Cascode substitutions

On 1/31/2016 7:02 PM, John Larkin wrote:
> On Sun, 31 Jan 2016 17:52:24 -0500, mixed nuts
> <melopsitticus@undulatus.budgie> wrote:
[...]
>>
>> Less pain.
>
> LT Spice has the ltline element, a classic Heaviside lossy
> transmission line. It doesn't do higher-order things like skin effect.

If I get accurate loss numbers, I can predict performance.  Sometimes 
you have to tell the sales guy where the limits and tradeoffs are before 
people sign contracts.

>
> At some point, prototype it or get some big-bucks EM modeling
> software.

Prototypes take time and can't be shipped.  The idea is to prototype on 
the computer, generate BOMs and assembly drawings, order parts, kit, 
build, tune, test and ship (~16 weeks release to finish).  Any 
mechanical iterations (longer lines for instance) can add a week or 
more.  If 50% of the orders are for custom or S/N 001 widgets, you don't 
want to add 50 weeks to your production load if you can avoid it.

These things are complicated enough that an EM package can't reasonably 
handle it all in one chunk.  We do model things like couplers and the 
coil module but those get turned into a component model subcircuit and 
go into a library.  As we gain bench experience, library models are 
refined accordingly.

If I can get within 3% on a simulation, it's a near guarantee that the 
production kit will yield a deliverable in a week.

-- 
Grizzly H.

"John Larkin"  wrote in message 
news:s4atabllmsscoihqglh5cumstf9dfcjsli@4ax.com...
>>You can tweak a "narrow" amp into a "wide" one by playing with the 
>>filter
>>topology and impedances.
>
>And gain drops as you do it. Lots of RF amps will drop below unity
>power gain as you try to widen their bandwidth. That's why people are
>using exotica like GaN to make really wideband amps.

Yeah, like I said -- Z and C trade off for GBW.  It's "more or less" 
because not all devices are pure C.  There's often a diffusion 
(distributed RC) aspect present (like a MOS gate), or complexities from 
incidental matching (or mismatching) due to lead/electrode geometry (stray 
L/C, transmission line effects).

>>As you reduce Q (assuming you are able to, by increasing Z and L, and
>>reducing C; normally C has a minimum due to the amplifying device, which
>>trades off with Z as an approximate GBW limitation), the bandwidth 
>>spreads
>>out, and the impedance ratios available via reactive matching networks 
>>get
>>closer to 1.  (...)
>
>For a given device, you generally have a limiting GBW, so wideband
>tweaks eventually kill all the gain. RF mosfets and HF mmics are
>usually internally tuned, to get gain at high frequencies, and there's
>nothing you can do to make them wideband.

Yup.

Of course, internal tuning means two things:
- It's made for convenience
- It's so fast that you can't practically tune it on a PCB anyway, so they 
have to do it for you

I've got to imagine those 60GHz transistors are darn tricky to get to 
behave, unless their properties are such that they're a natural fit for a 
planar circuit and tend to match nicely to, well, even just package and 
PCB pad impedances, really!

Of course, a dash of base/gate resistance goes a long way, as pretty much 
everyone is fond of doing. :)


>Low noise figures are assiciated with tuned circuits, too.
>

Yes, for a couple reasons:
- Tuned circuits are at HF, well away from flicker noise (usually)
- Simply bandwidth reduction: less bandwidth is less noise volts or amps 
(but same nV/rtHz)
- In comparison with wideband circuits that might be unmatched (that is, 
not in a meaningful RF sense), they'll likely be designed much closer to a 
noise or power match, or something around there.

The goal of low noise wideband (in the sense of, freeform, transistor to 
transistor, wide impedance circuits) still includes matching, the algebra 
of matching up noise voltages/currents/resistances, and also of taking 
advantage of correlations if possible.  (Not having much experience with 
low noise circuits, I don't fully understand/internalize all that's going 
on in Phil Hobbs' circuits, but they're excellent examples.)

