RF switches

[John Larkin]

Question for the RF guys here.

We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
GHz or better, switching in well under 10 ns.

ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
MHz lower signal bw makes no sense.

HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
obsolete.

Who else makes parts like this?

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

[Rich S]

On Saturday, February 26, 2022 at 12:03:45 AM UTC, John Larkin wrote:
> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.
>
>

> HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
> obsolete.

Well the HMC347B seems to be "in production"
https://www.analog.com/en/products/hmc347b.html#product-overview

and in stock
https://www.digikey.com/en/products/detail/analog-devices-inc./HMC347B/9484664?utm_source=505&utm_medium=supplier&utm_campaign=buynow
https://www.mouser.com/ProductDetail/Analog-Devices/HMC347B?qs=sGAEpiMZZMv0NwlthflBi%252BwqWTQIZySMXYxFWpKgzJk%3D

cheers, RS

[jla...@highlandsniptechnology.com]

347B is a bare die. Mouser seems to have some of the only surviving
packaged part, HMC347ALP. We could scoop up a lifetime supply, for $70
each.

Thanks.



--

I yam what I yam - Popeye

[Clifford Heath]

On 26/2/22 11:03 am, John Larkin wrote:
> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.
>
> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
> MHz lower signal bw makes no sense.
>
> HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
> obsolete.
>
> Who else makes parts like this?

We've used Skyworks SKY13453 (2-way) and SKY13317 (3-way) but they're
only 6GHz. Check their other parts?

And Qorvo of course:
<https://www.qorvo.com/products/switches/discrete-switches>
<https://store.qorvo.com/products/switches/rf-switch?att_4714=1&att_4719=6%2c7>

Clifford Heath

[John Walliker]

On Saturday, 26 February 2022 at 00:03:45 UTC, John Larkin wrote:
> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.
>
> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
> MHz lower signal bw makes no sense.
>

The data sheet does say:
"All of the RF ports (RFC, RF1, and RF2) are dc-coupled to 0 V,
and no dc blocking is required at the RF ports when the RF line
potential is equal to 0 V."
There is a similar device, the ADRF5025 that goes down to 9kHz, but that has
much slower switching.
There must be resistive filtering on the gate drive and a resistive path from the
FET sources to ground to bias the FETs. The time constant must be such that
there is not enough time for the gate charge to leak away during each half-cycle of
the switched signal. If so, the tradeoff between minimum operating frequency
and switching time would make complete sense, as would the power derating
at low frequencies.
The data sheet also mentions that there are no low-frequency spurious signals, so
they are not using a charge pump for FET bias.

John

[jla...@highlandsniptechnology.com]

On Sat, 26 Feb 2022 03:22:53 -0800 (PST), John Walliker
<jrwal...@gmail.com> wrote:

>On Saturday, 26 February 2022 at 00:03:45 UTC, John Larkin wrote:
>> Question for the RF guys here.
>>
>> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
>> GHz or better, switching in well under 10 ns.
>>
>> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
>> MHz lower signal bw makes no sense.
>>
>The data sheet does say:
>"All of the RF ports (RFC, RF1, and RF2) are dc-coupled to 0 V,
>and no dc blocking is required at the RF ports when the RF line
>potential is equal to 0 V."
>There is a similar device, the ADRF5025 that goes down to 9kHz, but that has
>much slower switching.
>There must be resistive filtering on the gate drive and a resistive path from the
>FET sources to ground to bias the FETs. The time constant must be such that
>there is not enough time for the gate charge to leak away during each half-cycle of
>the switched signal. If so, the tradeoff between minimum operating frequency
>and switching time would make complete sense, as would the power derating
>at low frequencies.

Maybe. I suspect the gate drivers are slow to save power.

>The data sheet also mentions that there are no low-frequency spurious signals, so
>they are not using a charge pump for FET bias.
>
>John

The RF people seem to pick some arbitrary low frequency limit on
parts, possibly based on the blocking caps on their eval board.

Some people apparently work with spectrum analyzers that commonly have
a 9 KHz low end, so spec their parts to 9 KHz.

I did once use a Hittite 8-to-1 RF mux that was spec'd "DC-to-12 GHz."
It didn't work right below about 50 MHz. I eventually got the chip
designer on the phone but he wouldn't explain it because it was
proprietary.

The RF world is weird.

