SI4732-A10 VCO Detuning

[G8PTN]

Hi,
For general information, an example of the SSB pitch (tone) change I have observed on the ATS-Mini when using a whip antenna and touching near the base is shown below.

The effect is more noticeable on the higher frequency bands (10MHz and above). For example, listening to RTTY on 10.1MHz, the change
is evident.

I believe that the internal VCO operates at a high frequency (maybe GHz as implied in one of the documents) and the synthesiser uses this to generate the clocks used by the I/Q mixer.

I think stray capacitance coupling to pins on the SI4732-A10 (or the device itself) can cause the VCO to change frequency. The momentary change in VCO frequency means that the synthesised (divided) clocks and demodulated audio tone in SSB will also change.

From the “spectroid” App waterfall the behaviour is as follows:

• Adding capacitance moves the VCO in one direction.
• The PLL then brings the VCO back to nominal frequency.
• Removing capacitance moves the VCO in the other direction.
• The PLL then brings the VCO back to nominal frequency.

Since the SI4732 was never really designed for SSB, the frequency changes would not cause any issues for AM and FM reception.

Whilst the effect is not a major problem, I have not noticed it on other hardware that uses the SI4732-A10. This may be due to the circuit arrangement and/or lack of screening.

73, Dave

[G8PTN]

Hi,
Other users have now started to notice the effect that was identified some time ago on the SI4732 Mini Receiver hardware.
https://t.me/talkradio/174172/202729

Initially, I suspected the XTAL was moving frequency, but I do not believe this to be the case, since the same effect occurs with an external RCLK to the SI4732, and the momentary frequency shift occurs as described above.

As stated previously, I think the effect is related to VCO detuning in the SI4732-A10, perhaps due to the single antenna arrangement and lack of screening.

On other receivers that use the same chip, they are generally in metal enclosures and/or have some sort of isolation between AMI and FMI inputs (e.g. RF switching, RF filters), and do not exhibit the same issue from an external point of view.

I did remove the lid on an ATS-20+ and the same effect occurs when there are stray capacitance placed near the SI4732-A10 or its pins.

73, Dave

[max-arnold]

Found some experiments with adding a screen:
https://t.me/talkradio/174172/204333
https://t.me/talkradio/174172/204721

[G8PTN]

Found some experiments with adding a screen:

Hi Max,
Thanks for the links. The headphone amplifier version may have an influence, but I do not believe it will be the cause.

The ATS20+ shows the same behaviour if you remove the lid and move you hand near the SI4732-A10 or RF input filters.

I still think the effect is due to detuning of the VCO inside the device, which operates at a high frequency.

Both ATS-MINI units that I have are V1, and show the same detuning effect when touching the antenna. It is only noticeable when listening on SSB, because the VCO frequency change results in the change in tone on the audio.

73, Dave

[DB9JG]

Hi,
I own hardware versions V1 and V3.
Both exhibit the detuning effect:
V1 shows detuning:
• When touching the AM input or FM input
• When bridging the gap between the telescopic antenna and the SMA socket ground
V3 shows detuning:
• When touching the FM input
• When bridging the gap
• NOT(!!) when touching the AM input
In V3, a FET now decouples the AM input from the antenna, while the FM input is still connected directly to the SMA socket. This suggests that a load change at the FM input is likely to detune the VCO/PLL circuit.

I was looking for a software method to deactivate the FM stages in SSB mode but did not find such a command.

[G8PTN]

Hi Peter,
Thank you very much for the information.

I did a little more experimentation on the ATS20+ that I have. On this hardware, when the AMI is used, the FMI path is completely isolated.

With the antenna switch set to AM and listening to an RTTY signal (mode SSB), the following observations were made.

• The VCO does not appear to be detuned when touching the AMI (Pin 8) input.
• The VCO is detuned when touching the FMI (Pin 6) input, and to some extent adjacent pins 5 and 4.
• The VCO is detuned with non-contact coupling around the area shown below. This implies that the embedded VCO is close to this area.

In summary, for the ATS-MINI:

• Isolating the FMI input signal path to the antenna connector (at the high VCO frequencies) should reduce the VCO detuning effect.
• Close coupling effects to SI4732-A10 could still be an issue because of the mechanical construction.

I think this aligns with your tests. Unfortunately, I do not think this can be fixed by firmware.
73, Dave

[DB9JG]

Hello Dave,
Thank you very much for the information!
Your experiments with hand sensitivity also match my own experiences. In the MiniRx version 1, I even installed a small shielding plate to prevent interference from the speaker, the battery, or the wiring. The reason: at high volume, I noticed a kind of FM microphony.

Thank you also for the circuit detail of the ATS20+, which very much confirms our assumptions! I agree with you that internal interference could only be reduced by separating the FMI from the antenna. In the ATS20, this is achieved by the additional amplifier with its reverse attenuation.
However, I wonder if a simple (pin)diode switch on the FMI might already be sufficient.
One could use an output pin of the ESP32-S3 to isolate the FMI during SSB operation (and possibly even short it additionally).
This could even be advantageous if it leads to a reduction of the capacitance at the AMI during HF operation, since the HF/VHF diplexer could then be designed differently.
This could further improve the active antenna, as the more favorable voltage division at the now smaller input capacitance and the capacitance of the antenna probe would result in better performance!

