Pages: [1] 2 3   Go Down

Author Topic: Kokusai mechanical filters  (Read 1380 times)

HAMHOCK75

  • Member
  • Posts: 1297
Kokusai mechanical filters
« on: November 25, 2019, 04:11:13 PM »

Kokusai mechanical filters are well known to have a problem that foam packing for the elements will deteriorate into a gooey mess.

This link shows how to repair them.

http://jlkolb.cts.com/site/koku.htm

My Yaesu FLDX400 transmitter has the same filter described in the link above. I took it out recently for examination.




I used a nanoVNA to scan the filter in both directions. Interesting that the characteristics are not exactly the same probably due to differences in the transducers. The results were a center frequency of 454.600 kHz, 6 dB bandwidth about 1.8 kHz, and an input impedance of about 900 ohms. The frequency accuracy of the nanoVNA was verified with an HP5340A counter. The two where within 8 Hz of each other. Also found that at 455 kHz, the nanoVNA can step in 1 Hz steps.




There is considerable mismatch loss from a 50 ohm source into a 900 load, and from a 900 source to a 50 ohm load. I calculated about 14 dB so the real passband loss of the filter is about 21 dB - 14 dB = 6 dB.

Some of the specifications for this filter can be found at this link but it is not clear what the expected loss from these filters should be by themselves. The "gain" they show is for a test circuit which includes the tubes.

http://www.kwradios.ngunn.net/IF%20FILTERS/Kokusai_Filter_Datasheet.pdf

This Collins catalog shows the FA-21 SSB filter can have an transmission loss of 9.5 dB.

http://jptronics.org/Collins/MECH_FILTERS/collins.filter_catalog.pdf



Logged

G3RZP

  • Member
  • Posts: 2254
Re: Kokusai mechanical filters
« Reply #1 on: November 26, 2019, 02:34:09 AM »

How accurate is the bandwidth, filter shape and ripple when measured in 50 ohms rather than in the designed impedance? I remember W6NL telling me that mismatching crystal filters can give a response with more loss but a flatter group delay characteristic.
Logged

W9IQ

  • Member
  • Posts: 8866
Re: Kokusai mechanical filters
« Reply #2 on: November 26, 2019, 03:15:08 AM »

One approach to get around this limitation in the lab is to use series resistors to correct the source and load impedances. After taking the measurements, the effect of these load resistors are mathematically backed out of the results.

The primary disadvantage of this approach is that your instant VNA plots need to be regenerated in Excel or similar to reflect the corrections.

- Glenn W9IQ
Logged
- Glenn W9IQ

God runs electromagnetics on Monday, Wednesday and Friday by the wave theory and the devil runs it on Tuesday, Thursday and Saturday by the Quantum theory.

VK6HP

  • Member
  • Posts: 1222
Re: Kokusai mechanical filters
« Reply #3 on: November 26, 2019, 05:40:54 AM »

Glenn

I've found that to be a practical approach.  For a quick result, I've not even bothered to apply point by point corrections - just a representative global shift to estimate whether the insertion loss is about what it should be.   For example, looking at a 6 kHz 'red series' Collins AM filter (50 k ohm input/output impedances) I just grabbed from the cupboard, I see that I tested it with 47 k ohm series resistors and tweakable parallel capacitors in the 130 pF region.  The saved response from the Rigol spectrum analyser/TG is almost copy-book.

Hamhock75, regarding your pictures, I was surprised to see that the Kokusai filter does not appear to be hermetically sealed.  Is that actually the situation?  If it is unsealed I imagine the ravages of time work faster.

I have not found any Collins filters that have failed spontaneously but some have been done to death by shorted blocking capacitors putting lots of dc through matching networks.  I can't quite bring to mind the receiver series where that was a problem but perhaps Carl recalls the details.

Incidentally, I notice that Owen Duffy is doing some interesting characterization of the NanoVNA: https://owenduffy.net/blog/?p=16193   Out of interest I bought a NanoVNA-F, the larger model chosen because the bigger display and more robust connector mounting work better for me.  My first impressions are mostly positive, but mirror some of the observations made by Owen about the smaller unit.

