Matching a monolithic xtal filter

[RST Engineering \(jw\)]

Excuse me, you wanna rethink that answer before I give you the sophomore
engineering lecture on dB?

Jim

 

[Tony Williams]

That would be just fine, too, 'cept that in my quick look, it
looked like the 4:1 turns ratio from MiniCircuits was much
cheaper than any suitable 5:1 in their catalog.

Thanks Tom. As you know rf is not my bag so I'm an
interested lurker on these threads, trying to follow
the sums..... and where the sums produce something
that may not be practical to implement. That was the
reason for the question.

M/A-com has some similar small transformers and may have
something in a 5:1 turns ratio that would be appropriate.

A 1:4 transformer could be used as a 1+4 auto of course.

I'm not sure how sensitive these particular filters are to load
impedance. With many filters, if you're not trying for the
absolute best conformance to the specified filter shape, a modest
mismatch from the recommended load and source impedances doesn't
really matter that much. It's one of those "YMMV" things--test
to be sure you get what you want.

AFAIR from the data sheet, that filter is about 75KHz
bandwidth, but they also specify a guaranteed attenuation
out at +/- 1MHz from the centre frequency. So presumably
the source/load impedance has to be somewhere near the
req'd 1200//1.8pF over that range.

 

[Fred Bloggs]

Thanks Tom. As you know rf is not my bag so I'm an
interested lurker on these threads, trying to follow
the sums..... and where the sums produce something
that may not be practical to implement. That was the
reason for the question.




A 1:4 transformer could be used as a 1+4 auto of course.




AFAIR from the data sheet, that filter is about 75KHz
bandwidth, but they also specify a guaranteed attenuation
out at +/- 1MHz from the centre frequency. So presumably
the source/load impedance has to be somewhere near the
req'd 1200//1.8pF over that range.

It's like all the other crystal parameters, that impedance derives from
the motional parameters near resonance and in most applications
designers use bandpass impedance matching to avoid exciting spurious
responses and non-linear mixing in the crystal.

 

[Tony Williams]

It's like all the other crystal parameters, that impedance
derives from the motional parameters near resonance and in most
applications designers use bandpass impedance matching to avoid
exciting spurious responses and non-linear mixing in the crystal.

I take that to mean that the 1200//1.8pF need
only be maintained for not much more than the
75KHz passband. Thanks Fred.

 

[Tom Bruhns]

I take that to mean that the 1200//1.8pF need
only be maintained for not much more than the
75KHz passband. Thanks Fred.

A bit more than that, actually, Tony. It's good to not hit a crystal
filter with out-of-band signals that are too high in amplitude,
because crystal filters (a) have spurious responses (should be pretty
low if it's a good design, and well implemented), and (b) are not
strictly linear devices and thus will allow mixing of two out-of-band
signals to produce an in-band signal. Crystal filters are generally
pretty low distortion, but people going for the ultimate in receiver
performance end up paying a lot of attention to their crystal
filters. So anyway, it helps to have some filtering in front of the
filter, to avoid those problems, though there's an obvious limit to
how much you can do. At a 45MHz center frequency, without going
really overboard with the LC filter, you probably will end up with a
3dB bandwidth at least a couple MHz wide. To do much better while
keeping the filter loss low requires coils with high Q, which get
physically large.

On the other hand, in a lot of applications, the crystal filter
distortion and spurious responses are low enough that relatively
broadband coupling is not a problem. You need to look at the whole
system to decide what's appropriate.

Cheers,
Tom

 

[Tom Bruhns]

Excuse me, you wanna rethink that answer before I give you the sophomore
engineering lecture on dB?

