Crystal ladder networks

[Wendymott]

Hi Peeps.....In my new receiver, I have selected to make a Crystal ladder filter using 4.432 Mhz crystals.... two reasons.. 1. I have 13... and 2..there was a circuit suggesting the values of the loading tap capacitors.
I built the filter as shown on the schematic, but I had a double peak about 500 hz apart.
I made a test jig and used variable capacitors to adjust the cap values to hopefully get a reasonably flat top response.
This was not 100% successful, but near enough, the top is not flat but a slight dip in the centre of about 3 db.
...as all the 4 crystals had varying resonant frequencies, spreading about from. 4.432991 to 4.433210, as tested in a fet oscillator and frequency counter. I selected 4 with the nearest frequencies, 4.433056,4.43069,4.435059 and 4.43072. My question is this... in what order should the crystals be fitted, or does it not matter...as can be seen the tap between xtals 2 & 3 dont have a capacitor to ground... adding one makes a nasty peak above the main peak, thus was left out.
I could use one of my 9 Mhz filters but, it means the LO will be + 9 mhz and to keep things simple I would like the LO to be lower than high.

[Terry_VK5TM]

Can't remember where I saw it recently (possibly SPRAT magazine), but lower frequency xtals at the ends of the filter and consecutively building up to higher value ones in the middle.

So low-middle-highest-middle-low xtal frequencies.

Although theoretically, if the differences are 100Hz or less between xtals, it shouldn't have much if any effect.

[MrBungle]

The order probably doesn’t make a lot of difference here. I experimented with this recently and found no practical difference if there was less than about 50Hz unless you’re building narrow CW filters. This probably explains why they’re so expensive. I got through 200 crystals (thank you aliexpress!) before I had a good set for a CW filter!

The matching is the most important thing for flatness and insertion loss rather than overall shape. It’s pretty difficult to do this in circuit so I bought a NanoVNA for £30 off eBay and build filters to match 50 ohms using matching transformers (just two FT37-43). I keep everything 50 ohms - so much easier! freedom to fiddle with the thing in isolation then.

I would in your case attempt to work out the optimum filter impedance out of circuit then build matching transformers for the Plessey ICs based on their impedance spec. No idea what that is. NE602 is about 1.5K.

Edit: to note you need crystals with the same series and parallel resonant frequency for really effective filters. I worked out I couldn’t be bothered to go that far so stuck with two at 500Hz and 1.2Khz ones for my CW receiver.

[Radio Wrangler]

The resistive part of the impedances presented to each end of the filter set the Q's of the end resonators. The Q's of inner resonators derive from this, set by the coupling factors between stages.

So termination impedance is very very important.

Note that the Q's of inner stages will have to be higher than those of outer stages. So the frequencies of inner stages are more important than those of outer stages.

David

[G6Tanuki]

As noted, the terminating-impedances of these filters are very important. In your design you've got a parallel resistor terminating the input, but this is shunted by the winding of your transformer, which will present a reflected version of the mixer-output-stage's impedance in parallel with your shunt-resistor.

When I've built these sorts of filters I've usually driven them from an emitter-follower, whose emitter-resistor has a value to present the 'right' load to the filter input.

At the output-end, I use another emitter-follower - the trick here to present the 'right' impedance is to provide base-bias to the emitter-follower by using two identical resistors in series, each of 2x the reaquired impedance - the centre-point then sits at 1/2 ths supply voltage and also presents a good match to the filter's output.

[David G4EBT]

Back in June 1984 Radcom featured the fist part of several by the late Lorin Knight, MIEE (G2DXK), entitled 'A Transceiver for the HF Bands'. The first part covered the crystal ladder filiter in some detail. The bandwidth was switchabe by reed relays from 2.4kHz to 300Hz (for CW).

The first part is here, and might be of interest, as might other parts of the project in subsequent Radcom issues:

https://www.americanradiohistory.com...om-1984-06.pdf

He also designed a valve linear amplifier. using a pair of 6146Bs. I greatly enjoyed reading both articles and at the time, was still interested in amateur radio so came close to building them, but later embarked on the well known G3TSO rig, which used the Plessey series of communications ICs - long since obsolete.

[Radio Wrangler]

There were also articles by Jack Knight in Radcom
Then Wes Hayward's writings in his amateur radio design books.
Article by M Pochet in wireless world was rather short but gives a nice recipe pattern of shunt capacitors.
Then you're back inro Dishal and those working at Bell labs.

Oh, QEX is worth searching.

Finally, you can do a lot if you add parallel inductors to cancel the case capacitance of the crystal. Have a play on LT spice, it handles DIY crystal models rather well.

David

[G6Tanuki]

Here, from "Amateur Radio Techniques" is an example of what I was describing about matching a crystal-ladder-filter; in this case the filter has a required termination of 800-900 Ohms.

There are various easily-found articles by F6BQP which are worth reading.

[Wendymott]

Hi Peeps..... And there's me thinking.. it was going to be simple..the reason being.. I took a couple of circuits and a filter design article from the tinterweb, thinking it was all going to be ok. Thanks David X 2 and Tanuki and Mr Bungle.. it looks like my pcb ver 3 is going to be re designed..I will add the buffers in and out to my test jig and see how it changes things. The crystals were from the colour TV ref oscillator, thus were reasonably plentiful.
I have looked at the article of Lorin Knight and screen shot your circuit Tanuki.. THanks.