[jla...@highlandsniptechnology.com]

Wow, a lot of near-misses.

I'm amplifying an arbitrary waveform as the seed of a largish laser.
And I want to inject a 100 ps fiducial pulse. We can make a nice
programmable 100 ps gaussian pulse, but can't passively mix it with
the arb without wrecking both. So I was thinking that a switch could
select the impulse and then immediately switch over to the arb. But
the customer wants at most 5 ns between the fid and the arb.

Maybe we'll tell them they have to wait longer. The 5 ns requirement
might not be absolute.

I could delay the fiducial some, to start it after we've told the
pokey analog switch to go. Maybe.

[Joe Gwinn]

On Fri, 25 Feb 2022 18:02:56 -0800, jla...@highlandsniptechnology.com
wrote:

Probably why I mostly see PIN diodes used for such things.

Joe Gwinn

[John Larkin]

On Sat, 26 Feb 2022 13:20:54 -0500, Joe Gwinn <joeg...@comcast.net>

PINs are cheap and can switch lots of power. They are terrible for
time domain signals. At really high frequencies, their off capacitance
can be a problem. 0.2 pF is about 50 ohms at 14 GHz.

[John Walliker]

There are two almost identical parts (ADRF5024 and ADRF5025) with different
tradeoffs between switching speed and low-frequency cutoff. That isn't going
to be explained by limitations of the test gear or blocking caps on the eval
board - especially when the eval board doesn't even have any dc blocking caps on it!
The data sheet is quite specific that there is a dc path to ground on all the rf pins.
It is a real tradeoff and there is nothing arbitrary about it.

John

[Chris Jones]

On 27/02/2022 03:16, jla...@highlandsniptechnology.com wrote:
> On Sat, 26 Feb 2022 13:08:15 +1100, Clifford Heath
> <no....@please.net> wrote:
>
>> On 26/2/22 11:03 am, John Larkin wrote:
>>> Question for the RF guys here.
>>>
>>> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
>>> GHz or better, switching in well under 10 ns.
>>>
>>> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
>>> MHz lower signal bw makes no sense.
>>>
>>> HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
>>> obsolete.
>>>
>>> Who else makes parts like this?
>>
>> We've used Skyworks SKY13453 (2-way) and SKY13317 (3-way) but they're
>> only 6GHz. Check their other parts?
>>
>> And Qorvo of course:
>> <https://www.qorvo.com/products/switches/discrete-switches>
>> <https://store.qorvo.com/products/switches/rf-switch?att_4714=1&att_4719=6%2c7>
>>
>> Clifford Heath
>
> Wow, a lot of near-misses.
>
> I'm amplifying an arbitrary waveform as the seed of a largish laser.
> And I want to inject a 100 ps fiducial pulse. We can make a nice
> programmable 100 ps gaussian pulse, but can't passively mix it with
> the arb without wrecking both.

Why? Does each one produce interfering crud when it should produce
nothing, or are you just not able to tolerate the amplitude loss of a
passive resistive combiner?

[jla...@highlandsniptechnology.com]

Amplitide. Neither the arb nor the fiducual generator can make over
about 0.75 volts peak, and the distributed amp needs all of that to
drive the modulator. DA's invert, so adding another $300 chip in the
signal chain creates new tangles.

The many available RF switches are various flavors of weird and all
are poorly spec'd. I've never seen any mention of the capacitance of
the mux control pins, for instance. Or of switching glitches. OIr,
heaven forbid, any actual waveforms. The really fast ones need a pair
of maybe-big poorly-specified negative switch control voltages.

Analog Devices acquired Hittite, which had the fastest RF switches.
"Support" is mostly now forums. Some issues that I care about are
asked about on the forum; some haven't been answered in four years.

TI had the same problem when they acquired Burr-Brown. Nobody knew
much about the parts.

[Gerhard Hoffmann]

Am 27.02.22 um 16:35 schrieb jla...@highlandsniptechnology.com:

> Analog Devices acquired Hittite, which had the fastest RF switches.
> "Support" is mostly now forums. Some issues that I care about are
> asked about on the forum; some haven't been answered in four years.

> TI had the same problem when they acquired Burr-Brown. Nobody knew
> much about the parts.

But this is not only for aquired products. My questions about
the bleeding controls in the ADF5356 synthesizer are unanswered,
too, after 2 years. Some 500 reads.