>>There was one RF part that you found rather lazy, an HBT I think? 
>>That's
>>probably such an example. Relatively high Cout, but fT through the roof?
>
>Lazy? I don't recall that. Phemts and GaN (and, to a lesser extent
>SiC) parts have way higher Gm/c ratios than silicon, so make much
>better wideband amps.


Lazy, in a relative sense.  Something about the differences between 
GaAsFETs, GaNFETs, HBTs and such; some years ago, you measured a bunch and 
found some gave relatively poor pulse outputs despite their bandwidth 
(fT?) specs.

>We do work in time domain. One way to get outrageous gain-bandwidth is
>with distributed amplifiers. They are hard to do discretely, so are
>generally ICs. You can buy a LF-to-40 GHz amp, for example, from
>Hittite or Triquint or whoever they are these days.


Was that one of those parts where you saw unspecified "DC" in the 
datasheet, and it turned out their "DC" was 100MHz? ;-)

Distributed amplifiers are interesting tech, and there's no shortage of 
freely available articles and patents on them.  If it works at 20GHz, 
it'll work at 20MHz too.  Just adjust the impedances for your devices, and 
transmission line delays (or equivalent L/C) for the bandwidth.

I've learned one thing, though: adjusting a 9th+ order, all-pole filter, 
with 16 adjustable controls, is nothing any tech should have to be 
subjected to.  My pity goes out to the first generation Tektronix 
people...

>OK, but RF power amps are usually tuned. That is a lot more practical
>(lower cost, fewer stages, higher efficiency) than going wideband.
>

Yes, but impractical for lab applications, say a continuous sweep from 
150kHz to 80MHz. :)

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com 

Am 01.02.2016 um 01:15 schrieb Phil Hobbs:

>> lol, want talk about universal concepts of "fairness"!
>
> Again, I don't know anybody who advances the view you recite.  Do you?

Sounds like Kant's "behave in a way that it would work if everybody
behaved like you."

I we did that, we all would be quite nice people.
No gods needed.

Gods are only needed if you want a free house with a view to the
Jordan, just for killing the non-believers on this or that side.
Ok, the promise of a paradise is slightly more shabby since a
return of investment right now is considered better by most
people.

The simple minded who had never anything to do with r.o.i are
natural candidates for a combined shabby / short term solution.
It combines the drawbacks of both with one's 15 minutes of fame.
All it takes is a C4 belt and firm persuasion. "Paradise NOW!"

Our local mutation of that god: if you want to be a king
and need a reason, then "by grace of God" is always a good starter.
Heretics will be handed over to the "weltlichen Arm" for stake
processing, which is convenient. (I just can't find an acceptable 
translation.)

Or He could bring you in a position as a pope where you can
challenge every European king for > 1500 years. (Pope Julius in
1510: Now let's check out who has the thicker balls: The king of
France or me.")

Or our nordic gods here: Unlimited beer supply in Walhalla and
flogging our enemies all day long, now that's a word!

And they have the same drawbacks as other gods: not obeying
the standards that they expect from others: denying the right of
hospitality, betraying the giant who built their castle, getting
f*cked and even giving birth to a 6-legged horse, for a warrior!
(Loki) or crossdressing to get Donar's hammer back.
That's what we admire in gods: They may be a*holes, but
they get away with it. And if we are nice, they might do
this & that for us.


eehhk,  what has all of that to do with cascodes?

regards, Gerhard

On Sun, 31 Jan 2016 20:29:03 -0600, "Tim Williams"
<tiwill@seventransistorlabs.com> wrote:

