TCXO upgrade
TCXO upgrade
Hi,
I am very new to the (tr)uSDX but already looking for ways to modify and tinker with mine. One upgrade I identified right off the bat is a TCXO upgrade. I know that the PCB is not set up to support one, but perhaps it can be bodged in.
One thing I discovered while exploring this possibility is that the PLL chip on my board is not a genuine Si5351, but a MS5351M. This article suggests that it performs well and is a suitable substitute, but it was an interesting discovery. It shouldn't have any impact on a TCXO mod.
I designed a small PCB (15mm x 10mm) that I plan to test for a TCXO mod. My thought is to remove the 27 MHz crystal and then use two of its four pads for the clock signal input (to Xa on the MS5351M) and ground. VDD for the TCXO will be connected at the cathode of D2 where the 3.3V supply for the PLL chip is generated. That diode pin and the two necessary pads from the crystal footprint are all in line on the main board. My small PCB has castellated vias that will be placed next to / above those pads. Connection will be made with a tiny bodge wire or perhaps just a blob of solder, depending on how things line up. The mod PCB will be glued to the top of two larger, nearby ICs for support. I don't like that very much, but there's really no where to secure the board otherwise.
The TCXO I picked is a 27 MHz +/-500ppb part from Digikey that costs about $3 USD. The entire assembled mod PCB would cost perhaps $4 USD in parts, ignoring shipping and such.
If my idea works out I will update. It probably won't in its initial form but tinkering is the fun part! Any input is welcome.
Thank you.
Dan Watson
N4XTL
I am very new to the (tr)uSDX but already looking for ways to modify and tinker with mine. One upgrade I identified right off the bat is a TCXO upgrade. I know that the PCB is not set up to support one, but perhaps it can be bodged in.
One thing I discovered while exploring this possibility is that the PLL chip on my board is not a genuine Si5351, but a MS5351M. This article suggests that it performs well and is a suitable substitute, but it was an interesting discovery. It shouldn't have any impact on a TCXO mod.
I designed a small PCB (15mm x 10mm) that I plan to test for a TCXO mod. My thought is to remove the 27 MHz crystal and then use two of its four pads for the clock signal input (to Xa on the MS5351M) and ground. VDD for the TCXO will be connected at the cathode of D2 where the 3.3V supply for the PLL chip is generated. That diode pin and the two necessary pads from the crystal footprint are all in line on the main board. My small PCB has castellated vias that will be placed next to / above those pads. Connection will be made with a tiny bodge wire or perhaps just a blob of solder, depending on how things line up. The mod PCB will be glued to the top of two larger, nearby ICs for support. I don't like that very much, but there's really no where to secure the board otherwise.
The TCXO I picked is a 27 MHz +/-500ppb part from Digikey that costs about $3 USD. The entire assembled mod PCB would cost perhaps $4 USD in parts, ignoring shipping and such.
If my idea works out I will update. It probably won't in its initial form but tinkering is the fun part! Any input is welcome.
Thank you.
Dan Watson
N4XTL
Re: TCXO upgrade
Hello Dan,
the mod sounds interesting. I'm suffering from frequency drift and would like to upgrade to TCXO as well.
