Comparison of Transmit Phase Noise With Various Oscillators on the TS-590S by Mark Goldberg, W7MLG.

Introduction


The transmit phase noise of a TS-590S was measured with a Perseus SDR, with various oscillators installed in the TS-590S. The standard Crystal Oscillator (XO), a Reference TCXO and three different TCXOs from eBay identified as “Cheap TCXO (M)”, “Cheap TCXO (A)” and “Cheap TCXO (I)” were tested. The phase noise of the bare TCXOs was also measured. 

Update December 2016: I was also able to borrow a real Kenwood SO-3 TCXO and make some of the same measurements with it.

If you are interested in the Reference TCXO, see Mark's TCXO Board.

Andy, K3WYC has done some testing with two TS-590s with various combinations of these oscillators that show similar results.
See Andy's TS-590 Oscillator Evaluation.

Transmit Phase Noise

Over a wide frequency range, 10 Hz to 1 MHz offset from a 14.1 MHz carrier, the TS-590S transmit phase noise is shown in Figure 1 for the four types of oscillator. This was measured using a narrow bandwidth on the Perseus. 

Update December 2016: I added the Kenwood SO-3 TCXO to this chart and made some measurement improvements at 10 and 100 Hz for all oscillators. Lower phase noise is better.


Figure 1: TS-590 Transmit Phase Noise Comparison 10 Hz to 1 MHz

Update December 2016: All four types match pretty closely at 100 Hz, 1000 Hz and above 100 kHz, possibly within the measurement error, but, when fitted with cheap TCXO (A), the TS-590S exhibits a phase noise bump at about a 10 kHz offset. The real SO-3 was better than the others at 10 Hz. The reference TCXO was a little better than the others at 1 MHz. This was further investigated using a wide resolution bandwidth technique with the Perseus. The transmit phase noise of the TS-590 when fitted with two samples of the cheap TCXO, (A) and (I), the Reference TCXO and the Standard XO is shown in Figure 2.


Figure 2: TS-590 Phase Noise Comparison 100 Hz to 100 kHz


The various traces in Figure 2 are as follows.

TS-590 Standard XO – This is the measurement, using the Perseus, of the TS-590S Transmit Phase Noise at 14.1 MHz using the standard XO. It is somewhere between the two ARRL measurements for the TS-590S and TS-590SG.

TS-590 Ref TCXO (M) – This is the measurement, using the Perseus, of the TS-590S Transmit Phase Noise at 14.1 MHz using a reference 15.6 MHz TCXO. It is virtually the same as the measurement with the standard XO. This confirms what certain experts tell me, that the master oscillator is not the major phase noise factor with standard synthesized transmitters. The design of the synthesizer is the major factor.

TS-590 Cheap TCXO (A) – This is the measurement, using the Perseus, of the TS-590S Transmit Phase Noise at 14.1 MHz using one sample of a cheap 15.6 MHz TCXO. There is a significant increase in phase noise between about 1 and 15 kHz offset from the carrier. Above that, it becomes very similar to the standard XO, suggesting that a bandpass filter in the radio comes into play.

TS-590 Cheap TCXO (I) – This is the measurement, using the Perseus, of the TS-590S Transmit Phase Noise at 14.1 MHz using a second sample of a cheap 15.6 MHz TCXO. There is the same significant increase in phase noise between about 1 and 15 kHz offset from the carrier.

ARRL Review TS-590S – These are the numbers pulled from the chart in the ARRL’s review of the TS-590S. This is here as a sanity check. The results above from the standard XO and the Reference TCXO are at or somewhat worse than this.

ARRL Review TS-590SG – These are the numbers pulled from the chart in the ARRL’s review of the TS-590SG. This is here as a sanity check. The results above from the standard XO and the reference TCXO are at or somewhat better than this. There is some concern why this is so much different than the TS-590S.



Bare TCXO Phase Noise

To check my results, I made additional measurements of the Bare TCXOs on the Perseus and on a very expensive ($100k USD) Agilent 5052B Signal Source Analyzer. I was able to borrow some time on one; no I do not have one! The results of these measurements of the bare TCXOs are seen in Figure 3.


Figure 3: Bare TCXO Phase Noise Comparison 100 Hz to 100 kHz

The various traces in Figure 3 are as follows.