About as helpful as the Altium forum.

Gerhard

[jla...@highlandsniptechnology.com]

On Sun, 27 Feb 2022 18:15:57 +0100, Gerhard Hoffmann <dk...@arcor.de>
wrote:

We had a lot of hassle using the LMX2571 synth, that TI made necessary
by being coy. Makes no sense.

>
>About as helpful as the Altium forum.
>
>Gerhard

It must be difficult to find and keep good application and support
engineers. They have a life that is, basically, continuous job
interviews. It's a great job for a recent grad.

I guess that the chip designers, if they are even still around, need
to be defended against customer questions, or they wouldn't get work
done.

We have to test a lot of parts ourselves to understand them.

We need a web site where we share insights, measurements, code, bugs.

[Phil Hobbs]

And the 1/f corner of the parts.

>
> Some people apparently work with spectrum analyzers that commonly have
> a 9 KHz low end, so spec their parts to 9 KHz.

HP 8566B, for instance. That isn't a coupling cap, it's just the wings
of the huge peak at DC. The HP 8568B has a coupling cap, which makes it
harder to blow up the first mixer.

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 Jones]

Could you increase the output of the fiducial generator somehow? Then a
combiner with unequal resistors could have low loss for the arb.

If you don't care too much about the shape of the fiducial, there are
probably also some less-broadband-on-one-port combiners that could help
with lower loss for the arb.

If you can't increase the output voltage of the fiducial generator, it
may be sufficient to provide multiple outputs with the same voltage, or
a single, lower-impedance output driving a combiner designed for
different impedances.

Could you yank the arb output high just during the fiducial pulse, with
the emitter of a fast NPN? The impedance would not stay constant, but
that mightn't matter.

[jla...@highlandsniptechnology.com]

On Mon, 28 Feb 2022 11:20:28 +1100, Chris Jones

We're using a Micrel laser driver chip, Sy88022. It makes 25 ps edges
and has beautiful linear amplitude control, but is intended to drive a
low impedance laser and won't swing even one volt.

Inspired by Leo Bodnar's pulse generator, which uses a similar Maxim
chip.

>
>If you don't care too much about the shape of the fiducial, there are
>probably also some less-broadband-on-one-port combiners that could help
>with lower loss for the arb.

The shape really matters. It will be used to test the response of
laser bits downstream. The customer wants a 100 ps gaussian.

>
>If you can't increase the output voltage of the fiducial generator, it
>may be sufficient to provide multiple outputs with the same voltage, or
>a single, lower-impedance output driving a combiner designed for
>different impedances.
>
>Could you yank the arb output high just during the fiducial pulse, with
>the emitter of a fast NPN? The impedance would not stay constant, but
>that mightn't matter.
>
>

It's hard enough to make a 100 ps programmable gaussian pulse. The
analog switch looks like the thing we need.

There are some amazing RF switches around. ADRF5025 is rated for 9 KHz
to 44 GHz. It switches slow so we'd have to deal with that.

The fast switches all look like nightmares. Some need 7 volt (or 40
volt!) gate drives into unspecified capacitances.

[Gerhard Hoffmann]

Am 26.02.22 um 16:07 schrieb jla...@highlandsniptechnology.com:

> The RF people seem to pick some arbitrary low frequency limit on
> parts, possibly based on the blocking caps on their eval board.
>
> Some people apparently work with spectrum analyzers that commonly have
> a 9 KHz low end, so spec their parts to 9 KHz.

The 9 KHz come from some EMC norms. If a lab wants to certify
compliance to these norms, they must measure to 9 KHz.
Everything better might cost one more synthesizer loop aka money
or lose dynamic range because of 1/f, filter bandwidth or
over all synthesizer quality.

>
> I did once use a Hittite 8-to-1 RF mux that was spec'd "DC-to-12 GHz."
> It didn't work right below about 50 MHz. I eventually got the chip
> designer on the phone but he wouldn't explain it because it was
> proprietary.

The 8720A network analyzer goes from 20 GHz down to 50 MHz.
They won't specify anything they cannot measure.
And 50 MHz is just nervous DC.

Many s-parameter files end at either 3 or 6 GHz because
the 8753 A/B/SE/WHATEVER ends at 3 or 6 GHz without/with
the option -6.