My Hi-band version of the (tr)uSDX is using the FDT86256 as final MOSFET (Q5). The FDT86256 is located on the backside of the RF-board (compared to the 3x BS170s) and is directly facing the 27Mhz crystal.
Even with good efficiency (75%-82% on the 5 upper bands), the heat radiated by the MOSFET results in a frequency shift most pronounced when running FT8.
As a containment I experimented with a heat shield between RF- and main board and also tested a small fan. At the moment I run FT8 with the sidewall housings removed to get better heat removal and frequency stability.
See pic: https://www.qrz.com/db/K2RMM
So, a TCXO solution would be appreciated, please keep us updated.
73 de Jens DL2RMM (K2RMM)
the mod sounds interesting. I'm suffering from frequency drift and would like to upgrade to TCXO as well.
My Hi-band version of the (tr)uSDX is using the FDT86256 as final MOSFET (Q5). The FDT86256 is located on the backside of the RF-board (compared to the 3x BS170s) and is directly facing the 27Mhz crystal.
Even with good efficiency (75%-82% on the 5 upper bands), the heat radiated by the MOSFET results in a frequency shift most pronounced when running FT8.
As a containment I experimented with a heat shield between RF- and main board and also tested a small fan. At the moment I run FT8 with the sidewall housings removed to get better heat removal and frequency stability.
See pic: https://www.qrz.com/db/K2RMM
So, a TCXO solution would be appreciated, please keep us updated.
73 de Jens DL2RMM (K2RMM)
Re: TCXO upgrade
Hi Jens,
could you please be more specific about the frequency drift ?
When does it show ? Under what conditions ? I never experienced it myself, that´s why I´m asking.
Not even when TXing in WSPR mode for 2 minutes straight.
73 Mnauel; DL2MAN
Re: TCXO upgrade
Hi Manuel,
the drift only occurs when I transmit. When just receiving there is no drift at all.
Attached is an example of FT8 on 10meters today. When I start the 15sec transmitting cycles the waterfall shifts to the right. About 50Hz in 3minutes.
As mentioned before, I run the FDT86256 as Q5 on the backside of the RF-board. It is in close proximity to the 27Mhz crystal. Efficiency on 10m is close to 80%. The picture was taken when operating with the side walls removed. When adding my improvised "heat shield" between RF- and mainboard, the drift is cut in half. I did not recognize such a strong drift when I operated the Lo-band RF-board with BS170's.
73 de Jens, DL2RMM
the drift only occurs when I transmit. When just receiving there is no drift at all.
Attached is an example of FT8 on 10meters today. When I start the 15sec transmitting cycles the waterfall shifts to the right. About 50Hz in 3minutes.
As mentioned before, I run the FDT86256 as Q5 on the backside of the RF-board. It is in close proximity to the 27Mhz crystal. Efficiency on 10m is close to 80%. The picture was taken when operating with the side walls removed. When adding my improvised "heat shield" between RF- and mainboard, the drift is cut in half. I did not recognize such a strong drift when I operated the Lo-band RF-board with BS170's.
73 de Jens, DL2RMM
Re: TCXO upgrade
Hi Jens,
thanks for the Hint. I actually never checked it with FDT86256, in deed.
Always with BS170´s.
Will check that..... Yes, there´s still bugs to be discovered
73 Manuel; DL2MAN
thanks for the Hint. I actually never checked it with FDT86256, in deed.
Always with BS170´s.
Will check that..... Yes, there´s still bugs to be discovered

