I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

If it does finally work, I'm left feeling nothing like an exacting
engineer, and more like an artist that has piles on layers and layers
of oil paint till the right effect was achieved.

I now find any mention of the 612 mixer and its (non)osclillator gives
me panic attacks. (well i'm currently trying to get it to work at
130Mhz with voltage tuning).

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

bigorangebus wrote:
I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

If it does finally work, I'm left feeling nothing like an exacting
engineer, and more like an artist that has piles on layers and layers
of oil paint till the right effect was achieved.

I now find any mention of the 612 mixer and its (non)osclillator gives
me panic attacks. (well i'm currently trying to get it to work at
130Mhz with voltage tuning).

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

* Model your parasitics, including circuit-layout induced ones.
* Remember that _everything_ has parasitics if the frequency
is high enough.
* Transistor models are never perfect; you need ones intended for
use at your frequencies. RF transistors should have models that
include package effects.
* Tightly laying out your circuit board will reduce inductances
but increase capacitances.
* There are (very spendy) packages that will do parasitic
analysis of your PCB layouts that you can then run through
spice.

It is an art, and one that I haven't mastered myself, but you can put a
lot of science into it if you're careful.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

On Mon, 30 Apr 2007, bigorangebus wrote:

I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I am a real _amateur_ ham operator. I _love_ to make homebrew stuff but
with tubes (not solid state), and will offer MY version of your story.

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

I do tinkering (not _engineering_), and do pretty sloppy, ugly, breadboard
construction of:

1. the real thing
2. only a small part of the whole project at a time
3. turn on "the juice" for the smoke test: results: 1-smoke and no
action (then scratch head), 2-no smoke but action (cigars for all), or
3-no smoke & no action (also scratch head). There are variations on
this. Beyond the scope of this post.

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

Decades ago I built, in a physics research lab on a university campus,
several dual-delay line pulse amplifiers (5-6 tubes, fancy delay lines).
All identical parts, identical schematic, identical tubes. One of these
amplifiers put out a trailing spike that was not supposed to be there. I
tried everything to figure out where it came from. I spent weeks on this.
Finally, I put in a bypass cap that was not supposed to be there and that
fixed it. Told the boss about all this. He shrugged his shoulders.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

I have found commercially made components that were crap. eg. Precision
resistors 20% off value. Parts dead out of the box.

If it does finally work, I'm left feeling nothing like an exacting
engineer, and more like an artist that has piles on layers and layers
of oil paint till the right effect was achieved.

The object of the game is to get stuff that works and forget the castles
in the sky. Oh, yes, I worked at a place where another guy--making very
very very fancy-schmantzy stuff--built a circuit according to "plan" and
it didn't work either. Much cussing, grunting, cussing,
yelling-and-shouting-and-huffing-and-puffing and in the end, I recall that
they pulled a kludge fix, too.

I now find any mention of the 612 mixer and its (non)osclillator gives
me panic attacks. (well i'm currently trying to get it to work at
130Mhz with voltage tuning).

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

I don't know about _you_, but for me, I do better when I build small parts
of the whole thing, then test them individually, and I have minimum test
equipment (oscilloscope, meters, signal generators) and you can put
inputs, measure outputs, measure voltage & currents, and tinker and adjust
as you go.

You sometime have to test the part separately from the circuit to actually
prove to yourself that it really does work. Its a big pain, but it makes
you check everything.

You might ask yourself if you are assembling things with good techniques
(yes--I kid you not--I have actually run into people who did not know
that you had to strip the insulation, whether enamel or plastic, off of
wires before putting them into sockets or clips) including soldering
technique.

Oh, yes, since I work with tubes that run hundreds of volts on the
plates-screens, yes, I've gotten a few electrical shocks (they hurt, and
at higher current, burn flesh) but also be careful about not having a
finger from each hand in contact with "hot" metal conductors. High power
RF also burns flesh and feels like heat rather than pain and you might
smell burnt flesh (its terrible) before you feel anything.

Excellent...!! Professional engineers also spend a lot of the time in
the lab waving their arms in hopes that they can dissipate the smoke
before someone runs in to see where the bang came from!

