Topics

#ubitx LTspice shootout results. #ubitx

Bo Barry <bobarr@...>
 

Getting back in the groove with my years of SPICE, PSPICE ,  etc here's what I found testing the 7 3904s with possible replacements. 
Tested were the following   3904, 5109, 2369, 3866, 2219, 2222, BSX-20.  The A versions were not tested. 

Hats off to the original designer,  the winner is the 3904 ! 

I looked at the gain at 21MHz. I figured it will be years before 10 meters is useful again. 
The BSX-20 came in a close second. 

I've only tested the first three above & have the next two ordered.  I doubt if I'll bother with the last two. 

I used the 3904 for years and years in my college Labs,  thinking it was just a cheap,  lowly,  audio transistor -the only frequencies we used.

I initially thought a simple tweak/mod would boost things up to 10 watts on all bands,  but my excitement was busted when someone said it could easily be done by increasing the supply from 12 to 15  volts.  :) 
Enjoying the cute little rig.  73, Bo W4GHV

Bo Barry <bobarr@...>
 

One update. This SPICE models aren't exact. 
A  transistor beta may be listed as 100-300.  Not the single value in the SPICE model.

Long ago I built a fantastic heathkit MOSFET receiver, used sockets, and moved the ones provided around until the sensitivity was FANTASTIC. .2uV or less I believe.

Might get a handful of 3904s and plug and pray. 😂
Bo

Bo Barry <bobarr@...>
 

One more comment.
I only looked at the output level at 21Mhz with an AC sweep.
Enjoying all the comments on other's REAL experiences.
As a college prof we stayed in the audio range all the time in the labs. 😊

ajparent1/KB1GMX <kb1gmx@...>
 

Bo,

Spice may say one winner but your model is wrong.

The issues
Too low FT, 3904 has ft of 300mhz,  some seemed less so.
HFE that depresses at higher currents.  Spice models are not valid for that if they have it at all.
Not all parts used have uniform HFE
Transformers were likely "perfect" not real.

I don't hate the 3904, I use them a lot but trying to get 16db of gain at even 14mhz
where the beta is only 21 (Gain of about 13db) isn't a successful path.  If the stage
was designed for 10DB it would still be doing well at 30mhz.   Then you need more
stages of lower gain.  Its just decades of design and build you find out things.

The most common issue seen is the drive to the gates of the IRF510s easily hit 400-500mw at 80m
and were down to 100mW at 18mhz.    This is why turning up the voltage helps.

The winner in actual radios was 2n2222A (Not PN2222) and the 2n5109 was better above 20mhz.
Why is 2n2222A(to18) better, it has the same FT?  Its HFE increases with higher currents and it
can actually handle higher currents.  It also exhibits a increasing current bandwidth rather than 
a decreasing one(3904).  So its better, but not best.

Also Q90 in many had very low gain in that stage, replaced with BFR106, even 2n2369 was better.
Some of the bias values were limiting performance and many of the transformers favored lower HF 
and taking a turn off really helped in the midband and higher.

Lets take the Q90 stage its feedback and layout suggests a 18db amplifier, the 3904 will produce
that to about 5mhz and begin showing decreasing gain from that point with increasing frequency.
If we want constant gain to 30mhz we need a transistor with a 1500mhz Ft, that is not a 3904.

There are other things but detailing them is pointless as Pspice lacks facility for including board
level parasitics unless you measure the board (at RF) and include the results as parasitic components.

IF the argument is above 20 is dead...  I don't care.  The problem will still be there.

There a string from May/June of about 300 postings about all this.  The string is forgotten the
problems persist.

Allison

Ralph Mowery
 

Spice and other programs and manual calculations may be exact, but components are not.  Reminds me of when I was in college and doing labs.  The school had boxes that had components on bread boards. We would calculate circuits and then build them.  Some worked and some did not.  If you get a 'lucky box' as we called it the components would mostly match the nominal values used in the calculations.  Some boxes seemed to have values out of tolerance  or stacked  to one side of the limits.  They were not made that way, it just happened. 

We learned to try and design for the worst case with the active components and let the passive components set the parameters for the most part.


On Mon, Aug 20, 2018 at 10:19 AM, Bo Barry <bobarr@...> wrote:
One update. This SPICE models aren't exact. 
A  transistor beta may be listed as 100-300.  Not the single value in the SPICE model.

Long ago I built a fantastic heathkit MOSFET receiver, used sockets, and moved the ones provided around until the sensitivity was FANTASTIC. .2uV or less I believe.

Might get a handful of 3904s and plug and pray. 😂
Bo
_._,_._,_

Arv Evans
 

Advantage of Spice simulators is that you can define, or re-define, the parameters of
a specific component, or components.  Almost everything is programmable.  This lets
us use Spice simulations for generic parts with Monty-Carlo runs to see the effect of
specification variations, and still use it for very specific tests where we take extra care
to define tight specifications for particular components.  Real challenge with Spice
runs is to know what you don't know and thus compensate for not knowing what we
don't know but need to know.  8-)
Spice also lets one define components that do not yet exist so we can try new ideas
without having to design and build these non-existent  parts.  This lets us evaluate uBITX
output filters before they have been invented by Allison or Warren.
Arv
_._


On Mon, Aug 20, 2018 at 8:27 AM Bo Barry <bobarr@...> wrote:
One more comment.
I only looked at the output level at 21Mhz with an AC sweep.
Enjoying all the comments on other's REAL experiences.
As a college prof we stayed in the audio range all the time in the labs. 😊