73, Peter.
« Last Edit: November 26, 2019, 05:43:53 AM by VK6HP »
Logged

W1BR

  • Member
  • Posts: 4422
Re: Kokusai mechanical filters
« Reply #4 on: November 26, 2019, 07:02:49 AM »

Some of the mechanical filters in my JRC-515ND suffered the same foam degradation.  Fortunately, they were repairable.

Regarding 50 ohm terminations,  as I recall the Collins FA series (used in S Lines) were close to being
50 ohms Z when used with a series resonating capacitor.

Pete W1BR
Logged

KM1H

  • Member
  • Posts: 11155
Re: Kokusai mechanical filters
« Reply #5 on: November 26, 2019, 05:03:38 PM »

Quote
I have not found any Collins filters that have failed spontaneously but some have been done to death by shorted blocking capacitors putting lots of dc through matching networks.  I can't quite bring to mind the receiver series where that was a problem but perhaps Carl recalls the details.

The R-390A built by several companies but using Collins mechanical filters. All the paper caps developed sufficient leakage to cause performance problems throughout but the one in the IF module killed filters.....usually all as the owner switched thru them.

Carl
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #6 on: November 27, 2019, 02:11:10 AM »

820 ohm resistors were added in series with the output and S21 input of the nanoVNA to match the approximate input and output impedance of the filter. Other than the expected additional mismatch loss, the center frequency, bandwidth, and general shape of the pass band did not change.

These MF-455-10AZ28 filters are, as VK6HP noticed, not well sealed. The metal cover simply has four metal tabs that fold over the plastic bottom. Regarding Owen Duffy's blog. I enjoy testing tube gear but have always been hesitant to connect up an expensive VNA due to the possibility of accidental damage from the voltages in tube gear, but due to the low cost of the nanoVNA, I feel much more comfortable doing that.

The carrier oscillator is at 453.5 kHz, the filter is centered at 454.6 kHz ( which is about 400 Hz low and out of specification by 100 Hz ) with a 1.8 kHz bandwidth ( the minimum bandwidth at -6 dB is specified at 2 kHz ) so the SSB audio passed is 200 Hz to 2 kHz. In spite of all the out spec conditions, it still works although the audio is a bit narrow.

It will be interesting to look inside next.
« Last Edit: November 27, 2019, 02:23:11 AM by HAMHOCK75 »
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #7 on: November 27, 2019, 11:46:29 PM »

Here are photos of the filter with the 820 ohm resistors added and the nanoVNA result.




Next begins the disassembly of the filter. Four tabs hold the cover on. Once those tabs are bent open, the top metal cover will slip off. Inside the metal cover is a cardboard piece. In the photos below, it is clear that the foam has become brittle and pieces come off just from the disassembly.




The foam is dry however until near the base. Some authors believe that the foam disintegrates into a gooey mess. A gooey mess was found at the base but I am not convinced that it is from the deterioration of the foam. It has the consistency of rubber cement and is not hard but is pliable. Based on the last photo the six center elements appear deliberately glued to the base with the end transducers floating.




Does anyone have any thoughts about that?
Logged

KM1H

  • Member
  • Posts: 11155
Re: Kokusai mechanical filters
« Reply #8 on: November 28, 2019, 07:37:33 AM »

Great and informative post.

Wish I had an answer to the floating question. What does the VNA do when you tap them lightly?  Maybe just some new foam will help.

In Collins filters Ive opened the goo is thruout and I use pieces of seat cushion foam to replace. Maybe the originals were glued in place and the resultant goo took decades to affect all the foam. Since there are numerous styles of their filters produced over many, many decades I can only speak to what I have actually seen.

Carl
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #9 on: November 28, 2019, 01:44:43 PM »

Hope all are enjoying a good Thanksgiving day.