73 Manuel; DL2MAN
-
- Posts: 105
- Joined: 01 Jan 2022, 03:47
Re: TCXO upgrade
Do you have an OSH Park file for the board? And what are you going to use for C1andC2?
Also, the the processor seems to connect to two sides of the crystal with two lines but the oscillator only has one "out" point, so how do you hook it up?
Also, the the processor seems to connect to two sides of the crystal with two lines but the oscillator only has one "out" point, so how do you hook it up?
Re: TCXO upgrade
Hi, today I´ve tried out WSPR with my FDT86256 Rig, and could not see any drift at all.DL2RMM wrote: ↑11 Jul 2022, 17:48 Hi Manuel,
the drift only occurs when I transmit. When just receiving there is no drift at all.
Attached is an example of FT8 on 10meters today. When I start the 15sec transmitting cycles the waterfall shifts to the right. About 50Hz in 3minutes.
As mentioned before, I run the FDT86256 as Q5 on the backside of the RF-board. It is in close proximity to the 27Mhz crystal. Efficiency on 10m is close to 80%. The picture was taken when operating with the side walls removed. When adding my improvised "heat shield" between RF- and mainboard, the drift is cut in half. I did not recognize such a strong drift when I operated the Lo-band RF-board with BS170's.
73 de Jens, DL2RMM
drift when transmitting FT8 example 50Hz in 3minutes.jpg
That was with Full 5W Out, and as you know WSPR Transmissions have a duration of 2 Minutes:
Code: Select all
Timestamp Call MHz SNR Drift Grid Pwr Reporter RGrid km az Mode
2022-07-17 11:40 DL2MAN 7.040098 -1 0 JN59xh 5 S51RS JN86cq 430 131 2
2022-07-17 11:40 DL2MAN 7.040077 -12 0 JN59xh 5 OE9BKJ JN47ul 263 220 2
2022-07-17 11:40 DL2MAN 7.040073 +7 -1 JN59xh 5 DL9GCW JN48vc 208 231 2
2022-07-17 11:40 DL2MAN 7.040086 -28 -1 JN59xh 5 DF8OE JO42jr 440 331 2
2022-07-17 11:40 DL2MAN 7.040079 -20 0 JN59xh 5 SM7KHA JO65kv 735 4 2
2022-07-17 11:40 DL2MAN 7.040065 -23 0 JN59xh 5 SA6BSS/FLY JO68sc 983 5 2
2022-07-17 11:40 DL2MAN 7.040063 -22 -1 JN59xh 5 M0MVA IO81 1088 288 2
2022-07-17 11:40 DL2MAN 7.040064 +3 -1 JN59xh 5 DL4TOM JN59it 106 302 2
2022-07-17 11:40 DL2MAN 7.040073 -20 0 JN59xh 5 HA8HL/P JN96ov 601 114 2
2022-07-17 11:40 DL2MAN 7.040063 0 0 JN59xh 5 SQ3E JO71rq 362 42 2
2022-07-17 11:40 DL2MAN 7.040067 -16 -1 JN59xh 5 G4HZX IO91xk 882 290 2
2022-07-17 11:40 DL2MAN 7.040069 +1 -1 JN59xh 5 DJ6DK JN48fw 258 262 2
2022-07-17 11:40 DL2MAN 7.040098 -7 -1 JN59xh 5 SWL1/JO21 JO21jn 566 299 2
2022-07-17 11:40 DL2MAN 7.040079 +6 0 JN59xh 5 PH4RTM JO31ai 477 301 2
2022-07-17 11:40 DL2MAN 7.040086 -10 -1 JN59xh 5 DD5XX JN48ps 203 254 2
2022-07-17 11:40 DL2MAN 7.040064 -13 0 JN59xh 5 G4FLQ/1 IO95ga 1111 310 2
2022-07-17 11:40 DL2MAN 7.040067 -7 -1 JN59xh 5 F4VTQ JN28ux 456 268 2
