82uzxhpyi/Sig_Indicator_2.JPG?dl=0


Nice but that's a lot of parts.

With ECL swings, about 800 mV p-p, it works better with fewer diodes:

https://www.dropbox.com/s/xndruojyp819042/Sig_Indicator_4.JPG?dl=0

Then an opamp to drive the LED.

That works. Still feels like cheating not to do it the old-school way.

If I use a Hittite HMC987 fanout chip, I can use one of its diff ECL
outputs just for the monitor LED, and I can load that as hard as I
like.

https://www.dropbox.com/s/avw42lia7ejayug/Sig_Indicator_3.JPG?dl=0

That Hittite chip is $20.


Ouch!

It's wonderful. If it helps me sell a bunch of $1500 boxes, it's a
bargain. The CML comparator ahead of it costs almost as much.

We're looking at some distributed amplifier chips in the $200-400
range, and some fast logic gates for $70 each.

Yes, but to say it like a greedy Wall Street banker, how much higher
could the profit have been?

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

On 2019-05-05 07:14, Cursitor Doom wrote:

On Sun, 05 May 2019 07:10:00 -0700, Joerg wrote:

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.

...and the following week he came back and said, "Could you do it for
50c?"

Oh, we could probably push that another buck or so but there comes a
point where the amortization of the NRE no longer happens.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

On 2019-05-05 08:12, John Larkin wrote:

On Sun, 05 May 2019 07:10:00 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 13:58, John Larkin wrote:
On Sat, 04 May 2019 12:13:26 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 08:31, John Larkin wrote:
On Sat, 04 May 2019 07:03:47 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-03 08:18, John Larkin wrote:
On Fri, 3 May 2019 06:41:33 -0000 (UTC), Jasen Betts
jasen@xnet.co.nz> wrote:

On 2019-05-03, John Larkin <jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

maybe use transistors instead of diodes?

Yes. BFT25 maybe, 5 GHz Ft.


Probably an ordinary BFS17 would also do. RC series in front of the
base, base at near zero DC via another resistive path. The BE junction
rectifies some and then an LED in the collector path is lit. Very little
base current will do because modern LED in the collector path at 1mA can
generate light that is uncomfortably bright. It'll pulse but very fast.

It is not very elegant but less parts.

BFS17 is great precisely because it's fast but not too fast. But
probably a tad slow to detect 2 GHz. Ft is just about 2 GHz.


That's ok, you aren't relying on it being a perfect amplifier. Only to
have the BE junction rectify a little and cause enough CE current to
keep an LED lit. The LED won't see much of the RF. In essence the same
that happens in EMI cases but where this isn't desired, where cell phone
RF >1GHz rectifies itself at the first BE junction of even slow opamps
and causes a demodulated signal in the output.


The dual doubler thing looks good, from a few hundred KHz to over 2
GHz, with 5 pF loading on each side of my diff ECL logic.

https://www.dropbox.com/s/axonmf82uzxhpyi/Sig_Indicator_2.JPG?dl=0


Nice but that's a lot of parts.

With ECL swings, about 800 mV p-p, it works better with fewer diodes:

https://www.dropbox.com/s/xndruojyp819042/Sig_Indicator_4.JPG?dl=0

Then an opamp to drive the LED.


That works. Still feels like cheating not to do it the old-school way.

Solid-state diodes were invented before tubes!

But the opamp, the opamp. When I was young, around 20, I never used
opamps. It was all discretes. Tubes for the big stuff.

If I use a Hittite HMC987 fanout chip, I can use one of its diff ECL
outputs just for the monitor LED, and I can load that as hard as I
like.

https://www.dropbox.com/s/avw42lia7ejayug/Sig_Indicator_3.JPG?dl=0

That Hittite chip is $20.


Ouch!

It's wonderful. If it helps me sell a bunch of $1500 boxes, it's a
bargain. The CML comparator ahead of it costs almost as much.

We're looking at some distributed amplifier chips in the $200-400
range, and some fast logic gates for $70 each.


Yes, but to say it like a greedy Wall Street banker, how much higher
could the profit have been?

