Cavity Filter Episode 3

A quick update while I'm in the lab to do some preliminary tests. I just did the very first frequency sweep on an RF network analyser and I'm quite satisfied so far. Although the top plate, that holds the tuning screws, is not soldered in place yet it seems as if everything works.

I still wonder why I chose male SMA for the cavity filter connectors, since cables usually have male connectos too. Luckily I found one gender changer and I'm also abusing an SMA T-connector.

One of the centre resonator pins was a bit lopsided, which I corrected by simply bending it back into place. I noticed that this particular pin is longer than the natural resonance length of 71.3 mm (1013.3 MHz) while I managed to keep the other pins at 69.0 mm. I will have to shorten this pin otherwise it will be useless.

Here's a picture of the first sweep with a span from 0.3 MHz to 3000 MHz. Apparently the filter resonates at 1029.6 MHz without the top plate and tuning screws. I was expecting a slightly higher frequency like 1060 MHz or so but 1030 MHz is nice.

Centring on that peak and narrowing the span to 20 MHz gave the above transmission plot. This is still without the top plate and tuning screws in place. I'm impressed by the quality factor of almost 1000. Granted, the skirt to the right has a hump and the insertion loss is rotten but this made me curious to see how the filter would work out with the top plate in place.

This is the transmission plot with the top plate attached and held in place with duct tape. A first attempt at tuning confirmed that the lopsided centre resonator is too long and adjusting its tuning screw does not change the filter curve at all. The other screws work very well although I think the tuning process is a bit too sensitive. With one of the cavities not tuned this filter has a very nice insertion loss of -6.1 dBm and a Q of roughly 600.

Curious to see the rejection at 1024 MHz (the second local oscillator) and 1034.7 MHz (the in-band image frequency) I changed the span to 45 MHz and did a new, slower sweep. The result is shown in the above picture.

I found -54 dBm at 1024 MHz and about -66 dBm at 1034.7 MHz, which is a lot less than what I want it to be (-120 dBm). But given the fact that one of the cavities is off-resonance and that this first test was not intended to be very accurate, I am looking forward to doing the measurement again later.

I haven't decided yet whether I need to re-solder the resonator pin or simply shorten it by removing about 2 mm of copper, but maybe re-flowing the solder joint is the proper solution. I still need to write episode #2 as well, which will tell more about how I soldered the base plate to the cavities.

Cavity Filter Episode 2

I was unable to gain access to a drill press so I had to use a cheap hand-drill to drill the holes in the tubes. I hated every minute of it! Our workshop has a drill press, unfortunately I didn't get access to it.

After I finished soldering the individual cavities together it was time to prepare the connectors and hairpin couplers. For that purpose I used a few pieces of 0.141" hard-line and extracted the centre conductor. The PTFE dielectric was cut into short pieces and provided spacers to keep the hairpin couplers from touching the cavity walls.

I used a hot air gun to solder the tubes together and got very nice joints using this method.

Instead of drilling holes in the end plate I notched the bottom of the tubes, bent the hairpins through the notch and soldered them before soldering the base plate in place.

The thick copper bars have almost been a show-stopper for me since it was impossible to drill holes for the hair pin couplers. Luckily, Sam recommended me to not drill holes at all and rather notch the tube and bend the hair pin through the notch. This worked very well.

The tubes were then placed on top of the bottom plate. I placed a ring of solder wire on the inside of each cavity while and simply fed solder to the outside joints as the bottom plate was hot enough to melt it. Added lots of flux and the result was excellent.

The second resonator from the right sticks too far into the cavity and will probably render it useless. I'll have to see how it affects the amplitude response.

If any of the resonators is lopsided, it can easily be bent back into the centre by sliding a drill bit into it and gently pulling it to where you want it to be.

I think I forgot to mention that I threaded the holes for the tuning screws. That way I will only need one nut to fixate them once the cavity is tuned. The screws I used are M4 x 40 but I think 25 mm  ones would have done the trick too.

The annoying part began when I wanted to attach the top plate next evening.

More components arriving

Excellent! I got my feed-through capacitors on Monday and the VCO kit from Mini-Circuits. Additionally, I'll have the day off on Wednesday so there's lots of time to grab the solder iron and errr...write some reports! I just realised that I still have work to do.