2022-07-17 11:40 DL2MAN 7.040059 +9 -1 JN59xh 5 PI4THT JO32kf 482 314 2
2022-07-17 11:40 DL2MAN 7.040090 -16 0 JN59xh 5 DN4BAS JO62sq 391 16 2
2022-07-17 11:40 DL2MAN 7.040107 -10 0 JN59xh 5 PD1V JO21pi 525 298 2
2022-07-17 11:40 DL2MAN 7.040065 -23 -1 JN59xh 5 YO3WL KN34ah 1205 112 2
2022-07-17 11:40 DL2MAN 7.040067 -7 -1 JN59xh 5 DK8JP/1 JO31gk 451 304 2
2022-07-17 11:40 DL2MAN 7.040064 -13 0 JN59xh 5 G4FLQ/2 IO95gb 1113 310 2
2022-07-17 11:40 DL2MAN 7.040092 -10 -1 JN59xh 5 DB9SH JN58tc 137 190 2
2022-07-17 11:40 DL2MAN 7.040069 +8 0 JN59xh 5 DL4RU JN69cr 50 21 2
2022-07-17 11:40 DL2MAN 7.040066 -17 -1 JN59xh 5 DC7TO JO62qk 361 15 2
2022-07-17 11:40 DL2MAN 7.040163 -23 -1 JN59xh 5 MW0CWF IO81ik 1105 288 2
2022-07-17 11:40 DL2MAN 7.040067 -22 0 JN59xh 5 GM4SJB/H IO88ba 1417 319 2
2022-07-17 11:40 DL2MAN 7.040087 +1 0 JN59xh 5 PA0SLT/4 JO33kg 564 323 2
2022-07-17 11:40 DL2MAN 7.040111 -4 -1 JN59xh 5 DB9OH JO52ji 348 347 2
2022-07-17 11:40 DL2MAN 7.040067 +7 -1 JN59xh 5 DK8JP JO31gk 451 304 2
2022-07-17 11:40 DL2MAN 7.040069 +4 0 JN59xh 5 DL2JA JN58vk 98 187 2
2022-07-17 11:40 DL2MAN 7.040064 -27 0 JN59xh 5 M0LMH IO93gx 1059 304 2
2022-07-17 11:40 DL2MAN 7.040061 -28 -1 JN59xh 5 2E0DSS IO82xl 1041 295 2
2022-07-17 11:40 DL2MAN 7.040098 +2 0 JN59xh 5 DL0HMK JO54af 562 347 2
2022-07-17 11:40 DL2MAN 7.040064 -14 0 JN59xh 5 MM3NDH IO86ha 1270 312 2
2022-07-17 11:40 DL2MAN 7.040073 -3 -1 JN59xh 5 DK8JP/2 JO31gk 451 304 2
2022-07-17 11:40 DL2MAN 7.040069 -6 0 JN59xh 5 OK2IP JN89ef 321 90 2
2022-07-17 11:40 DL2MAN 7.040065 -9 0 JN59xh 5 SA7LIL JO66ib 752 4 2
2022-07-17 11:40 DL2MAN 7.040071 +11 0 JN59xh 5 DL/PA0EHG JO32sq 486 322 2
2022-07-17 11:40 DL2MAN 7.040064 +4 0 JN59xh 5 DC1RDB JN58ss 67 207 2
2022-07-17 11:40 DL2MAN 7.040065 -24 -1 JN59xh 5 HB9GVC/RX JN47ki 315 227 2
2022-07-17 11:40 DL2MAN 7.040061 -15 -1 JN59xh 5 2E0ILY/KIWI IO82qv 1091 297 2
2022-07-17 11:40 DL2MAN 7.040064 -15 -1 JN59xh 5 DL2ZZ JO31lo 436 308 2
2022-07-17 11:40 DL2MAN 7.040068 -16 0 JN59xh 5 GM4SJB/V IO88ba 1417 319 2
2022-07-17 11:40 DL2MAN 7.040069 -20 -1 JN59xh 5 PA3WLE JO22is 636 310 2
2022-07-17 11:40 DL2MAN 7.040068 +11 -1 JN59xh 5 DJ9PC JN59po 58 304 2
2022-07-17 11:40 DL2MAN 7.040067 +10 0 JN59xh 5 DK8FT JN58oe 137 204 2
2022-07-17 11:40 DL2MAN 7.040064 +1 -1 JN59xh 5 ON4WS JO20hi 538 285 2
2022-07-17 11:40 DL2MAN 7.040056 -1 -1 JN59xh 5 DG2NPE JN59tw 74 341 2
2022-07-17 11:40 DL2MAN 7.040050 -17 -1 JN59xh 5 DL5XL JO43gi 508 333 2
2022-07-17 11:40 DL2MAN 7.040071 -7 0 JN59xh 5 PA0RWT/3 JO33lg 561 324 2
2022-07-17 11:40 DL2MAN 7.040064 -8 -1 JN59xh 5 I0UVN JN61nl 876 174 2
2022-07-17 11:40 DL2MAN 7.040069 -24 0 JN59xh 5 G8NXD IO70je 1234 281 2
2022-07-17 11:40 DL2MAN 7.040068 +1 -1 JN59xh 5 DK6UG JN49cm 272 276 2
