Band-stop filter

[Skywave]

A need has arisen to introduce a small amount of attenuation in a wide-band amplifier chain. The centre freq. is of the order of 50 MHz and an approx. 1 to 1.5 dB attenuation between 35 and 65 MHz is required. A conventionally-designed L/C band-stop filter springs to mind, but the attenuation at the corner freqs. will be excessive. However, what does spring to mind is to proceed with that approach but to bridge the input and output terminals of that filter with a suitably-valued resistor, thus reducing the attenuation in the stop-band to the required degree. The iterative impedance is 50 Ohms.

I'm not a filter guru, so I have to ask: does that idea make sense?

Al.

[Terry_VK5TM]

This may give you some ideas/inspiration.

Originally made for use on the output of DDS modules, but might be adaptable to your use.

http://www.vk5tm.com/gallery/g3oag/g3oag.php

Terry

[Radio Wrangler]

Bandstops can be very devils to design. In theory an ideal bandstop should pass DC to daylight and beyond with just the wanted notch cut out. But this is impossible bandwidth.

You can't design a bandstop without deciding on limits that its passband must cover. The components which givethe required response at 60MHz will cause trouble a decade lower and a decade higher.

One trick is to plant a bandstop within a planned wider bandpass or lowpass.

If you only want a controlled and small amount of attenuation in the 'stopband' then you might find more of interest looking for equaliser designs.

The seminal paper on equalisers was by Harry Kimball in the journal of the society of motion picture and television engineers in about 1938.

If I had to design one, I'd probably go for a tranversal filter architecture, though.

It's not a small job.

David

1 to 1.5 dB isn't much, I am intrigued as to why?

[Peter.N.]

In the early days of band 3 someone made a cheap selective band 1 attenuator for areas where the band 1 signal was excessive and they worked to well, you could attenuate the sound or vision but not all that effectively both, they consisted of an adjustable iron core in a coil in series with the feed, all housed in a small aluminium box.

Peter

[G0HZU_JMR]

Originally Posted by merlinmaxwell

1 to 1.5 dB isn't much, I am intrigued as to why?

Yes, it does seem an odd spec. I've designed all manner of 'odd' RF filters over the years from LF through many GHz but this is one of the 'oddest' specs I've seen. In a design review the inevitable question would be asked "is this odd filter requirement really just a solution looking for a problem to solve?" In other words, could the filter requirements be removed if you altered some other aspect of the system?

However, I don't think it would be that difficult to make such a filter as long as you can withstand some overall insertion loss to help with lossy matching etc. It does seem that this would be possible because it looks like this is going to be fitted somewhere in between some gain stages. Also, a lot depends on how steep you want those 1.5dB shoulders to be at 35MHz and 65MHz.

A conventionally-designed L/C band-stop filter springs to mind, but the attenuation at the corner freqs. will be excessive. However, what does spring to mind is to proceed with that approach but to bridge the input and output terminals of that filter with a suitably-valued resistor, thus reducing the attenuation in the stop-band to the required degree. The iterative impedance is 50 Ohms.

I wouldn't use this approach even though I'm not entirely sure how you would connect it up in order to define your 1.5dB of (extra) flat loss across 35-65MHz.

[G0HZU_JMR]

I'm a bit poorly with a bug at the minute so I can't concentrate very well but here's a simulation plot of a quick and dirty filter that assumes you can tolerate about 6dB of loss overall.

You can see it is flat across a lot of the log frequency scale of 1-500MHz and it is flat inside the -1.2dB stopband too. The VSWR is pretty good. The overall differential group delay is about 8ns although I could probably tidy that up a bit. This is just a quick (default) design although it does use a lot of lumped parts. This is the deluxe default version with 10 inductors and all of the inductors and caps are sensible values with sensible Q values as well for this frequency range. No scary unrealisable lumped parts anywhere so I think the filter will be realisable. I could probably get it down to 6 inductors if you can live with droopy shoulders at 35 and 65MHz. it only takes me a few seconds to make a different version. This lossy filter topology isn't exactly rocket science

Can you live with the 6dB insertion loss? If not then I guess this filter can at least act as a benchmark. You would probably have to go to somewhere professional like BSC filters to get something markedly better than this.

I would also think that even a weeny bit off the 50 Ohm perfect impedance for both input and output would make a bigger than 1.5dB difference. I measured the input impedance of an ICOM HF radio once, it was all over the place, I haven't bothered since (mainly for my sanity) as the radio worked fine.

