From: Jim Thompson on
On Thu, 10 Jun 2010 09:56:30 -0700, Jim Thompson
<To-Email-Use-The-Envelope-Icon(a)On-My-Web-Site.com> wrote:

>On Thu, 10 Jun 2010 09:36:33 -0700, John Larkin
><jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>
>[snip]
>>>>>>>> John Larkin wrote:
>>>>>>>>
>>>>>>>> [snip]
>>>>>>>>>
>>>>>>>>> The cool thing is that the collector swing is almost exactly 2xVcc
>>>>>>>>> peak-to-peak. As the amplitude builds up, at the negative swing peak
>>>>>>>>> the emitter goes a little bit negative, to get out of the way, and the
>>>>>>>>> collector swings to just about ground. That forward-biases the c-b
>>>>>>>>> junction and discharges the base cap, reducing transistor base current
>>>>>>>>> hence gain. So it has a built-in peak detecting AGC amplitude
>>>>>>>>> leveling loop with close to zero TC. All from 5 parts. Or sometimes
>>>>>>>>> six.
>>>>>>>
>[snip]
>
>Really ?:-)
>
>This post is only to record Larkin's asinine statement, "...reducing
>transistor base current hence gain..." for posterity ;-)
>
> ...Jim Thompson

Oh, yes, I forgot to emphasize that engineering term, "...almost
exactly..." :-)

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

The only thing bipartisan in this country is hypocrisy
From: VWWall on
John Larkin wrote:
> On Thu, 10 Jun 2010 09:09:55 -0500, John Fields
> <jfields(a)austininstruments.com> wrote:
>
>> On Thu, 10 Jun 2010 06:56:15 -0700, John Larkin
>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>>
>>> On Thu, 10 Jun 2010 07:27:47 -0500, John Fields
>>> <jfields(a)austininstruments.com> wrote:
>>>
>>>> On Wed, 09 Jun 2010 21:01:19 -0700, John Larkin
>>>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>>>>
>>>>> On Wed, 09 Jun 2010 16:02:57 -0500, John Fields
>>>>> <jfields(a)austininstruments.com> wrote:
>>>>>
>>>>>> On Wed, 09 Jun 2010 10:30:37 -0700, John Larkin
>>>>>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>>>>>>
>>>>>>> On Wed, 09 Jun 2010 11:31:31 -0500, John Fields
>>>>>>> <jfields(a)austininstruments.com> wrote:
>>>>>>>
>>>>>>>> On Tue, 08 Jun 2010 15:09:51 -0700, John Larkin
>>>>>>>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>>>>>>>>
>>>>>>>>> If I claimed that there was nitrogen in the air, he and
>>>>>>>>> JF would hack a Spice simulation and prove me wrong.
>>>>>>>> ---
>>>>>>>> You're being absurd, as usual, but it seems you lucked out this time
>>>>>>>> and your oscillator works in LTspice.
>>>>>>>>
>>>>>>>>
>>>>>>> Since we manufactured and sold lots of them before Spice was
>>>>>>> available, and they worked just fine, the luck is on Spice's part. Or
>>>>>>> yours.
>>>>>>>
>>>>>>> This will shock the kiddies, but it *is* possible to design circuits
>>>>>>> without using Spice. Usually it's faster and better.
>>>>>> ---
>>>>>> You're preaching to the choir, bucko.
>>>>>>
>>>>>> In your world, maybe, but when you're talking circuits with hundreds
>>>>>> of thousands or millions of transistors, it's not possible.
>>>>>>
>>>>>> This may come as a surprise to you, but many (if not most) of the
>>>>>> circuits which you buy and incorporate into your products were
>>>>>> designed using SPICE, so the fact that you assemble them into working
>>>>>> product that you don't simulate doesn't mean it's free of SPICE.
>>>>> SPICE 1 was introduced at a conference in 1973. It wasn't very good,
>>>>> and SPICE 2, 1975, was better. DRAMS were introduced in 1970.
>>>> ---
>>>> Red herring, cheater, or, at the very least, ignoratio elenchi.
>>>
>>> The first few generations of RAM were designed before Spice existed,
>>> which you declared to be "not possible."
>>>
>>> I don't know how to say that in Latin.
>> ---
>> Pity, as is your lack of Englishy reading comprehension, since the
>> point which was being made was that even though you pooh-pooh SPICE,
>> your livelihood, today, depends on it.
>
> I don't think so. People designed radar, magnetrons and klystrons and
> waveguides and servos and all that, without computers. They did the
> math. Early computers were obviously designed without help from
> computers.
>
I worked on the design of the magnetron for the Nike Zeus radar in 1950.
There were no computers useful for this.

Bell Telephone at that time was considering the use of two inch diameter
waveguide for wide-band telephone use. In the Te0 mode it has very low
losses, when operated at a frequency far above cutoff.

I recall using a Freiden mechanical calculator to find the
characteristics of the forty or so modes that could propagate in the
guide. This particular calculator did square roots, which was a big
help. There was no easy way to do this on the main frame.

APL and then Fortron were the first "engineering languages" I used with
main-frame computers. Waiting for the results limited the usefulness of
computer simulations. The old slide rule was very useful, as were
graphical devices like the "Smith chart".