I dont mean to be too highbrow...its just Ive spent so much of my
working life ensuring that if you build something on a production line
100,000 times 5% failure is pretty bad. And people obviously build
small transistor FM/AM/shortwave radios in china with the worst cheap
components...and most of them seem to work of sorts. Though i have
designed high end audio amplifiers and each time a new batch of
capacitors came in there was lots of "listen tests" as amazingly you
really could here 100nF placed on a supply line (dont ask me how...I
still dont understand it).

If the pro rf engineers do too much of the "try it and see" approach
they deserve their sky high salaries for the loss of hair and frayed
nerves waiting for the customer returns from their first 10,000 units!

I did wonder whether selling the car and buying some impressive test
equipment would make me feel better


On 30 Apr, 20:47, Straydog wrote:
On Mon, 30 Apr 2007, bigorangebus wrote:
I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I am a real _amateur_ ham operator. I _love_ to make homebrew stuff but
with tubes (not solid state), and will offer MY version of your story.

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

I do tinkering (not _engineering_), and do pretty sloppy, ugly, breadboard
construction of:

1. the real thing
2. only a small part of the whole project at a time
3. turn on "the juice" for the smoke test: results: 1-smoke and no
action (then scratch head), 2-no smoke but action (cigars for all), or
3-no smoke & no action (also scratch head). There are variations on
this. Beyond the scope of this post.

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

Decades ago I built, in a physics research lab on a university campus,
several dual-delay line pulse amplifiers (5-6 tubes, fancy delay lines).
All identical parts, identical schematic, identical tubes. One of these
amplifiers put out a trailing spike that was not supposed to be there. I
tried everything to figure out where it came from. I spent weeks on this.
Finally, I put in a bypass cap that was not supposed to be there and that
fixed it. Told the boss about all this. He shrugged his shoulders.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

I have found commercially made components that were crap. eg. Precision
resistors 20% off value. Parts dead out of the box.

If it does finally work, I'm left feeling nothing like an exacting
engineer, and more like an artist that has piles on layers and layers
of oil paint till the right effect was achieved.

The object of the game is to get stuff that works and forget the castles
in the sky. Oh, yes, I worked at a place where another guy--making very
very very fancy-schmantzy stuff--built a circuit according to "plan" and
it didn't work either. Much cussing, grunting, cussing,
yelling-and-shouting-and-huffing-and-puffing and in the end, I recall that
they pulled a kludge fix, too.

I now find any mention of the 612 mixer and its (non)osclillator gives
me panic attacks. (well i'm currently trying to get it to work at
130Mhz with voltage tuning).

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

I don't know about _you_, but for me, I do better when I build small parts
of the whole thing, then test them individually, and I have minimum test
equipment (oscilloscope, meters, signal generators) and you can put
inputs, measure outputs, measure voltage & currents, and tinker and adjust
as you go.

You sometime have to test the part separately from the circuit to actually
prove to yourself that it really does work. Its a big pain, but it makes
you check everything.

You might ask yourself if you are assembling things with good techniques
(yes--I kid you not--I have actually run into people who did not know
that you had to strip the insulation, whether enamel or plastic, off of
wires before putting them into sockets or clips) including soldering
technique.

Oh, yes, since I work with tubes that run hundreds of volts on the
plates-screens, yes, I've gotten a few electrical shocks (they hurt, and
at higher current, burn flesh) but also be careful about not having a
finger from each hand in contact with "hot" metal conductors. High power
RF also burns flesh and feels like heat rather than pain and you might
smell burnt flesh (its terrible) before you feel anything.

On Mon, 30 Apr 2007, bigorangebus wrote:

Excellent...!! Professional engineers also spend a lot of the time in
the lab waving their arms in hopes that they can dissipate the smoke
before someone runs in to see where the bang came from!

