System dynamic range improvement

[blackberryer]

In the end, it was only inspired by the nanovna V2 project. The dynamic range of the system can be improved by redesigning the directional bridge. The MMIC-based differential amplifier instead of the balun can reduce the cost and increase the SNR, so as to achieve a faster scanning rate. Simply use an BJT LNA, which is much cheaper than a broadband balun. I am here to make suggestions, not sure if it is feasible in principle.

[jankae]

I have to admit that your message is a bit confusing for me. It sort of feels like a first part is missing or it is an answer to another post? Could you elaborate a bit more and maybe add a part number for the amplifier? This could be an interesting idea.

[blackberryer]

I’m sorry for not being very clear, maybe it’s just my own guess. In fact, what I mean is that I want to make some modifications to the following circuit. Converted into a circuit like this:

[jankae]

Thank you for the schematic, I think I do understand the idea now. At first glance I don't really see the benefit of this approach. Sure, you would be able to measure signals with smaller amplitudes because of the additional amplifier but the mixer (or more accurately with the gains as they are now in the schematic: the ADC driver) would also saturate earlier for high amplitude signals. As far as I can see, this would just shift the allowed signal amplitude to lower levels instead of increasing the dynamic range. Or am I missing something?

Or is this simply about replacing the balun without any additional amplification? In this case the amplifier probably needs to have pretty good CMRR but this could be worth looking into.

[hugen79]

Essentially the Agilent 1GM1 and ADI's newly released ADL5960 both have integrated differential amplifiers, I'm having a hard time finding a suitable standalone differential amplifier, do you have one you can recommend?

[blackberryer]

Essentially the Agilent 1GM1 and ADI's newly released ADL5960 both have integrated differential amplifiers, I'm having a hard time finding a suitable standalone differential amplifier, do you have one you can recommend?

I think you are interested in the solution used on nanovna V2 plus4? It does not use a pure differential amplifier, but a BJT LNA. There are many types, such as spf5043z

[zsellera]

No RF engineer here, but using such amps makes sense if:

1. want to match different impedances (buffer)
2. signal conditioning (amplification before the mixer stage)
3. single-ended to double-ended conversion, or vice-versa

Here the bridge provides a differential output, and the mixer offers a 50 ohm differential input, which is perfect for the bridge (as it offers the same output). The balun here may act as a filter for the common-mode component coming from the bridge (guess), but I don't know why is it needed - maybe @jankae can shed some light on it.

The nanovna v2 is a switch-porn, where the bridge's differential 50 ohm output must be matched to a single-ended 50-ohm line. The baluns seen on the original (open) v2.2 design seem to me bad design choice, probably driven by cost reasons only. Having a differential input amp there makes perfect sense.

The ADL5960 has those preamps there probably for all three reasons (impedance matching, preamp and SE to differential conversion).

[blackberryer]

No RF engineer here, but using such amps makes sense if:

1. want to match different impedances (buffer)
2. signal conditioning (amplification before the mixer stage)
3. single-ended to double-ended conversion, or vice-versa

Here the bridge provides a differential output, and the mixer offers a 50 ohm differential input, which is perfect for the bridge (as it offers the same output). The balun here may act as a filter for the common-mode component coming from the bridge (guess), but I don't know why is it needed - maybe @jankae can shed some light on it.

The nanovna v2 is a switch-porn, where the bridge's differential 50 ohm output must be matched to a single-ended 50-ohm line. The baluns seen on the original (open) v2.2 design seem to me bad design choice, probably driven by cost reasons only. Having a differential input amp there makes perfect sense.

The ADL5960 has those preamps there probably for all three reasons (impedance matching, preamp and SE to differential conversion).

I tried this circuit, which is basically effective, but the AGC part of V2 version is not open source, resulting in the noise of more than 1G in my actual test, but the approximate track is correct. In addition, the author of V2 chose balun, which is unreasonable. He may not have seen jankae's design. If he has, the design of balun should be better

[YuryW]

An improvement in the frequency dependence of the dynamic range of S11 and S22 (at the same time as S12 and S21) can be obtained by replacing two baluns at the meter ports in LibreVNA with wider ones, for example, MABA-011108 ( https://www.macom.com/products/product-detail/MABA-011108#part-tech-order ) . A further increase in the dynamic range of S12 and S21, in my opinion, is achieved by continuing to apply the already used idea - to install an additional QPC6324 between the switch and the ports themselves. Then the noise floor of S21 and S12 will be determined by the bandwidth and shielding of the individual LibreVNA nodes.