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.


We compute "direct cost" of a product: parts cost + unburnened
assembly and test labor cost. Selling price has to be at least 3x,
preferably 6x to the occasional 10x. That works in our business.

Sound like good margins. For us guys it's ok as long as it pays the tab
at Zeitgeist and similar places.

There are businesses that charge 50x.

That is why there is (was?) a Ferrari dealership in the Bay Area.

For low volume stuff, the development and support costs are serious.

True. I don't get involved in too much low volume stuff. In the
beginning it may be but then it usually takes off. So far the topper is
a design that runs off the assembly line unchanged since 1994. At one
point this client asked me to indentify all resistors than could be
changed to 10% and to untrimmed (typically meaning 30%). Cost them 1/4h
of my time and they made that back probably within the month.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Cursitor Doom]

On Sun, 05 May 2019 08:12:36 -0700, John Larkin wrote:

> Solid-state diodes were invented before tubes!

??????????????????????????????????????????????????!!!

Citation, please.



--
This message may be freely reproduced without limit or charge only via
the Usenet protocol. Reproduction in whole or part through other
protocols, whether for profit or not, is conditional upon a charge of
GBP10.00 per reproduction. Publication in this manner via non-Usenet
protocols constitutes acceptance of this condition.

 

[Joerg]

On 2019-05-05 08:15, John Larkin wrote:

On Sun, 05 May 2019 07:24:02 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 19:51, Phil Hobbs wrote:
On 5/4/19 11:27 AM, Joerg wrote:
On 2019-05-04 07:39, Jeroen Belleman wrote:
On 2019-05-04 14:41, Cursitor Doom wrote:
On Fri, 03 May 2019 11:30:58 -0700, John Larkin wrote:

This is RF, and I don't do RF!

Can't say I blame you. It's a bit of a black art which gets blacker the
higher in frequency you go. Best avoided!

I'd hesitate to call a simple diode detector 'RF'. It's
about as simple as it gets. But why is RF a black art?


It isn't. It's fun. Whenever I do RF at a client that is about the
only time where I really regret getting older and retiring.

So don't. Beer you can buy.


However, you can't buy the fitness that a full day of mountain biking
brings. Like I did on Friday. You just can't get that at a gym.

We live in a beautiful place with hills and trails and views, so why
are people paying to sweat on boring treadmills in grim storefront
gyms?

That is what I'll never understand. Though you do need the time to just
take off. I do that on at least one weekday, taking the road bike or the
mountain bike and I am gone all day. In the evening I come back drenched
in sweat no matter what the temperature is, tired and totally happy.

I have
seen enough people who worked into the 70's, never had time for this
sports stuff and then paid a heavy price via a failing body. Assisted
living many years earlier than usual. This doesn't only apply to members
of the "0.1-ton class". I know people who are wiry and thin yet I can
see the effects of not doing any long endurance activities. The folks
that start huffing after a couple of briskly walked miles with hills in
there.

The other thing is volunteer work, in my case lay caregiving for our
church and, privately, nursing home visits. That kind of activity only
works with strict and pretty full schedules which would interfere with
consulting.

Also, bought beer isn't as good as homebrew almost no matter how
expensive. Not even close. So now I only take on some hardcore work,
enticing clients to hire their own engineers of finding another
consultant and then jumping in when they become stuck. Keeps the brain
cells going, creates income and retains a dose of EE fun for me.

Yes. Find some kids to do the grunt work.

Absolutamente

Main objective though is that they learn the ropes over time so that
they can do the whole enchilada. We don't live forever,

--
Regards, Joerg

http://www.analogconsultants.com/

 

[John S]

On 5/5/2019 10:15 AM, John Larkin wrote:

On Sun, 05 May 2019 07:24:02 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 19:51, Phil Hobbs wrote:
On 5/4/19 11:27 AM, Joerg wrote:
On 2019-05-04 07:39, Jeroen Belleman wrote:
On 2019-05-04 14:41, Cursitor Doom wrote:
On Fri, 03 May 2019 11:30:58 -0700, John Larkin wrote:

This is RF, and I don't do RF!