[MrBungle]

Meant to post this earlier. Worth a read from an amplifier termination perspective. These will give you easy to match amplifiers with consistent impedance: http://w7zoi.net/bidirectional_matched_amplifier.pdf

Prototype:



Measurements:



I’ve got one of these each end of my crystal filters as part of the IF strip. It also makes a good wideband termination for the first mixer output without having to futz with diplexers.

W7ZOI on filter design: http://w7zoi.net/genfil.pdf

Genuinely this took me longer than putting the rest of the receiver together! (Which is still unfinished after two years but does RX well on 40m, close to my K2’s performance)

[Argus25]

Originally Posted by G6Tanuki

When I've built these sorts of filters I've usually driven them from an emitter-follower, whose emitter-resistor has a value to present the 'right' load to the filter input.

This doesn't seem to make sense because in the case of the emitter follower, it is not the emitter resistor that primarily sets the impedance (looking into the emitter) It is instead the resistance in the base circuit, which includes the Thevenin resistance of the often two resistors there and the source impedance driving the base, all divided by the transistor's hfe. You can add the emitter resistor in parallel with that, for an accurate value, but it is usually an order of magnitude higher in value. The point being that the emitter resistor value you see is usually much higher in value than the small signal output impedance of the emitter follower

At the output-end, I use another emitter-follower - the trick here to present the 'right' impedance is to provide base-bias to the emitter-follower by using two identical resistors in series, each of 2x the required impedance - the centre-point then sits at 1/2 ths supply voltage and also presents a good match to the filter's output.

This is about right if you ignore the emitter follower's base impedance, which is about roughly its emitter resistance multiplied by the hfe, so usually in this case the impedance is set by the parallel values of the two resistors.

[Wendymott]

Well peeps.... got it sorted...I had totally got it wrong.. no surprise there then....
I buffered the output with a SL612 amp... and the response was improved so I could see what was really going on..... The values of the caps I had fitted resulted in an IF pass band of about 350 hz....Oops...I wanted 2.5 Khz ish.... re fitted the first and last caps as 47pf, and the centre one was re fitted at 68pf... now a nice flat topped response with a passband of 2.5 khz "ish". However thats where the problem really starts.
For the BFO I tried an LC circuit tuned with a varicap diode..a mechanical cap is out of the question...this drifted considerably, so I checked out my remaining 4.43 crystals,, I need 4.4320 and 4.4340 Mhz. I could pull one to the upper frequency but not to a lower frequency... so I contacted IQD crystals.... they were very helpful in pointing out that unless I wanted 1000 units minimum, they couldnt help, which is quite reasonable.
Any suggestions ??

[peter_scott]

Originally Posted by Argus25

This doesn't seem to make sense because in the case of the emitter follower, it is not the emitter resistor that primarily sets the impedance (looking into the emitter)

I suspect what was intended was an emitter follower (offering a very low impedance) followed by a series resistor to define the source impedance.

Peter

[G6Tanuki]

I've seen VXO circuits where they have used two crystals (of slightly different frequency) wired in parallel, and it's claimed that such circuits give a 'pulling' range greater than twice the pulling range of either of the two crystals. Don't understand why though!

Perhaps you could try this 'frankencrystal' approach?

[Argus25]

Originally Posted by Wendymott

Well peeps.... got it sorted...I had totally got it wrong.. no surprise there then....
I buffered the output with a SL612 amp... and the response was improved so I could see what was really going on..... The values of the caps I had fitted resulted in an IF pass band of about 350 hz....Oops...I wanted 2.5 Khz ish.... re fitted the first and last caps as 47pf, and the centre one was re fitted at 68pf... now a nice flat topped response with a passband of 2.5 khz "ish". However thats where the problem really starts.
For the BFO I tried an LC circuit tuned with a varicap diode..a mechanical cap is out of the question...this drifted considerably, so I checked out my remaining 4.43 crystals,, I need 4.4320 and 4.4340 Mhz. I could pull one to the upper frequency but not to a lower frequency... so I contacted IQD crystals.... they were very helpful in pointing out that unless I wanted 1000 units minimum, they couldnt help, which is quite reasonable.
Any suggestions ??

...hmmmm. I would assume being in the UK the likely 4.43 MHz crystals you would have lying about are actually 4.433619 about. So you pulled one up from that value to 4.434000.

To pull it down from 4.433619 to 4.432000 didn't work.

Trying to get 4.432000 from readily available Xtals seems a tall order for an off the shelf Xtal.

The closest I could find aside from the usual 4.433619 MHz, was a 4.423500
at Surplus Sales Nebraska.

[MrBungle]

This is where I get utterly lazy and put an arduino and Si5351a in mine for both VFO and BFO. Any form of VXO seems to hate me

[Wendymott]

Hi Mr Bungle and others..The idea of this "restoration" was to re model the Eddy EC10, as an "Analog" receiver, using the original case and dial.. and trying to keep digital out of it..fortunately .. with percy and verance.. I now have two xtals on frequency..so I am quite happy.. for the moment..just finished the PCB artwork for an etching session sunday, not using the iron on method.. too much area to cover.