|
In the spirit of posting efforts currently underway, I have been working on a new RF board and associated driver software. In a nutshell, the new board uses the Skyworks SI5340 instead of the SI5351. This change addresses several issues. The SI5340 offers rock solid performance from 100 KHz to 250 MHz with LVCMOS outputs (employed by the prototype RF board), dramatically reduced phase noise and jitter, much improved clock symmetry (upon which the performance of Tayloe mixers depends), and much simpler application-level control software (no mucking about with the PLL, Multisynth, or phase at the application level). The new RF board replaces the 5351 with the 5340, and restores true split operation (the SI5340 has four outputs, three of which are used for RX, TX, and CWTx). The most challenging part of the software was the algorithm for setting output frequency, that is, given a desired frequency, find near-optimal numerator and denominator values for the various dividers used within the SI5340. My software employs a mediant-based algorithm attributed (incorrectly) to Farey, and efficiently produces results comparable to those produced by ClockBuilder Pro (the stand-alone program for configuring Skyworks parts). Current status:
1) Driver and driver example code complete and tested with main board and SI5340 EVB
2) T41 software (forked from V66.9) patched to ifdef choice of SI5340 or SI5351 RF Board compiling and in test
3) New RF board hardware under construction (PCBs back from China; parts on hand)
For the curious, prototype schematic and board pic are attached.
73,
John Bennett
AE0AM
|
Very interesting, John. Do you plan to release the Gerbers for the board and the supporting software? Please keep us posted on your progress going forward.
Jack, W8TEE
On Monday, May 12, 2025 at 02:12:04 PM EDT, John Bennett via groups.io <jkb@...> wrote:
In the spirit of posting efforts currently underway, I have been working on a new RF board and associated driver software. In a nutshell, the new board uses the Skyworks SI5340 instead of the SI5351. This change addresses several issues. The SI5340 offers rock solid performance from 100 KHz to 250 MHz with LVCMOS outputs (employed by the prototype RF board), dramatically reduced phase noise and jitter, much improved clock symmetry (upon which the performance of Tayloe mixers depends), and much simpler application-level control software (no mucking about with the PLL, Multisynth, or phase at the application level). The new RF board replaces the 5351 with the 5340, and restores true split operation (the SI5340 has four outputs, three of which are used for RX, TX, and CWTx). The most challenging part of the software was the algorithm for setting output frequency, that is, given a desired frequency, find near-optimal numerator and denominator values for the various dividers used within the SI5340. My software employs a mediant-based algorithm attributed (incorrectly) to Farey, and efficiently produces results comparable to those produced by ClockBuilder Pro (the stand-alone program for configuring Skyworks parts). Current status:
1) Driver and driver example code complete and tested with main board and SI5340 EVB
2) T41 software (forked from V66.9) patched to ifdef choice of SI5340 or SI5351 RF Board compiling and in test
3) New RF board hardware under construction (PCBs back from China; parts on hand)
For the curious, prototype schematic and board pic are attached.
73,
John Bennett
AE0AM
-- Jack, W8TEE
|
I like that part a lot and have used it for several things. ?The only downside side is that you must iss 7474’s to generate the quadrature, and while it covers the MF and LF frequencies in a better way, the 7474’s are not reliable above 30 Mhz (from my testing and basis their datasheet). ??
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch – K9ZC Staunton, Illinois ? Owner – Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On May 12, 2025, at 1:37?PM, John Bennett via groups.io <jkb@...> wrote:
? In the spirit of posting efforts currently underway, I have been working on a new RF board and associated driver software. In a nutshell, the new board uses the Skyworks SI5340 instead of the SI5351. This change addresses several issues. The SI5340 offers rock solid performance from 100 KHz to 250 MHz with LVCMOS outputs (employed by the prototype RF board), dramatically reduced phase noise and jitter, much improved clock symmetry (upon which the performance of Tayloe mixers depends), and much simpler application-level control software (no mucking about with the PLL, Multisynth, or phase at the application level). The new RF board replaces the 5351 with the 5340, and restores true split operation (the SI5340 has four outputs, three of which are used for RX, TX, and CWTx). The most challenging part of the software was the algorithm for setting output frequency, that is, given a desired frequency, find near-optimal numerator and denominator values for the various dividers used within the SI5340. My software employs a mediant-based algorithm attributed (incorrectly) to Farey, and efficiently produces results comparable to those produced by ClockBuilder Pro (the stand-alone program for configuring Skyworks parts). Current status:
1) Driver and driver example code complete and tested with main board and SI5340 EVB
2) T41 software (forked from V66.9) patched to ifdef choice of SI5340 or SI5351 RF Board compiling and in test
3) New RF board hardware under construction (PCBs back from China; parts on hand)
For the curious, prototype schematic and board pic are attached.