Can't say I blame you. It's a bit of a black art which gets blacker the
higher in frequency you go. Best avoided!

I'd hesitate to call a simple diode detector 'RF'. It's
about as simple as it gets. But why is RF a black art?


It isn't. It's fun. Whenever I do RF at a client that is about the
only time where I really regret getting older and retiring.

So don't. Beer you can buy.


However, you can't buy the fitness that a full day of mountain biking
brings. Like I did on Friday. You just can't get that at a gym.

We live in a beautiful place with hills and trails and views, so why
are people paying to sweat on boring treadmills in grim storefront
gyms?

Indeed. I cannot exercise just for the sake of exercising. I played
racquet ball and hiked many trails in Belize. I got the exercise but
with great entertainment.

 

[John Larkin]

On Sun, 05 May 2019 07:24:02 -0700, Joerg <news@analogconsultants.com>
wrote:

On 2019-05-04 19:51, Phil Hobbs wrote:
On 5/4/19 11:27 AM, Joerg wrote:
On 2019-05-04 07:39, Jeroen Belleman wrote:
On 2019-05-04 14:41, Cursitor Doom wrote:
On Fri, 03 May 2019 11:30:58 -0700, John Larkin wrote:

This is RF, and I don't do RF!

Can't say I blame you. It's a bit of a black art which gets blacker the
higher in frequency you go. Best avoided!

I'd hesitate to call a simple diode detector 'RF'. It's
about as simple as it gets. But why is RF a black art?


It isn't. It's fun. Whenever I do RF at a client that is about the
only time where I really regret getting older and retiring.

So don't. Beer you can buy.


However, you can't buy the fitness that a full day of mountain biking
brings. Like I did on Friday. You just can't get that at a gym.

We live in a beautiful place with hills and trails and views, so why
are people paying to sweat on boring treadmills in grim storefront
gyms?


I have

seen enough people who worked into the 70's, never had time for this
sports stuff and then paid a heavy price via a failing body. Assisted
living many years earlier than usual. This doesn't only apply to members
of the "0.1-ton class". I know people who are wiry and thin yet I can
see the effects of not doing any long endurance activities. The folks
that start huffing after a couple of briskly walked miles with hills in
there.

The other thing is volunteer work, in my case lay caregiving for our
church and, privately, nursing home visits. That kind of activity only
works with strict and pretty full schedules which would interfere with
consulting.

Also, bought beer isn't as good as homebrew almost no matter how
expensive. Not even close. So now I only take on some hardcore work,
enticing clients to hire their own engineers of finding another
consultant and then jumping in when they become stuck. Keeps the brain
cells going, creates income and retains a dose of EE fun for me.

Yes. Find some kids to do the grunt work.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[John Larkin]

On Sun, 05 May 2019 07:10:00 -0700, Joerg <news@analogconsultants.com>
wrote:

On 2019-05-04 13:58, John Larkin wrote:
On Sat, 04 May 2019 12:13:26 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 08:31, John Larkin wrote:
On Sat, 04 May 2019 07:03:47 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-03 08:18, John Larkin wrote:
On Fri, 3 May 2019 06:41:33 -0000 (UTC), Jasen Betts
jasen@xnet.co.nz> wrote:

On 2019-05-03, John Larkin <jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

maybe use transistors instead of diodes?

Yes. BFT25 maybe, 5 GHz Ft.


Probably an ordinary BFS17 would also do. RC series in front of the
base, base at near zero DC via another resistive path. The BE junction
rectifies some and then an LED in the collector path is lit. Very little
base current will do because modern LED in the collector path at 1mA can
generate light that is uncomfortably bright. It'll pulse but very fast.

It is not very elegant but less parts.

BFS17 is great precisely because it's fast but not too fast. But
probably a tad slow to detect 2 GHz. Ft is just about 2 GHz.


That's ok, you aren't relying on it being a perfect amplifier. Only to
have the BE junction rectify a little and cause enough CE current to
keep an LED lit. The LED won't see much of the RF. In essence the same
that happens in EMI cases but where this isn't desired, where cell phone
RF >1GHz rectifies itself at the first BE junction of even slow opamps
and causes a demodulated signal in the output.


