Help with high insertion loss in my 112Mhz bandpass filter

[G0HZU]

Toroids are 'self shielding' which is a plus but getting the inductance exactly right with a toroid is not easy because you basically have no in between turn values available.

Whatever...

My main point is that you all seem to be helping the OP polish a turd because the filter schematic in the OP shows a filter with inferior performance to Hayward's filter(s) in terms of stopband performance. It should have lower insertion loss than the Hayward filter but that comes at the price of inferior selectivity compared to the original Hayward design.

[T3sl4co1l]

Toroids are 'self shielding' which is a plus but getting the inductance exactly right with a toroid is not easy because you basically have no in between turn values available.

They aren't, and you do!  We're talking VHF here, where you'll be using a few turns each -- squish the turns closer together for more L, or stretch apart for less.  The near-field around the individual turns is quite "interesting", not well shielded.  It takes a large, evenly distributed winding to get the "self shielding" effect that a smooth, symmetrical toroid has.

(A pot core on the other hand is quite well shielded, but those are hard to come by for VHF purposes!)

Tim

[G0HZU]

They aren't, and you do!  We're talking VHF here, where you'll be using a few turns each --

How about 30 turns of skinny wire on a T37 toroid like the image in my earlier post? 

That's how 'unspecial' the 5dB insertion loss performance is of the first two Hayward filters despite the sexy looking gimmick caps and impressive looking solenoids. I think I could get similar performance with a filter made using those T37 toroids (shown below) with about 28-30 turns and it could be built in minutes ugly style (point to point)using SMD caps and toroids with no special layout requirements. Yes, it looks wrong to use toroids like this at 112MHz but I'm trying to show that there isn't any voodoo magic required to get that shape factor and insertion loss at 112MHz.

[T3sl4co1l]

Yeah, 30 turns is enough for reasonable shielding.

Of note, #0 (air core), #1 and #2 are about the only powder materials suitable for VHF use.  Permeability isn't much above air, so you still need a lot of turns and inductance is that much more sensitive to bunching of turns.  But it's worthwhile if you need the boost in size or packing density without shields.

Tim

[G0HZU]

Yes, I would use phenolic #0 toroids. If I'm allowed to 'cheat' and use a more sensible size toroid such as T50 I'd expect to be able to beat Hayward's filter for insertion loss whilst matching it for bandwidth and selectivity. Such a filter could be tacked together in maybe two minutes with a soldering iron. but this would be a different filter topology that didn't need the exotic gimmick caps. Just tack together the coils and caps with no layout issues and it will work once the toroids are adjusted for lowest insertion loss. No need to adjust any caps, just use fixed values. Only three inductors to adjust...

However, if you gave the toroid filter to 100 novices in an untuned state I doubt many would be able to tune it correctly even if the correct number of turns is on each toroid. My guess is nearly all of them would declare it impossibly frustrating to tune up and they would give up after maybe 15 minutes of squishing around the toroid windings. Toroids are not nice to tune, especially on a narrow filter like this. But with 30 seconds of training and the right tool(s) they could all do it in about a minute or two. Suddenly it becomes a piece of cake

[cdev]

since we're talking about resonators what about having a tap instead of a capacitor.

[G0HZU]

Here's a simulation of the original Hayward filter using his inductor Q and cap models. It shows a 5dB loss and pretty good selectivity across 60-160MHz. Much better than the filter design in the OP of this thread.

However, there are alternative means to design Hayward's filter and one way is to transform away the problem components like the gimmick caps and I tried to design an equivalent filter with a topology that gave the same bandwidth and the same shape factor. But a key design aim was to make it easier to make with tame toroids and no daft component values like 0.092pF. I built it today and with little T37 toroids and some decent SMD caps I got about the same loss as Hayward did at just over 5dB. But with T50 toroids I get 3.8dB loss and a plot that looks very close to the simulation of Haward's filter below.
It took about 2 minutes to solder together and about the same to tune it up.  Note that the circuit of my filter is different to the parallel resonator circuit below and this is how I managed to get away from the gimmick caps. I also chose to use toroids because they are much less touchy and sensitive to strays from nearby metalwork etc.

[G0HZU]

If built correctly with decent components, the filter in the OP of this thread would look something like the simulation below. You can see that it isn't an equivalent to the original Hayward design because it doesn't have the same selectivity. So is it worth making if you want/expect the same performance as the Hayward filter(s)?

[cdev]

It depends on where the interference you want to eliminate is.

[G0HZU]

If the filter is going to be used in a spectrum analyser then I would have thought that rapid rejection of the swept first LO is one reason to try for a very narrow IF1 filter here. The VU2ESE filter has about twice the bandwidth and this also lets in more signals to the second mixer.

I think the VU2ESE filter is here:
http://hfsignals.blogspot.com/p/specan-reboot-of-w7zoi.html

Using a symmetric design with solenoids for the input 70MHz filter isn't ideal either. It could be made much better with a few subtle changes to that filter to improve the stopband performance up into UHF.

[cdev]

That looks like a great project.

[sab]

So, I followed the advice (single sided board, short leads, no unnecessary pads) from many generous people and, lo and behold, things certainly improved. 6.5db insertion loss and a pretty sharp frequency response (using my BG7TBL noise source which is at -10dbm). I've included some images.

Thank you very much for all your help.






Sanjay

[T3sl4co1l]

A bit lopsided, looks like one of the resonators is high or low?

Tim

[cdev]

You shouldn't assume the top and bottom of the PCB are both grounded unless you have lots and lots of vias between them. So many it looks like a checkerboard of vias.   More around the traces.

Otherwise they are a capacitor/inductor combination, not a ground.

[David Hess]

Admittedly I'm being picky but I was wondering if anyone had suggestions on how to improve my 112 Mhz band pass filter.

Build a helical resonator instead.  112 MHz is right in the range where this is very practical.  The one shown below is for 2 meters and has a tuning range exceeding 144 to 148 MHz.  The coupling window shown is way too large but suitable for higher power (coupling loss in this case is less than 0.3dB which is not ideal for a narrow filter) and covering a good part of the entire 2 meter band.  When I designed and build it, I did not have good information on how large to make the coupling window.

Alternatively, use better (higher Q) capacitors and piston trimmers or air capacitors for adjusting the sections.

[sab]

Yes, a helical resonator looks quite promising from everything I've read.
Thanks,