EMRFD Message Archive 11639

Message Date From Subject
11639 2015-09-21 23:57:46 kerrypwr Bridge-T Diplexer
I've built a bridge-T diplexer for 12.8 MHz; Q=10.

This is the schematic;

http://s8.postimg.org/uuj83l0ut/Diplexer_12_8_MHz_Q10.png

Initial tests of the entire module (mixer-diplexer-amplifier) showed less gain than I calculated; this was traced to the diplexer.

I set-up the diplexer in a fixture and tested it section-by-section; first the series-resonant section was installed, then the two 51R resistors were added and, finally, the parallel-resonant section was added to complete the device.

The parallel-resonant section was a little tricky as 62 nH is difficult to make accurately (the L & C values of this section are probably something like but not exactly the same as the design values) but I tinkered with the turns & their spacing on the toroid and eventually produced this sweep;

http://s17.postimg.org/ykjlcz1sf/Copy_of_Diplexer_12_8_MHz_Q_10.gif

 

It's fairly self-explanatory; the blue trace is the series-tuned circuit, the purple trace is the series-tuned circuit plus the two 51R resistors and the black trace is the complete diplexer.

The loss is low with the first two sections but increases greatly when the parallel section is added.

I then tested the parallel-tuned section on its own;

http://s15.postimg.org/ahxiyjiqz/Parallel_Tuned_Circuit.gif

 

Allowing for the disturbance to this rather critical circuit by removing it from the diplexer and putting it into a fixture, all seems OK.

Has anyone any comments on why the parallel-tuned circuit makes the attenuation so high and how/if this can be corrected?

Kerry VK2TIL.

 



11640 2015-09-22 00:15:12 Russell Shaw Re: Bridge-T Diplexer
11641 2015-09-22 01:50:24 kerrypwr Re: Bridge-T Diplexer
Thanks Russell.

The parallel-tuned circuit was in series in the test fixture for the sweep shown in the third image; there were no resistors or any other components present.

It therefore shows its highest impedance, at resonance, as a loss or a negative gain; its low Q would, I think, account for the relatively-small attenuation at resonance.

I did not find any resonances, although the small-value components were, as expected, sensitive to lead length.

Initial sweeps of the complete diplexer (the black trace) gave a two-peak response; one was the resonance of the parallel-resonant circuit and the other was at the 12.8 MHz centre frequency.

I "fiddled" with turns & spacing on the 62.2 nF inductor until I achieved the response shown by the black trace but the attenuation seems excessive to me.

I have studied Todd Gale's work (that has considerable input from Wes) on diplexers; it's wonderful work but it doesn't help me with this problem.

Kerry VK2TIL.
11642 2015-09-22 03:07:51 i7swx Re: Bridge-T Diplexer

Hi Kerry,


Why you inserted the diplexer after the mixer?

I guess it is to terminate the mixer output or to work as a resonant filter?


If the diplexer is to terminate the mixer IF to a wide band of frequencies, then I would suggest a Q=1. Attenuation is very low and no Tricky adjustments. Sporty adjustments should not give you more than 0.5 loss, fine tuning can help to 0.1/0.2dB


If you want to terminate and also work as a some bandpass filter, I suggest to select a Q of no more than 5, high Q means higher selectivity and higher attenuation.


Q=1  L=0.62uH C= 249pF


Q=5 Ls=3.11uH  Cs=50pF   Lp=0.124uH  Cp= 1244pF


73


Gian

I7SWX



11643 2015-09-22 03:20:20 kerrypwr Re: Bridge-T Diplexer
Thanks Gian.

Why does " ... high Q mean .. higher attenuation".  ?

Kerry VK2TIL.
11644 2015-09-22 04:49:14 Russell Shaw Re: Bridge-T Diplexer
11645 2015-09-22 05:05:20 kb1gmx Re: Bridge-T Diplexer
Check your math.  That 62nh seems excessively low for 12.8mhz.  Just by 
eyeball I'd think more like .62uH is closer to what is needed.  

Values I arrived at were 249pf and .62uH.    Try that.
The series leg runs more like 6.22uh and 25pf.  You may have to tweak
a little to allow for parasitics.  Those sounds reasonable for HF and 
comparable to those I've used in the past at 9 and 8mhz.

The idea is that 12.8mhz the circuit should be near transparent (low loss).
At either side of that looking into the circuit it should appear as a roughly 
50 ohm absorptive load (or at least stay inside the 2:1 circle on the 
smith chart). 