73,
John Bennett
AE0AM
<AE0AM_RF_Bd.jpg> <AE0AM_T41_RF_Prototype.pdf>
|
You know, the other benefit of this approach is that you can use the fourth output for RX2 and that could easily be a second RX chain placed as a daughter card. ?
Has me thinking. ?
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch – K9ZC Staunton, Illinois ? Owner – Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On May 12, 2025, at 1:37?PM, John Bennett via groups.io <jkb@...> wrote:
? In the spirit of posting efforts currently underway, I have been working on a new RF board and associated driver software. In a nutshell, the new board uses the Skyworks SI5340 instead of the SI5351. This change addresses several issues. The SI5340 offers rock solid performance from 100 KHz to 250 MHz with LVCMOS outputs (employed by the prototype RF board), dramatically reduced phase noise and jitter, much improved clock symmetry (upon which the performance of Tayloe mixers depends), and much simpler application-level control software (no mucking about with the PLL, Multisynth, or phase at the application level). The new RF board replaces the 5351 with the 5340, and restores true split operation (the SI5340 has four outputs, three of which are used for RX, TX, and CWTx). The most challenging part of the software was the algorithm for setting output frequency, that is, given a desired frequency, find near-optimal numerator and denominator values for the various dividers used within the SI5340. My software employs a mediant-based algorithm attributed (incorrectly) to Farey, and efficiently produces results comparable to those produced by ClockBuilder Pro (the stand-alone program for configuring Skyworks parts). Current status:
1) Driver and driver example code complete and tested with main board and SI5340 EVB
2) T41 software (forked from V66.9) patched to ifdef choice of SI5340 or SI5351 RF Board compiling and in test
3) New RF board hardware under construction (PCBs back from China; parts on hand)
For the curious, prototype schematic and board pic are attached.
73,
John Bennett
AE0AM
<AE0AM_RF_Bd.jpg> <AE0AM_T41_RF_Prototype.pdf>
|
Hi John-
?
That looks really excellent!
?
I note that you have independent outputs from the 5340 to the QSD and QSE.? That means there is a good chance that automated transmit calibration will work!
The exact mechanism for the current V12 design TX auto-calibration failure is unknown.? It is possibly an isolation issue, in which case there would still be an issue,
because the QSD and QSE are still in close proximity.? BUT if it is a different mechanism, then perhaps it has a better chance to work.
Also, since you have indendent control of the LOs to QSD and QSE, you can try the 48kHz offset rather than direct conversion.? That may be a big improvement in itself.
?
So this will be an interesting test!? It's also good news for T41-2, which uses a PLL module rather than integrating the function into one of the primary radio boards.
I did that with the intent of upgrading to a higher performance LO in the future.? The Si5340 may be it! ?Modularization is a good thing! ?
?
--
73 Greg KF5N
|
If the Si5340 has a mode with 180 degree outputs, then a very trivial change to the flip-flop circuits will allow divide-by-2.
That gets you to 6M with ease.? Also note that John is using the LVC variant, which has a specified higher upper frequency.
The other thing is the changes to the code for low-frequency operation are not necessary.? The requirement for quadrature code management in the other bands also goes away.
?
--
73 Greg KF5N
|
Just had a look at what the Si5340 is. It will have a place in high performance radios, but it's not going to replace the Si5351 in all applications. It's a much more expensive part; the cheapest version is over $16 in quantity one (the Si5351 is a bit over $2) and doesn't do fractional dividers so its frequency flexibility will be more limited. The least expensive one that does is over $21. It's in a 44-QFN?package that isn't as friendly to hand soldering as the Si5351 is (10-MSOP isn't ideal for hand soldering but 44-QFN is worse), and takes up more board real estate so it may not be well suited to small portable designs. Finally, it appears to be a much more power hungry device; the typical power consumption is listed on the datasheet as 680 mW. It requires both 1.8V and 3.3V power rails.
The Si5341 is basically the same device, but with 10 outputs. But not ten fully independent outputs, as there are only five MultiSynth blocks. The Si5340 has four, so each output can be programmed completely independently.