Jim

Let's see here...if I didn't miss anything, to begin with, the loss
will be quite a bit more than 6dB. The filter load/source is supposed
to be 1200 ohms in parallel with 1.8pF. I believe Joerg left off the
1.8pF in his suggestion of a series 1150 ohm resistor. Then the
output is to be terminated in net 1200 ohms and 1.8pF. For the
purpose of calculation, assume in the passband that the filter looks
like a simple short circuit. So the 50 ohm source now is delivering
power to 2400 ohms in parallel with 0.9pF. How does the power
delivered to the output 1200 ohms || 1.8pF compare with the power the
50 ohm source could deliver to a 50 ohm load? THAT's how much you are
giving up. And a bonus question: what other problems are caused by
not terminating the 50 ohm source in 50 ohms, and how much MORE loss
is there if you do add a resistor there to yield a net 50 ohm load?
(Hint: double-balanced mixers generally perform much worse with
respect to distortion products if they are not properly terminated...)

Cheers,
Tom

 

[RST Engineering \(jw\)]

That wasn't the point, Tom. You are absolutely correct, the total loss will
be a great deal more than 6 dB; my ORIGINAL point said "...right off the
crack of the bat..." with resistive matching. If I had carried it through
to a logical conclusion it would have been much worse.

No, my point to the person who said that if you lose 6 dB in voltage you
have lost 3 dB in power was the ultimate decibel freshman student blunder.
I'm sure you will agree that if you lose 6 dB, you lose 6 dB measured in
voltage (2:1) or power (4:1).

Jim

 

[Joel Kolstad]

I actually had a homework question that asked you to describe why resistive
matching was generally undesirable for low-level, RF signals. It almost
seemed insulting by then, as this was a senior-level RF class...

Of course I also had a freshman or sophomore calculus question asked you to
painfully describe why, if a cop times you traversing a mile in, say, 40
seconds, he knows you've exceeded the speed limit of 60Mph... aiee....

 

[Joerg]

Let's see here...if I didn't miss anything, to begin with, the loss
will be quite a bit more than 6dB. The filter load/source is supposed
to be 1200 ohms in parallel with 1.8pF. I believe Joerg left off the
1.8pF in his suggestion of a series 1150 ohm resistor. Then the
output is to be terminated in net 1200 ohms and 1.8pF. For the
purpose of calculation, assume in the passband that the filter looks
like a simple short circuit. So the 50 ohm source now is delivering
power to 2400 ohms in parallel with 0.9pF. How does the power
delivered to the output 1200 ohms || 1.8pF compare with the power the
50 ohm source could deliver to a 50 ohm load? THAT's how much you are
giving up. And a bonus question: what other problems are caused by
not terminating the 50 ohm source in 50 ohms, and how much MORE loss
is there if you do add a resistor there to yield a net 50 ohm load?
(Hint: double-balanced mixers generally perform much worse with
respect to distortion products if they are not properly terminated...)

Thomas wrote he had an active mixer in there ;-)

If it's got enough gain like >8dB or so you are usually fine with a
passive scheme plus a follower at the output.

 

[Joerg]

That wasn't the point, Tom. You are absolutely correct, the total loss will
be a great deal more than 6 dB; my ORIGINAL point said "...right off the
crack of the bat..." with resistive matching. If I had carried it through
to a logical conclusion it would have been much worse.

No, my point to the person who said that if you lose 6 dB in voltage you
have lost 3 dB in power was the ultimate decibel freshman student blunder.
I'm sure you will agree that if you lose 6 dB, you lose 6 dB measured in
voltage (2:1) or power (4:1).

The old saga of voltage dBs and power dBs. Or the perpetuum mobile

Once when someone "higher up" brought up power dBs in a seminar back in
college I asked him whether power dBs get sick more often than voltage
dBs. I got kicked out of the room ...

 

[Joerg]

A bit more than that, actually, Tony. It's good to not hit a crystal
filter with out-of-band signals that are too high in amplitude,
because crystal filters (a) have spurious responses (should be pretty
low if it's a good design, and well implemented), and (b) are not
strictly linear devices and thus will allow mixing of two out-of-band
signals to produce an in-band signal. Crystal filters are generally
pretty low distortion, but people going for the ultimate in receiver
performance end up paying a lot of attention to their crystal
filters. So anyway, it helps to have some filtering in front of the
filter, to avoid those problems, though there's an obvious limit to
how much you can do. At a 45MHz center frequency, without going
really overboard with the LC filter, you probably will end up with a
3dB bandwidth at least a couple MHz wide. To do much better while
keeping the filter loss low requires coils with high Q, which get
physically large.