73&55
Peter

[max-arnold]

Maybe some circuit can be added to isolate the FMI? It can be controlled by the gpio1 pin: G8PTN/ATS_MINI#1 (comment)

The goio1 pin is already supported in the firmware

[G8PTN]

Maybe some circuit can be added to isolate the FMI? It can be controlled by the gpio1 pin: G8PTN/ATS_MINI#1 (comment)

The goio1 pin is already supported in the firmware

Hi Max,
I suspect most of the ATS-MINI users may not want to make any hardware modifications.
73, Dave

[G8PTN]

Hi Peter,
Yes, I agree with all your comments.
Perhaps previous versions with the simple switch or separate RF connectors had solved the problem, although I have not seen anyone mention the issue before.

Ideally, there should be better front-end filtering for both AM and FM inputs, but I suspect to minimise cost, nothing has been implemented.
PS: I also added a couple of diodes on the antenna port to provide some protection.

The other issue with the SI4732-A10 solution is the accuracy of the reference clock (32.768kHz). I did experiment using the LEDC hardware block in the ESP32-S3, but the solution is not ideal. Whilst doing this test a few months ago, I noticed the VCO detuning effect.

Probably the easiest workaround is to add a trimmer on the XTAL, but I don’t expect this to be done because it would require factory test and adjustment, which is not viable for such a low-cost item.

The "Calibration" option in firmware allows +/- 2000 Hz adjustment, but this is not always enough on the higher HF bands and it needs to be maintained on a per band basis (ideally per frequency).

I did also find by experimentation that the synthesised frequency which is used by the internal mixers appears to have a variable error of several hundred Hz depending on the set frequency. I don't think the SI473X device family was designed for SSB, so most of the issues may not have considered.

73, Dave

[DB9JG]

Hi Dave and Max,
Thanks again.
This is a beautiful RX. Having started SWLing centuries ago with a retired monsterous Geloso-RX, I now literally hold the waves in my hand. The frequency “shift” is a very minor effect and only occurs in situations that can be avoided. The intellectual attraction here is to find its cause. So, I feel like I’m complaining at a very high level ;-)
Even the reference-based frequency deviation is astonishing — I would have expected MUCH more deviation and drifting with THIS simple hardware!
(I wrote/built an ESP32-based DCF77 time signal generator and was quite surprised findig it acceptably close to and stable at the right frequency.)

All the tiny little cases that have gathered on my workbench these days —like the VNA, spectrum analyzer, signal generator, digital oscilloscope, etc — will drift or jitter in one way or another, but for my purposes, that’s like being in electronic heaven :-)
Nevertheless, the idea of decoupling the FMI is still meandering in my mind ;-)

73&55
Peter

[DB9JG]

Hi Dave,
The detuning coupling between the VCO and the antenna pins suggests two possible ways the frequency could be affected:
• Indirectly, by a change in any DC parameters (which is unlikely, since the FM input is pulled to ground via an inductor)
• Directly, through capacitive coupling.
The second mechanism could allow the VCO signal to leak to the FM antenna pin. To investigate this, I used a spectrum analyzer at the input of the MiniRX and observed the following:

As expected, there is a signal from the VCO between 3 GHz and 3.4 GHz at the antenna connector.

The current FM input circuit only features a high-pass filter, which cannot suppress this outgoing signal.

Adding a simple low-pass filter at the FM input pin (and optionally at the AMi pin, if no decoupling high-Z amplifier is present) might effectively block this feedback path to the VCO circuit.

I will test this solution.

73
Peter, DB9JG

[Cormel980]

The gain is slightly less than 1. I was talking about something else, the 5m cable itself is an antenna, so the signal is higher, and not because the probe is in the right place.

[DB9JG]

That thesis is wrong. The active antenna (one of two) is well matched and has additional ferrite cores around the coax and a separate good ground connection. And you additionally overlooked the bias-T with its ground connection.
It has the advantage of not being near to any electric fog, as it is placed 3.5m higher on the roof.
Believe me: If You were sitting on my roof with the RX in your Hand you should agree! But I will stay at the workbench ;-)

[EmEiBee]

I was talking about something else, the 5m cable itself is an antenna, so the signal is higher, and not because the probe is in the right place.

How is a matched coaxial cable supposed to function as an antenna? If you use a single bell wire as the "antenna cable," I would agree, but even with a symmetrical, two-wire bell wire and power matching, the wire doesn't function as an antenna (at least in theory), and a power-matched coaxial cable certainly doesn't.
According to your statement, I wouldn't need any antennas on the roof, since I have at least 15 meters of coaxial cable to each of my antennas. But I can refute that by the fact that my receiver suddenly goes dead as soon as I turn off the remotely powered loop line driver.