Reference TCXO Perseus – This is the measurement, using the Perseus, of the bare Reference 15.6 MHz TCXO. This is significantly better than the TS-590S transmit phase noise measurements, again suggesting that the master oscillator is not the major phase noise factor with standard synthesized transmitters.

Reference TCXO 5052B – This is a measurement of the bare Reference TCXO using the 5052B. It can be seen that the 5052B has the capability to measure phase noise well below what the Perseus can do, but it costs 100X more. The reference TCXO is shown to have much lower phase noise than the Perseus is capable of measuring.

Cheap TCXO (I) Perseus – This is a measurement of one of the cheap TCXOs using the Perseus. It can be seen that this TCXO has a huge increase in phase noise up to 100 kHz. This is very unusual and a significant concern. The bandpass filter in the TS-590S is the only thing saving the rig from having even more phase noise at 100 kHz out and higher. For those that say you can’t design a bad TCXO, well, yes you can. The low frequency (100 Hz and down) measurements are as good or better than the others, which suggests that the basic crystal is good, the design of what is around it is likely the issue.

Cheap TCXO (I) 5052B – This is a measurement of one of the cheap TCXOs using the 5052B. It can be seen that this is very similar to the Perseus measurement, within about 3 dB. This confirms that the increase in phase noise is real. It also confirms that the Perseus is good enough to make measurements in this phase noise range, as it agrees with the device costing 100X more.

The measurement of these two devices, using two different measurement devices, shows an example of how phase noise measurement is very dependent on the measuring device. When you measure phase noise, you will get results that are dominated by the device under test, or the measuring device, whichever has the highest phase noise. The explanation of why is beyond the scope of this paper, but Google searches of “phase noise hf radio” and “phase noise Perseus” will provide you hours of reading.


Perseus Only Measurements


Figure 4: Bare TCXO Phase Noise Comparison 100 Hz to 100 kHz Perseus Only

In Figure 4 I measured all the TCXOs available to me on the Perseus alone. This is quick and easy to do, as I made up a connector to do so. It can be seen that all of the cheap TCXOs are pretty similar, and have significantly more phase noise than the Reference TCXO.

Example Plots From Measurement Devices

Example Perseus measurements of TX Phase Noise with the Reference and cheap TCXOs are shown in Figure 5 and Figure 6. There are some significant spurs that are similar with all the measurements, likely due to the radio or measurement system.

Example 5052B measurements of the bare Reference and cheap TCXOs are shown in Figure 7 and Figure 8. It can clearly be seen in this figure that the cheap TCXO phase noise goes up at 100 kHz. There are also a lot of spurs. Some of the spurs are the same in both measurements, due to the measuring system. Even $100k does not get you perfection, but some of the spurs are only seen on the cheap TCXO.

Update December 2016: I added figure 9 for the real Kenwood SO-3. It is better at 10 Hz than the others. There are a few spikes that I think are from the measurement system, but I can't be sure.


Figure 5: Example Perseus Measurement of TS-590 TX Phase Noise with Reference TCXO



Figure 6: Example Perseus Measurement of TS-590 TX Phase Noise with Cheap TCXO (I)



Figure 7: Example 5052B Measurement of Bare Reference TCXO



Figure 8: Example 5052B Measurement of Bare Cheap TCXO (I)




Figure 9: Example 5052B Measurement of Real Kenwood SO-3

Receiver Performance

MDS and RMDR Testing showed no difference between any of the oscillators in any of the tests when the TS-590S was doing down-conversion, but a significant reduction in performance with the cheap TCXO in the 6 meter band, where the TS-590S is doing up-conversion and does not have the narrow roofing filters. I only have one signal generator, so the one generator test method was used.

In the 20 Meter Band, With a 500 Hz DSP bandwidth (down-conversion with 500 Hz Roofing filter), MDS was measured at -138 dBm. Close spaced (2 kHz) RMDR was measured as -93 dB. Wide spaced (20 kHz) RMDR was measured as -117 dB. RMDR with a 15 kHz offset( matching the phase noise bump) was -116 dB. Switching oscillators made no measurable difference.

