**EMRFD Message Archive 11006**

MessageDateFromSubject11006 2015-04-27 04:51:48 k1rf_digital_stev... Question on hybrid connected crystal filters Hi all. I'm considering building a hybrid connected roofing filter at 45 MHz for a high performance up-conversion receiver. I have not found any really good description of why it is so much better than simply using a diplexer before a single filter. The EMRFD book section 3.38 says that the two filters must both be identical. How identical? What error(s) in shape or center frequency can be tolerated? Besides reducing the power that each filter sees by 3dB, I think an advantage might be more constant impedance close-in, especially around the crystal filter edges while the diplexer provides constant impedance further out so that the first mixer sees a constant impedance from zero to light. Not sure if it is worth going through the effort of a hybrid connected roofing filter. Any insights appreciated.

Thanks,

"Digital Steve", K1RF11007 2015-04-27 09:29:27 billw77aaz Re: Question on hybrid connected crystal filters The diplexer takes care of impedance far, far from the filter passband.

IE, at the DC end and at the VHF end of the spectrum. But the diplexer Q

is 1, so the impedance of the filter, both its passband (good match) and

the stopband (NOT good match), are seen by the mixer.

You have the concept: the driving point impedance of two diplexed filters

looks like 50 ohms both in the passband and the stopband. When the filters

are identical then the passband of the combined pair is unchanged and the

impedance is 50+j0 "everywhere". That, and the 3 dB reduction in power

seen by the filters is pretty powerful, especially if you're using a diode

mixer.

When making filters you can make well matched pairs. I've done that at

4.434, 5.200 and 9 MHz and the result was very, very close to 50+j0

everywhere. I have two surplus 45 MHz filters, 5 KHz wide, designed for 50

ohm impedance. They are NOT precisely identical and sealed so untweekable.

I combined them with hybrid networks anyway, and as I recall the passband

of the diplexed filters is the COMMON passband of both, so the passband of

the combination was slightly narrower than 5 KHz. And there is some small

deviation from 50 ohms in the region where they differed, but it was not

the extreme impedance of the individual filters.

Someone else may have a very precise answer of EXACTLY what happens where

they are mismatched.

I found it useful to make the two hybrid couplers and instead of filters.

I connect them with two identical resistors (not 50 ohms), terminate the

output end and adjust the capacitors and turns on the inductors to get a

precise 50 ohm input impedance. Then swap input and output and again

adjust for precise 50 ohm impedance (just the capacitors this second

time). After the networks looked like good 50 ohm impedances from both

ends, then I connect the filters.

Bill W7AAZ

> Hi all. I'm considering building a hybrid connected roofing filter at 45

> MHz for a high performance up-conversion receiver. I have not found any

> really good description of why it is so much better than simply using a

> diplexer before a single filter. The EMRFD book section 3.38 says that

> the two filters must both be identical. How identical? What error(s) in

> shape or center frequency can be tolerated? Besides reducing the power

> that each filter sees by 3dB, I think an advantage might be more constant

> impedance close-in, especially around the crystal filter edges while the

> diplexer provides constant impedance further out so that the first mixer

> sees a constant impedance from zero to light. Not sure if it is worth

> going through the effort of a hybrid connected roofing filter. Any

> insights appreciated.

>

> Thanks,

> "Digital Steve", K1RF11008 2015-04-27 10:35:19 k1rf_digital_stev... Re: Question on hybrid connected crystal filters Thanks Bill. That was very helpful. I gathered several groups of crystals from various places, including ECS fundamental mode crystals, part number ECS-400-20-3X-TR. But then I scored ten Vectron 45 MHz 7.5 KHz monolithic filters, P/N MQF45.0 0750/21 at a very good price. See http://www.vectron.com/products/saw/pdf_mqf/MQF45.0-0750_21.pdf

These are fairly good, as they are spec'd both above and below the in-band area and have reasonable insertion loss. Many filters of this ilk are only spec'd on one side of the in-band area, not both sides. I need to make a test fixture up to try to find two filters that are as closely matched as possible. But you convinced me to go ahead and make a quadrature coupled dual filter configuration. I would have to use matching transformers and matching L-networks on each side of the filters in addition to the hybrids. This should keep me busy for awhile.

My front end will consist of six 7th order Butterworths similar to the "Peaberry SDR" filters but re-designed for better overlaps at the low end for continuous coverage followed by a 30 MHz lowpass for inproving image rejection at the upper end. The filters will use T44 sized cores for better linearity at high signal levels. This is driving a Clifton Labs push-pull Norten amp and a "Kiss" mixer.

Thanks again!!