Allison

11646 2015-09-22 16:23:30 Dennis Czelusniak Re: Bridge-T Diplexer
Is this the calculator you used?
>>> http://www.changpuak.ch/electronics/calc_16a.php <<<

Just curious!

Dennis


--------------------------------------------
11647 2015-09-22 19:34:50 kb1gmx Re: Bridge-T Diplexer
I can't say yes to that one.  Had to look at it and played with it some.
Problem is that it does gigo (garbage in, garbage out).  Unrealistic 
values are the result.  That's numerically correct and totally nuts
on the practical level.    

I used Pickett, 12inch aluminum.  However a pocket calc works too.

That and a Q of more than 2 for functions like that is always problematic
as it produces unworkable values.

Allison


11648 2015-09-23 15:56:05 kerrypwr Re: Bridge-T Diplexer
Thanks to all for the replies.

I have used that on-line calculator as well as Todd Gale's downloadable one; the equations are in the pdf of Todd's fine website and also in EMRFD.

The equations are simple, almost mental arithmetic; all the above sources give the same results.

Component values, ie that tiny inductance, become more difficult as Q and frequency increase; for instance, this is the test result for a 5 MHz/Q=10 bridged-T diplexer I built a few years ago;

http://s15.postimg.org/kmv2z9umj/Diplexer_Response.jpg


http://s24.postimg.org/atulgci1x/Diplexer_SWR.jpg


Quite acceptable.


At 12.8 MHz that small-value inductor is equivalent to about 2 1/2 inches of wire in air so getting the exact value is a problem in a "real" circuit.


But I did that (in the test fixture at least, it may not be reproducible in the "real world") and the response as seen in my earlier post is good save for the loss.


None of my references addresses the loss of a bridged-T diplexer; one, I think, says that loss should be as little as 0.5 dB but that seems to refer to a Q=1 device.


My initial question remains unanswered;


"Has anyone any comments on why the parallel-tuned circuit makes the attenuation so high and how/if this can be corrected?".


Incidentally, I have moved from 5 MHz IF to 12.8 MHz because "good" crystals are very difficult to find; I think that applies world-wide these days but it's certainly a problem here in Australia.


I have some good 5 MHz ones still but have acquired a batch of 12.8 MHz ones that test quite well, albeit not brilliantly.


The higher IF has some advantages but ease of high-Q diplexer construction is clearly not one of them.  :)


I will probably try something like Q=5 for the diplexer; the inductor will still be small in value and I might wind it on a Teflon toroid for stability and adjustment.


Kerry VK2TIL.

 

11650 2015-09-23 17:55:01 K5ESS Re: Bridge-T Diplexer

>> I used Pickett, 12inch aluminum.

 

Ahh. A true Renaissance man.  I got a few stares and inquiry’s  when I took my Extra class two years ago using my bamboo Post Versalog.

Mike K5ESS

 

11651 2015-09-23 17:57:15 Russell Shaw Re: Bridge-T Diplexer
11652 2015-09-24 05:03:31 kb1gmx Re: Bridge-T Diplexer
Makes that renaissance woman.

I have the K&E bamboo as well.  Left over from the day you had 
to have one.  I moved early to pocket calc, but for some things the 
slide rule still allows me to better visualize the problem.  That and 
no batteries to fail, though neurons may be at risk. 

At higher frequencies the energy storage issue (Q) starts to 
be very visible.  Hence my comment on unworkable values.

Since the whole goal of the network is to route unwanted out of 
band energy into resistors high Q is not a requirement.  


Allison


11655 2015-09-24 09:20:40 kb1gmx Re: Bridge-T Diplexer
An alternate I've used is the BP-BS thats band pass and band stop diplexer.  
Tonne software (free) has the model for that and at higher frequencies the values
and loss model is more reasonable.



Allison

11656 2015-09-24 13:16:10 K5ESS Re: Bridge-T Diplexer

Whoops.  Sorry ‘bout that.  A little unintended gender bias creeping in.

Mike K5ESS

 

11657 2015-09-24 13:59:04 kb1gmx Re: Bridge-T Diplexer
No harm no foul.

I can come up with a list of names and people that are atypical
as in name gives a false clue.  

But in electronic communications this mode is too often clue free.
I'm used to it.