Anyhow, these are super cheap Russian 100 pF feed-through capacitors. The date code on them suggests that they were manufactured in 1984. They're in excellent shape though and the one I measured was spot on. I'll use them to prevent the RF energy from escaping my prototype aluminium enclosures, since I have to get the DC power lines in there similar to how it's done in the next picture:

This is one of BG6KHC's aluminium enclosures. The leads you can see on the left side are feed-through capacitors that connect to the power lines and the serial interface of the 16-bit ADC inside. I like the ruggedness of this type of enclosure and intend to use a similar approach for my prototype modules.

For anyone who is looking at building Scotty's Spectrum Analyser and lives in Europe: the kit that contains all the necessary parts from Mini-Circuits is now available from Mini-Circuits Europe and I'd recommend you to contact one of their account managers for a quote. Their service is splendid and it took less than a week before I had the parts arrive at my office.

The kit contains:

• 1x ROS-1500+ 
• 2x ROS-2150VW+
• 1x TC16-161TG2+ 
• 4x ADE-11X+ 
• 10x ERA-33SM+

SMA Connectors

I received the SMA connectors from Hong Kong today and a couple of SMA male to SMA male RG316 cables. Shipping was relatively fast this time - only about 24 days.

The price for one cable was cheaper than the price for two connectors plus cable, so there is really no point in making them myself. Besides, these are just for testing purpose since I did not want to waste my hard pipe and semi-rigid cable for this.

The gold plating of the connectors does not have the same "feel" to it like good quality connectors from for example Suhner or Amphenol, but considering a price ratio of 1 to 8 and given the fact that I will not remove the connections once everything is installed, I guess I can live with it.

Resolution Bandwidth Filter #5

I found a screenshot that I took during the prototype test. It resembles the frequency response of the NDK 10F15DG after successfully tuning the matching circuit.

Really nice ripple of approximately 0.4 dB and the insertion loss is low too with only -2.4 dB. The picture actually states a loss of -7.9 dB but the input power was -5.5 dB.

The matching circuit was 3.3 uH in series and 10 pF + 5...25 pF trimmer in parallel,  if I remember correctly:

I recorded the 2-port touchstone data for this filter with 1600 points. You can download it as a zip file from my Google docs: NDK10F15DG_plain.zip

Input Switch

Slowly getting there.

This is the input switch of the network analyser. Its purpose is to switch between transmission/reflection measurement and it also provides a switch to choose between spectrum analyser (SA) and vector network analyser mode (VNA).

I decided to mount the SMA connectors in the same style as I did with the other modules instead of letting them protrude from the front panel.

I use 0.2 mm (50 mil) copper sheet for the  perimeter fence. It requires a bit of practise but I found that using one single strip works better than joining two brackets. I used a pair of scissors to cut the sheet.

 Internal shielding. It requires a lot of patience to install the fences. Some of them are made of double sided copper-clad FR4 in order to keep ground planes separate.

 Almost finished. I need to seal the individual sections with copper lids as soon as I get my hand on the missing capacitors. You can see the RF-relay that switches between VNA and SA mode in the lower left corner.

Bottom view. A few power supply jumpers are missing. The three SMA connectors on the top are for the  transmission, reflection and spectrum analyser inputs, the SMA connector to the left below them is the selection output. The semi-rigid coax jumper on the bottom connects two sections.

The thing needs a box!

I've been searching for a nice enclosure and it's really hard to find something nice. I was thinking of some kind of CNC-milled aluminium frame but since I still relocate modules every now so often I will postpone that decision until I actually finish all the modules.

For now, I will use a simple box made of FR4 and mount the modules on a PCB frame.

 The box is made of FR4. I took two 200x300 mm single-sided sheets and cut them into half. The width of the box on the inside is exactly the width of a DIN A4 sheet.

The bottom of the box is double sided FR4.

The sides are simply soldered together. I will use this to shield the finished device.

 Not pretty but extremely practical and cheap.

This is the FR4 frame on which I will mount the individual modules. Its dimensions are exactly the same as the bottom of the box.

I've been peeking to the left and right at what other people are doing. All of my SLIM modules have a 0.08" copper strip perimeter fence on the outside of the PCB that connects top and bottom ground. The milled holes in my base plate are exactly the size of a SLIM, i.e. 1.2" x 1.2" etc. That way I can place the shielded SLIMs on top of the base plate and tack them in place.