[Skywave]

Thank you everyone: a few useful leads to investigate.

Al.

[G0HZU_JMR]

I would also think that even a weeny bit off the 50 Ohm perfect impedance for both input and output would make a bigger than 1.5dB difference.

Yes, my filter was designed with this in mind and this explains some of the 6dB insertion loss. I can get the loss down to 3dB but it makes the filter more prone to the source/load mismatch issues you would see when fitted to a real amplifier circuit. I designed it to cope OK with up to 1.5:1 source and load VSWR. You can see a plot that explores all phase angles of a 1.5:1 VSWR at the source and the load. It looks a bit messy but this is actually quite good

It would never have this rapid zigzag ripple in reality but the plot explores very rapid rotations around the 1.5:1 VSWR circle at both the source and the load ports. So you see the peak to peak effect across a very short frequency interval. So it looks fuzzy. But the fuzz is really meant to graphically show the 'window' inside which any mismatch ripple could live.

[G0HZU_JMR]

Hope nobody minds but notch/bandstop filters are a bit of a passion of mine. I've designed loads of them over the years across very wide frequency ranges.

The plot below is a 60dB notch filter I designed many years ago at work. I actually did most of the concept work here at home. This is a plot of a real production filter, not a simulation. It works from DC to just over 6GHz and gives a deep notch at about 1.5GHz. This filter caused a sensation at work when I designed it because no filter design company at the time could compete with it. The insertion loss scale below is 1dB/div!

You probably need to know a bit about RF to understand why it is so special but it achieves low loss right up to 7GHz with no suckouts or re entry modes anywhere in the band. I designed this using unconventional means, you probably won't find this filter type in any textbook even today. The loss at 6GHz is just 0.6dB and this includes connector losses. The passband return loss is >=17dB across the whole range. It needed to be this good because it is used for a wideband transmitter. I think we have produced many hundreds of them at work, possibly >1000

[G0HZU_JMR]

I'm a bit bored and feeling a bit better too so I had a go at finding a more efficient filter transform.

The plot below uses just 4 inductors and 4 caps and 5 resistors. Not bad!

If you are interested I could build this one over the weekend. It shouldn't take long to test it.

[Skywave]

Thank you, Jeremy: I'll accept your offer. Presumably, that refers to a cct. diag.

One Q. arising: if that filter has a resistor connected 'across' it, will doing that reduce the level of the stop-band attenuation? And its effect on Zit?

In addition to all that, I should now state that today, a lot of things associated with some serious house renovations have landed in my in-tray, so it could be some time before I get more seriously involved with this.

Al.

[G0HZU_JMR]

Yes, I'll try and build it tomorrow or Saturday and post up a schematic. However, I must repeat that this solution isn't an exotic one. It is a hammer and nail solution and you might be surprised to see how simple the concept is and I don't think it will be what you are expecting in terms of how the whole thing operates.

One Q. arising: if that filter has a resistor connected 'across' it, will doing that reduce the level of the stop-band attenuation? And its effect on Zit?

I'm still not sure what you mean here. Maybe when you see my (hopefully working) circuit it might fit what you mean here.

[G0HZU_JMR]

I knocked up a quick and dirty version 'ugly' style on the back of some spare PCB and it seems to work just like the simulation. But ideally I need to know your upper frequency limit because this dictates how much I tidy it up to get the best passband performance up at UHF and it also dictates what components I use in the resistive part of this filter.

The low side limit is no problem, it will work down to DC but can you give me an upper frequency limit? I'm going to guess 500MHz or less but I can get it to go higher than this.

I also need to know if you don't mind using SMD parts for the capacitors. I don't think ordinary leaded ceramic caps can be used here unless you can tolerate some response ripple. The SMD caps I've used here are quite big. eg they are like little cubes, 3mm x 3mm x 3mm and they are really easy to work with.

[G0HZU_JMR]

Can you also let me know if you want 1dB, 1.25dB or 1.5dB depth? You can see by the image below that the depth can be set quite flat.

This is a plot of my first lashup filter taken with my network analyser. It's built in dead bug style and I haven't bothered to try to get optimal performance from it because I need to know the answers about the upper frequency limit and also the depth you want in the dip. But the plot below shows that the filter can be built and the performance is realisable

Would you want a steeper shoulder on the high side? I think I can do this but it will need one extra capacitor. This cap wouldn't have to be SMD. The penalty for this would be a blip of a few ns of group delay up around 200MHz but I suspect that you won't be bothered by group delay specs? You would normally only be interested in group delay if you wanted to send an ultra wideband digital signal through the filter that covered the whole frequency range.

