开云体育

David Platt,

First the input has a L-network coil plus input C of the Jfet.
So the input impedance is likely under 2000 ohms maybe 75 ohms.

The gray areas on the drawing modules are triple tuned band
pass circuits, a real pain to dial them in but better than commonly
used transformers (only double or single tuned) of the day.
You will need a output pad as that likely has gain in excess of
18-40DB to protect the VNA input. Also a attenuator at the input
as over driving it will cause errors. Marantz did a lot of neat and
odd things over the years.

The 10.7 ceramic filters (not the narrower crystal filters) are typically
300 or 1500 ohm load.

Ceramic filter do not like DC applied to them ( lifespan issue).

The usual way to jig them is two back to back broad band transformers
to go from 50 to user defined impedance and the exact same back to
50 ohms. Hooked up on a board the networks can be "calibrated" for
loss and band width.

The filters are inserted and tested. Generally they are much wider than
the desired FM bandwidth so the critical item is center frequency.

Allison
----------------
No direct email, it goes to bit bucket due address harvesting in groups.IO

Here's a few graphics from the Murata Ceramic filter manual. The first one shows the change in response, both amplitude and group delay, versus impedance load. The next one is Murata's suggested filter test setup.
Bob

Right.
Small comment: Be careful on layout around the in/out pins on the filter. I can see on your picture you might have quite some capacitance from in/out pins to ground.
Normally ground is keept away also on inner layers.

Dave,

Some folks are getting all worked up over the small stuff.

Reminds me of what I was taught in the Air Force:? measure with a micrometer, mark with a grease pencil,

cut with a chain saw.

I remember a trip I took for several weeks to VA to stay with my grandson while his mother jetted off to EU

for 10 days.? I brought my handheld 2M, but could not hit the repeaters inside with the rubber duckie.? Found

some #22 wire, made a dipole for 2M, thumb tacked it to the wall up as high as I could, made "coax" with

the #22 as twisted pair.? Worked like a champ!

73

Glen K4KV

In that respect, after working decades above 50 MHz, 630 and 2200 meters is
a relief. Clip leads are not even seen by the RF energy.

Dave - W?LEV

--

*Dave - W?LEV*
*Just Let Darwin Work*
*Just Think*

I solved the problem like this.

Aha! Great - I had wondered at the time I read it about the power capacity of the N type. Since I don't have any, though, I wasn't motivated to research it. Glad to know the truth now - thanks.

I know a lot of older (tube-type) test equipment used twin banana plugs. Maybe the development of o'scopes that reached higher frequencies caused the development of the shielded coaxial connector.

And I'm not sure I should even post this, drifted so far off topic. I do think the original question was important, and the info about HF measurements not being so critical of lead lengths for calibration.

Doug

On Tue, Mar 3, 2020 at 04:12 PM, David Platt wrote:

This will give you a high S11 return loss, and
provide plenty of signal to drive the JFET. You _might_ need to pad down the
NanoVNA drive signal to keep from overloading the JFET and saturating the
whole IF chain.
I wouldn't personally want to try fiddling with those filters without a _lot_ of
careful thinking up front... the alignment procedure is probably a whole cave-full
of hungry bears. Mark the starting positions carefully!!!

As you say, fiddling with something this complex is unwise; it's said they
had to invent a stereo sig-gen good enough to get the lowest distortion and highest
stereo separation it could achieve over the full audio spectrum.

I want to look at the 19, but reserve an 18 for tweaking, one I did more than 30 years
ago. At that time anything worth listening to was over 150 miles away, and it needed
better adjacent channel selectivity. I narrowed the IF substantially using its built-in
scope, which displays bandpass shape and (with a constant scope horizontal gain) was
useful to gauge the change. Steep sides and flat top over the IF passband for a wide
range of input signals - the only way to gauge flatness of each stage, as they
gain-compressed.

When I look
at the next stage of the schematic (the limiter assembly) I see this signal
driving right into the base of a BJT, whose emitter is AC-bypassed to
ground...

So, you have that schematic? Let me know if not; I have service manuals
for both the 18 and 19.

A safety "gotcha" here, though... this IF strip has some gain in it due to the
input and interstage-buffer transistors! The signal coming out of the output
will probably be larger than the signal going in - I'm not sure how much.

Agreed! I'm most concerned about the VNA output amplitude, which almost
certainly far exceeds even the strongest off-the-air signal. Pad input and
output, then reduce as required.

Appreciate your feedback, Dave.
--
I_B_Nbridgema

Hi John,

This matter has already been discussed by those skilled in the art of calibration.
What's important is that you can use it in short waves, but you said both 2M and 70cm.
Therefore, I would recommend that you purchase a N dad calibration kit.
I bought it on aliexpress for 18 usd at a reasonable price here:

It's cheaper now :D

I hope I could help.
73, Gyula HA3HZ

Newbie question. I run 2 x 10cm sma pigtails from my nano, onto an external frame. They connect to a back to back SMA <> N Socket. My reason for this is I will use the nanovna-f outdoors a lot, connecting 10mm dia and up coax, so I'm not keen on direct connection to an SMA, with or without pigtails.