>"John Larkin"  wrote in message 
>news:s4atabllmsscoihqglh5cumstf9dfcjsli@4ax.com...
>>>You can tweak a "narrow" amp into a "wide" one by playing with the 
>>>filter
>>>topology and impedances.
>>
>>And gain drops as you do it. Lots of RF amps will drop below unity
>>power gain as you try to widen their bandwidth. That's why people are
>>using exotica like GaN to make really wideband amps.
>
>Yeah, like I said -- Z and C trade off for GBW.  It's "more or less" 
>because not all devices are pure C.  There's often a diffusion 
>(distributed RC) aspect present (like a MOS gate), or complexities from 
>incidental matching (or mismatching) due to lead/electrode geometry (stray 
>L/C, transmission line effects).
>
>>>As you reduce Q (assuming you are able to, by increasing Z and L, and
>>>reducing C; normally C has a minimum due to the amplifying device, which
>>>trades off with Z as an approximate GBW limitation), the bandwidth 
>>>spreads
>>>out, and the impedance ratios available via reactive matching networks 
>>>get
>>>closer to 1.  (...)
>>
>>For a given device, you generally have a limiting GBW, so wideband
>>tweaks eventually kill all the gain. RF mosfets and HF mmics are
>>usually internally tuned, to get gain at high frequencies, and there's
>>nothing you can do to make them wideband.
>
>Yup.
>
>Of course, internal tuning means two things:
>- It's made for convenience
>- It's so fast that you can't practically tune it on a PCB anyway, so they 
>have to do it for you
>
>I've got to imagine those 60GHz transistors are darn tricky to get to 
>behave, unless their properties are such that they're a natural fit for a 
>planar circuit and tend to match nicely to, well, even just package and 
>PCB pad impedances, really!
>
>Of course, a dash of base/gate resistance goes a long way, as pretty much 
>everyone is fond of doing. :)
>
>
>>Low noise figures are assiciated with tuned circuits, too.
>>
>
>Yes, for a couple reasons:
>- Tuned circuits are at HF, well away from flicker noise (usually)
>- Simply bandwidth reduction: less bandwidth is less noise volts or amps 
>(but same nV/rtHz)
>- In comparison with wideband circuits that might be unmatched (that is, 
>not in a meaningful RF sense), they'll likely be designed much closer to a 
>noise or power match, or something around there.
>
>The goal of low noise wideband (in the sense of, freeform, transistor to 
>transistor, wide impedance circuits) still includes matching, the algebra 
>of matching up noise voltages/currents/resistances, and also of taking 
>advantage of correlations if possible.  (Not having much experience with 
>low noise circuits, I don't fully understand/internalize all that's going 
>on in Phil Hobbs' circuits, but they're excellent examples.)
>
>>>There was one RF part that you found rather lazy, an HBT I think? 
>>>That's
>>>probably such an example. Relatively high Cout, but fT through the roof?
>>
>>Lazy? I don't recall that. Phemts and GaN (and, to a lesser extent
>>SiC) parts have way higher Gm/c ratios than silicon, so make much
>>better wideband amps.
>
>
>Lazy, in a relative sense.  Something about the differences between 
>GaAsFETs, GaNFETs, HBTs and such; some years ago, you measured a bunch and 
>found some gave relatively poor pulse outputs despite their bandwidth 
>(fT?) specs.

I tried some 45 GHz SiGe bipolars as pulse amplifiers and was
disappointed with the speed. Maybe I did something wrong; Joerg seems
to have had better results.

Phemts seem to have no Ft equivalent. They seem to act like an
infinitely fast gadget with parasitic capacitance and lead inductance.
They switch *fast*

The Avago enhancement phemts can be handy.



>
>>We do work in time domain. One way to get outrageous gain-bandwidth is
>>with distributed amplifiers. They are hard to do discretely, so are
>>generally ICs. You can buy a LF-to-40 GHz amp, for example, from
>>Hittite or Triquint or whoever they are these days.
>
>
>Was that one of those parts where you saw unspecified "DC" in the 
>datasheet, and it turned out their "DC" was 100MHz? ;-)

That was a Hittite absorptive analog mux. It's supposed to terminate
unselected input ports, and does above 100 MHz. The data sheet says
"DC TO 4 GHz" in huge type.  

>
>Distributed amplifiers are interesting tech, and there's no shortage of 
>freely available articles and patents on them.  If it works at 20GHz, 
>it'll work at 20MHz too.  Just adjust the impedances for your devices, and 
>transmission line delays (or equivalent L/C) for the bandwidth.
>
>I've learned one thing, though: adjusting a 9th+ order, all-pole filter, 
>with 16 adjustable controls, is nothing any tech should have to be 
>subjected to.  My pity goes out to the first generation Tektronix 
>people...