> I don't use Spice a lot, and could certainly get along without it. It
> is helpful when evaluating nonlinear systems, where math solutions get
> messy.
>
> I think that most board-level designers use, and trust, Spice too
> much. Given the cost and time for IC fab, it makes sense for IC
> designers to use it to check their work before baking silicon. I don't
> think that Widlar used Spice.
>

--
Virg Wall
From: dagmargoodboat on
On Jun 10, 11:56 am, Jim Thompson <To-Email-Use-The-Envelope-I...(a)On-
My-Web-Site.com> wrote:
> On Thu, 10 Jun 2010 09:36:33 -0700, John Larkin
>
> <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote:
>
> [snip]>>>>>>> John Larkin wrote:
>
> >>>>>>> [snip]
>
> >>>>>>>> The cool thing is that the collector swing is almost exactly 2xVcc
> >>>>>>>> peak-to-peak. As the amplitude builds up, at the negative swing peak
> >>>>>>>> the emitter goes a little bit negative, to get out of the way, and the
> >>>>>>>> collector swings to just about ground. That forward-biases the c-b
> >>>>>>>> junction and discharges the base cap, reducing transistor base current
> >>>>>>>> hence gain. So it has a built-in peak detecting AGC amplitude
> >>>>>>>> leveling loop with close to zero TC. All from 5 parts. Or sometimes
> >>>>>>>> six.
>
> [snip]
>
> Really ?:-)
>
> This post is only to record Larkin's asinine statement, "...reducing
> transistor base current hence gain..." for posterity ;-)

Reducing i(b) makes i(e) fall, r(e) rises, stage gain falls. Seems
okay--have I missed something?

James Arthur
From: John Larkin on
On Thu, 10 Jun 2010 09:46:13 -0700, BlindBaby
<BlindMelonChitlin(a)wellnevergetthatonethealbumcover.org> wrote:

>On Wed, 09 Jun 2010 10:30:37 -0700, John Larkin
><jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote:
>
>>This will shock the kiddies, but it *is* possible to design circuits
>>without using Spice. Usually it's faster and better.
>
>
> It is usually faster, but these days, not usually better. And the
>speed gain is only about the operator of the sim package, not the sim.
>A good sim app user can beat you, hands down, and have reliable numbers
>to compare with real builds as well.
>
> Sim apps have gotten orders of magnitude tighter in their iterative
>analysis and inclusion of parasitics, etc.
>
> Far better than you, with or without your bench.

Simulation is like breadboarding. Neither is designing.

John

From: John Larkin on
On Wed, 9 Jun 2010 21:23:04 -0700 (PDT), dagmargoodboat(a)yahoo.com
wrote:

>On Jun 9, 9:18�pm, Winfield Hill <Winfield_mem...(a)newsguy.com> wrote:
>> Jim Thompson wrote...
>>
>> >>> I know John won't respond, but could someone, perhaps Win, tell me
>> >>> how the "AGC" works?
>>
>> >> I'm sure he would, but why should I, over the years you've insulted
>> >> me at least as much as him, and perhaps more aggressively?
>>
>> >> Anyway, he did explain it, SFAICT.
>>
>> >> Note the BJT is over-biased - plenty of base current, that if left
>> >> unchecked would charge the base-to-ground capacitor and over-current
>> >> the transistor. �So the oscillator runs and examining cycle-by-cycle,
>> >> the collector swings higher and higher until it goes negative with
>> >> respect to the base voltage, close to saturating the transistor,
>> >> and turning on the base-collector diode a bit, robbing current from
>> >> the base capacitor. �This process servos the BJT current to just the
>> >> right level to sustain an oscillation collector-voltage level where
>> >> just the right amount of current is robbed each cycle to control the
>> >> base voltage. �Thereby insuring that the collector goes close to the
>> >> emitter on each cycle, establishing a tightly-controlled amplitude,
>> >> which as John pointed out, is temperature independent to first order
>> >> since Vce(sat) is relatively temperature independent.
>>
>> >> John said Vcc peak, but actually it must be closer to Vcc - Vce(sat).
>>
>> > I'll buy that the collector forwards biases, and you enter a limit
>> > cycle. �Thus I'd call it ALC. �I don't see any _gain_ variation that
>> > "AGC" would imply.
>>
>> �Picky, picky. �To my mind, the base current robbed by the
>> �collector starves the base, lowering the CE stage's gain,
>> �until the exact equilibrium is achieved. �ALC, AGC, pick
>> �your name as you like. �Either way it gets the job done
>> �rather nicely, and is a bit different from what we've seen
>> �elsewhere, such as in old radio circuits. �I see that it
>> �has been analyzed as a possible RF oscillator technique.
>> �But it seems to me that, working as we imagine, Vce(sat)
>> �and all, this trick would be limited to far far below fT.
>
>Just to clarify, the RF versions I posted are similar to, but not the
>same as John's. They're standard UHF designs, Class A, without John's
>precision AGC. I don't think they can use John's AGC method directly--
>if saturated, the transistors would be too slow--but maybe a Baker-ish
>clamp thing would do the job.

In my oscillator, a c-b schottky diode would keep the transistor c-b
junction from conducting, and keep the transistor out of saturation.
Tempco would still be low. That simplifies things considerably. Not
bad.


>Oh, and John's oscillator really swings ~ 2* (Vcc + Vbe), not 2* (Vcc
>- Vbe). Reason being, the AGC operates as the average base voltage
>gets sucked down to near 0v, killing the gain.

I seem to recall the DC base voltage being about +.6. So the collector
swings to just about zero, and the AC output is 2*Vcc p-p. Somebody
could Spice this, if they were interested, and see exactly what
happens. The transformer ratio gets involved some, too.

John