Question is, for calibration, can I just use the 3 x SMA calibration connectors by placing these onto an SMA <> N Plug adapter on the other side of the frame, or will the additional distance inside the adapter affect this?

I'd be mainly using this in HF and 2m if this is more critical at higher frequencies, although ideally, I may also need to use on 70cm at times as well.

73
John MM0CCC

I found a copy of a 1998 Murata databook for these sorts of filters. One set of graphs shows the effect of having source/load impedances which diverge from 330 ohms. Higher impedances seem to create the sort of effects we can see in my most recent graphs above... gain droop on the high-frequency side of the curve, and a group delay curve where the low-frequency peak is higher than the high-frequency peak.

I don't know yet whether the filter is seeing an impedance with an R that's too high, or one with an X that's too far from zero... could be both of course.

I've moved the case over to Cults for a paltry donation($2) to help with my time and costs. You can print this in any material you want. TPU is a softer material with a cushion.

I have the Murata data sheet here, and for a 180 ceramic filter, your group
delay plot looks normal. Murata does not show data beyond +/- 150 kHz, though,
and the y axis scale is 1 dB / divison, 6 dB total, so be careful making
comparisons to other plots not scaled the same. I can post that pdf page for
the GDT filters, showing a small version of all the plots, in you want.
PS thanks for the heads up to see what's going on here Dave.

I'd very much appreciate postings of the PDFs of the filter data sheets, Bob. A lot of the information on the older parts is getting hard to locate.

I have a few Model 18s and a Model 19 receiver. The Model 19 schematic is

attached; note no transformers. What do you make of it?

The 18 has a BJT input stage vs. the 19's JFET, and the 18 has no AGC for its
passive front end.

Oh, that's a beauty! Individual IF filters tunable inductors and caps - five adjustments per section, each. That really is something.

I wouldn't personally want to try fiddling with those filters without a _lot_ of careful thinking up front... the alignment procedure is probably a whole cave-full of hungry bears. Mark the starting positions carefully!!!

The good news is, this is going to be pretty easy to drive and measure, I think, as long as you look at the whole IF strip as a unit, because you're not actually driving any of the filters directly. The input side of each filter bank is driven from its own transistor amp and isolated from whatever you might do outside, and all but the last of the filter bank outputs is also isolated from the outside.

Take a look at the "IF input from front end". This has a very high input impedance... 470 kOhms, in parallel with the input impedance of the Q301 JFET (which will be extremely high at DC, and somewhat capacitive at 10.7 MHz). The L332 input inductor (between the jack and the gate of the JFET) might be there to cancel out the capacitive reactance... I haven't crunched the numbers to see.

Anyhow, I suspect that the best way to drive this input is to use a simple three-way T connector - NanoVNA drive port on one leg, the tuner's IF input connected to the other, and a good 50-ohm resistor connected to the third to terminate the drive signal. This will give you a high S11 return loss, and provide plenty of signal to drive the JFET. You _might_ need to pad down the NanoVNA drive signal to keep from overloading the JFET and saturating the whole IF chain.

On the output side... well, this may be a bit trickier. R317 (coming out of the last filter) is 2.7k, so the filter won't see less than that. When I look at the next stage of the schematic (the limiter assembly) I see this signal driving right into the base of a BJT, whose emitter is AC-bypassed to ground... so, the input impedance of the limiter is going to be quite low (a few tens of ohms at a guess - I'd have to evaluate the limiter schematic and figure out what sort of current they're running through its first transistor). So, odds are, you could just run this output into the 50-ohm input of a NanoVNA's second port, and the impedance would be OK.

A safety "gotcha" here, though... this IF strip has some gain in it due to the input and interstage-buffer transistors! The signal coming out of the output will probably be larger than the signal going in - I'm not sure how much. You don't want to risk overloading the bridge and mixer in your NanoVNA's second port. It would probably be wise to stick something like a 20 dB 50-ohm attenuator pad between the IF strip output and the NanoVNA input at first... do a sweep and see what your peak gain is. Then, consider reducing the amount of padding until you're close to a 0 dB insertion loss.

Fortunately there isn't any DC imposed on either the input or output of this IF strip (according to the schematic) so you don't need to worry about that... I don't think any of the +/-12 can get through to the analyzer in either direction...

... unless you've got a fault on the board. Wouldn't hurt to isolate both the input and the output from the NanoVNA with .1 uF caps, just to be safe.