> The RF world is weird.

No, just costly.

And PINs will switch even slower.
They start to distort below a few MHz.

Maybe you can build something discrete with CEL 20 GHz GaAsFets.

< https://www.digikey.de/de/products/detail/cel/CE3520K3-C1/6165462 >

Cheers, Gerhard

[jla...@highlandsniptechnology.com]

On Mon, 28 Feb 2022 02:49:53 +0100, Gerhard Hoffmann <dk...@arcor.de>
wrote:

>Am 26.02.22 um 16:07 schrieb jla...@highlandsniptechnology.com:
>
>> The RF people seem to pick some arbitrary low frequency limit on
>> parts, possibly based on the blocking caps on their eval board.
>>
>> Some people apparently work with spectrum analyzers that commonly have
>> a 9 KHz low end, so spec their parts to 9 KHz.
>
>The 9 KHz come from some EMC norms. If a lab wants to certify
>compliance to these norms, they must measure to 9 KHz.
>Everything better might cost one more synthesizer loop aka money
>or lose dynamic range because of 1/f, filter bandwidth or
>over all synthesizer quality.
>
>>
>> I did once use a Hittite 8-to-1 RF mux that was spec'd "DC-to-12 GHz."
>> It didn't work right below about 50 MHz. I eventually got the chip
>> designer on the phone but he wouldn't explain it because it was
>> proprietary.
>
>The 8720A network analyzer goes from 20 GHz down to 50 MHz.
>They won't specify anything they cannot measure.
>And 50 MHz is just nervous DC.

The Hittite mux thing was real. It is supposed to terminate the
unselected inputs with 50 ohms, but there is apparently a cap in
series with each terminator resistor. So much for "DC."

>
>Many s-parameter files end at either 3 or 6 GHz because
>the 8753 A/B/SE/WHATEVER ends at 3 or 6 GHz without/with
>the option -6.
>
>> The RF world is weird.
>
>No, just costly.
>
>And PINs will switch even slower.
>They start to distort below a few MHz.
>
>Maybe you can build something discrete with CEL 20 GHz GaAsFets.
>
>< https://www.digikey.de/de/products/detail/cel/CE3520K3-C1/6165462 >

We use a lot of MiniCircuits enhancement phemts, the SAV series. I
considered making my own RF switch from a couple of them, but it would
be risky. Gate drive would be interesting.

We did make a nice switchable attenuator with a SAV541.

[John Miles, KE5FX]

On Friday, February 25, 2022 at 4:03:45 PM UTC-8, John Larkin wrote:
> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.

PE42525, maybe. 9 kHz-60 GHz, SPDT, 8-12 ns switching time to
10%/90%, 48-60 ns settling time to 0.05 dB.

I'd be surprised if you find anything faster than that. PSemi makes
fantastic stuff (at least, when you can buy it.)

-- john, KE5FX

[Chris Jones]

If the fiducial generator will drive a low impedance, and if the width
of the pulse is only 100ps, I'm sure there is some tricky transmission
line transformer way to impose it in series with your arb output and
maintain decent pulse shape. It would take some experimentation though.

Whatever switch you choose, I wonder what amount of glitch gets injected
into the signal output when it switches (charge injection or whatever
you want to call it).

[jla...@highlandsniptechnology.com]

On Mon, 28 Feb 2022 16:55:18 +1100, Chris Jones

One would think that would be specified on the data sheet of an analog
switch. It isn't.

I guess that RF people don't have oscilloscopes.

[jla...@highlandsniptechnology.com]

Interesting, but bare die. It reminds me of some of the Hittite parts:
no power pins. And no capacitance spec on the control pins.

[mixed nuts]

On 2/25/2022 19:03, John Larkin wrote:
> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.
>
> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
> MHz lower signal bw makes no sense.
>
> HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
> obsolete.
>
> Who else makes parts like this?

Start here:

Warren, W. S. (1987).
Effects of Pulse Shaping in Laser Spectroscopy
Laser Applications to Chemical Dynamics.
doi:10.1117/12.966903

WSW is still active - Duke U.


--
Grizzly H.

