开云体育

Hello

I do have some experience with both the SI5351A in the shape of the RFzero: www.rfzero.net and the AD9850 and AD9912 used in the Next Generation Beacons platform www.rudius.net/oz2m/ngnb/dds.htm

The AD9850 is a truly first generation DDS and is, to be honest, not good for anything but experimentation. NEC used an AD9850 in one of their receivers but disabled it above 5 MHz due to spurious. The AD99-series are, as mentioned, much cleaner but also in a different price range. I believe that most of the DDSes can do phase shift.

The Si5351A can also do phase shift but much more static. For a low cost and fair performance VFO you can easily setup a SI5351A with no, 90 deg.(I/Q but lower freq is about 4,7 MHz without under clocking) or 180 deg phase shift.

All this being said please also take a look at the close in spurs which many tends to neglect.The close in performance is much more important than the harmonics.

73
Bo, OZ2M

Bob:
? The Si570 is very popular.? Low phase noise and the non-cmos options(LVECL, etc) will go up to 1400 MHz.? The Si564 just came out.? Lower noise than the Si570 and covers 100 KHz to 3 GHz.

Dave - WB6DHW

While the 9850is common thee are other AD part like the 9851 and many more
all more costly but cleaner.

The 5351 with the right programming for the microprocessor can do ritty direct
by simply changing programmed frequency it woud require the micro to monitor
an IO pin with the serial data. No need for two VFOs to do that.

PSK with inverter however is not workable do to clicks and spikes. It still
requires a SSB radio to get get the cosine amplitude moduleation.

Allison
--
Please reply on the group, no private emails we as a group get to share info that way.

I know about the AD9850 and SI-5351.

I like the SI-5351 because it can do 3 VFOs.

You could use 2 VFOs for RTTY.

I think you could do PSK with a simple inverter. (180degree phase shift)

Are there other viable DDS options?

Bob Macklin
Seattle, Wa

I've also built dozens of these over the decades. From a lot of experience and a lot of LTSpice modeling, I've come to the conclusion that frequency matching is the most important crystal parameter by a wide margin. For SSB-bandwidth filters especially, if you get the crystals frequency matched to roughly 5% of the filter bandwidth (less than 100 hz), then you should be able to get a reasonably good passband, assuming you match the filter impedance properly. This latter step is probably the most important implementation parameter after crystal matching. I generally don't even measure Q any more; the most pronounced affect that it has on performance is thru loss.

My frequency matching procedure uses a stable station receiver that has warmed up for a while. I build a simple logic-gate oscillator with a header-pin socket to test the crystals. I run one of those audio spectrum analyzer programs on my computer connected to my station receiver audio output with the span set to about 500 hz and the center frequency around 1 Khz. I put a crystal in the oscillator and tune the receiver to a 1 khz beat note as indicated on the spectrum analyzer display. Call this the reference. Then, without touching the station receiver frequency, go through the remainder of the crystals, observing their frequency relative to the reference crystal's frequency on the spectrum display. Notice that just handling the crystals produces a noticeable change in frequency, so you might want to wear cotton gloves.

I prefer the "QER" filter topology (Google it) as it is much flatter in the passband and is much more forgiving than traditional typologies, but it does require 2 additional crystals in the end sections. It also has equal value inter-crystal capacitors, which should be matched for best results. You can determine the optimum termination impedance by putting variable resistors between your 50 ohm test signal source and the filter terminations Adjust these pots for best flatness, measure them and add 50 ohms to determine the optimum filter termination impedance. I usually build tuned L-networks for minimum loss matching.

Joe
W3JDR

Thanks for posting and sharing, Nick. Some good info.

72, Paul NA5N

OK, after saying

"Oh, I did a presentation on some of this stuff at OzarkCon a few years
back. My presentation can be found here. It gives some of the references I
mention above and describes techniques. "

I forgot to past in the link. Here it is:



More 72,

Nick, WA5BDU

On Wed, Jul 31, 2019 at 10:56 AM Bob Macklin <macklinbob@...>
wrote:

How do you do the crystal testing?

Does Elecraft sell sets of matched crystals?

Bob Macklin
K5MYJ
Seattle, Wa.
"Real Radios Glow In The Dark"
-

All,

The goals of testing are to:
Get matched crystals,
Remove those with low Q or other defects,
Create a group that are more or less the same.
Do it simply, but good enough to get reasonable and consistent results.

Elecraft may sell matched set for repair purposes but not standalone, Intek sells simple filter kits
and I believe Diz sells matched crystals. I'd bet there are others. You have to look!

That said, I've been making filters for many decades. My approach is buy a dozen or more
cheap rocks (for larger filter I may need as many as 10 matched) and test them for
frequency in a oscillator circuit (simple).

My experience for the last 30 years if I'm paying more than 1$ a a microprocessor rock its too costly!
Those are mass produced in a wide range of frequencies and typically 20 of them now are between
39 and 69 cents each. What to do with the excess, consider this a good commercial filter, is typically
100$ maybe less if removed from a radio. The excess crystals are cheap enough to consider as trash
or use for other projects like VXOs.

