FSK Radio design

WW II IFF used much the same system as the current aircraft transponder design, which is a hyperfast takeoff on the on-off keying similar to Morse Code. Digital Amplitude Modulation.

Jim

It should be possible to make a DC restore circuit that works for all data patterns. It might get confused by a noise glitch inside a sustained long run, but would soon fix itself when any real data came in.

John

Teletype over phone lines or RF was often BFSK, at screaming rates like 75 baud. It's apparently early-1900s technology.

John

I looked it up in Taylor & Westcott, "Principles of Radar" (Cambridge,

1948), and you're quite right. PAM all the way. John's example of teletypes is more to the point--FSK is really ancient.

T&W is one of the "Modern Radio Technique" series published after the War to collect the declassified radio stuff developed during the Later Unpleasantness of 1939-45. I collect them--the best one is "Superregenerative Receivers" by Whitehead. The superregen is really an amazing device in its way--it has a logarithmic response, and enough gain to go from thermal noise to driving headphones with one tube.

And it was invented by Major Edwin H. Armstrong, who also invented the superhet and frequency modulation. (Another one of my technical heroes--there were giants in those days.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics Electro-optics Photonics Analog Electronics

55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Sounds interesting.

Do you have the RadLab books? I eventually managed to collect the whole, if motley, set.

I went to a used technical bookstore just outside the MIT campus, and asked if they had any RadLab books. They'd never heard of them.

The superregen is really an

Yeah. Somebody did a 555 based radio, pretty nasty, and that suggested this, not quite as nasty:

ftp://jjlarkin.lmi.net/Schmitt_Radio.JPG

John

That's a PWM rather than a superregen, though, right? The RF envelope just makes a small change in the average pulse width. (Maybe you could use the other three Schmitt sections to make an RF amplifier.) ;)

I have only one RadLab book in hardcopy, but I do have what I think is the whole set in softcopy. Can't say I've read them, though. I have read all through both of Terman's radio books, which are excellent.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics Electro-optics Photonics Analog Electronics

55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

On Jun 25, 11:00=A0am, Phil Hobbs wrote: .> I never said that you weren't good at your one trick, just that you had

You have yet to discuss the selectivity of the coaxial discriminator approach that Larkin proposed and you have endorsed. Since you have such broad experience how about commenting on the selectivity characteristics of that design?

As far as "enforcing" my approach on others, by what means can I possibly enforce my proposed approach on the OP? I just am warning him to not go down the path of using a coaxial FM discriminator at 978MHz and warning him no to explore a PLL demodulation approach for this design. He can choose to waste his time exploring those two approaches. If you both think he should explore those approaches then you two are the ones that really do not understand receiver design. Why would you seriously propose that he waste his time on such non practical approaches?

d

=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0845-480-2058=A0=A0=A0=A0=A0=A0end_of_the_s= kype_highlighting

t

s

data

tell

O

n though

low

to an

s

rovide

ated

going

cy to

_is_ your

vely

ing

would,

ng

lator

the

perhaps

00KHz, three

able

mean

n

is

y - z

book

p

he

M.

hat.

t

h a

s

error.

of

rror

like

you

. d

i,

n a

in

d

. a

s y

But hysterisis is one possible solution ...

If you've designed it right in the first place,so you are begging the question.

But still scew-up-able, if you do it wrong.

Huh? The distinction I was making between the PD1 and PD2 phase detectors on the 4046 was that the regular one is slow to lock if the phase shifts are outside the the regular +/-90 degree range, while PD2 always moves the VCO in the right direction.

Gardner didn't come into that - I was making a point about tracking the signal oscillator during the 0 to 1 and 1 to 0 transistions. How the hell you have managed to conflate the two issues escapes me. If you really foul up on tracking the signalling oscillator for a number of bit periods, the phase deviation can exceed 90 degrees, and PD2 at least retrieves that situation, but Gardner's third order phase-pocked loop would be an over-kill way of avoiding that.

I was trying to get the OP to think about what he was doing.

You seem to have a fixed idea about what the OP ought to be doing, and seem to be interpreting everything I've posted in terms of the way it fits with your - not explicitly defined - solution.

-- Bill Sloman, Nijmegen

If that upside down ground symbol connected to the center tap of the input LC is supposed to be an antenna, the antenna symbol connects at the apex of the triangle, not the midpoint.

The old Heatkit twoer and sixer (aka "Benton Harbor Lunchboxes" used a superregen on two and six meters. When driving around town, you could hear a superregen a half mile away.