The dual doubler thing looks good, from a few hundred KHz to over 2
GHz, with 5 pF loading on each side of my diff ECL logic.

https://www.dropbox.com/s/axonmf82uzxhpyi/Sig_Indicator_2.JPG?dl=0


Nice but that's a lot of parts.

With ECL swings, about 800 mV p-p, it works better with fewer diodes:

https://www.dropbox.com/s/xndruojyp819042/Sig_Indicator_4.JPG?dl=0

Then an opamp to drive the LED.


That works. Still feels like cheating not to do it the old-school way.

Solid-state diodes were invented before tubes!

If I use a Hittite HMC987 fanout chip, I can use one of its diff ECL
outputs just for the monitor LED, and I can load that as hard as I
like.

https://www.dropbox.com/s/avw42lia7ejayug/Sig_Indicator_3.JPG?dl=0

That Hittite chip is $20.


Ouch!

It's wonderful. If it helps me sell a bunch of $1500 boxes, it's a
bargain. The CML comparator ahead of it costs almost as much.

We're looking at some distributed amplifier chips in the $200-400
range, and some fast logic gates for $70 each.


Yes, but to say it like a greedy Wall Street banker, how much higher
could the profit have been?

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.

We compute "direct cost" of a product: parts cost + unburnened
assembly and test labor cost. Selling price has to be at least 3x,
preferably 6x to the occasional 10x. That works in our business.

There are businesses that charge 50x.

For low volume stuff, the development and support costs are serious.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[John Larkin]

On Sun, 5 May 2019 15:34:33 -0000 (UTC), Cursitor Doom
<curd@notformail.com> wrote:

On Sun, 05 May 2019 08:12:36 -0700, John Larkin wrote:

Solid-state diodes were invented before tubes!

??????????????????????????????????????????????????!!!

Citation, please.

https://en.wikipedia.org/wiki/Rectifier#Solid_state





--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[John Larkin]

On Sun, 05 May 2019 08:40:10 -0700, Joerg <news@analogconsultants.com>
wrote:

On 2019-05-05 08:12, John Larkin wrote:
On Sun, 05 May 2019 07:10:00 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 13:58, John Larkin wrote:
On Sat, 04 May 2019 12:13:26 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 08:31, John Larkin wrote:
On Sat, 04 May 2019 07:03:47 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-03 08:18, John Larkin wrote:
On Fri, 3 May 2019 06:41:33 -0000 (UTC), Jasen Betts
jasen@xnet.co.nz> wrote:

On 2019-05-03, John Larkin <jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

maybe use transistors instead of diodes?

Yes. BFT25 maybe, 5 GHz Ft.


Probably an ordinary BFS17 would also do. RC series in front of the
base, base at near zero DC via another resistive path. The BE junction
rectifies some and then an LED in the collector path is lit. Very little
base current will do because modern LED in the collector path at 1mA can
generate light that is uncomfortably bright. It'll pulse but very fast.

It is not very elegant but less parts.

BFS17 is great precisely because it's fast but not too fast. But
probably a tad slow to detect 2 GHz. Ft is just about 2 GHz.


That's ok, you aren't relying on it being a perfect amplifier. Only to
have the BE junction rectify a little and cause enough CE current to
keep an LED lit. The LED won't see much of the RF. In essence the same
that happens in EMI cases but where this isn't desired, where cell phone
RF >1GHz rectifies itself at the first BE junction of even slow opamps
and causes a demodulated signal in the output.


The dual doubler thing looks good, from a few hundred KHz to over 2
GHz, with 5 pF loading on each side of my diff ECL logic.

https://www.dropbox.com/s/axonmf82uzxhpyi/Sig_Indicator_2.JPG?dl=0


Nice but that's a lot of parts.

With ECL swings, about 800 mV p-p, it works better with fewer diodes:

https://www.dropbox.com/s/xndruojyp819042/Sig_Indicator_4.JPG?dl=0

Then an opamp to drive the LED.