I dont mean to be too highbrow...its just Ive spent so much of my
working life ensuring that if you build something on a production line
100,000 times 5% failure is pretty bad. And people obviously build
small transistor FM/AM/shortwave radios in china with the worst cheap
components...and most of them seem to work of sorts. Though i have
designed high end audio amplifiers and each time a new batch of
capacitors came in there was lots of "listen tests" as amazingly you
really could here 100nF placed on a supply line (dont ask me how...I
still dont understand it).

If the pro rf engineers do too much of the "try it and see" approach
they deserve their sky high salaries for the loss of hair and frayed
nerves waiting for the customer returns from their first 10,000 units!

I did wonder whether selling the car and buying some impressive test
equipment would make me feel better

$200 at local hamfests and some cheap gear will get you most of what you
need. I bought a decent audio gen for $15, a decent RF gen for $3, and an
old tektronix scope for $45 (but it crapped out just recently [or, I
should say that it craps out after it warms up]). I do fine.


===== no change to below, included for reference and context =====


On 30 Apr, 20:47, Straydog wrote:
On Mon, 30 Apr 2007, bigorangebus wrote:
I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I am a real _amateur_ ham operator. I _love_ to make homebrew stuff but
with tubes (not solid state), and will offer MY version of your story.

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

I do tinkering (not _engineering_), and do pretty sloppy, ugly, breadboard
construction of:

1. the real thing
2. only a small part of the whole project at a time
3. turn on "the juice" for the smoke test: results: 1-smoke and no
action (then scratch head), 2-no smoke but action (cigars for all), or
3-no smoke & no action (also scratch head). There are variations on
this. Beyond the scope of this post.

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

Decades ago I built, in a physics research lab on a university campus,
several dual-delay line pulse amplifiers (5-6 tubes, fancy delay lines).
All identical parts, identical schematic, identical tubes. One of these
amplifiers put out a trailing spike that was not supposed to be there. I
tried everything to figure out where it came from. I spent weeks on this.
Finally, I put in a bypass cap that was not supposed to be there and that
fixed it. Told the boss about all this. He shrugged his shoulders.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

I have found commercially made components that were crap. eg. Precision
resistors 20% off value. Parts dead out of the box.

If it does finally work, I'm left feeling nothing like an exacting
engineer, and more like an artist that has piles on layers and layers
of oil paint till the right effect was achieved.

The object of the game is to get stuff that works and forget the castles
in the sky. Oh, yes, I worked at a place where another guy--making very
very very fancy-schmantzy stuff--built a circuit according to "plan" and
it didn't work either. Much cussing, grunting, cussing,
yelling-and-shouting-and-huffing-and-puffing and in the end, I recall that
they pulled a kludge fix, too.

I now find any mention of the 612 mixer and its (non)osclillator gives
me panic attacks. (well i'm currently trying to get it to work at
130Mhz with voltage tuning).

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

I don't know about _you_, but for me, I do better when I build small parts
of the whole thing, then test them individually, and I have minimum test
equipment (oscilloscope, meters, signal generators) and you can put
inputs, measure outputs, measure voltage & currents, and tinker and adjust
as you go.

You sometime have to test the part separately from the circuit to actually
prove to yourself that it really does work. Its a big pain, but it makes
you check everything.

You might ask yourself if you are assembling things with good techniques
(yes--I kid you not--I have actually run into people who did not know
that you had to strip the insulation, whether enamel or plastic, off of
wires before putting them into sockets or clips) including soldering
technique.

Oh, yes, since I work with tubes that run hundreds of volts on the
plates-screens, yes, I've gotten a few electrical shocks (they hurt, and
at higher current, burn flesh) but also be careful about not having a
finger from each hand in contact with "hot" metal conductors. High power
RF also burns flesh and feels like heat rather than pain and you might
smell burnt flesh (its terrible) before you feel anything.

We dont seem to have places where you can get cheap stuff in Britain.
I tried to buy a spectrum analyser from ebay last week, good job I
didnt pay it was a con artist. Problem is with kit like spectrum
analysers is that you need something capable like an HP, otherwise its
can be next to useless. And everyone seems to know the value of an
HP856xrange unit, being $4000+!