That's where the old concept of the Q-multiplier comes in. After that it
only boils down to how good you are able to control the CF of a resonant
circuit up front. But shhht, don't tell anyone. The younger lads out
there don't have the foggiest idea what that is.

[For the uninitiated: No, it has nothing to do with Q-Tips or luxury
cars of the Infinity brand ...]

On the other hand, in a lot of applications, the crystal filter
distortion and spurious responses are low enough that relatively
broadband coupling is not a problem. You need to look at the whole
system to decide what's appropriate.

Yep.

 

[Fred Bloggs]

A bit more than that, actually, Tony. It's good to not hit a crystal
filter with out-of-band signals that are too high in amplitude,
because crystal filters (a) have spurious responses (should be pretty
low if it's a good design, and well implemented), and (b) are not
strictly linear devices and thus will allow mixing of two out-of-band
signals to produce an in-band signal. Crystal filters are generally
pretty low distortion, but people going for the ultimate in receiver
performance end up paying a lot of attention to their crystal
filters. So anyway, it helps to have some filtering in front of the
filter, to avoid those problems, though there's an obvious limit to
how much you can do. At a 45MHz center frequency, without going
really overboard with the LC filter, you probably will end up with a
3dB bandwidth at least a couple MHz wide. To do much better while
keeping the filter loss low requires coils with high Q, which get
physically large.

On the other hand, in a lot of applications, the crystal filter
distortion and spurious responses are low enough that relatively
broadband coupling is not a problem. You need to look at the whole
system to decide what's appropriate.

Cheers,
Tom

He should at least have a trap for the mixer LO feedthrough, that will
be the biggest undesirable in the mix and catches up with most designs.
Of course if it's tunable, he has problems.

 

[Fred Bloggs]

No, my point to the person who said that if you lose 6 dB in voltage you
have lost 3 dB in power was the ultimate decibel freshman student blunder.

Yeah, well no person said that. I did say that half-power is 3dB. You
have a problem with that?

 

[Tom Bruhns]

Tom Bruhns wrote: ....

That's where the old concept of the Q-multiplier comes in. After that it
only boils down to how good you are able to control the CF of a resonant
circuit up front. But shhht, don't tell anyone. The younger lads out
there don't have the foggiest idea what that is.

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

Cheers,
Tom

 

[Joerg]

Tom Bruhns wrote:
...


That's where the old concept of the Q-multiplier comes in. After that it
only boils down to how good you are able to control the CF of a resonant
circuit up front. But shhht, don't tell anyone. The younger lads out
there don't have the foggiest idea what that is.

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

What made you gun-shy here? Got hurt by them? Nowadays you can create
nice gain controlled amps and in most of my cases this is under full
computer-control. That was way different when I started as a teenage
hobbyist where the price tag of an Apple II would make you cringe. Now
you can buy a uC for a buck.

 

[Tom Bruhns]

On Sep 11, 11:15 am, Joerg <[email protected]>
wrote:

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

What made you gun-shy here? Got hurt by them? Nowadays you can create
nice gain controlled amps and in most of my cases this is under full
computer-control. That was way different when I started as a teenage
hobbyist where the price tag of an Apple II would make you cringe. Now
you can buy a uC for a buck.

Prove to me you can add one and maintain +55dBm IIP3 and I might think
about it. Well, heck, prove to me that you even NEED it in front of a
GOOD crystal filter, too.

Cheers,
Tom

 

[RST Engineering \(jw\)]

Yeah, well no person said that. I did say that half-power is 3dB. You have
a problem with that?

No, I have no problem with that. What I have a problem with is when you
said...

Jim said: > "Except that method loses you 6 dB right off the crack of the
bat in a

place where loss adds directly to noise figure."