[Simon S Aysdie]

On Friday, February 25, 2022 at 4:03:45 PM UTC-8, John Larkin wrote:

> Question for the RF guys here.
>
> We want a wideband RF switch IC that switches fast. Bandwidth lf to 10
> GHz or better, switching in well under 10 ns.
>
> ADRF5024 is 100 MHz to 44 GHz, but switches in about 17 ns. The 100
> MHz lower signal bw makes no sense.
>
> HMC347 looked pretty good, DC-14 GHz and 10 ns typ, but it's gone
> obsolete.
>
> Who else makes parts like this?

https://www.psemi.com/pdf/datasheets/pe42525ds.pdf

and other switches.

I'd check Macom too. We've been designing some of our own lately.

Just a note: there is a difference between amplitude settling and switching speed. I think you do care about switching speed.

[Skittles]

It amazes me that so many people spec RF switches without asking the
power of the signal to be switched. Pretty fundamental.

What is the typical coherence length for light in chemical solutions?


Skittles

[jla...@highlandsniptechnology.com]

I'm working with short, shaped pulses, so I want the switch to have
good pulse fidelity within some fast time of switching.

ADRF5024 looks good. Its gross switching happens in under 10 ns and is
pretty clean. In fine RF tradition, it's underspecified and specs are
basically dishonest.

[Skittles]

On Tue, 08 Mar 2022 07:59:10 -0800, jla...@highlandsniptechnology.com
wrote:

You may want to look the the switch through a TDR (Time domain
reflectomter). That will give you the best measurement for switch's
ability to maintain pulse fidelity.

Skittles

[jla...@highlandsniptechnology.com]

On Tue, 08 Mar 2022 09:04:16 -0700, Skittles<Skit...@nowhere8.org>
wrote:

We've been testing a number of RF switches with pulse generators and
TDR. It has been interesting.

Several of the eval boards have the no-solder screw-on microwave
connectors. The eval boards are free or less if you account for those.

[mixed nuts]

>10^10 light years for dilute homogeneous solutions, <1 um for dense
anisotropic dispersive solutions.


--
Grizzly H.

[John Miles, KE5FX]

On Monday, February 28, 2022 at 2:33:54 AM UTC-8, jla...@highlandsniptechnology.com wrote:
> Interesting, but bare die. It reminds me of some of the Hittite parts:
> no power pins. And no capacitance spec on the control pins.

I see they now have a packaged version:

https://www.psemi.com/products/rf-switches/broadband-rf-switches/pe42546

Only rated for 45 GHz, though, so there's that. No actual data yet other than the
"product brief." Their parts generally do what they say, though.

-- john, KE5FX

[Skittles]

On Tue, 08 Mar 2022 08:58:31 -0800, jla...@highlandsniptechnology.com

What kind of transmission line are you running into the switches?
Sometimes that can really affect your measurements.

For instance if you are running coplanar or microstrip and you have
some components from the line to ground, they should be split equally
to both sides of the line. Say your circuit had 100 ohm to ground, it
should be 2 each 200 ohm to ground on both edges of the line so the
currents split evenly. Make sure the ground current follows the same
path as the signal, each being in different planes.

[jla...@highlandsniptechnology.com]

On Tue, 08 Mar 2022 21:09:03 -0700, Skittles<Skit...@nowhere8.org>

The eval boards usually have CPW

https://www.dropbox.com/s/rbo4mzo640z43t6/ADRF5024-EVALZTOP-web.png?raw=1

which they can do if all you need is one tiny part on a relatively
giant board.

Our boards are most always microstrip, because we need to pack a lot
of parts onto a small board.

https://www.dropbox.com/s/bb16beiwymbu28a/T502A.jpg?raw=1

>
>For instance if you are running coplanar or microstrip and you have
>some components from the line to ground, they should be split equally
>to both sides of the line. Say your circuit had 100 ohm to ground, it
>should be 2 each 200 ohm to ground on both edges of the line so the
>currents split evenly. Make sure the ground current follows the same
>path as the signal, each being in different planes.

In real life, we need to jam parts very close together, and that
dominates layout.

Things that go to ground usually do that to a big layer 1 ground pour.
Vias down to layer 2 have too much personality.

We don't think much about "return currents." Actually, not at all.