This receiver really is a beautiful example of a fully-discrete transistor design from its era.

Do you know a Marantz receiver/tuner model number? Is a schematic available?
or part numbers for the transformers?

I have a few Model 18s and a Model 19 receiver. The Model 19 schematic is attached; note no transformers. What do you make of it?

The 18 has a BJT input stage vs. the 19's JFET, and the 18 has no AGC for its passive front end.

--
I_B_Nbridgema

I have the Murata data sheet here, and for a 180 ceramic filter, your group delay plot looks normal. Murata does not show data beyond +/- 150 kHz, though, and the y axis scale is 1 dB / divison, 6 dB total, so be careful making comparisons to other plots not scaled the same. I can post that pdf page for the GDT filters, showing a small version of all the plots, in you want.
PS thanks for the heads up to see what's going on here Dave.
Bob

There is for sure a problem with this measurement. When the two GD-peaks in
pass band looks like this something is wrong. That is confirmed by the pass
band curve that is too un-symmetric - rolling off in pass band.

Thanks, SJ - I rather suspected something of the sort.

Something that's certainly going on here, is that the measurement plots are for the test-jig-plus-filter, not just for the filter... I calibrated the "through" signal using a direct connection between the two cables. So, the behavior of the filter, and the behavior of the jig with its L-match and pad, are being conflated together.

I rather strongly suspect that reactance from the L-match is at least partly to blame. The match might be perfectly resistive at the design frequency, but with a Q of around 2.8 it's going to be significantly capacitive off to one side and signlficantly reactive to the other side. This is very probably affecting the phase quite a bit and could account for the asymmetries we see.

As a short-term tweak I'll try re-running the calibration, over a narrower frequency span, with the "through" cal being done through the whole test jig (with the filter position shorted). Then, see what a measurement of the filter looks like. It may be better but I don't expect really great results.

Although the L-match looks good enough for coarse sorting of these filters (to find the center frequency) a better matching arrangement is going to be required for reliable phase measurements. It'll need to be a broader-band match.

My next thought is to grab a couple of ferrite beads and some magnet wire, and wind up a couple of bifilar 4:1 broadband transformers and make another jig. One of these on each end, plus a 130R in series with each, ought to let the filter look into the impedance it wants without significant reactance appearing in the pass-band. The S11 return loss won't be wonderful, but probably not terrible either (about 1.55:1 SWR) and likely not enough to seriously affect the measurement.

Great to hear it works for you!
OTG cables, which are typically microUSB, always have pins 4 & 5 connected at the micro plug end.
Pin 5 is GND and when 4 is pulled to low on a device that supports OTG, its interface switches to USB host mode and supplies +5V at 500mA max.
As for the USB-C socket on your Nano - there have been a number of comments about the quality of the part and my unit has the same issue.
One user found it was the cable that was shipped with the unit - they bought a good cable and the problem went away.
Cheers,
Larry

UPDATE

Got it working. Looks like the usb-c socket on the nano may be a
little sloppy.

First, I used the USB Serial Console app and was able to send
commands that worked such as "help" and "pause" and resume". So OK
this is communicating.

Then I opened the NanoVNA android "webapp" app, and it indicated it
was connecting but wouldnt respond, moving the unit I noticed it
asking to connect again, and figured out that (I think) the usb-c port
on the nano, or possibly cable end, is sloppy.

Played with it and now the android app runs the nano.

INTERESTINGLY, the web site "app" approach still generates the "fail
to claim interface". But, who cares since the android app now works.'

I'll try another USB cable before exchanging the nano itself.

Lessons learned: mainly, the OTG cable must be on the phone, not the
nano. They dont work both ways. also, the console is a big help to
test communications because u know instantly if it is responding.


thanks to all who helped

I'm not sure whether this filter was specified to be linear-phase, but I believe that was probably part of its goals. It's an IF filter specified and sold for use in FM stereo tuners. In these signals, the peak FM signal deviation is 75 kHz. The baseband signal extends out to about 53 kHz (L+R audio bandwidth is nominally 15 kHz and there's a double-sideband-modulated L-R subcarrier centered at 38 kHz). So, Carson's Rule suggests that we'd want at least 128 kHZ of bandwidth on either side of the carrier. More bandwidth and more-linear phase behavior results in better demodulation of the FM signal, and thus lower distortion (particularly in the stereo L-R signal).

Yes, I plan to re-sweep these filters with a narrower bandwidth, and multiple sweep segments in order to get a more accurate look at the passband behavior.

I played around with a spreadsheet copy of one sweep's S2P data last night - wrote the formulae to plot the amplitude, and much of the math to calculate the group delay. I need to implement the phase-unwrapping step next... less easy in a spreadsheet than in something like Python. In any case I should soon have a package put together to go from an S2P file, to a nice dual-axis Gnuplot PNG file.