That works. Still feels like cheating not to do it the old-school way.

Solid-state diodes were invented before tubes!


But the opamp, the opamp. When I was young, around 20, I never used
opamps. It was all discretes. Tubes for the big stuff.

I made my own little opamps, about a square inch of PCB with wires
sticking out one side, a SIP basically. Started with a pair of
brand-new GE NPN plastic transistors. By selecting one resistor, I
could trim them to 1 uV/degC.

If I use a Hittite HMC987 fanout chip, I can use one of its diff ECL
outputs just for the monitor LED, and I can load that as hard as I
like.

https://www.dropbox.com/s/avw42lia7ejayug/Sig_Indicator_3.JPG?dl=0

That Hittite chip is $20.


Ouch!

It's wonderful. If it helps me sell a bunch of $1500 boxes, it's a
bargain. The CML comparator ahead of it costs almost as much.

We're looking at some distributed amplifier chips in the $200-400
range, and some fast logic gates for $70 each.


Yes, but to say it like a greedy Wall Street banker, how much higher
could the profit have been?

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.


We compute "direct cost" of a product: parts cost + unburnened
assembly and test labor cost. Selling price has to be at least 3x,
preferably 6x to the occasional 10x. That works in our business.


Sound like good margins. For us guys it's ok as long as it pays the tab
at Zeitgeist and similar places.

Alas, Z is a long walk from our new place. But there is a great dive
bar 2 blocks away.

There are businesses that charge 50x.


That is why there is (was?) a Ferrari dealership in the Bay Area.


For low volume stuff, the development and support costs are serious.


True. I don't get involved in too much low volume stuff. In the
beginning it may be but then it usually takes off. So far the topper is
a design that runs off the assembly line unchanged since 1994. At one
point this client asked me to indentify all resistors than could be
changed to 10% and to untrimmed (typically meaning 30%). Cost them 1/4h
of my time and they made that back probably within the month.

Wow, untrimmed resistors. What do they cost in the millions?


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Phil Hobbs]

On 5/5/19 10:10 AM, Joerg wrote:

On 2019-05-04 13:58, John Larkin wrote:
On Sat, 04 May 2019 12:13:26 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-04 08:31, John Larkin wrote:
On Sat, 04 May 2019 07:03:47 -0700, Joerg <news@analogconsultants.com
wrote:

On 2019-05-03 08:18, John Larkin wrote:
On Fri, 3 May 2019 06:41:33 -0000 (UTC), Jasen Betts
jasen@xnet.co.nz> wrote:

On 2019-05-03, John Larkin <jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the
lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd
rather just
use fairly ordinary parts.

Any other ideas?

maybe use transistors instead of diodes?

Yes. BFT25 maybe, 5 GHz Ft.


Probably an ordinary BFS17 would also do. RC series in front of the
base, base at near zero DC via another resistive path. The BE junction
rectifies some and then an LED in the collector path is lit. Very
little
base current will do because modern LED in the collector path at
1mA can
generate light that is uncomfortably bright. It'll pulse but very
fast.

It is not very elegant but less parts.

BFS17 is great precisely because it's fast but not too fast. But
probably a tad slow to detect 2 GHz. Ft is just about 2 GHz.


That's ok, you aren't relying on it being a perfect amplifier. Only to
have the BE junction rectify a little and cause enough CE current to
keep an LED lit. The LED won't see much of the RF. In essence the same
that happens in EMI cases but where this isn't desired, where cell phone
RF >1GHz rectifies itself at the first BE junction of even slow opamps
and causes a demodulated signal in the output.


The dual doubler thing looks good, from a few hundred KHz to over 2
GHz, with 5 pF loading on each side of my diff ECL logic.

https://www.dropbox.com/s/axonmf82uzxhpyi/Sig_Indicator_2.JPG?dl=0


Nice but that's a lot of parts.

With ECL swings, about 800 mV p-p, it works better with fewer diodes:

https://www.dropbox.com/s/xndruojyp819042/Sig_Indicator_4.JPG?dl=0

Then an opamp to drive the LED.


That works. Still feels like cheating not to do it the old-school way.