Allison
11658 2015-09-24 18:01:34 kerrypwr Re: Bridge-T Diplexer
The inductor is iron-powder; I used both T37-2 & T37-6 with the same (poor) results.

The point about inductance -v- resistance is a very interesting one that is borne-out by my test results.

I think that I will try a Q;5 design; Q;10 is just asking too much at 12.8 MHz.

The objective is, of course, maximum selectivity at this early point in the chain.

Thanks to all for the input; it's been a good learning experience that drove me to delve deeper into the operation of diplexers.

Kerry VK2TIL.
11659 2015-09-25 04:57:39 kiheiman_98 Re: Bridge-T Diplexer
Interesting discussion.  I have been working on a similar project to install a diplexer on the IF output of a MCL ADE-1 mixer in an Atlas 210X transceiver.
My goal was to reduce the number of spurs in the RX that were being generated in conjunction with an AD9951 DDS VFO.  I used a diplexer calculator from this site:

 

 
I selected a Q value that would allow me to use standard value silver mica caps.  I ended up with a Q of 3.9 and used T50-6 toroid cores.
The series resonant circuit was 5.67 uh and 140 pf.
The parallel resonant circuit was 361 nh and 2200 pf.
 
This circuit added about 0.5 db decrease in the receiver’s 10 db S+N/N and the number of DDS spurs were greatly reduced.
I would like to try some T37-61 and T30-2 cores per the W8 DIZ design:
 
 
73
Clint
W7KEC
11660 2015-09-25 05:19:12 kb1gmx Re: Bridge-T Diplexer
That's the problem.  Your asking for selectivity and not trying to terminate the mixer.

At 12.8mhz a Q of 10 is still a very broad bandpass.

The bridged Tee is for insuring the mixer has a good (less than 2:1) termination. If 
you need selectivity you may want to rethink that and use a different network.  At 
best it will have enough selectivity to attenuate any LO leakage assuming a 
reasonable IF.

More likely a band pass with band stop is more what may help you.

No matter what network you use more selectivity means generally more loss.


Allison
11661 2015-09-25 05:28:29 kb1gmx Re: Bridge-T Diplexer
First you have DDS with spurs in its signal that are maybe 60-70db down at best.  That you can't fix
with changes elsewhere.  The only fix is to use a better DDS.

A bridge Tee works fine with a Q of 1.  The IF filter is you selectivity so all its doing is 
terminating the mixer and keeping LO out of the amp before the filter.

DDS is nice tech but far from clean. The 9851 has a SFDR of only 43DB at 70mhz using a 
180mhz clock. That device 9851 is only a hair better than the old 9850.  Since then the 
tech has gotten better but a considerable amount.  But they still have spurs.


Allison

11662 2015-09-25 05:58:49 kb1gmx Re: Bridge-T Diplexer
Thinking on this and knowing the Atlas radio.

I"d bet the DDS was to fix the common complaint drift.  The problem is while DDS doesn't drift
it has spurs, and not just a few.  So the problem is to fix one problem you introduce another.
Its important to note the Atlas had a very high dynamic range for its day and using a weak
DDS very definitely degrades the radio by introducing spurs.


Not the answer you'd like but read the 9851 datasheet and ap notes.

Allison
11663 2015-09-25 07:50:40 Lasse Moell Re: Bridge-T Diplexer
The proposed chip, 9951 do have a spurious free dynamic range close to 90 dB up to 1 MHz from the carrier. This is at higher frequencies than used in the Atlas, so it seems realistic to approach close to 100 dB at the needed frequency! Not bad at all, given the very nice phase noise too! Draw-backs? Yeah... needing a 400 MHz clock with good p/n for a starter. And likely need for a bandpass filter bank to clean up far out spurs. Still should put the old Atlas in high performance class :D

I bet the old vfo/xo mixing scheme does not offer a dynamic spur free  range of 100 dB...

/Lasse SM5GLC

25 september 2015 14:28:28 +02:00, kb1gmx@arrl.net [emrfd] skrev:

 

First you have DDS with spurs in its signal that are maybe 60-70db down at best.  That you can't fix

with changes elsewhere.  The only fix is to use a better DDS.

A bridge Tee works fine with a Q of 1.  The IF filter is you selectivity so all its doing is 
terminating the mixer and keeping LO out of the amp before the filter.

DDS is nice tech but far from clean. The 9851 has a SFDR of only 43DB at 70mhz using a 
180mhz clock. That device 9851 is only a hair better than the old 9850.  Since then the 
tech has gotten better but a considerable amount.  But they still have spurs.


