NanoVNA Custom Software

[ogden]

Sigilent's website shows them taking orders for their new VNA.   $3400 or so USD, without the cal kit, no TDR and no VNA software included.  Look like you more into the $5K when it's all said and done.    Plus you still have the problem of it not being a full 2-port system.

You get what you pay for. Anyway standalone SA with S11+S21 3GHz VNA for $4000 is hellova good offer.

Hoping this time around that they give the unit to people who can actually show the practical justification of the higher cost when compared with this $50 unit.

Perhaps manufacturers which are looking for dependable instrument having better than 60 70dB dynamic range? Not even mentioning freq range. Of course - many hobbyists are fine with nanoVNA, but not every buyer of VNA is hobbyist.

[edit] Comparison of $50 vs $50000 VNA: https://nuclearrambo.com/wordpress/comparing-nanovna-with-the-keysight-fieldfox-n9952a/

[hendorog]

Sigilent's website shows them taking orders for their new VNA.   $3400 or so USD, without the cal kit, no TDR and no VNA software included.  Look like you more into the $5K when it's all said and done.    Plus you still have the problem of it not being a full 2-port system.   Hoping this time around that they give the unit to people who can actually show the practical justification of the higher cost when compared with this $50 unit.    

https://siglentna.com/spectrum-analyzers/sva1000x-spectrum-vector-analyzer/

You are an optimistic man if you think the Nano will get to 3.2 GHz
The nano is a nice 300MHz device.

Also you might have missed this:
"Vector Network Analysis from 10 MHz – 1.5 GHz / 100 kHz – 3.2 GHz (Now included as standard)"

[radiolistener]

3.2 GHz is good. But 1.5 GHz for 50 USD is good enough for a radio amateurs

[hendorog]

[edit] Comparison of $50 vs $50000 VNA: https://nuclearrambo.com/wordpress/comparing-nanovna-with-the-keysight-fieldfox-n9952a/

That review raises an interesting point. They observe that the nano performs worse when used in full span. I've noticed this as well.

A user on another forum explained it in SA terms. Where a narrow span on an SA has a lower noise floor due to a narrow RBW being used.

That theory makes no sense. There is no variable filtering in the nano and the sampling is the same regardless of span as far as I can tell.

However wider spans with the same number of points take larger frequency steps. I wonder if the Si chip needs a longer time to settle when making larger frequency jumps? Maybe simply increasing the wait time prior to sampling will improve things in wide spans?

[OwO]

Yes, all PLL synthesizers take longer to settle when the frequency jump is larger. And yes, the nanoVNA will eventually reach 3GHz (and at a similar price I've heard). I know definitely that it's going to be based on the adf4350 + si5351.

[hendorog]

Yes, all PLL synthesizers take longer to settle when the frequency jump is larger. And yes, the nanoVNA will eventually reach 3GHz (and at a similar price I've heard). I know definitely that it's going to be based on the adf4350 + si5351.

Good point - I was referring to the current hardware obviously.

I expect the adf4350 version will be far superior to the current one above 300MHz.
Edit: But still not in the same class as the SVA family.

[ogden]

And yes, the nanoVNA will eventually reach 3GHz (and at a similar price I've heard). I know definitely that it's going to be based on the adf4350 + si5351.

To me "same price" seems to be way too optimistic. ADF4350 is expensive, VNA needs two. SA612 (3pcs) also have to be replaced with better mixers, like LT5560 or similar. Where we can follow information about next generation of nanoVNA? Any pointers?

[radiolistener]

I wonder if the Si chip needs a longer time to settle when making larger frequency jumps?

yes, large frequency jump needs to reset PLL, it takes some time for PLL lock. But it works fast enough.

[OwO]

ADF4350 costs about $0.4 each. I've seen the design and the 3 mixers are replaced with one higher spec mixer that is switched between the 3 channels. A variable gain amplifier is added at baseband using one opamp and switched feedback resistors for improved dynamic range. Audio codec is removed and the stm32 built in ADC is used instead. So overall there are quite a few components removed compared to V1 but I heard the performance should be comparable or better.

EDIT: mixer is probably going to be the ad8342.

[radiolistener]

Audio codec is removed and the stm32 built in ADC is used instead.