If I use a Hittite HMC987 fanout chip, I can use one of its diff ECL
outputs just for the monitor LED, and I can load that as hard as I
like.

https://www.dropbox.com/s/avw42lia7ejayug/Sig_Indicator_3.JPG?dl=0

That Hittite chip is $20.


Ouch!

It's wonderful. If it helps me sell a bunch of $1500 boxes, it's a
bargain. The CML comparator ahead of it costs almost as much.

We're looking at some distributed amplifier chips in the $200-400
range, and some fast logic gates for $70 each.


Yes, but to say it like a greedy Wall Street banker, how much higher
could the profit have been?

No kidding, I had projects where stuff retailed well north of $100, the
per unit cost was $10 and the client asked "Could we do it for $5?". So
we did.

We do an ultralow noise digital laser controller with a BOM cost of $37
including the board and connectors, excluding the diode laser and the
optomechanics. It replaces a $3500 purchased laser. It's way more fun
doing stuff for cheap like that.

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

 

[whit3rd]

On Sunday, May 5, 2019 at 8:15:59 AM UTC-7, John Larkin wrote:

We live in a beautiful place with hills and trails and views, so why
are people paying to sweat on boring treadmills in grim storefront
gyms?

Rainy season September-May. "The bluest skies you've ever seen are in Seattle."

 

On Saturday, May 4, 2019 at 1:54:11 PM UTC-4, whit3rd wrote:

On Saturday, May 4, 2019 at 6:22:54 AM UTC-7, bloggs.fre...@gmail.com wrote:
On Thursday, May 2, 2019 at 11:15:28 PM UTC-4, John Larkin wrote:
I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz.
I'd like to light an LED if there is signal present.

You could unbalance the pulldown loading on that buffer by 10-20% and detect that. Differential drive are the original wideband rectifier.

Good plan, but that mainly detects duty cycle unless the load is frequency-dependent.
And, over a thousand-to-one range of frequencies, at that. Am I being
unreasonable in wanting to get an indication of amplitude or frequency from a detector?

An oldschool plan: two termination resistors, one capacitor-coupled. Thermal-couple both resistors
to diodes, and detect any difference in forward drop on the diodes. The only RF components
are resistors and a blocking capacitor to reject DC. A humble LM324 jellybean can handle
the fine details of driving an LED.

Box it up with a couple of penlight cells, and you have a low-Z RF probe.

Okay that's the premium performance solution. He just wants to know if a signal is there, so he's detecting a change is DC quiescent to something else, anything else.

 

[whit3rd]

On Sunday, May 5, 2019 at 8:34:39 AM UTC-7, Cursitor Doom wrote:

On Sun, 05 May 2019 08:12:36 -0700, John Larkin wrote:

Solid-state diodes were invented before tubes!

??????????????????????????????????????????????????!!!

Citation, please.

<https://en.wikipedia.org/wiki/Coherer>

 

On Sat, 4 May 2019 22:56:48 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2019-05-04 08:59, John Larkin wrote:
On Sat, 04 May 2019 08:27:53 -0700, Joerg
news@analogconsultants.com> wrote:

On 2019-05-04 07:39, Jeroen Belleman wrote:
On 2019-05-04 14:41, Cursitor Doom wrote:
On Fri, 03 May 2019 11:30:58 -0700, John Larkin wrote:

This is RF, and I don't do RF!

Can't say I blame you. It's a bit of a black art which gets
blacker the higher in frequency you go. Best avoided!

I'd hesitate to call a simple diode detector 'RF'. It's about as
simple as it gets. But why is RF a black art?


It isn't. It's fun. Whenever I do RF at a client that is about the
only time where I really regret getting older and retiring. There
is so much to do and it seems the next generations aren't supplying
enough engineers for this field.


Logic and even opamps are so fast these days that they work well
into the RF domain.


Yeah, but "real men's RF" is discrete. Just like real cars have a
clutch. Which I was really happy to have on a drive in the
mountains yesterday.

I work in time domain, picoseconds some times, but I don't think
that's RF. RF, to me, is LC-tuned narrowband circuits and antennas.
DC-to-10 GHz is more macho than 9.9 to 10.1 GHz.