Those that match to 100hz are set aside for further testing. A high rock and a low rock are marked
(for BFO use). Of the frequency matched crystals I screen for Q, low Q rocks are also set aside
for wider filters or other uses. Up to this point there may have been failures due to damage or
excessively low Q. Those are discarded. Hopefully I get enough crystal for the filter planned
and usually not less than 4 and often 6 or more. Why more? I've found the common everyday
filter of 4-5 crystal tend to have poor (2:1 or worse) skirt selectivity for SSB bandwidths.
However the simpler 4 crystal filter should if done right be an adequate performer.

Testing used only simple fixtures and one of the high rocks for a VXO to sweep the filter with
a adjustable but decently stable signal. These days a DDS or SI5351 NCO is a good way too.
The only gear of significance is a frequency counter good enough to measure to the nearest
1 or 10hz. The rest is calculator or spread sheet and my time.

Having a spectrum analyzer (or network analyser) is nice but I've not yet built a filter that was
better than the hand tested ones. Just prettier pictures.

In the end its not rocket science. It should not be!

FYI this is for the common ladder filter, I've also made half lattice filters
to 8 poles and that is a bit more demanding due to crystals of multiple
frequencies plus required tuning.


Allison
---
Please reply on the group, no private emails, we as a group share info that way.

I currently use a system called PHSNA. It's been described in QQ and has a
Yahoo discussion group with a lot of files and descriptions.

There are much simpler ways though, generally starting with some methods
described by Hayward in QST and in one or more ARRL QRP books.

You generally need a stable and clean signal source that can be controlled
to 1 Hz. A PVXO by K8IQY published on one of the QRP quarterlies works
well. I've used it as well as a DDS and a MultiPig PLL.

Then there's a crystal fixture that typically puts the crystal in a 12.5
ohm environment and uses either transformers or resistors. It's pretty
simple. They outboard end of this is terminated in 50 ohms and then
measured with something like a diode RF probe, oscilloscope or log power
meter.

There's also the shifted frequency method by G3UUR which is pretty simple
but I'm not sure it gives Q & Rs. I think Chuck Adams may have worked that
out though, in a recent QEX article. Or maybe it was an entirely different
technique.

Oh, I did a presentation on some of this stuff at OzarkCon a few years
back. My presentation can be found here. It gives some of the references I
mention above and describes techniques.

I'm not sure if Elecraft sells matched crystals or not. Having the
frequencies well matched is a good starting point, but knowing that the
crystals came from a batch that generally have good Q values would be
important too if you're not going to do the measurements.

You can build good filters without doing all those measurements (except
matching) if you know they come from a good batch.

73-

Nick, WA5BDU

How do you do the crystal testing?

Does Elecraft sell sets of matched crystals?

Bob Macklin
K5MYJ
Seattle, Wa.
"Real Radios Glow In The Dark"

I got my 50 ea. 4.9152 MHz crystals a day or so ago so I've fired up my testing stuff and did measurements on ten of them.

Average series resonance was 4,913,519 Hz.

Q values went from 18,000 to 103,000, so a bit low but with a lot of variation. The average Q was 57,000.

I'd be interested in comparing my results with anyone else's measurements.

72-

Nick, WA5BDU

Soda Pop II sold

--
N8DAH

Shop.Kit-Projects.com ( HTTPS://Shop.Kit-Projects.com )

Kit-Projects.com ( )

HT20 Sold pending funds
Soda Pop II still up
TNC still up accepting offers

--
N8DAH

Shop.Kit-Projects.com ( HTTPS://Shop.Kit-Projects.com )

Kit-Projects.com ( )

Tuner sold

--
N8DAH

Shop.Kit-Projects.com ( HTTPS://Shop.Kit-Projects.com )

Kit-Projects.com ( )

--
N8DAH

Shop.Kit-Projects.com ( HTTPS://Shop.Kit-Projects.com )

Kit-Projects.com ( )

Due to financial reasons these items must go:

4SQRP Hilltopper 20 with LCD: 110$ shipped

SodaPopII with case ready to go 80$ shipped

4SQRP 4S tuner 50$ shipped

MFJ-TNC x 1270-x with 2 yaesu cables 110 shipped

Private msg me and good on the zed for email
--
N8DAH

Shop.Kit-Projects.com ( HTTPS://Shop.Kit-Projects.com )

Kit-Projects.com ( )

I mailed an envelope with a flat 5x7 inch piece of cardboard and an embedded transistor, total weight of 0.7 ounces.

The charge was $0.55 for first class mail PLUS a surcharge of $0.15 for special handling.

The PO does sell a 70 cent "forever" stamp...

-Diz

No, I was just curious if you had any problems with the clerk at the PO trying to charge more than 70 cents.

Ray,
AB7HE

Yes...you need one?

Diz,

Did the PO let you mail the envelope for 70 cents?

Ray,
AB7HE