[YuryW]

Good day everyone. If we assume that the spurious signal penetrates through the paths indicated in the figure in red, then I propose one of the solutions as follows. At the points marked in the figure, as an example, put a 12dB attenuator and a TRF37A73 amplifier. S12 of such an amplifier decreases with increasing frequency and reaches -19...-21 dB at 4...6GHz. That is, if the assumption and paths of spurious signal penetration are correct, we will get an improvement in isolation between LibreVNA ports in the upper part of the range of at least 31 dB. Or the second option is to put a separate MAX2871 synthesizer on each ADL5801 mixer.

[sp9bsl]

Hi Yury, you cant't use separate MAX2871 because there will be no information about its phase in two port measurement. Anyway putting an attenuator and amplifier is interesting idea.

[YuryW]

Hi Slawek. If we assume that LO-to-RF Input Leakage = RF-to-LO Input Leakage for ADL5801, then at 5500 MHz they are equal to 29 dB. A 25 dB resistive coupler has 25 dB isolation. The simulation showed that for a TCM1-63AX+ bridge with an input resistive attenuator, the reflected wave is at best 19 dB at 6 GHz. On two QPC6324 switches, 1.35*4=5.4dB will be lost in both directions. Then we get mathematically only along this path we have an isolation between the LibreVNA ports 19+5.4+25+29+29=107.4 dB with a calibrated load on port No. 1. If we take into account the reflections only from the input of the QPC6324 17 dB at 6 GHz, then we get an isolation between the LibreVNA ports when port No. 1 is loaded (according to the diagram above) to a calibrated load of 50 Ohms: 17+25+29+29=100 dB. So no, this option is not needed if the goal is to get isolation within 100 dB. Best regards.

[YuryW]

Good day everyone. Assuming that the isolation in the ADL5801 LO -> RF is approximately equal to RF -> LO only at frequencies closer to 5...6 GHz, and at low frequencies RF -> LO is much less than LO-> RF, which is indicated in the datasheet approximately -29 dBc, then the problem of improving the dynamic range at high LibreVNA frequencies can still be solved by adding a 12 dB attenuator and a TRF37A73 amplifier before the LO inputs of the first mixer (see post on Nov 3, 2022 above). Indirectly, this is evidenced by the changes that the 6-layer version received and which, in my opinion, led to an improvement in the dynamic range compared to the previous version - namely, in the 6-layer version, a 10 dB attenuator appeared before the LO ADL5801 input (except Port 2) .

In the absence of a signal at the RF input of the ADL5801, it is probably enough to have the first local oscillator signal at the LO input ADL5801 and the HF Source signal reduced by about 70 dB to get an IF of about -70 dB at the mixer output.

And perhaps it is enough to put a chain of 12 dB attenuator and TRF37A73 amplifier just before the LO input of the ADL5801 reference receiver. And for Port 1 and Port 2, before the LO input of the ADL5801, install only 10 dB attenuators. Moreover, the frequency response of the dynamic range in the formula 25 + 29 (frequency) +10 +10 will determine the value of 29 dB in the current LibreVNA scheme, which will change quite strongly when the frequency changes up to 65 dB at low frequencies. Similarly, in the concept of PCB layout, it must be taken into account that LO 1 and HF Source are shielded and separated from each other as best as possible. Not so with the 6 layer version where the two signals get pretty close together especially through the vias which at high frequencies can be micro radiating antennas on each PCB layer, reducing the shielding of the signal layers between them. If we add a 10 dB attenuator to the Lo input of the first mixer in Port 2 and a 12 dB attenuator and a TRF37A73 amplifier to the Lo input of the first mixer of the reference receiver (instead of the current 10 dB attenuator) to the version of the 6-layer LibreVNA wiring, then according to this theory, we get a gain in dynamic range for Port1 is about 22 dB, and for Port2 is about 32 dB. And if the Lo level at the input of the first mixer of the reference receiver is made the same as the level at the Lo inputs of the first mixers Port1 and Port2, that is, if we put a 10 dB attenuator after the new TRF37A73 amplifier, then we get another 10 dB gain in dynamic range. Best regards.

[YuryW]

A little continuation. I do not have a 3D electromagnetic simulator, but in my opinion in the layout of the 6-layer LibreVNA printed circuit board there are several dangerous approaches (from the point of view of mutual interference through vias less than necessary at -100 dB) of the 1.Lo and HF Suorce signal lines. Points 1 and 4 are especially questionable. In my opinion, there are no restrictions to remove these lines with signals from each other (at points 1, 2, 3, 4) to the maximum possible distance in this case without changing the dimensions of the case.

To view the traditional gerber and exellon files, I used the wonderful program for CNC machine tools (manufacturing of simple printed circuit boards) GGEasy.

https://github.com/XRay3D/GERBER_X3/releases

Best regards.