I did once go to a ham radio show just outside of london, nothing
cheap there though. Though I will remember two enthusiastic bearded
men who were each looking through boxes of junk at opposite ends of a
stand, they had their head in their boxes and were talking to each
other on radios..saying "bill theres some good stuff in this box,
over", "george..yes Ive found some great stuff in here, over"!

Meanwhile...I have a 612 based colpitts oscillator which on test works
at 110Mhz with 10pf base/emmiter and 22pF to ground on the emmiter.
But curiously jumps up to 210Mhz when I double the 22pF cap to 44pF.
Ive got a frequency counter via a FET buffer showing those
frequencies, and (with the counter turned off) I can verify the
frequency with my portable scanner set close by. So increasing the
capacitance doubles the freq of operation!? I do wonder whether the
spectrum has just spread into harmonics...but i cant find them on my
scanner...and wheres that HP spectrum analyzer..hmm

If anyone out there knows why increasing the emitter ground capacitor
in a colpitts will increase the frequency I would be very happy to be
educated! Given that the freq should be the emitter ground cap in
series with the emmiter base, in parallel with the inductance. Which
is 100nH, with a 1/2PI ROOTLC tuned circuit relationship.

Happy happy happy...

On Apr 30, 12:59 pm, bigorangebus wrote:
I'm a professional electronics engineer (not rf enginneer though), and
what I really find frustrating about radio projects is the
uncertainty!

I spend hours designing and analysing filters. Designing oscillators
and minimising harmonics, noise etc. Then I cheerfully commit it to a
PCB (I tightly laid out surface mount PCB to minimise layout
inductances etc).

Nothing ever works as simulated. Transistors never have the rf gain
as SPICE seems to suggest.

And local oscillators...well carefully chosen components just get
thrown out the window to be replaced by more and more random changes
in component values.

Oscillators are very tricky to design using tools like SPICE. Most
simple oscillators by definition build up in amplitude until the
device goes nonlinear and loop gain becomes one. SPICE models outside
the linear range are often very poor.

And a very important factor for oscillators, is that we often require
that the oscillator start reliably at all corners of the gain/
temperature space, but SPICE models usually sit squarely in the center
of gain/temperature space. SPICE models for crystal parameters are
largely nonexistent or very proprietary and again don't cover all
corners of gain/temperature space.

So I wonder, what would be your top tips for someone moving into the
radio design arena? Are there hidden secrets that nobody tells and
the books omit? I'm not thinking about PCB layout here, more things
like, are simulation programs of any use (if so which ones) and what
kind of design procedures can result in predicatable results?

Canned crystal oscillators are a good choice if they really, really
have to start reliably. Many handbook recipes for oscillators are not
necessarily reliable starters.

Simulation for oscillator design was a laughable concept when I first
learned SPICE in the 80's. Today we have more CPU power, but the
models are the weakness.

If you are working in the VHF/UHF range, amplifier design usually
starts with a transistor model for the frequency range. It's rare to
have to get into SPICE.

Tim.

"bigorangebus" wrote in message
oups.com...
We dont seem to have places where you can get cheap stuff in Britain.
I tried to buy a spectrum analyser from ebay last week, good job I
didnt pay it was a con artist. Problem is with kit like spectrum
analysers is that you need something capable like an HP, otherwise its
can be next to useless. And everyone seems to know the value of an
HP856xrange unit, being $4000+!

I did once go to a ham radio show just outside of london, nothing
cheap there though. Though I will remember two enthusiastic bearded
men who were each looking through boxes of junk at opposite ends of a
stand, they had their head in their boxes and were talking to each
other on radios..saying "bill theres some good stuff in this box,
over", "george..yes Ive found some great stuff in here, over"!

Meanwhile...I have a 612 based colpitts oscillator which on test works
at 110Mhz with 10pf base/emmiter and 22pF to ground on the emmiter.
But curiously jumps up to 210Mhz when I double the 22pF cap to 44pF.
Ive got a frequency counter via a FET buffer showing those
frequencies, and (with the counter turned off) I can verify the
frequency with my portable scanner set close by. So increasing the
capacitance doubles the freq of operation!? I do wonder whether the
spectrum has just spread into harmonics...but i cant find them on my
scanner...and wheres that HP spectrum analyzer..hmm

If anyone out there knows why increasing the emitter ground capacitor
in a colpitts will increase the frequency I would be very happy to be
educated! Given that the freq should be the emitter ground cap in
series with the emmiter base, in parallel with the inductance. Which
is 100nH, with a 1/2PI ROOTLC tuned circuit relationship.