The filter was tested with the nanoVNA after most the foam but not the cement at the bottom was removed. The results were the same as before. I decided to remove the cement. The following scan shows significant change. The center frequency has shifted up and is now in specification, the bandwidth increased to about 2.4 kHz and is also now in specification, and the insertion loss decreased.

As shown in the photo, the nanoVNA has a nice feature that if you select one marker, then another, it displays the frequency difference between them. In this case 2.4 kHz between markers 2 and 3. The scan below is with the filter directly connected to the VNA without any 820 ohm resistors.



The above image with pass band ripple is what had been expected from a non-Butterworth multi-pole filter design. To explore a bit more, the impedance varies considerably between peak ( 410 ohms ) and valleys ( 985 ohms ) in the pass band as shown in the images below.




I also explored how much pressure could be exerted against the filter elements before a noticeable change occurred which would help determine how much the foam could press against those elements. I pressed against the elements with my fingers and found they could tolerate more pressure than I expected. The harder I pressed the more the VNA display moved towards the images before the filter was restored.

I have begun exploring how to replace the foam. One obstacle was how to create an end cap for the filter where the wires attach. In the photo below, I found it was very easy to create a depression in a piece of foam for the end cap by pushing the tip of a soldering iron into the foam.

« Last Edit: November 28, 2019, 02:04:14 PM by HAMHOCK75 »
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #10 on: November 28, 2019, 06:19:34 PM »

Load resistors were added in these two photos after the filter had been reassembled with new foam. The first load is 680 ohms  which is actually 680 plus 50 ohms from the VNA for 730 ohms total. The insertion loss is mostly from the resistors. The second photo has load resistors of 360 ohms plus 50 ohms from the VNA for a total of 410 ohms.




In the first photo above, the measured input resistance is 1,320 ohms. Subtract 730 ohms for an actual input resistance of 590 ohms. In the second photo, the measured input resistance is 977 ohms. Subtract the 410 ohms for an actual input resistance of 567 ohms.
« Last Edit: November 28, 2019, 06:26:33 PM by HAMHOCK75 »
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Kokusai mechanical filters
« Reply #11 on: November 28, 2019, 09:16:53 PM »

HH75,

You beat me to it in suggesting some experimentation with the termination.  Often, the passband ripple (or variation) can be optimized by fine tuning the termination, including adjustment of a trimming capacitor.  Your filters are low impedance types and are probably a bit less critical but you could try some fine adjustments.  For example, here's the response of that Collins AM filter I mentioned earlier.  It would meet spec with just the recommended fixed parallel input capacitors but some adjustment makes for a much nicer response.



Ignore the details of the spectrum analyser settings - it was an early experiment with the instrument.

Regarding the attachments of the intermediate resonators, I can't offer anything beyond the obvious that it probably provided some form of soft boundary condition useful to the designers,

Knowing your interest in technology history, you might enjoy the story of Roshan L. Sharma, a young Indian graduate of UCLA in the early 1950s, who was hired by Collins at the time he was in danger of being deported.  His appointment was on the strength of a college course on the acoustics of tin plates.  His first task a few days later was to specify millions of dollars worth of lapping equipment for the new Collins filter R&D program.  Ben Stearns, in his book "Arthur Collins Radio Wizard" outlines the successful lengths that Arthur went to in order to keep Sharma, an undoubted genius, working for the company and the US.

73, Peter.

« Last Edit: November 28, 2019, 09:25:15 PM by VK6HP »
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #12 on: November 30, 2019, 02:49:58 AM »

Thank you for the name Roshan L. Sharma! His is a not only a fascinating story but it opened up a treasure trove of information about mechanical filters.

https://www.amazon.com/Roshan-L.-Sharma/e/B001K8CF4K%3Fref=dbs_a_mng_rwt_scns_share

With his name, I started following his patents which include a mechanical filter that resembles an assembly of flat plates and a patent on how to test mechanical filters to the earliest Collins mechanical filter patents. The first was filed by Melvin L. Doelz. They date from January 14, 1949.