2022-07-17 11:40 DL2MAN 7.040065 +2 -1 JN59xh 5 PA3ABK JO21it 582 301 2
2022-07-17 11:40 DL2MAN 7.040071 +5 0 JN59xh 5 DF1DR JO31rm 402 309 2
2022-07-17 11:40 DL2MAN 7.040064 +8 -1 JN59xh 5 HB9TMC JN46lj 394 216 2
2022-07-17 11:40 DL2MAN 7.040065 -4 0 JN59xh 5 PA4HJH JO22su 596 314 2
2022-07-17 11:40 DL2MAN 7.040064 +3 -1 JN59xh 5 OE9TAV JN47tk 270 220 2
2022-07-17 11:40 DL2MAN 7.040083 -14 0 JN59xh 5 IZ6QQT JN63hr 623 175 2
2022-07-17 11:40 DL2MAN 7.040064 -14 -1 JN59xh 5 PA1JMS JO21cx 621 301 2
2022-07-17 11:40 DL2MAN 7.040064 -7 -1 JN59xh 5 IZ2442SWL JN45ns 447 209 2
2022-07-17 11:40 DL2MAN 7.040077 +4 0 JN59xh 5 ON4AWM0 JO20ot 510 292 2
Where did you get your Kit/Rig from. Maybe the selected crystal is just a bad one.
73 Manuel; DL2MAN
Re: TCXO upgrade
Hi Manuel,
my kit (=crystal) was part of the first german group buy (DC8LZ).
I continued some FT8 "drift testing" on the upper bands and it seems the drift gets stronger with higher frequency:
20meters => 2Hz/min
17meters => 3Hz/min
15meters => 5Hz/min (worst efficiency, stronger heat generation by FDT86256)
12meters => 4Hz/min
10meters => 8Hz/min
The testing was done with 15sec on/off FT8 cycles for 10minutes each.
The biggest "jump" happens after the first 15sec transmit cycles and stabilizes after ~5 min at 1-2Hz/min.
There is absolutely no drift when only receiving.
Do you have a high band kit with FDT86256 that you could test on 10meters (for comparison) ?
73 de Jens, DL2RMM
Re: TCXO upgrade
Once I test the mod and verify it works, I will post links and pictures. I don't want to recommend this until I know it works.Ohwenzelph wrote: ↑12 Jul 2022, 16:04 Do you have an OSH Park file for the board? And what are you going to use for C1andC2?
Also, the the processor seems to connect to two sides of the crystal with two lines but the oscillator only has one "out" point, so how do you hook it up?
The TCXO only has a single output, which will be fed into the XA pin of the Si5351. C1 and C2 are both 100 nF. C1 is just a bypass cap for the power rail, and C2 is a series capacitor on the clock signal line going into XA as recommended in the Si5351 datasheet. (The XB pin on the Si5351, the other pin used for a normal crystal, will be unused and left open as recommended in the datasheet)
I have the TCXOs on hand already and the PCBs shipped today. Looking forward to trying this out.
Re: TCXO upgrade
I've had some success with my TCXO upgrade board. There are a few things I want to revise, but the upgrade is installed and working well in one of my radios.
This is the assembled board. I ordered the PCB from OSHPark with their 0.8mm option, so the complete board is very low profile and poses no risk of shorting against the RF board when installed.