Time domain is usually harder and it's real RF. TDR, for example, where
you can get weird kinds of reflections and you have to understand and
diagnosed where they could be coming from. A slight connector or via
mismatch may not even show up in a 10GHz narrowband circuit but in
pulse-echo it can cause a lot of frustration.


My Audi is the first automatic I've ever owned, and I think manuals
are silly now. The 6-speed dual-clutch tranny is faster and smarter
than I am, and it's great on the hills here. I can manual shift it
if I want, and occasionally do for engine braking, like on the crazy
downhill after Donner Summit, where grannies in SUVs (and truck
drivers) are smoking their brakes.


I drove a rental car like that in Germany and it's ok. 7-speed automatic
when the stick was pulled towards one side, hand-shifting when pushed to
the other side, no clutch pedal. I still don't want it on ice, snow or
in mud. Or when you have to rock the car out of a nasty diagonal rut
with shifting rocks in there.

My wife's '19 Mustang convertible has a 10-speed automatic with paddle
shifters on the steering wheel. Right thumb shifts up, left thumb
shifts down. I wouldn't want it in a snowstorm, either. AAMOF, it
just gets in the way anytime. Autos are so good anymore that it
doesn't matter.

When I later rented a stick-shift box truck with clutch and all I was
surprised nobody came out to the truck with me despite my California
license. Maybe they thought if someone can speak German he knows how to
drive those.

Cars and computers should do as they're damn well told. I drive a 2012
Mustang convertible that I bought new in 2017. (It got snagged in the
Takata air bag recall.)

She had a '14 Mustang before this one that had the Takata issue. When
we traded it last year, the passenger side airbag cartridge still
hadn't been replaced.

 

krw@notreal.com wrote in
news:n27vcep1150qhfhkhsthf1c4khgm4kqm0f@4ax.com:

She had a '14 Mustang before this one that had the Takata issue.
When we traded it last year, the passenger side airbag cartridge
still hadn't been replaced.

It figures that you are a retarded fucking ford owner.

 

On Friday, 3 May 2019 10:25:55 UTC+1, tabby wrote:

On Friday, 3 May 2019 04:15:28 UTC+1, John Larkin wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

just add Rs in series with your 10pF pickoffs.

Oh, you can also simplify it if you don't mind it missing a single transition one way only. Signal thru C, R to ground, through R into tr. Add collector led/r. Job done.


NT

You do realise the more complex options offered won't get every first pulse either.


NT

 

On Mon, 6 May 2019 03:02:31 +0000 (UTC),
DecadentLinuxUserNumeroUno@decadence.org wrote:

krw@notreal.com wrote in
news:n27vcep1150qhfhkhsthf1c4khgm4kqm0f@4ax.com:

She had a '14 Mustang before this one that had the Takata issue.
When we traded it last year, the passenger side airbag cartridge
still hadn't been replaced.


It figures that you are a retarded fucking ford owner.

It must be a real bitch to be AlwaysWrong, bitch.

 

[John Larkin]

On Thu, 02 May 2019 20:15:21 -0700, John Larkin
<jjlarkin@highlandtechnology.com> wrote:

I'll have a LVDS signal with data or a square wave, in the range of
maybe 1 MHz to 2 GHz. I figure I can pump it through an 10EP89 ECL
buffer to give it some heft. Each output will become almost 2 volts
p-p.

I'd like to light an LED if there is signal present. How does this
look?

https://www.dropbox.com/s/bvyjs4zgyprx3tu/Sig_Indicator.JPG?dl=0

The 10 pF caps shouldn't load my signals much, especially in the lower
signal path that doesn't need to be super fast.

I suppose I could buy some fancy RF detector chip, but I'd rather just
use fairly ordinary parts.

Any other ideas?

If I have an unused differential PECL output, I could do this:

https://www.dropbox.com/s/khafe0dvqmz0l8x/ECL_Sig_Det_1.JPG?dl=0

Maybe I'll stick to RF-style diode detection. Less tricky.


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com