Fred said: "That would be 3dB in power and who says it has any effect on
total NF at
this stage of IF processing anyway."

If there is any way of interpreting that NOT to mean that you said a 6 dB
voltage loss is a 3 dB power loss, then I will stand corrected.

Jim

 

[Joerg]

On Sep 11, 11:15 am, Joerg <[email protected]>
wrote:

Tom Bruhns wrote:

how much you can do. At a 45MHz center frequency, without going
really overboard with the LC filter, you probably will end up with a
3dB bandwidth at least a couple MHz wide. To do much better while
keeping the filter loss low requires coils with high Q, which get
physically large.

That's where the old concept of the Q-multiplier comes in. After that it
only boils down to how good you are able to control the CF of a resonant
circuit up front. But shhht, don't tell anyone. The younger lads out
there don't have the foggiest idea what that is.

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

What made you gun-shy here? Got hurt by them? Nowadays you can create
nice gain controlled amps and in most of my cases this is under full
computer-control. That was way different when I started as a teenage
hobbyist where the price tag of an Apple II would make you cringe. Now
you can buy a uC for a buck.

Prove to me you can add one and maintain +55dBm IIP3 and I might think
about it. ...

Should not be a problem. It's all a matter of supply voltage and how far
you have to go in BW reduction ;-)

... Well, heck, prove to me that you even NEED it in front of a
GOOD crystal filter, too.

I've never said that, only responded with a suggestion because someone
wrote it might be needed. It's a habit of my trade, if a client
absolutely wants BW reduction then the client gets the BW reduction.

 

[Simon S Aysdie]

On Sep 11, 11:15 am, Joerg <[email protected]>
wrote:

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

What made you gun-shy here? Got hurt by them? Nowadays you can create
nice gain controlled amps and in most of my cases this is under full
computer-control. That was way different when I started as a teenage
hobbyist where the price tag of an Apple II would make you cringe. Now
you can buy a uC for a buck.

Baaackk iiinnnn theeee oolldddd ddaaaaysss wwhhhhheenn IIII
wwwwwwwaaaaaaaaasssssss yyoouuuunnnggg, I had the bright idea to add
negative resistance to improve the Q of an otherwise passive bandpass
circuit, since the coil was limiting the performance due to its Q. An
experienced engineer told me not to bother trying for commercial grade
equipment that had to operate across a wide temperature range. I never
revisited it.

So you've had luck with this technique at RF? Seems more a ham/garage
thing to me.

 

[Joerg]

On Sep 11, 11:15 am, Joerg <[email protected]>
wrote:

Tom Bruhns wrote:

how much you can do. At a 45MHz center frequency, without going
really overboard with the LC filter, you probably will end up with a
3dB bandwidth at least a couple MHz wide. To do much better while
keeping the filter loss low requires coils with high Q, which get
physically large.

That's where the old concept of the Q-multiplier comes in. After that it
only boils down to how good you are able to control the CF of a resonant
circuit up front. But shhht, don't tell anyone. The younger lads out
there don't have the foggiest idea what that is.

Ouch! Not around my designs, thank you. :-(

[For the uninitiated: just stay away from them.]

What made you gun-shy here? Got hurt by them? Nowadays you can create
nice gain controlled amps and in most of my cases this is under full
computer-control. That was way different when I started as a teenage
hobbyist where the price tag of an Apple II would make you cringe. Now
you can buy a uC for a buck.

Prove to me you can add one and maintain +55dBm IIP3 and I might think
about it. Well, heck, prove to me that you even NEED it in front of a
GOOD crystal filter, too.

Cheers,
Tom

Just as an example which you can still occasionally buy but someone
would almost have to die first because they tend not to part with it,
scroll down to the Drake 2B:

http://home.earthlink.net/~kf6gk/

One of the finest receivers ever made, have used it myself. For
something like $40 extra you could upgrade to a Q-multiplier inside the
speaker cabinet (early 60's pricing, gets me drooling). This puppy has a
dynamic range from here to the Klondike.