I tested the board before installing it in my radio. I verified the output waveform on an oscilloscope, a nice 1.2 Vpp clipped sine wave. I also measured the frequency on a calibrated counter. This TCXO is about 30 Hz or 1.1 ppm low. TCXOs aren't perfect but that's much better initial frequency accuracy than I got from the crystal.

Next step is removing the default 27 MHz crystal, Y1. This is the scary part of course but I got the crystal off without causing any damage. Labeled are the three pads that the upgrade PCB connects too for power, output and ground.

I used some double sided tape to secure the upgrade PCB to the main board, aligning the castellated vias with the three pads they connect to. I wanted something easily reversible in case I had to troubleshoot. I think some epoxy or hot melt glue could be used to further secure the PCB. Notice that the PCB is slightly crooked due to interference from U9. I will fix this in the next revision with a small cutout there.

Next I soldered the castellated vias to the three pads using small tinned wire.

Finally I powered on the radio and tested it with no issues. A frequency calibration value of 27000000 was perfectly acceptable, though 26999950 was slightly more accurate. This made sense given my measured value of the TCXO output before install. Either way, much better initial frequency accuracy than the crystal my board came with.
Of course the true value of the TCXO is stability over a wide temperate range. This will take some more testing and comparison of a stock unit versus this "upgraded" unit.
I need to revise the upgrade PCB before sharing it, but I'm happy with where I'm at right now. Do I have the only (tr)uSDX with a TCXO right now?
Who knows.
This is the assembled board. I ordered the PCB from OSHPark with their 0.8mm option, so the complete board is very low profile and poses no risk of shorting against the RF board when installed.

I tested the board before installing it in my radio. I verified the output waveform on an oscilloscope, a nice 1.2 Vpp clipped sine wave. I also measured the frequency on a calibrated counter. This TCXO is about 30 Hz or 1.1 ppm low. TCXOs aren't perfect but that's much better initial frequency accuracy than I got from the crystal.

Next step is removing the default 27 MHz crystal, Y1. This is the scary part of course but I got the crystal off without causing any damage. Labeled are the three pads that the upgrade PCB connects too for power, output and ground.

I used some double sided tape to secure the upgrade PCB to the main board, aligning the castellated vias with the three pads they connect to. I wanted something easily reversible in case I had to troubleshoot. I think some epoxy or hot melt glue could be used to further secure the PCB. Notice that the PCB is slightly crooked due to interference from U9. I will fix this in the next revision with a small cutout there.

Next I soldered the castellated vias to the three pads using small tinned wire.

Finally I powered on the radio and tested it with no issues. A frequency calibration value of 27000000 was perfectly acceptable, though 26999950 was slightly more accurate. This made sense given my measured value of the TCXO output before install. Either way, much better initial frequency accuracy than the crystal my board came with.
Of course the true value of the TCXO is stability over a wide temperate range. This will take some more testing and comparison of a stock unit versus this "upgraded" unit.
I need to revise the upgrade PCB before sharing it, but I'm happy with where I'm at right now. Do I have the only (tr)uSDX with a TCXO right now?

Re: TCXO upgrade
Very nice. Thanks for the update.
Being extra cautious, I'd be inclined to put a piece of Kapton tape on the top and bottom. Then there is no chance of anything shorting. But I have plenty of Kapton tape.
Being extra cautious, I'd be inclined to put a piece of Kapton tape on the top and bottom. Then there is no chance of anything shorting. But I have plenty of Kapton tape.
Re: TCXO upgrade
Just a thought - is it too much to ask if those three solder pads are 0.1" apart (or close enough)? If so, use through hole instead of castellated, and you can use a 3-pin header for a more secure connection.
Re: TCXO upgrade
The spacing of the castellated vias currently is 2.2mm, so just a bit narrower than 0.1"/2.54mm. I checked with a piece of header and 0.1" spacing would just barely work with the pads, hitting the outer edges of VDD and GND for soldering. The problem is, standard straight through header would add a lot of thickness to the board, and the overhang of the mod PCB with through holes instead of castellations would make it difficult to solder to the pads.
I might be able to do a piece of right angle header, hanging over the side of the board, and soldered surface mount style to minimize height.
Re: TCXO upgrade
Hey, N4XTL
Your idea is great, is your latest PCB revision ready?
Also can I ask about the role of C1 and C2?
73!
Your idea is great, is your latest PCB revision ready?
Also can I ask about the role of C1 and C2?
73!
-
- Posts: 105
- Joined: 01 Jan 2022, 03:47
Re: TCXO upgrade
"C1 and C2 are both 100 nF. C1 is just a bypass cap for the power rail, and C2 is a series capacitor on the clock signal line going into XA as recommended in the Si5351 datasheet. (The XB pin on the Si5351, the other pin used for a normal crystal, will be unused and left open as recommended in the datasheet)"