Doing an RF board with the Si5340 is very much in the spirit of the T41 project. We'll get to see how much the higher performance synthesizer makes in real world operation. Well done, AE0AM.
Very interesting, John. Do you plan to release the Gerbers for the board and the supporting software? Please keep us posted on your progress going forward.
Jack, W8TEE
On Monday, May 12, 2025 at 02:12:04 PM EDT, John Bennett via <jkb= [email protected]> wrote:
In the spirit of posting efforts currently underway, I have been working on a new RF board and associated driver software. In a nutshell, the new board uses the Skyworks SI5340 instead of the SI5351. This change addresses several issues. The SI5340 offers rock solid performance from 100 KHz to 250 MHz with LVCMOS outputs (employed by the prototype RF board), dramatically reduced phase noise and jitter, much improved clock symmetry (upon which the performance of Tayloe mixers depends), and much simpler application-level control software (no mucking about with the PLL, Multisynth, or phase at the application level). The new RF board replaces the 5351 with the 5340, and restores true split operation (the SI5340 has four outputs, three of which are used for RX, TX, and CWTx). The most challenging part of the software was the algorithm for setting output frequency, that is, given a desired frequency, find near-optimal numerator and denominator values for the various dividers used within the SI5340. My software employs a mediant-based algorithm attributed (incorrectly) to Farey, and efficiently produces results comparable to those produced by ClockBuilder Pro (the stand-alone program for configuring Skyworks parts). Current status:
1) Driver and driver example code complete and tested with main board and SI5340 EVB
2) T41 software (forked from V66.9) patched to ifdef choice of SI5340 or SI5351 RF Board compiling and in test
3) New RF board hardware under construction (PCBs back from China; parts on hand)
For the curious, prototype schematic and board pic are attached.
73,
John Bennett
AE0AM
--
Jack, W8TEE
|
Thanks to everyone for the fast feedback. To answer questions in order:
1) Yes, I plan to release hardware and software files to this site as others have done. If there is demand, perhaps Bill can add this RF board to his stock.
2) My experience with the 74ALVC74D is much better than Bill's (my observed performance > 100 MHz; clean power is critical); I expect the '3233's to poop out before the 7474's. That said, the next version will use the SI5340 LVPECL differential output option to drive TI CDCM1802's. This is a slick part that operates up to 800MHz, divides by 1, 2, 4, or 8, and offers an LVCMOS output up to 200MHz. If the mixer could support it, this would put operation from 2200M to 2M within easy reach (see below).
3) I currently use the Out0 for Rx and Out2 for TX. Having an unused output between TX and Rx during split operation is intended to cut down on cross-talk during split operation.
4) For the TX and RX clocks, I do indeed use the inverted Si5340 output capability. These outputs are *exactly* 180 degrees out of phase (at least beyond my ability to measure any difference with a 2 GHz scope) when terminated correctly. I think the external FF's are important, because phase symmetry is critical in a Tayloe mixer (which is switching FETs on and off at both edges).
5) The SI5340 is indeed more expensive, but the performance/cost benefit is pretty good (at least to my way of thinking).
6) Soldering a 44-QFN is a bit of a challenge, but a stainless stencil is ~$10, which makes the temp-controlled hot-plate soldering technique pretty reliable.
7) As you can see, I was able to replace the 5351 (and the MF logic) with a 5340 and all of its support circuitry. It's tight, but it all fits, and it's a clean layout.
8) Yes, power consumption is higher, but dependent on output configuration. I added the required clean 3.3 to 1.8V supply to the prototype RF board in the available space with no problems.
9) The SI5340 has four main versions. Two of these, the Si5340A-D (up to ~1GHz with some holes) and Si5340B-D (up to 350MHz; perfect for this application) do support both fractional and integer divides. The fractional divide feature is what makes the part a good fit.
?
Now I have a question. While I think replacing the current RF board is a win, I have been giving serious thought to creating a separate VFO card down the road. This would free up room on the RF board to do other things (like add an optional non-Tayloe second mixer (to support 2M). There are lots of options (old-school Minicircuits mixer, or something like the AD831) I am experimenting with both. I am also experimenting with replacing the Teensy with a Raspberry Pi 5, which would give HDMI outputs, and more performance and memory than the T41"X" is ever likely to need.