[Anthony William Sloman]

On Wednesday, March 9, 2022 at 4:50:13 PM UTC+11, jla...@highlandsniptechnology.com wrote:
> On Tue, 08 Mar 2022 21:09:03 -0700, Skittles<Skit...@nowhere8.org> wrote:
> >On Tue, 08 Mar 2022 08:58:31 -0800, jla...@highlandsniptechnology.com wrote:
> >>On Tue, 08 Mar 2022 09:04:16 -0700, Skittles<Skit...@nowhere8.org> wrote:
> >>>On Tue, 08 Mar 2022 07:59:10 -0800, jla...@highlandsniptechnology.com wrote:
> >>>>On Tue, 1 Mar 2022 20:10:05 -0800 (PST), Simon S Aysdie <gwh...@ti.com> wrote:
> >>>>>On Friday, February 25, 2022 at 4:03:45 PM UTC-8, John Larkin wrote:

<snip>

> >What kind of transmission line are you running into the switches?
> >Sometimes that can really affect your measurements.
> The eval boards usually have CPW

It's hard to build anything but a co-planar waveguide on a printed circuit board.

>
> https://www.dropbox.com/s/rbo4mzo640z43t6/ADRF5024-EVALZTOP-web.png?raw=1
>
> which they can do if all you need is one tiny part on a relatively
> giant board.
>
> Our boards are most always microstrip, because we need to pack a lot
> of parts onto a small board.
>
> https://www.dropbox.com/s/bb16beiwymbu28a/T502A.jpg?raw=1

I have tried to get John Larkin to think about (buried) strip-lines, which you can bury inside multi-layer boards. Unlike structures on the surface, they can be non-dispersive. There's no way to suggest this in away that sounds flattering, so John isn't interested.

> >For instance if you are running coplanar or microstrip and you have
> >some components from the line to ground, they should be split equally
> >to both sides of the line. Say your circuit had 100 ohm to ground, it
> >should be 2 each 200 ohm to ground on both edges of the line so the
> >currents split evenly. Make sure the ground current follows the same
> >path as the signal, each being in different planes.
> In real life, we need to jam parts very close together, and that
> dominates layout.
>
> Things that go to ground usually do that to a big layer 1 ground pour.
> Vias down to layer 2 have too much personality.
>
> We don't think much about "return currents." Actually, not at all.

John Larkin doesn't like doing any thinking very much, and does much less of it than he should.

--
Bill Sloman, Sydney

[DecadentLinux...@decadence.org]

Anthony William Sloman <bill....@ieee.org> wrote in
news:6e6c0cc9-daf7-416d...@googlegroups.com:

> I have tried to get John Larkin to think about (buried)
> strip-lines, which you can bury inside multi-layer boards. Unlike
> structures on the surface, they can be non-dispersive. There's no
> way to suggest this in away that sounds flattering, so John isn't
> interested.

I have done some stripline designs. Before I would move forward with
your idea, which sounds good to me, btw, I would run it past my boss of
over 30 years, and he is one of the top RF engineers on the planet
(was).

[jla...@highlandsniptechnology.com]

Parts mount on the surface. Stripline traces are inherently several
layers down. The connection to a stripline trace involves at least two
vias. Vias are deadly for really fast signals.

Sloman pontificates and insults and hasn't designed actual electronics
in decades, and what he did decades ago sounds mostly like failures.

We do a lot of multilayer test boards and real production boards.
Experiment is a good check on guesswork theories.

And we're not designing RF, we're doing picosecond time domain stuff.
Working around 10 GHz is different from DC-to-10GHz. We measure signal
fidelity in PPM, not dB.

This drives the "slicer" section of a 2-stage e/o modulator:

https://www.dropbox.com/s/34j6pjg31qie2le/T503_FA3.jpg?raw=1

We've improved the monitor pickoff on the current rev. This fast stuff
depends a lot on instinct and experiment, mostly because we don't have
good, or usually any, time-domain part models. It's amazing how much
we don't know about the parts we use.

[John S]

On 3/9/2022 9:10 AM, jla...@highlandsniptechnology.com wrote:
> On Wed, 9 Mar 2022 10:16:16 -0000 (UTC),
> DecadentLinux...@decadence.org wrote:
>
>> Anthony William Sloman <bill....@ieee.org> wrote in
>> news:6e6c0cc9-daf7-416d...@googlegroups.com:
>>
>>> I have tried to get John Larkin to think about (buried)
>>> strip-lines, which you can bury inside multi-layer boards. Unlike
>>> structures on the surface, they can be non-dispersive. There's no
>>> way to suggest this in away that sounds flattering, so John isn't
>>> interested.
>>
>> I have done some stripline designs. Before I would move forward with
>> your idea, which sounds good to me, btw, I would run it past my boss of
>> over 30 years, and he is one of the top RF engineers on the planet
>> (was).
>
> Parts mount on the surface. Stripline traces are inherently several
> layers down. The connection to a stripline trace involves at least two
> vias. Vias are deadly for really fast signals.