Allison




11664 2015-09-25 07:54:42 kb1gmx Re: Bridge-T Diplexer
Yes the 9951 is very good, but compare to the 9851 used... not nearly as good.

The SPC SFDR at 70mhz is plain terrible for most RX/TX applications at -43dbc.

If you turn that around you need a SFDR approximating the dynamic range of 
the receiver or something over 80db for simple radios and higher for radios with
dynamic range approaching 100db or more.  If not the spurs are then hearable
as they are well above the noise floor.


Allison
11665 2015-09-25 08:13:46 Clint Re: Bridge-T Diplexer
For the 9951 in the 210X, I basically copied the design of the 9951 in the Argonaut 539.  I do have a bandpass filter (8 to 24 Mhz) in the output that feeds the mixer (same as the Argonaut).  My unit has a 100 Mhz osc that is multiplied to 400 Mhz internally.   I also have a 3 db attenuator at the LO input.  My 210X has more spurs that the Argonaut, but it is very usable.  Most of the spurs are masked by the antenna noise.  Also, most of the spurs can only be heard when using a 10 Hz tuning rate.  When using a 100 Hz or 1 Khz tuning rate, the spurs are not heard, even though they are still there.
 
And I did go with the 9951 to get a more stable VFO.  The 210X was terrible, even after a two hour warm-up.  I now see a max of + – 10 Hz change throughout the day.  Another benefit to performance is that the 2nd and 3rd harmonics on the old 210X VFO were only down 10 to 20 db from the fundamental.
 
Here is a list of radios that use the AD9951 (there are probably more):
 
Kenwood TS590S
Yaesu FTDX5000
Icom IC7000
Icom IC7100
Icom IC7200
TenTec 588 Omni VII
 
I did not check the schematic to see if they used diplexers (other than the Argonaut 539).
 
73
Clint
 
 
 
11666 2015-09-25 08:51:07 kb1gmx Re: Bridge-T Diplexer
Me culpa, next time wear glasses while reading.  read 9951 as the more common 
and no so good 9851.

Most did or selectivity or relied on mixer balance to keep the LO from going downstream.


Allison
11667 2015-09-25 09:03:21 kb1gmx Re: Bridge-T Diplexer
More... and more general.


The Atlas is one of many radios that were for the day better than some around them in 
the area of overload performance.  It was also very small.  Because of that its a hard 
radio to truly improve as good has to be very good.  

Alas atlas used a band switched VFO that even for its day was terrible for drift.
It would have been better if they did a low VF and premix but space and switching
and room for the needed filters for that was lacking.  Fixing the VFO is an area
needed but attention is needed with regard to unintentionally generating spurs
is required or otherwise there can be new problems.

Another area with that radio is the finals.  The radio was fairly rugged and did great 
for power but the power amp stability was marginal on some and that tended to kill 
finals which for us now means subbing or redesign.  The final problem plagues 
many radios of the day as that was what it was for the at time and they are gone.

I'm looking to find a ugly one with blown finals that modding is not seen as a 
bad thing.  then I may look at solutions to those two problems.

To me on the VFO DDS is a possible solution, but being a bit biased away from them
I would first look at other paths.  Finals a new module using new board and old metal
might be a path using latest power MOSFETs.

Allison

11668 2015-09-25 18:28:30 n2msqrp Re: Bridge-T Diplexer
I always like compact transceivers and was impressed when the 210 radio came out.

My friend NO2K had one of those sets. I remember drift being in issue. I was considering stabilizing the drift with a "Huff and Puff" frequency stabilizer circuit. Have you though of using one these stabilizing circuits?

Herb Johnson tried to develop a third generation set, the 310, which had a digital frequency display. I've read that Herb also considered using DDS technology but the design of the 310 never worked out. Hams who put down a deposit lost their money.

I have a introductory spec sheet for the 310 which includes a history of the Atlas Radio Company. Here is an excerpt:

"In 1974, Herb started his second company and named it Atlas Radio, after the 1924 vintage diesel engine in the 1924 motor vessel, "Westwind", owned and skippered by his friend, Don Gumpertz, K6OF). Atlas introduced the first really successful all solid state transceiver (no tubes). In his design he had the valuable assistance of Les Earnshaw, founder of Southcom International.

Mike N2MS

----- Original Message -----