In the 6 Meter Band, With a 500 Hz DSP bandwidth (up-conversion with 2.7 kHz Roofing filter), MDS was measured at -141.4 dBm. RMDR with a 15 kHz offset (matching the phase noise bump) was -98.9 dB with the Reference TCXO. Switching to the cheap TCXO reduced the RMDR to 90.6 dB, a significant reduction in performance. I did not have time to do more measurements, but I expect that there would also be RMDR degradation on other bands and using other spacing and bandwidths where the TS-590s does up-conversion and does not have narrow roofing filters.

These results do not exactly match other's measurements, but that may be due to differences in procedure or equipment. The bottom line is that there is a degradation in performance when using the Cheap TCXO where the TS-590S does up-conversion. I did not have time to do enough testing to completely evaluate the degradation under all conditions.


Update December 2016: The Real Kenwood SO-3 is definitely the best one, at a price of course. The improvement at 10 Hz will not be significant to normal operation, as you do not normally operate within 10 Hz of another station, except possibly for a few low bandwidth digital modes.

Conclusions


It is apparent that there is a significant problem with all three samples of cheap TCXOs that were tested. They increase transmit phase noise by about 12 dB in the 15 kHz range offset from the carrier. This is likely to be detected by someone operating in close proximity. Receiver performance when using up-conversion was also degraded.

This is a test of only three samples, it cannot be assumed that all cheap TCXOs have this issue, but since they are generic with no source markings there is no way to tell if what you get is like this or not.

As stated earlier, the basic crystal seems to be OK, there is something in the TCXO design of the parts around it that is causing this problem. I did some more studies on a spectrum analyzer and found a lot of spurs on the cheap TCXOs and no spurs on the Reference TCXO.

On the cheap TCXOs there are spurs at 1.2 MHz intervals below the 15.6 MHz output frequency. There are also spurs at Spurs at 9.6 MHz, 38.4 MHz, 57. MHz, 76.8 MHz, none seen at 19.2 MHz. The 31.2 MHz is the second harmonic of 15.6 MHz and is expected. See Figure 9. This suggests that the cheap TCXOs are actually a small synthesizer, not a true TCXO at 15.6 MHz.



Figure 10: Spurs of Cheap TCXO


My best guess as to how the cheap TCXOs are made is as follows.

Start with a 19.2 MHz or 38.4 MHz TCXO (Various ones at these frequencies are $0.25 on Alibaba). These are very common frequencies, easily available, while 15.6 MHz is a custom frequency.

Divide by 16 or 32 to get a 1.2 MHz reference for one input of a PLL Phase Detector. (4040, 4060 ripple counters are $0.15 at Digikey, 4046 PLL is $0.21 at Digikey)

The 13th Harmonic of that is 15.6 MHz. Use a few more gates, flip-flops, counters to divide the output by 13.

Run VCO at 15.6 MHz, use the divided by 13 signal for other input of PLL Phase Detector.

The whole thing may be made for $2 or so including caps, resistors, etc.

It is actually a clever design if all you care about is price and if it is on frequency, and don’t care about phase noise.

It is pretty clear from all this measurement and analysis that a good TCXO likely cannot improve the synthesizer phase noise significantly. A bad one can make it worse.

I would like to thank the following people who assisted me with samples, guidance and a sounding board for my rants.

Adam Farson VA7OJ
Rob Sherwood NC0B
Andy Durbin K3WYC
Israel Vicente AD7ND
Bob Doyle WA3TGF
Mike Allred AE4G


Raw Data







Here are some of my other sites:

Modifications to my RV, including solar power and extra storage:                   https://sites.google.com/site/marksrvmods/
My Controleo2 based SMT Reflow Oven:                                                          https://sites.google.com/site/markscontroleo2build/
My TS-590S MODs including a buffer board install for a panadapter:              https://sites.google.com/site/marksts590smods/
My TCXO Boards to replace the SO-3 in Kenwood TS-590 radios:                   https://sites.google.com/site/markstcxo/
Modifications to allow use of an external clock in a Perseus SDR:                    https://sites.google.com/site/perseusmods/
How I use Spectrum Lab Software to do frequency measurements:                  https://sites.google.com/site/spectrumlabtesting/
Pictures I took of the 2017 Total Solar Eclipse from Menan Butte, Idaho:         https://sites.google.com/site/marks2017eclipsephotos/


Revised December 15, 2017