Perhaps I'm misunderstanding you, due to my insufficient English skills; I don't want to rule that out.

[Cormel980]

Perhaps I wrote it incorrectly. With 2sk544 and bfg35 I used a probe ~5x10cm. Without a cable it doesn't work, there is no gain. With a 1.5m cable, without grounding, it works great. Conclusion: reception is, strictly speaking, on the cable braid, and the probe itself is a counterweight. The receiver is small, the scanner is the size of a pack of cigarettes.

[DB9JG]

A proper installation of an active antenna - taking care of the E-field probe being at a good location and having a good counterweight is indeed important for good results and not just picking up the noise from all the domestic devices. So it is indeed correct to have the active antenna not only being insulated at the end of the koax without a correct grounding of the mast.

http://www.sonic.net/~n6gn/OSHW/BB/Whip-Tipps-EN.pdf

https://www.pa3fwm.nl/technotes/tn09d.html

https://www.pa3fwm.nl/technotes/tn05.html#lpf

[Cormel980]

Hi Dave
In many receivers, for example, pl-330, D-808 and others, the FM input is switched off by a diode switch.
etonsatellitsch.pdf

[G8PTN]

Hi,
Thank you for the information.
Yes, diode switching would provide additional blocking.
There is also a 150nH inductor on the FMI input to provide low-pass filtering.
These radios have much better filtering on the AM input as well.
73, Dave

[EmEiBee]

...perhaps we should brainstorm a switched preselector with the same housing dimensions. The only question is whether the ESP still has an I²C, 1W, or something similar available for control.

But if so, it would have (almost) only advantages, even if the effort is, of course, not insignificant...

[Cormel980]

Hi Dave

PS: I also added a couple of diodes on the antenna port to provide some protection.
Using a 1 kOhm resistor is wrong, it is not needed. In outdoor AM active antennas, protective ~300 Ohm before the diodes are used in series with the input, but only in AM.
Good protective diodes BAV99 sot23.

There is also a 150nH inductor on the FMI input to provide low-pass filtering.
AN383.pdf recommends VCO filter bead Murata BLM15GA750SN1, Impedance 75 Ohm at 100 MHz and 1000 Ohm at 1 GHz, i.e. ~120 nH and 6 pF capacitor in parallel with the input in 3x3mm Schematic. Maybe 3-5 pF (~400 Ohm at 100 MHz) will improve VCO suppression.

[EmEiBee]

RSSI is NOT the point of RSSI. You will never see a difference with and without High-Z
The point is to convert the high impedance of a rod antenna to the low Z of the 4732- rf-input.
That helps to keep distortions and qrm lower

[DB9JG]

I did not want to talk about RSSI.
The SI4732's input is sensitive enough -- matching the input to the telescopic antenna probe is the aim.

The V3-Hardware already incorporates an impedance changing FET with about 5pF gate-capacitance (= violet part of my schematic). But the AM/FM antenna diplexer adds 18pF(C28) by its FM path in parallel to the AMi.
To avoid this effect, a first test could be to (temporarily) remove C28 from the original circuit of the V3-HW.

My changed diagram then may become the next stage of changing the given Antenna input stage of the V3-hardware, so that the FET with its 5pF becomes the only load to the telescopic antenna (~15pF)

The reason for these efforts is to improve SNR while using the telescopic antenna, while I see that an (0dB) external active antenna performs slightly better than the Mini-RX + telescopic antenna being at the same location.

[EmEiBee]

Ok, then I'm on the wrong path; sorry. My fault while reading to fast...

And now I see what you like to do. For a test it's the right way to remove C28.
However, I think the better approach would be to place the diplexer after the FET rather than in front of it. Since there's no pre-selection anyway, and the FET is essentially saddled with the entire bandwidth, a diplexer in front of the FET doesn't really make sense. At best, a bandpass filter to smooth out signals below 160m and above 2m would be better, and a diplexer with the appropriate crossover frequency would be placed after it.

[DB9JG]

That's exactly what my new schematic suggests. The blue parts are my changes to the original circuit.

Antenna >>> Bandpass for 100MHz and parallel a lowpass for <30MHz

The Diplexer at the output of the FET prevents AMI to see the FM-Band and the FMi to see shortwave signals - just like the datasheet of the SI4732 request.

And additionally the FMI gets decoupled, because of the VCO leakage issue.

[EmEiBee]

It's nice that we agree on that ;)

[Cormel980]

...I didn't see any improvements in RSSI...Have you had a chance to check this?
Unfortunately, the screen and receiver emit a lot of interference. In the firmware, you can completely disable the screen, the interference decreases, but does not disappear, on another receiver they are still audible, checking will not give anything, the improvement is drowned in interference. You can improve reception by removing the telescope with the amplifier from the case, while the field-effect transistor is not needed and can be dismantled.
Reference: adding C28 18pF to the 15pF telescopic antenna and 5pF transistor, according to calculations, reduces the output signal by ~5,5 dB. On a good receiver, this is noticeable on weak signals.