Patent US2615981A.

https://patents.google.com/patent/US2615981A/en?oq=us2%2c615%2c981

Those early patents look remarkably like the Kokusai filter photos above. Collins and I believe Kokusai use a wire wrapped around a piece of nickel. Nickel has magnetostrictive properties which is how the process starts.

The next patent filed September 24, 1951 most resembles the Kokusai filter

Patent US2717361A

https://patents.google.com/patent/US2717361

August 14, 1952 a patent ( US2656516A ) regarding supporting means.

https://patents.google.com/patent/US2656516A/en

Roshan's patent ( US2887877A ) describing how filters are tested and an method for identifying problem parts in an assembled filter is filed August 31, 1954.

https://patents.google.com/patent/US2887877

This article may explain the difference between the high vs low impedance transducers.

http://www.rfcafe.com/references/electronics-world/mechanical-filters-electronics-world-april-1969.htm
« Last Edit: November 30, 2019, 03:11:16 AM by HAMHOCK75 »
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Kokusai mechanical filters
« Reply #13 on: November 30, 2019, 06:23:27 AM »

You're welcome, and thanks also for the links mentioning Sharma's various books.  The one that started life as an outline for Arthur Collins of Sharma's Hindu religious beliefs fits well with stories of Collins' interest in the beliefs of some other company engineers.  Collins was a complex character by all accounts and, along with great confidence in his own abilities (as one might expect), seems to have had a genuine interest in the beliefs of others. 

With regard to filters, Stearns writes that Doelz was the early driver from about 1946 and made many successful prototypes.  Sharma codified the design equations and understood how to reconcile theory and production practice. In the early days about 15 filters per month was the limit but later the monthly total was in the thousands.  Reading the history of the filter development and the larger push to make SSB practical and attractive, one can't help but be struck by the sheer amount of engineering talent and the capacity of Collins personally to direct it.

I found the RFCafe Southworth article the other day when looking around and agree it's an excellent summary of the state of play in 1969, which was I guess just about the zenith of mechanical filter technology.

73, Peter.
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Kokusai mechanical filters
« Reply #14 on: December 03, 2019, 10:16:01 PM »

There are a total of four mechanical filters in the Yaesu FRDX400 receiver for CW, SSB, AM, and FM.

Two are made by Kokusai and two are by Toko. The Kokusai filters are serviceable but the Toko units appear to be potted.




One of the Kokusai filter has the labelling scrubbed off but a test shows it is likely the same filter repaired above but with slightly narrower bandwidth and more loss most likely due to some degradation. The MF-455-03AZ27 is a 600 Hz filter. It shows the most degradation with more than 20 dB loss. It was disassembled and repaired the same as above. Below is a photo taken during the clean up process. It was very similar in condition to the above filter in that the foam was crumbly but for the most part intact. Again the center elements of the filter were stuck to the plastic base.



Instead of a photo of the nanoVNA, the software nanoVNASaver version 1.4 was used. The software has the advantage that it allows more types of output for S11. The capture shows S11 as an r + jx output while S21 is log-magnitude.



There may be some issues with the nanoVNA or the software in narrow sweeps like that above. The step size says 20 Hz but the display appears to be 200 Hz steps. The frequency scale does not read out properly

Version 0.1.4 of nanoVNASaver was used because the two newer versions 0.1.5 and 0.2.0 had issues on my computer with markers. With 0.2.0 an attempt to use or enter a marker frequency crashed the program. With 0.1.5, I could enter the marker frequency but the marker would default to the left side of the screen and could not be moved.

The filter itself appears to be now in good working order. The 6 dB bandwidth is about 600 Hz and the insertion loss in the above capture of 34 dB is about the same as the filter repaired earlier with 680 ohm resistors in series with the source and load.

Working on these filters was pretty enjoyable. They are much bigger than SMD parts and can take a fair amount of handling without damage.
« Last Edit: December 03, 2019, 10:20:53 PM by HAMHOCK75 »
Logged
Pages: [1] 2 3   Go Up