"Digital Steve", K1RF11009 2015-04-27 11:03:46 iq_rx Re: Question on hybrid connected crystal filters Hi Bill,Thanks for some wisdom on this topic. You covered it well, and in passing raised the question about "exactly" what happens close in. The short and correct answer is provided by your experiments--it works pretty well, even with imperfectly matched filters.I explored this problem a few years ago using a circuit simulator. The math for even a single quadrature hybrid with errors and different impedances gets out of hand, and a circuit simulator is the right tool for the job. But it's a big job, and you end up with simulation results that remain to be tested.As one expects from the basic theory with perfect quadrature hybrids and identical filters, the match is flat over a very wide range. Since perfect quadrature hybrids and an identical filter pair are what you get automatically in the simulator, you have to really work for meaningful results with real components. Simply throwing a small offset in a single component can result in wild impedance swings in and near the passband, and might convince you that a common gate amplifier with 50 ohm input impedance and good reverse isolation is a better approach.But when you go to the next level and introduce a statistical spread of realistic component tolerances across everything in the schematic, including all the crystal parameters, parasitics and losses in the hybrids, the wild swings tend to even out. I expect there is some rigorous probability theory that covers this, but a common sense explanation is that if 100 variables all add up to some number (zero, for example) and are all statistically distributed around some mean value, a relatively large excursion by one variable has relatively less effect.I stopped there, and noted that clearly more work is needed to get the simulations to agree with reality. This is a wonderful example of the old saying "the simulation is the greater experiment." In the same number of hours doing careful experiments at the bench, I'd have deeper understanding and a piece of working hardware. In the engineer's toolbox of theory, experiment, and simulation, we often discover that simulation is the weak link.Thanks again for the wisdom on the subject.Rick KK7B11010 2015-04-27 13:32:36 billw77aaz Re: Question on hybrid connected crystal filters T'weren't wisdom, but after Wes pointed me at the ancient article in QST

I've used hybrid couplers with a handful of filters. While it does depend

upon very accurate component values to perform well, if you can perform

return loss measurements, better yet SWEPT return loss with an N2PK VNA,

adjustment of those components is a piece of cake.

It is truly a miraculous concept.

Bill

> Hi Bill,

>

> Thanks for some wisdom on this topic. You covered it well, and in

> passing raised the question about "exactly" what happens close in. The

> short and correct answer is provided by your experiments--it works pretty

> well, even with imperfectly matched filters.

>

> I explored this problem a few years ago using a circuit simulator. The

> math for even a single quadrature hybrid with errors and different

> impedances gets out of hand, and a circuit simulator is the right tool

> for the job. But it's a big job, and you end up with simulation results

> that remain to be tested.

>

>

> As one expects from the basic theory with perfect quadrature hybrids and

> identical filters, the match is flat over a very wide range. Since

> perfect quadrature hybrids and an identical filter pair are what you get

> automatically in the simulator, you have to really work for meaningful

> results with real components. Simply throwing a small offset in a single

> component can result in wild impedance swings in and near the passband,

> and might convince you that a common gate amplifier with 50 ohm input

> impedance and good reverse isolation is a better approach.

>

>

> But when you go to the next level and introduce a statistical spread of

> realistic component tolerances across everything in the schematic,

> including all the crystal parameters, parasitics and losses in the

> hybrids, the wild swings tend to even out. I expect there is some

> rigorous probability theory that covers this, but a common sense

> explanation is that if 100 variables all add up to some number (zero, for

> example) and are all statistically distributed around some mean value, a

> relatively large excursion by one variable has relatively less effect.

>

>

> I stopped there, and noted that clearly more work is needed to get the

> simulations to agree with reality. This is a wonderful example of the

> old saying "the simulation is the greater experiment." In the same

> number of hours doing careful experiments at the bench, I'd have deeper

> understanding and a piece of working hardware. In the engineer's toolbox

> of theory, experiment, and simulation, we often discover that simulation

> is the weak link.

>

>

> Thanks again for the wisdom on the subject.

>

>

> Rick KK7B

>

>

>

>

>11011 2015-04-27 13:57:53 billw77aaz Re: Question on hybrid connected crystal filters > Thanks Bill. That was very helpful. I gathered several groups of crystalsMy filters were made by CTS Knights, Inc. I bought them from a Ham Radio

> from various places, including ECS fundamental mode crystals, part number

> ECS-400-20-3X-TR. But then I scored ten Vectron 45 MHz 7.5 KHz monolithic

> filters, P/N MQF45.0 0750/21 at a very good price. See

> http://www.vectron.com/products/saw/pdf_mqf/MQF45.0-0750_21.pdf

> http://www.vectron.com/products/saw/pdf_mqf/MQF45.0-0750_21.pdf

Magazine advertisement. They have been in my junkbox a long time and have

a date code in March of 1978! They are designed for 50 ohm impedance, 5

Khz wide at -3 dB so pretty good as a "roofing filter" in an up conversion

scheme.

