EMRFD Message Archive 12842

Message Date From Subject
12842 2016-05-18 11:19:08 Jim Leslie Crystal filter insertion loss
I used Xlad08 to design a 5 crystal 9MHz/500Hz/50 ohm CW filter - my first xtal filter. Testing it, I found
the following and not really sure about what I see or if I'm measuring it properly. With an HP8656B set to
an o/p level of -30dBM to the input to the filter on it's center frequency, I see 6.65mV on my HP3400A at the
filter input and 4.95mV at the filter output. I think this would be an insertion loss of -2.56dB. gpla.exe says
about -3.46dB so it seems close. However if I add 50 ohm termination at the output the IL is -8.82 dB (the filter i/p
drops to 5.8mV and the output drops to 2.1mV.)
One other test I did was to set the i/p to the lowest level I could and see if I could measure the output. The minimum
I could go to was -54dBM i/p to the filter and still easily see an output of the filter on my meter. So I think it works ok
but I wonder about which test is the correct way. I'm thinking terminated is? I was going to try another one for 500 ohms
next to compare results. Is -8.82dB loss considered high since glpa indicated about -3.46dB?
thx -jim
12843 2016-05-18 12:08:43 Bill Carver Re: Crystal filter insertion loss
12844 2016-05-18 12:32:21 Bill Carver Re: Crystal filter insertion loss
12845 2016-05-18 13:32:50 Jim Leslie Re: Crystal filter insertion loss
That I don't recall and is not in my notes but yes it is definitely possible - I tried so many things. Thanks for all the input and I will do a recheck this evening and try to make better notes.
This is all helpful and good to know GPLA is close to predicted. One thing I didn't mention in that when I was building, I measured each cap on a Digibridge and was
within a few pf of where everything should be at each stage. I also used shielding between stages. The first filter is always a challenge and was being very picky about everything. I just got an old HP VNA so will see what it looks like on that next; getting all that going now....
thx Bill,

12846 2016-05-18 15:46:23 Bill Carver Re: Crystal filter insertion loss
12851 2016-05-19 12:40:51 Jim Leslie Re: Crystal filter insertion loss
Thanks Bill. I tried a few more things to try to understand this and here is what I have found so far.

Connect the siggen o/p to 18" RG58 with a BNC "T" at the end and that "T" connected to another 18" RG58 going
to the 3400A & with 50 ohm termination at the meter.
I get 8.3mV out of the siggen as measured at the 3400A.

Then connect a 50 ohm termination at the 3400A ( I have 2 Tektronix BNC 50 ohm terminators and measured each
on a Fluke, Simpson and a Keithly to be 50 ohm +/- a couple milliohms so I trust these parts)
I then get 4.0mV siggen o/p as measured at the 3400A.
Remove the 50 ohm termination and connect the filter to the BNC "T" with the 3400A at the end of the cable.
No termination connected anywhere here, just the filter and nothing connected to the filter o/p either.
I measure 3.71mV at the filter input now, down from 4.0mV above. Small drop when connecting the filter.

Add a 50 ohm termination at the filter output and now measure 3.2mV on the 3400A at the filter input (down from 3.71mV).
Also now measure 1.11mV at the filter output with the 50 ohm termination now connected there for -9.27dB

Disconnect all termination and measure 3.71mV at the filter input and 2.76mV at the filter output. -2.57dB
This is somewhat closer to GPLA at -3.46dB
Lastly I put it on an HP8505A VNA and see IL of -10.2dB [manual and very slow sweep to measure peak]. Not too far
from -9.27dB measured above but still more, and about a dB.

I have a precision attenuator and have next to verify the VNA for accuracy. I think it's ok as I checked it
with surplus Drake LP filter and it matches it's specs printed on the side. But to be sure, I will recheck later.
I think the results should match a bit better than a full dB difference.

In addition to the 3400A I also have a Ballantine 323 RMS meter. I tested with both and they agree within a fraction of a mV with
each other so I think can believe both meters. Also checked on a scope w/10X probe. Same results in all cases so I think
my mV measurements are ok.

Now I may just go ahead and build another filter except a 500 ohm one and compare results. Just that I am really very curious about
why there is so much apparent IL with this one and I hate to quit.

I'm looking at all this as a learning experience and taking my time trying to be as methodical as possible.
And heeding the old adage of Bob Pease - if something looks funny, record amount of funny...
So if I did something above that looks funny, please let me know!
73 jim

12852 2016-05-19 13:01:16 Bill Carver Re: Crystal filter insertion loss
I "don't like" the amount of capacitance of all that coax, about 100 pF
including the 3400A. But I don't think that has anything to do with what
you're seeing, just being a curmudgeon.

Sanity check: get one 18" cable with the terminator at the 3400A end.
Plug that onto the signal generator set at 9 MHz and whatever amplitude
you want, say 10mV. Now put your precision attenuator between signal
generator and the cable-terminator-3400A. If you set it to 3 dB do you
get a 3dB lower reading of 7.07 mV on the 3400A? I can't imagine you
won't, but this is a sanity check, you're verifying the insertion loss
of the attenuator matches its marked value.

If that looks OK, now insert the filter in place of the attenuator. I'm
assuming a lot.....connectors on the filter......but in any case no
additional 18" cable. The only change you're making is to replace the 3
dB attenuator with the filter. Insertion loss is what GPLA predicted:
what do you see for the INSERTION LOSS of the filter?

If that doesn't look right, here's what I'd suggest: box up the HP8505A
and send it to me with the filter. I'll take a look at the filter and
send them back to you in a decade or two. How's that for a helpful offer?