Look for a Tek 545 schematic!


-- 

John Larkin         Highland Technology, Inc

lunatic fringe electronics 

"John Larkin"  wrote in message 
news:crltablh6nhabaif4650cvpgb6stomvifd@4ax.com...
>>I've learned one thing, though: adjusting a 9th+ order, all-pole filter,
>>with 16 adjustable controls, is nothing any tech should have to be
>>subjected to.  My pity goes out to the first generation Tektronix
>>people...
>
>Look for a Tek 545 schematic!
>

Heck, it had almost nothing; the 517 was king:
http://w140.com/tekwiki/wiki/File:Tek_517_vert_output.jpg

The 585 peaked at 100MHz of mostly-hollow-state bandwidth, but by then 
they introduced distributed deflection CRTs, which did a much nicer job. 
Only needing tens of volts, they used five 6DJ8s (dual triodes, comparable 
to a BF862 with 1/3 the transconductance, lower Ciss, comparable Coss, 10 
times the operating voltage... and who knows how much more noise :) ) to 
drive a somewhat beefy pentode to finally drive the deflection plates.

Industrial art.  A shame they aren't useful anymore, but at least a few 
people appreciate them for what they are.

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com 

On Sun, 31 Jan 2016 22:48:33 -0600, "Tim Williams"
<tiwill@seventransistorlabs.com> wrote:

>"John Larkin"  wrote in message 
>news:crltablh6nhabaif4650cvpgb6stomvifd@4ax.com...
>>>I've learned one thing, though: adjusting a 9th+ order, all-pole filter,
>>>with 16 adjustable controls, is nothing any tech should have to be
>>>subjected to.  My pity goes out to the first generation Tektronix
>>>people...
>>
>>Look for a Tek 545 schematic!
>>
>
>Heck, it had almost nothing; the 517 was king:
>http://w140.com/tekwiki/wiki/File:Tek_517_vert_output.jpg

But the 547 was beautiful, the last of the big ole Tek plugin scopes.
The CRT electron optics was amazing. I have a few.

>
>The 585 peaked at 100MHz of mostly-hollow-state bandwidth, but by then 
>they introduced distributed deflection CRTs, which did a much nicer job. 
>Only needing tens of volts, they used five 6DJ8s (dual triodes, comparable 
>to a BF862 with 1/3 the transconductance, lower Ciss, comparable Coss, 10 
>times the operating voltage... and who knows how much more noise :) ) to 
>drive a somewhat beefy pentode to finally drive the deflection plates.
>
>Industrial art.  A shame they aren't useful anymore, but at least a few 
>people appreciate them for what they are.
>
>Tim

Here's the 547 CRT:

https://dl.dropboxusercontent.com/u/53724080/Tubes/547_crt.JPG

The 519 was a 1 GHz scope. They avoided the bandwidth limits of a
vertical amplifier by not having one.

https://dl.dropboxusercontent.com/u/53724080/Tubes/519_CRT.JPG




-- 

John Larkin         Highland Technology, Inc

lunatic fringe electronics 

On Monday, 1 February 2016 10:00:42 UTC+11, Phil Hobbs  wrote:
> On 01/31/2016 04:43 PM, bitrex wrote:
> > Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> Wrote in
> > message:
> >> On 01/31/2016 03:42 PM, bitrex wrote:
> >>>>
> >>>> You've already said that you're a moral relativist, so there's
> >>>>  no use appealing to your better nature.  ;)
> >>>
> >>> True, but it doesn't seem to stop people from trying...
> >>
> >> A lot of moral relativists are much better than their principles,
> >> and I hope you're one.  The principles themselves are so bad that
> >> it's scarcely possible for someone to be worse.
> >>
> >> I try very hard to deal only with straight shooters.  I've had to
> >> fire customers who weren't, and I certainly wouldn't knowingly work
> >> with someone whose moral ideas were controlled by his own
> >> convenience.
> >
> > The weird thing is, regardless of my moral relativism, I seem to
> > generally behave in a moral fashion. But I couldn't tell you exactly
> >  why. Guess I had those messages drilled into me real good and I
> > can't handle the cognitive dissonance. I catch spiders and gently put
> > them outside.
> 
> Which does make you better than your principles, at least part of the 
> time.  You might want to inquire--as a matter of some urgency--where 
> that higher moral law (for that is what it is) that judges your 
> intentions comes from.