In what way are they deadly, John? Would the result of being deadly show
up in a frequency domain test?

[John Larkin]

Time and (wideband) frequency domain tests are nearly interchangeable.
In a fast time domain test, a via typically looks like a capacitor.
Vias can be carefully tuned to match a trace impedance, but a thru via
overshoots the transition to a microstrip and makes a nasty little
stub with a pad on both ends. Blind vias are better but run up board
cost.

We usually keep our fast connections microstrip on layer 1, with
ground plane as layer 2, and run power planes and slow stuff on layers
below.

Ground vias are inductive, so we have big ground pours on layer 1 with
a lot of vias down to the layer 2 ground plane.

Layout becomes a puzzle like doing single-sided boards in ancient
times.

"RF" is usually narrowband, so parasitics can be tuned out.

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

[Anthony William Sloman]

On Thursday, March 10, 2022 at 2:10:27 AM UTC+11, jla...@highlandsniptechnology.com wrote:
> On Wed, 9 Mar 2022 10:16:16 -0000 (UTC), DecadentLinux...@decadence.org wrote:
> >Anthony William Sloman <bill....@ieee.org> wrote in news:6e6c0cc9-daf7-416d...@googlegroups.com:
> >
> >> I have tried to get John Larkin to think about (buried)
> >> strip-lines, which you can bury inside multi-layer boards. Unlike
> >> structures on the surface, they can be non-dispersive. There's no
> >> way to suggest this in away that sounds flattering, so John isn't
> >> interested.
> >
> > I have done some stripline designs. Before I would move forward with
> >your idea, which sounds good to me, btw, I would run it past my boss of
> >over 30 years, and he is one of the top RF engineers on the planet
> >(was).
>
> Parts mount on the surface. Stripline traces are inherently several
> layers down. The connection to a stripline trace involves at least two
> vias. Vias are deadly for really fast signals.

If "designed" - which is to say fudged - buy John Larkin. It's extra inductance which can be neurtalised.

> Sloman pontificates and insults and hasn't designed actual electronics in decades, and what he did decades ago sounds mostly like failures.

Cambridge Instruments fast stuff was mostly 1988 to 1991, and was a technical success and a commercial failure.
I did at bit more at Nijmegen University around 1997. I cleaned up an old nanosecond pulse generator by replacing some of the TTL with ECLinPS, and got rid of a nasty sub-nanosecond jitter, which prompted the user to get us to design an ECLinPS-based replacement. We spent about a year doing the detailed design of the hardware and the software to run it, but the user ran out of funding at the point when we were starting on the layout. Not a successful development, but not a technical failure either.

> We do a lot of multilayer test boards and real production boards.
> Experiment is a good check on guesswork theories.

The theories aren't guesswork, but experiment is a lot cheaper than simulations that are detailed enough to be all that reliable.

> And we're not designing RF, we're doing picosecond time domain stuff.

And so was I.

> Working around 10 GHz is different from DC-to-10GHz. We measure signal fidelity in PPM, not dB.
>
> This drives the "slicer" section of a 2-stage e/o modulator:
>
> https://www.dropbox.com/s/34j6pjg31qie2le/T503_FA3.jpg?raw=1

We were producing 500psec wide pulses from about 1985. Your scope image is of a 4nsec wide pulse.

> We've improved the monitor pickoff on the current rev. This fast stuff
> depends a lot on instinct and experiment, mostly because we don't have
> good, or usually any, time-domain part models. It's amazing how much
> we don't know about the parts we use.

And don't seem to be able to find out. Models are what you put together to fit experimental data. Manufacturers often do it for you.

LTSpice frequently gives access to some manufacturers part models, but when I've used the BFR92 I've had to import the NXP Gummel Poon model myself.

https://www.nxp.com/downloads/en/spice-model/spice_BFR92A.prm

It has shown up in ,asc files I've posted here.

--
Bill Sloman, Sydney