Jim

Actually, the antenna should tie to the hi-Z (top) end of the tank.

s

data

tell

O

n though

low

to an

s

rovide

ated

going

cy to

_is_ your

vely

ing

would,

ng

lator

the

perhaps

00KHz, three

able

mean

n

is

y - z

book

p

he

M.

hat.

t

h a

s

error.

of

rror

like

you

. d

i,

n a

in

d m

. a

s y

Or is low enough that the phase error can build up over a number of bit intervals.

I can't imagine where you got that bizarre idea from. I'm well aware what the 4046 PD1 and PD2 can and can't do and have known about both for a few decades now. Gardner's book doesn't mention the 4046 - or even Jim Thompson's MC4024 and MC4044 that preceded it, although the

4024/4044 combination was around when Gardner first published his book, but Gardner does explain the differences between simple multiplier-type phase detectors like the 4046's PD1 and the more complicated logic-based phase-frequency detectors like the 4046's PD2.

My point about the third order loop was that it could - in principle - track the rapid frequency changes at 0 to 1 and 1 to 0 transitions. This had nothing to do with the vices and virtues of the 4046's PD2, and I find it difficult to understand how you could have manged to conflate the two - very different - ideas.

-- Bill Sloman, Nijmegen

It won't work that way?

The schmitt thing isn't a superregen, in that there's no RF oscillation. It's a comparator with an exponentially decaying, hence adaptive, threshold, which it does have in common with a superregen.

Just a silly idea.

John

Depends on the impedances.

John

e:

s

data

tell

O

though

low

to an

rovide

ated

going

cy to

s_ your

vely

ing

would,

ng

lator

the

perhaps

ee

le

an

n

is

y - z

book

p

he

M.

hat.

t a s

error.

of

rror

ke

you

. d

i,

n a

in

d

. a

s y

=20

"The soultion"? There's only one? And you haven't pointed to it or spelled = it out?

=20

=20

A claim that is implementation dependent, and you haven't explicitly descri= bed your preferred implementation. =20

Since you seem to have adopted John Laarkin's "I'm a genius - buy stuff fro= m me" posting style, I suppose you had to make that claim, even though you = haven't bothered to tell the OP about your "not very hard" solution. =20

Which is where your pretensions to competence come unstuck. I wasn't saying= that the 4046 PD2 was not a phase detector - it's a phase-frequency detect= or, as I've known for a few decades now - Gardner's discussion of the vario= us sorts of "phase" detectors didn't tell me anything I didn't know when I = first read his book.

And the reference to Gardner was the - perfectly correct, if entirely acade= mic - claim that you'd need a third order loop to track the frequecy throug= h ramp at every bit edge transition. You got your knickers in a twist about= that too - presumably in hot pursuit of John Larkin's style of vacuously c= laimed omniscience. It didn't have anything to do with the comment about th= e phase-frequency detector, which might have been a good choice for a phase=

-locked loop with poor phase tracking.

--=20 Bill Sloman, Nijmegen

Are you confusing modulation index or modulating signal bandwidth with deviation?

I still do not remember deviation being stated, only bit rate.

tell

though

an

provide

going

your

would,

perhaps

Umm, no. Hetrodyne does not scale the frequency it shifts them. A 10.7 MHz IF may be troublesome due to bandwidth issues of available IF cans.

Do any of these help?

??*-)

Based only of these numbers, one would assume that the bandwidth is in the order of slightly over 1 MHz (as an order of magnitude), thus the bandwidth to carrier ratio is about 0.1 %.

It was later revealed that the deviation is +/-300 kHz, thus the deviation to carrier frequency ratio is even less than 0.1 %. This is a bit of a challenge for free running oscillators.

Assuming the 1 Mbit/s stream is simply NRZ coded and properly filtered prior to modulation, the highest frequency in the base band signal (...0101010.. sequence) is 500 kHz, thus the modulation index is 0.6. Looking at the Bessel graphs, the carrier is quite strong (0.9), the first sidebands at +/-500 kHz are weaker (about 0.3), the second pair of sidebands at +/1000 kHz are even weaker (0.06 amplitude) and the third pair at +/-1500 kHz are very weak (about 0.02).

Thus the receiver filter bandwidth should be slightly above 1 MHz wide, to let through the first significant side bands. Truncating the second pair of sidebands should not introduce too much ISI.

Maybe, squarish waves are odd harmonic dominant. I recommend keeping the the 3rd harmonic (1.5 MHz) within 3 dB of the strong component of 500 = kHz.

?-)

On Jun 23, 8:35=A0pm, John Larkin > >> John

How does this work without a delay? This sounds like the description of a FM delay line discriminator?