Evolution. Monkeys get upset when they are treated unfairly. Social animals need rules and mechanisms that stop free-loaders from exploiting everybody else.

> > But given some of the horrible experiences I've had in life, I find
> > it very difficult to believe that morality is anything innate or
> > universal. At the end of the day, the Universe doesn't care.

But the survival statistics do. For a lot of situations, you (and your off-spring) will only do well if the tribe cooperates for the good of all.
 
> One of the many problems with that view is the immense agreement between
> people of all eras and all places on what is and is not OK.  There are
> disagreements at the edges, e.g. whether you have to be unselfish
> towards everybody or just your own family/tribe/clan/nation, or whether
> you can have one wife or five.

It's mostly around the edges. Civilisation is all about expanding the group of people we treat as if they were members of our tribe - which should be everybody, these days, except possibly active members of ISIS and Donald Trump supporters.
 
> But I claim that you can't actually imagine a society that sincerely
> admired people who ran away in battle, or who double-crossed all their
> friends, for instance.

It wouldn't last long - though running away in battle does have it's up side, in that "he who fights and runs away, lives to fight another day".

Double-crossing all your friends isn't a way to sustain good social relationships, but if you are very good at moving to new places and making new friends it could work, but not very often, granting that six degrees of separation means that word travel fast.

-- 
Bill Sloman, Sydney

On 2016-01-31 12:34, Tim Williams wrote:
> Designing an RF amplifier.  Concept is, high power, complementary cascodes
> for the output stage (with heavy class A use, but being PP, class AB is an
> option).  50 ohm output, direct drive, say 10W level.
>
> I happen to have a complementary pair that's not too slow (2SC2690A and
> 2SA1220A), and I'd like to maximize the bandwidth around that.  The NPN
> side is fine, I have a 30V, 1A, 2GHz transistor that would pair very well
> with it.  Don't have any such thing for the PNP side.
>
> So, what if I fake a PNP, by wrapping, say, a BFT92 around the NPN?
> http://seventransistorlabs.com/Images/Sziklai_Cascode.png
> That'd be Q1 = BFT92, Q2 = 2SC4821, Q3 = 2SA1220A, and resistors for
> flavor, but probably roughly representative.  (Ground wouldn't actually be
> ground-ground, but probably something like +40V, and "+12V" would be
> +45V.)

This is not the way RF amplifiers are usually built. More
common is to split the input  with a 180-degree hybrid,
amplify each half with an NPN transistor, and then
recombine the outputs with another 180-degree hybrid.
No PNPs needed. Bandwidth can be of the order of three
decades.

Jeroen Belleman

On Monday, 1 February 2016 11:50:23 UTC+11, Gerhard Hoffmann  wrote:
> Am 01.02.2016 um 01:02 schrieb John Larkin:
> 
> >>>>>>> I happen to have a complementary pair that's not too slow (2SC2690A 
> >>>>>>> and 2SA1220A), and I'd like to maximize the bandwidth around that.  
> >>>>>>> The NPN side is fine, I have a 30V, 1A, 2GHz transistor that would 
> >>>>>>> pair very well with it.  Don't have any such thing for the PNP side.
> >>>>>>>
> >>>>>>> So, what if I fake a PNP, by wrapping, say, a BFT92 around the NPN?
> >>>>>>> http://seventransistorlabs.com/Images/Sziklai_Cascode.png
> >>>>>>> That'd be Q1 = BFT92, Q2 = 2SC4821, Q3 = 2SA1220A, and resistors for
> >>>>>>> flavor, but probably roughly representative.  (Ground wouldn't 
> >>>>>>> actually be ground-ground, but probably something like +40V, and 
> >>>>>>> "+12V" would be +45V.)
> >>>>>>>
> >>>>>>> The combination is still fast... ah, but Sziklai connections have a
> >>>>>>> propensity for oscillation all their own, let alone in a cascode, plus
> >>>>>>> whatever other machinations I might have for feedback around the 
> >>>>>>> thing. Game killer?
> 
> I must have lost the beginning of this thread on my server, so...
> 
> The biggest RF PNP that I have seen is the NXP BFG31, SOT223.
> (and I have secured a reel of it :-))