[Skywave]

Hello Jeremy: thanks for your reply.

I think it's time for a bit of background as from where all this is coming from. Initially, it was in association with my attempts to build a RF voltmeter, but such an amplifier has so many possible applications that my attempts at that have become a separate topic! All my various efforts have shown me that trying to produce such an amplifier with a 'flat' response, ±1.5 dB max. from 2.5 to 95 MHz, is far more difficult than I imagined. I did some experiments with the Analogue Devices AD 8055 chip (not the SMD variant), but performance was worse than my discrete component attempts!

Your Qs.:

For entire amp:
*HF limit: 100 MHz.
*LF limit: 2.5 MHz

For filter:
Depth of atten.: about 1.5 dB.
B/W: as stated: 35 - 65 MHz
Rate of atten. (slope): fairly gradual & shallow.

As for the design, construction and test of things like this, I am hampered by various factors. Lack of suitable test equipment (I don't own a spectrum analyzer with its in-built tracking generator); my 'scope is a Tek. 2465, so the 'flat' B/W runs to about 100 MHz (just); my eye-sight and age-related reduced dexterity makes working with SMD components difficult; there are many other things demanding of my attention these days: priorities dictate!

I think that says enough for now - and again, thanks for your work on my behalf.

Al.

[G0HZU_JMR]

I think the best thing to do would be abandon the notch filter and just have another go at the wideband amplifier design. I can vaguely recall that you wanted something around 20dB gain.

The simplest solution would be to use something like MiniCircuits ERA5-SM. It looks like MiniCircuits has recently updated their website but here is the link to the ERA5 amplifier.

https://www.minicircuits.com/WebStor...del=ERA-5SM%2B

Here's the datasheet:
https://www.minicircuits.com/pdfs/ERA-5SM+.pdf

I've probably got several ERA5 devices here along with most of the rest of the Minicircuits range and I've used amplifiers like this loads of times and they provide flat gain from LF though UHF. Provided that the PCB layout is good, with good grounding the amp will be unconditionally stable.

If you want to make something using discrete parts then 20dB gain is risky in terms of achieving unconditional stability. I would rather do it in two stages but a lot depends on what signal handling you want and what other specs you want. But these days everyone takes the lazy approach and chooses a MMIC device. There are loads to choose from depending on gain and frequency response. The downside with the MMIC amplifiers is that they can be relatively inefficient. The ERA5 typically runs at 65mA.

If you don't like the SMD package then you could make a lower bandwidth variant using discrete BJT parts in a darlington config. But you would be wise to aim for a lower gain than 20dB (eg 14dB) if you want to make it unconditionally stable. It will also take up a lot more space.

[G0HZU_JMR]

In case this is of use I knocked together a basic simulation of a BFR91 based 50R gain block with just over 12dB gain up to 100MHz. The design uses negative feedback to set the input and output to 50R and this also controls the gain. Compared to the proper/modern MMIC amplifiers this thing is a bit of a dog but I quickly built one on a test PCB and tested it and it works just the same as the simulation.

The gain is about 12.4dB at low frequencies and it droops about 0.25dB by 100MHz. The input and output VSWR is better than 1.25:1 across the 100MHz range and the noise figure is about 4dB on the simulation and I measured it at 4.1dB on the real amplifier.

See below for the built circuit and the schematic. It isn't a very strong amplifier, the OIP3 is only about +25dBm and the P1dB is going to be about +10dBm or 10mW.

There's also a plot of gain and input and output VSWR of the test PCB measured on my VNA. Also a plot showing the real part of the port impedance and the imaginary of the real PCB tested by the VNA. Ideally this should be 50R real (on the left scale) and 0R imaginary (on the right scale) for a 50R amplifier and it is pretty close.

This should be an easy and reliable (in terms of performance) amplifier to make if you don't want to use MMIC amplifiers and you only need good performance to about 100MHz. But you will have to use a tight layout and decent caps like I have used in the image below.

[Skywave]

Yes, thanks Jeremy. That cct. looks like it has promise in many applications. Right now, a lot of my time is being taken up by non-hobby / non-electronic interests & commitments, but I will squeeze in some more work on the project to which this thread is related as and when I can. Balancing one set of tasks / problems against other ones helps me keep a grip on the ol' sanity! And it keeps me out of the pub!

Al.