12854 2016-05-19 16:30:35 Jim Leslie Re: Crystal filter insertion loss
Well I just discovered I don't have adaptors for the attenuator - bad news. However I built a 3dB one and it does work ok as checked on my scope and 3400A.
The capacitance is not an issue as I get the same results even without the cable - I do have a back to back BNC male-male adaptor - so I can easily connect it
direct 1 inch long.
Set the siggen for 10mV output and measure 10mV at the 3400 with the 50 terminator.
Insert the 3dB attenuator and measure 7.07mV pretty much exactly too or as close as I can tell with an analog meter.
So far, so good. Remove the attenuator and insert the filter. 3400 shows about 2.85mV about -10.9dB which is consistent
with previous. Same on other meter.

Time to rebuild. I'll try a 500 ohm one next and hopefully have better results to post.
A generous offer re: the 8505A! Don't know what to say. I will seriously consider it for say....oh I don't know, several years. :)

Thanks for all of your help!
73 - Jim

12855 2016-05-19 19:13:59 Bill Carver Re: Crystal filter insertion loss
OK James.....guess I'll have to just live with my N2PK and DG8SAQ VNAs for a few more years.

Well.....the caps are known quantities, the GLPA is (to me) a known modelling program, so the only thing left is the crystals. Or their characterization. Or, I suppose, someone mis-entering the xtal model into GPLA. As I said, I have had my share of blunders! But I also have a handful of successful filters, mostly at 4434 KHz, one 14 poles (I don't claim to be completely sane).

A 500 ohm filter will require a couple 50-to-500 ohm pads to the 50 ohm generator and detector. But that works.........



12858 2016-05-19 19:37:50 Dana Myers Re: Crystal filter insertion loss
12859 2016-05-19 20:15:54 Jim Leslie Re: Crystal filter insertion loss
There is a sequel to this saga; I had built another filter quite a while ago, 6 poles on this old one vs 5 on the latest. Both  min/loss 9MHz/600Hz/50 ohm but never was able to test it back then. I dug it out tonight and used the same crystals from the one I had been working on the past few days. Got 6.66mV in and 2.41mV out for -8.83dB IL. So just now I removed all the crystals & ran them all through my newly built motional parameter test jig and noted all the frequencies of each. I will start from scratch and re-do everything.
I'll be QRT for a while.

Those VNA's you have are great! Bet they don't weigh 100+ pounds too. I had to reinforce my bench for this thing.  :)
14 poles? Seems completely normal to me.
thx again -

12862 2016-05-19 21:45:34 Cecil Lindsey Re: Crystal filter insertion loss
I've noticed that you have been verifying your measurements in milli-volts levels as Bill suggested, i.e. 10mV less 3dB as 7.07mV etc. When you are looking for insertion loss you seem to have shifted over to dB reference. The HP3400 meter is scales in reference to 0 dBm (1 milli-watt @600 ohms). Remember dBm refers to power measurements referenced to the expected 1 milli-watt @600 ohms. However you have a 50 ohm termination on he input of your 3400, not the expected scale reference of 600 ohms. In the case of my HP400EL meter (and I suspect a similar value for your HP3400) requires the addition ~10.5dB to correct the error caused by the miss-matched 50/600 ohm termination. So if we take your -10.9 db and add10.5dB we get -.4dB which is more likely to be what you were expecting.

Hopefully this was more helpful, than confusing ;)

Cecil NA6RF

12863 2016-05-19 22:47:35 Bill Carver Re: Crystal filter insertion loss
The 3400A meter face says "DECIBELS, 1 mW 600 ohms".  So if you trying to measure absolute power then yes, the 3400A meter scales are only directly readable with a 600 ohm impedance. And would have to be corrected for use in any other impedance.

But Jim's not trying to measure "dBm", an absolute power referenced to 1 mW. He's taking two voltage measurements and wants to know the RATIO of those two measurements. Expressing that ratio on the common logarithmic scale gives "dB", not dBm. If the impedance stays the same then the difference between two decibel scale readings is in fact the insertion loss of the filter.

If you DO choose to correct the reading fine.....you would add 10.5 dB to the two power readings (the power in 50 ohms will be 10.5 dB MORE than the same dBm reading in a 600 ohm system.). The decibel difference between those two correct readings is unchanged by doing that correction. In fact, it doesn't matter what switch position you're in. Could be 1 mV full scale or 10 volts full scale. Could be 50 ohms or 600 ohms to 10,000 ohms. If the meter pointer drops 3 dB when you insert the "unknown", it has a 3 dB insertion loss in that 50, 600 or 10,000 ohm system.


12865 2016-05-19 23:14:21 Bill Carver Re: Crystal filter insertion loss
Your tenacity is commendable. And, as you probably already know, figuring it out will be worthwhile.....in the long run!

12869 2016-05-20 13:38:32 Jim Leslie Re: Crystal filter insertion loss
Thanks! Success is the ability to go from one failure to another with no loss of enthusiasm I suppose.
Well after much investigation I may have found the root cause but will not be sure until I rebuild it. I rechecked the motional
parameters last night and discovered that I did not allow for frequency tolerance between crystals. Plugging in the same
offsets into GPLA it confirmed my 10dB IL. So I am now using MeshTune and think I may be finally on the right track.
Now with the offsets and MeshTune calculated series caps, I see 2.22dB loss in GPLA and not 10dB anymore. This
was likely my problem all along but must rebuild it to know for sure. The good news too would be if this theory is correct
that my measuring approach is ok.

12870 2016-05-20 14:21:19 Cecil Lindsey Re: Crystal filter insertion loss
Thanks Bill for extracting me from my own confusion! Your well written reply has been printed and posted on my workbench, least I have another decibel relapse.