I've been pondering a direct sampling i.f. system at 45 MHz. That will

need more stopband attenuation that these filters are likely to produce.

Were your filters a one-of-a-kind/surplus purchase, eBay, or ????

Bill W7AAZ11012 2015-04-27 19:48:46 k1rf_digital_stev... Re: Question on hybrid connected crystal filters I got them on EBAY. There are other auctions going on from time to time. Here's one in progress now:

10 pcs Vectron Monolithic Crystal Filter M45B4 45 MHz 7.5 kHz bandwidth. SMDYou would need at least a second conversion to provide additional filtering. My second conversion will be down to 15 KHz with an audio diplexer followed by a 24 bit A/D and DSP engine for downconversion and I/Q generation using the "Fs/4" technique and FIR filters for additional 90 degree phase shift of I with respect to Q, sideband selection, different filter widths, etc

11014 2015-04-28 04:54:22 Weddig, Henning-C... Re: Question on hybrid connected crystal filters from my experience:

the two filters should have similar transmission (Phi 21 from the S21

scattereing parameters) phase responses in order to get about -20 dB

return loss over the bandwidth of the hybrid which is in the range of a

few MHz but good enough to supplement the diplexer.

Diplexer:

use a loded Q of >1, preferrably about 3.

Higher loaded Q´s will give unfortunate values of the series and

parallel resonantor and make it more difficult to align.

The use of Q=1 gives a too large bandwidth, so in conjunction with the

hybrids the matching becomes worse at some offset frequencies.

With the design of a Watkin Johnson FH1 preamp, Watkin Johnson HMJ5

mixer, (a big pity that Triquint/Quorvo made it obsolete), diplexer, 90

deg hybrids and two four pole xtal filters 44.85 Mhz / 12 kHz wide (from

Telefilter) we were able to design a receiver for VHF FM martime

operation to give 85 dB of IMD performance (some ETSI requirement)

But: we had to select the filters for nearly similar transmission phase

slope using a VNA and testbed!

In general: the xtal filter is the bottleneck of a high IP receiver.

I doubt that SMD filters will be better than the "big" HC49 or even

larger xtals!

Henning Weddig

DK5LV

Am 27.04.2015 um 14:18 schrieb bcarver@safelink.net [emrfd]:

> The diplexer takes care of impedance far, far from the filter passband.

> IE, at the DC end and at the VHF end of the spectrum. But the diplexer Q

> is 1, so the impedance of the filter, both its passband (good match) and

> the stopband (NOT good match), are seen by the mixer.

>

> You have the concept: the driving point impedance of two diplexed filters

> looks like 50 ohms both in the passband and the stopband. When the filters

> are identical then the passband of the combined pair is unchanged and the

> impedance is 50+j0 "everywhere". That, and the 3 dB reduction in power

> seen by the filters is pretty powerful, especially if you're using a diode

> mixer.

>

> When making filters you can make well matched pairs. I've done that at

> 4.434, 5.200 and 9 MHz and the result was very, very close to 50+j0

> everywhere. I have two surplus 45 MHz filters, 5 KHz wide, designed for 50

> ohm impedance. They are NOT precisely identical and sealed so untweekable.

> I combined them with hybrid networks anyway, and as I recall the passband

> of the diplexed filters is the COMMON passband of both, so the passband of

> the combination was slightly narrower than 5 KHz. And there is some small

> deviation from 50 ohms in the region where they differed, but it was not

> the extreme impedance of the individual filters.

>

> Someone else may have a very precise answer of EXACTLY what happens where

> they are mismatched.

>

> I found it useful to make the two hybrid couplers and instead of filters.

> I connect them with two identical resistors (not 50 ohms), terminate the

> output end and adjust the capacitors and turns on the inductors to get a

> precise 50 ohm input impedance. Then swap input and output and again

> adjust for precise 50 ohm impedance (just the capacitors this second

> time). After the networks looked like good 50 ohm impedances from both

> ends, then I connect the filters.

>

> Bill W7AAZ

>

> > Hi all. I'm considering building a hybrid connected roofing filter at 45

> > MHz for a high performance up-conversion receiver. I have not found any

> > really good description of why it is so much better than simply using a

> > diplexer before a single filter. The EMRFD book section 3.38 says that

> > the two filters must both be identical. How identical? What error(s) in

> > shape or center frequency can be tolerated? Besides reducing the power

> > that each filter sees by 3dB, I think an advantage might be more constant

> > impedance close-in, especially around the crystal filter edges while the

> > diplexer provides constant impedance further out so that the first mixer

> > sees a constant impedance from zero to light. Not sure if it is worth

> > going through the effort of a hybrid connected roofing filter. Any

> > insights appreciated.