Resolution Bandwidth Filter #4

The fourth RBW filter. I'm beginning to like the smaller cans, this one is a Hy-Q10726 which is a 8-pole bandpass with a bandwidth of 7.5 kHz and a centre frequency of 10.7 MHz.

There's really not much to it. The filter can, two connectors, the PCB and matching components.

Long version SMA connectors.

The innards. I think I matched this particular filter with 5.6 µH and 33 pF + a 6...40 pF trimmer. Note how the input and output section is shielded with a fence in the middle of the board.

Update: Managed to take a picture. Ripple could be better. I also recorded the 2-port touchstone data for this filter with 1600 points. You can download it as a zip file from my Google docs: Hy-Q10726_matched.zip

More perimeter fence. Getting good at this :)

Direct Digital Synthesiser (DDS 1 and 3)

These are the synthesisers that steer the phase locked oscillators. Their heart consists of an AD9850 which is a 125 MHz complete CMOS DDS from Analog Devices.

The DDS 1 Module, SLIM-DDS-107, is the fine frequency "steering" source for PLO 1 in the MSA.

 Finished DDS1. There's not much in there, you can see a 78L05 voltage regulator to the left, the AD9850 itself in the centre and a 10.7 MHz crystal filter to the right.

The filter is shielded from the rest of the circuit with a fence.

Bottom view. The SMA connectors are not placed yet as they weren't in the mail. Connectors in China are incredibly cheap but you have to live with up to 30 days of shipping...

 Both synthesisers and a couple of extra perimeter shields.

A few copper strips prepared for the shields

The first few attempts were crap but his frame is actually almost a perfect fit. I'm bending the strips with small pieces of PCB.

Current status

I've completed another set of boards this weekend. The new PCBs are really small and the quality is alright. The complete lack of solder mask and silk screen can be annoying at times.

Boards that are fully populated at the moment are:
The Control Board, the four Mixers, the two DDS boards, the 64 MHz Master Oscillator, log detector, AD converter unit, Phase Detector and the Filter Bank Selector.

Boards missing so far:
The three Phase Locked Oscillators, I.F. amplifier.

Note to self: place an order for the VCO chips at Mini-Circuits.

New set of PCBs

An RF-engineer from the USA was kind enough to provide me with a full set of boards for the Vector Network Analyser project. This way I don't have to send an order to ExpressPCB who otherwise offer an excellent service. The shipping costs to Europe is a is $67.00 for a 2-3 day courier, which is the only shipping option their software let me choose.

This is what the PCBs look like when you order their standard service 2-layered boards. The cost for two of these panel is around $100 and the cost for 10 panels is $333. I highly recommend you to organise a group buy or simply buy 10 panels yourself and ask if anyone needs one. New people join every month and face the same problem of how to acquire the boards.

From the top left to the bottom: two PCBs for crystal ladder IF filters, ADC PCB for chip packages with 8 leads and the resolution bandwidth filter bank switch.

On a side note: with enough people for a group buy, you may as well order ROHS silver PCBs with silk screen and solder mask. The lack of solder mask can be a bit annoying at times.

Resolution Bandwidth Filter #3

Another filter update. I acquired a couple of monolithic crystal filters with the smaller D-151-D multi-pole package that fits the 1 x 1 SLIM modules. Most of the filters were manufactured by NDK with the exception of a single Hy-Q10726.

They are all 8-pole filters but their centre frequency varies. I now have

NDK 10F15DG:

has a centre frequency 10.7 MHz and a 6 dB bandwidth of 15 kHz

NDK 11F30D:

has a centre frequency of 11.5 MHz and a 6 dB bandwidth of 30 kHz

NDK 11A1.0C:

has a centre frequency of 11.4 MHz and a ? dB bandwidth of 1 kHz

Hy-Q 10726:

has a centre frequency of 10.7 MHz and a 3 dB bandwidth of 7.5 kHz

This is the NDK 11F30D. I managed to tune it and save a two port touchstone file but the data is still in the memory of the network analyser in the uni. It took me two diskettes to figure out that the floppy drive is broken. I'll have to repair it at some point.

Took a shot with my iPod, best I can do at the moment. I'm not too happy with the ripple.

Update: I recorded the 2-port touchstone data for this filter with 1600 points. You can download it as a zip file from my Google docs: NDK11F30D_matched.zip