It's been around for a while. It caught my attention around 1990, and I would have liked to design it into our 0.5nsec +7/-7V beam blanker (to drive the RF NPN's that  formed the output) but what we had was fast enough at the time.

http://www.nxp.com/documents/data_sheet/BFG31.pdf
 
> Anything that has a high gain as the product of the gains of two
> transistors is unfriendly to feedback since its combined gain
> falls at 12 dB/octave and at gain = 1, Mr. Bode will get you.

We couldn't get anything to behave sensibly without "base-stoppers" - 20R to 30R of surface mount resistor in series with the base, and close to it.

The right value of ferrite bead would presumably have been better (and we knew it at the time), but back then there weren't all that many ferrite beads around and they weren't all that well characterised.
  
> Maybe one can do the same to the Sziklai?

Don't know. The beam blanker that worked used BFT93 PNP drivers, to drove NPN HP RF power transistors at the output, and the BFG31's would have replaced them, if they'd worked. The whole circuit was about charging and discharging Miller capacitance fast, rather than linear amplification as such, though it looked linear enough and we had a low current mode for moving the electron beam around a bit without actually turning it fully off.

-- 
Bill Sloman, Sydney

On Mon, 1 Feb 2016 04:21:21 +0100, Gerhard Hoffmann
<ghf@hoffmann-hochfrequenz.de> wrote:

>Am 01.02.2016 um 01:15 schrieb Phil Hobbs:
>
>>> lol, want talk about universal concepts of "fairness"!
>>
>> Again, I don't know anybody who advances the view you recite.  Do you?
>
>Sounds like Kant's "behave in a way that it would work if everybody
>behaved like you."

Or the "Prisoner's Dilemma".
>
>I we did that, we all would be quite nice people.
>No gods needed.
>
>Gods are only needed if you want a free house with a view to the
>Jordan, just for killing the non-believers on this or that side.
>Ok, the promise of a paradise is slightly more shabby since a
>return of investment right now is considered better by most
>people.
>
>The simple minded who had never anything to do with r.o.i are
>natural candidates for a combined shabby / short term solution.
>It combines the drawbacks of both with one's 15 minutes of fame.
>All it takes is a C4 belt and firm persuasion. "Paradise NOW!"
>
>Our local mutation of that god: if you want to be a king
>and need a reason, then "by grace of God" is always a good starter.
>Heretics will be handed over to the "weltlichen Arm" for stake
>processing, which is convenient. (I just can't find an acceptable 
>translation.)
>
>Or He could bring you in a position as a pope where you can
>challenge every European king for > 1500 years. (Pope Julius in
>1510: Now let's check out who has the thicker balls: The king of
>France or me.")
>
>Or our nordic gods here: Unlimited beer supply in Walhalla and
>flogging our enemies all day long, now that's a word!
>
>And they have the same drawbacks as other gods: not obeying
>the standards that they expect from others: denying the right of
>hospitality, betraying the giant who built their castle, getting
>f*cked and even giving birth to a 6-legged horse, for a warrior!
>(Loki) or crossdressing to get Donar's hammer back.
>That's what we admire in gods: They may be a*holes, but
>they get away with it. And if we are nice, they might do
>this & that for us.
>
>
>eehhk,  what has all of that to do with cascodes?
>
>regards, Gerhard