73,
j
|
Interesting points!? What I have observed with the 3253 is that degradation is noticeable at 10M.? Still good enough though.? It looks worse at 6M
What I am referring to is the 3253 in the role as a quadrature modulator.? Spurs start to creep up with increasing frequency.? Also carrier rejection
gets worse and worse.? My QSE2DC design has the capability to null the carrier, which mitigates this problem.? But I think this system will run out of
gas before it gets to 2M.? I would guess that performance as a demodulator is also falling off versus frequency.
?
Early in my experiments with the T41, I built a Gilbert cell based modulator using a very old part.? The MC1496.? This is one of the oldest semiconductor
devices still in production.? It worked, but it was already running out of gas at 10M.? I? haven't found a reasonable replacement for the MC1496 yet.
A basic Gilbert cell using modern devices at reasonable cost is what I am looking for.
?
--
73 Greg KF5N
|
Hi John
?
Looking forward to your performance results...
?
I haven't yet built an RF board, been stuck on toroids and other things, so would consider your RF board.
?
Regards
Phil VK2KKZ
|
Hi Greg
?
Is there any difference in performance of QSE2DC and the V12 circuit other than the nulling capacity? Nulling is a software task?
I notice your use of the differential "op amps" for the filters, is that a performance thing?
?
Regards
Phil VK2KKZ
?
On Tue, May 13, 2025 at 09:41 AM, Greg KF5N wrote:
Interesting points!? What I have observed with the 3253 is that degradation is noticeable at 10M.? Still good enough though.? It looks worse at 6M
What I am referring to is the 3253 in the role as a quadrature modulator.? Spurs start to creep up with increasing frequency.? Also carrier rejection
gets worse and worse.? My QSE2DC design has the capability to null the carrier, which mitigates this problem.? But I think this system will run out of
gas before it gets to 2M.? I would guess that performance as a demodulator is also falling off versus frequency.
?
Early in my experiments with the T41, I built a Gilbert cell based modulator using a very old part.? The MC1496.? This is one of the oldest semiconductor
devices still in production.? It worked, but it was already running out of gas at 10M.? I? haven't found a reasonable replacement for the MC1496 yet.
A basic Gilbert cell using modern devices at reasonable cost is what I am looking for.
?
--
73 Greg KF5N
?
?
|
Hi Phil-
?
I don't know if there is a performance difference between QSE2DC and the V12 circuit.? There is insufficient data available for comparison.
Yes, carrier nulling is a software task which has been available in T41EEE for quite a while.
?
With regards to the design using differential circuits, this was an attempt to gain all of the benefits of a symmetrical circuit, and
in particular noise immunity.? The T41, in its 3D printed plastic case, is lacking for shielding and grounding.? Now the problem remains
with the Audio Adapter, which has single-ended outputs.? What I've got now is working well enough to continue development.?
In the future, I would like to get a fully differential signal chain in both transmit and receive signal paths.
?
Another goal was squeezing the maximum potential performance out of the 3253 multiplexer.? This device is not specifically intended for
this application; it has been pressed into service as a modulator and demodulator.
?
There is more information on differential and instrumentation amplifiers in the references in the README for the QSE2DC:
?
--
73 Greg KF5N
|
No one seems to consider this but you COULD build analogs of these devices using high frequency discrete parts. ?I have something like breadboarded as a replacement for the mixer… it works but needs a lot more work. ?Its pretty simple. ?
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch – K9ZC Staunton, Illinois ? Owner – Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On May 12, 2025, at 6:08?PM, John Bennett via groups.io <jkb@...> wrote:
? Thanks to everyone for the fast feedback. To answer questions in order:
1) Yes, I plan to release hardware and software files to this site as others have done. If there is demand, perhaps Bill can add this RF board to his stock.
2) My experience with the 74ALVC74D is much better than Bill's (my observed performance > 100 MHz; clean power is critical); I expect the '3233's to poop out before the 7474's. That said, the next version will use the SI5340 LVPECL differential output option to drive TI CDCM1802's. This is a slick part that operates up to 800MHz, divides by 1, 2, 4, or 8, and offers an LVCMOS output up to 200MHz. If the mixer could support it, this would put operation from 2200M to 2M within easy reach (see below).
3) I currently use the Out0 for Rx and Out2 for TX. Having an unused output between TX and Rx during split operation is intended to cut down on cross-talk during split operation.
4) For the TX and RX clocks, I do indeed use the inverted Si5340 output capability. These outputs are *exactly* 180 degrees out of phase (at least beyond my ability to measure any difference with a 2 GHz scope) when terminated correctly. I think the external FF's are important, because phase symmetry is critical in a Tayloe mixer (which is switching FETs on and off at both edges).