Happy happy happy...



10p-22p-100nH should run 200MHz!. The 44pF cap sounds a bit dodgy for HF
osc operation.
Oscillator caps I use for the (later version) SA602 mixer are usually
5pF-5pF.
A 110MHz-210Mhz design earlier this year used 5pF-5pF-100nH and 2 back to
back SMV1259 varicaps.
And yes, above about 30MHz concrete account has to be taken of the layout
wiring and parasitics. Modelling will get you maybe 70% of the way there,
invisible factors the other 30% .
Most important of all is minimising or allowing for the effect of test gear
and it's connections. From bitter experience I realise it's a skilled
artform that can only be acquired through doing.
And yes, the more test gear the better. Come what may, buy that speccy
analyser!.
If you ever come across designs where the author indicates possession of
only basic test gear then be wary. Very wary!.






--
Posted via a free Usenet account from http://www.teranews.com

This is really useful. I'd neglected to mention the other 22pF cap in
parallel with the inductor(in the test circuit, to be replaced by a
varicap with 1nF DC block in actual design), and with circuit
parasitics it runs at about 110Mhz. I actually doubled the 22pF cap
in the emitter/ground (two 22pF caps in parallel). So one 22pF was
110Mhz, and two in parallel was over 200Mhz! So theres the quandry.

My main circuit uses one varicap diode, can't use back to back as I
need the whole cap range. Ive realised that just connecting a 10k pot
(in my test circuit) to a 47 series resistor onto the diode (as seen
on several circuits on the web) is a bit dodgy as the effect is not
linear (presumably because the ends of the 10k pot create a lower rf
impedance).

I was also wondering whether my surface mount 100nH inductor has too
low a Q to be stable (49). Might be causing some of my more random
affects.

"bigorangebus" wrote in message
oups.com...
This is really useful. I'd neglected to mention the other 22pF cap in
parallel with the inductor(in the test circuit, to be replaced by a
varicap with 1nF DC block in actual design), and with circuit
parasitics it runs at about 110Mhz. I actually doubled the 22pF cap
in the emitter/ground (two 22pF caps in parallel). So one 22pF was
110Mhz, and two in parallel was over 200Mhz! So theres the quandry.

My main circuit uses one varicap diode, can't use back to back as I
need the whole cap range. Ive realised that just connecting a 10k pot
(in my test circuit) to a 47 series resistor onto the diode (as seen
on several circuits on the web) is a bit dodgy as the effect is not
linear (presumably because the ends of the 10k pot create a lower rf
impedance).

I was also wondering whether my surface mount 100nH inductor has too
low a Q to be stable (49). Might be causing some of my more random
affects.


At 44pF the emitter feedback factor is only about 0.2 and moving into an
area where a Colpitt may fail to oscillate and all bets are off.
Could be the osc' is running into a on/off squegging mode and the counter is
picking up the 2nd harmonic of a grossly distorted waveform. (a speccy would
show this instantly).
The 47k series R is fine. The diode can see as loading a min of 42k and a
max of 52k, which essentially is a zero load on the dynamic resistance of
the tuned circuit, which may only be a few kohm at best. (a variable R
loading of this nature has only trivial effect on an oscillator)
A Q of 49 is perfectly OK to run the Colpitt. In practice, higher Q's will
offer maybe a 10%-20% increase in maximum attainable oscillation frequency.
The 100nH coil I prototyped was 5 turns of thin tinned wire wound round a
pencil. Doesn't sound much but cost nothing and gave a Q of 180!.

(as an isolated side note, the same coil using Gold plated wire gave Q =180
and the same with Silver plated wire Q=200)




--
Posted via a free Usenet account from http://www.teranews.com