> >

> > Thanks,

> > "Digital Steve", K1RF

>

>11015 2015-04-28 05:10:08 Lasse Moell Re: Question on hybrid connected crystal filters Even if the frequency is different, there is a lot of information on this topic here:

http://martein.home.xs4all.nl/pa3ake/hmode/roofer_intro.html

It seems good crystals / filters make all the difference in high DR design.

/Lasse SM5GLC28 april 2015, 'Weddig, Henning-Christof' Henning-Christof.Weddig@desy.de [emrfd]

skrev: from my experience:

the two filters should have similar transmission (Phi 21 from the S21

scattereing parameters) phase responses in order to get about -20 dB

return loss over the bandwidth of the hybrid which is in the range of a

few MHz but good enough to supplement the diplexer.

Diplexer:

use a loded Q of >1, preferrably about 3.

Higher loaded Q´s will give unfortunate values of the series and

parallel resonantor and make it more difficult to align.

The use of Q=1 gives a too large bandwidth, so in conjunction with the

hybrids the matching becomes worse at some offset frequencies.

With the design of a Watkin Johnson FH1 preamp, Watkin Johnson HMJ5

mixer, (a big pity that Triquint/Quorvo made it obsolete), diplexer, 90

deg hybrids and two four pole xtal filters 44.85 Mhz / 12 kHz wide (from

Telefilter) we were able to design a receiver for VHF FM martime

operation to give 85 dB of IMD performance (some ETSI requirement)

But: we had to select the filters for nearly similar transmission phase

slope using a VNA and testbed!

In general: the xtal filter is the bottleneck of a high IP receiver.

I doubt that SMD filters will be better than the "big" HC49 or even

larger xtals!

Henning Weddig

DK5LV

Am 27.04.2015 um 14:18 schrieb bcarver@safelink.net [emrfd]:The diplexer takes care of impedance far, far from the filter passband.

IE, at the DC end and at the VHF end of the spectrum. But the diplexer Q

is 1, so the impedance of the filter, both its passband (good match) and

the stopband (NOT good match), are seen by the mixer.

You have the concept: the driving point impedance of two diplexed filters

looks like 50 ohms both in the passband and the stopband. When the filters

are identical then the passband of the combined pair is unchanged and the

impedance is 50+j0 "everywhere". That, and the 3 dB reduction in power

seen by the filters is pretty powerful, especially if you're using a diode

mixer.

When making filters you can make well matched pairs. I've done that at

4.434, 5.200 and 9 MHz and the result was very, very close to 50+j0

everywhere. I have two surplus 45 MHz filters, 5 KHz wide, designed for 50

ohm impedance. They are NOT precisely identical and sealed so untweekable.

I combined them with hybrid networks anyway, and as I recall the passband

of the diplexed filters is the COMMON passband of both, so the passband of

the combination was slightly narrower than 5 KHz. And there is some small

deviation from 50 ohms in the region where they differed, but it was not

the extreme impedance of the individual filters.

Someone else may have a very precise answer of EXACTLY what happens where

they are mismatched.

I found it useful to make the two hybrid couplers and instead of filters.

I connect them with two identical resistors (not 50 ohms), terminate the

output end and adjust the capacitors and turns on the inductors to get a

precise 50 ohm input impedance. Then swap input and output and again

adjust for precise 50 ohm impedance (just the capacitors this second

time). After the networks looked like good 50 ohm impedances from both

ends, then I connect the filters.

Bill W7AAZHi all. I'm considering building a hybrid connected roofing filter at 45

MHz for a high performance up-conversion receiver. I have not found any

really good description of why it is so much better than simply using a

diplexer before a single filter. The EMRFD book section 3.38 says that

the two filters must both be identical. How identical? What error(s) in

shape or center frequency can be tolerated? Besides reducing the power

that each filter sees by 3dB, I think an advantage might be more constant

impedance close-in, especially around the crystal filter edges while the

diplexer provides constant impedance further out so that the first mixer

sees a constant impedance from zero to light. Not sure if it is worth

going through the effort of a hybrid connected roofing filter. Any

insights appreciated.

Thanks,

"Digital Steve", K1RF

------------------------------------

------------------------------------

------------------------------------

Yahoo Groups Links

<*> To visit your group on the web, go to:

http://groups.yahoo.com/group/emrfd/

<*> Your email settings:

Individual Email