5) The SI5340 is indeed more expensive, but the performance/cost benefit is pretty good (at least to my way of thinking).
6) Soldering a 44-QFN is a bit of a challenge, but a stainless stencil is ~$10, which makes the temp-controlled hot-plate soldering technique pretty reliable.
7) As you can see, I was able to replace the 5351 (and the MF logic) with a 5340 and all of its support circuitry. It's tight, but it all fits, and it's a clean layout.
8) Yes, power consumption is higher, but dependent on output configuration. I added the required clean 3.3 to 1.8V supply to the prototype RF board in the available space with no problems.
9) The SI5340 has four main versions. Two of these, the Si5340A-D (up to ~1GHz with some holes) and Si5340B-D (up to 350MHz; perfect for this application) do support both fractional and integer divides. The fractional divide feature is what makes the part a good fit.
?
Now I have a question. While I think replacing the current RF board is a win, I have been giving serious thought to creating a separate VFO card down the road. This would free up room on the RF board to do other things (like add an optional non-Tayloe second mixer (to support 2M). There are lots of options (old-school Minicircuits mixer, or something like the AD831) I am experimenting with both. I am also experimenting with replacing the Teensy with a Raspberry Pi 5, which would give HDMI outputs, and more performance and memory than the T41"X" is ever likely to need.
73,
j
|
Hi Bill, I would love to see your simple design. I too have thought about replacing the '3253 with discrete FET switches, but getting very fast switching speed *and* suitable RF performance appears pretty challenging (at least to me). Many parts don't even specify turn-on/off time, or when they do, it's way too slow. Commercial RF switches do not emphasize switching speed (no one cares about 10+ns switching speed in an antenna switch or video multiplexer). I have limited experience designing with GaAs parts, but that might be a possibility. A pair of Mini-Circuits mixers followed by well-designed diplexers is where I am leaning, although an AD831 design would be much simpler, but not as performant.
73,
john
AEOAM
|
John, look up UHFSDR (WB6DHW design ) for another approach. I built one, have?another in a partly built state. It works well, I get UHF performance. I paired it with an EdenDSP 3.0 board.?
Larry KB3CUF?
Hi Bill, I would love to see your simple design. I too have thought about replacing the '3253 with discrete FET switches, but getting very fast switching speed *and* suitable RF performance appears pretty challenging (at least to me). Many parts don't even specify turn-on/off time, or when they do, it's way too slow. Commercial RF switches do not emphasize switching speed (no one cares about 10+ns switching speed in an antenna switch or video multiplexer). I have limited experience designing with GaAs parts, but that might be a possibility. A pair of Mini-Circuits mixers followed by well-designed diplexers is where I am leaning, although an AD831 design would be much simpler, but not as performant.
73,
john
AEOAM
|
You gonna be at Hamvention?? If so, stop by Booth #2607.? Al may be there too depending when people come. ? ? Dr. William J. Schmidt - K9HZ J62K(J68HZ) 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ VP2EHZ ? Owner - Operator Big Signal Ranch – K9ZC Staunton, Illinois ? Owner – Operator Villa Grand Piton – J68HZ Soufriere, St. Lucia W.I. Rent it: ? Moderator: North American QRO Group at Groups.IO. Moderator: Amateur Radio Builders Group at Groups.IO. ? email:? bill@... ? ?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]> On Behalf Of John Bennett via groups.io
Sent: Tuesday, May 13, 2025 10:45 AM
To: [email protected]
Subject: Re: [SoftwareControlledHamRadio] New SI5340-based RF Board in development? Hi Bill, I would love to see your simple design. I too have thought about replacing the '3253 with discrete FET switches, but getting very fast switching speed *and* suitable RF performance appears pretty challenging (at least to me). Many parts don't even specify turn-on/off time, or when they do, it's way too slow. Commercial RF switches do not emphasize switching speed (no one cares about 10+ns switching speed in an antenna switch or video multiplexer). I have limited experience designing with GaAs parts, but that might be a possibility. A pair of Mini-Circuits mixers followed by well-designed diplexers is where I am leaning, although an AD831 design would be much simpler, but not as performant.
|
Larry, thanks for the pointer. The UHFSDR design is highly informative.
j
|
jjpurdum
K9HZ
Shirley Dulcey KE1L