This change will reduce ADC dynamic range from 100 dB to 70 dB. So, it seems that the new NanoVNA will be worse...

[ogden]

ADF4350 costs about $0.4 each.

Where exactly you can get (genuine & new) ADF4350 for 0.4$? AD web: $6.05 @1000+

I've seen the design and the 3 mixers are replaced with one higher spec mixer that is switched between the 3 channels.

Right, MiniVNA Tiny. Existence of such design does not mean it is good. Problem with such approach - leakage through switches. Search this forum to see how bad this VNA actually is.

A variable gain amplifier is added at baseband using one opamp and switched feedback resistors for improved dynamic range. Audio codec is removed and the stm32 built in ADC is used instead.

Again - do not agree to such design decision. ADC of stm32 have barely 11 bit ENOB and 69dB SNR. Such VGA-augmented ADC will have worse linearity and temperature stability comparing to literally any generic audio ADC. Not to mention that VGA acting as part of ADC may slow sampling speed down because some/many points needs to be sampled at least 2 times while correct VGA gain is found.

[radiolistener]

ogden, there is also nightmare to capture samples with proper sample rate and low jitter with integrated ADC.

[ogden]

ogden, there is also nightmare to capture samples with proper sample rate and low jitter with integrated ADC.

This is not audio application, thus "right" sample rate is not required. Just pick closest to target number sample rate (round to ADC clock) and that's it. Yes, clock PLL of stm32 have inherent jitter which may ruin your day in high sample rate ADC applications, yet in this case it is low <= 50KSPS, ENOB of ADC is low. So ADC aperture jitter is not an issue. Yet anyway it is good idea to use new generation "stm32" IC's with improved clock jitter specs, rather than for example stm32l072 with it's +/- 600ps.

[joeqsmith]

Sigilent's website shows them taking orders for their new VNA.   $3400 or so USD, without the cal kit, no TDR and no VNA software included.  Look like you more into the $5K when it's all said and done.    Plus you still have the problem of it not being a full 2-port system.

You get what you pay for. Anyway standalone SA with S11+S21 3GHz VNA for $4000 is hellova good offer.

Hoping this time around that they give the unit to people who can actually show the practical justification of the higher cost when compared with this $50 unit.

Perhaps manufacturers which are looking for dependable instrument having better than 60 70dB dynamic range? Not even mentioning freq range. Of course - many hobbyists are fine with nanoVNA, but not every buyer of VNA is hobbyist.

[edit] Comparison of $50 vs $50000 VNA: https://nuclearrambo.com/wordpress/comparing-nanovna-with-the-keysight-fieldfox-n9952a/

For my hobby, I would most likely buy a used 8753.  Siglent isn't on the radar yet.     

I would have expected the people making reviews for the Siglent products to compare them against these low cost analyzers.   Instead, all the videos I saw for Siglent were fairly basic.   I am suggesting they up their game. 

[ogden]

For my hobby, I would most likely buy a used 8753. Siglent isn't on the radar yet.     

Oh.. Fine choice indeed. I believe that you have to get extremely lucky to find such in working condition for a "hobbyist-friendly price".

I would have expected the people making reviews for the Siglent products to compare them against these low cost analyzers. Instead, all the videos I saw for Siglent were fairly basic.

Considering that Siglent is new kid in the VNA block, perhaps their reviewers are such as well

[joeqsmith]

3.2 GHz is good. But 1.5 GHz for 50 USD is good enough for a radio amateurs

It does seem that most people who have written me are indeed amateur radio operators.   I would have guessed 300MHz would have covered most of their needs.  No doubt that the Nano does seem to fill some void for that group.   When Dave made his Siglent review, he makes a point about the amateurs being a potential market.   

If Siglent wants to gain some of that market, they need to make videos showing practical reasons why an amateur would buy their product over a $50 Nano. 

Sigilent's website shows them taking orders for their new VNA.   $3400 or so USD, without the cal kit, no TDR and no VNA software included.  Look like you more into the $5K when it's all said and done.    Plus you still have the problem of it not being a full 2-port system.   Hoping this time around that they give the unit to people who can actually show the practical justification of the higher cost when compared with this $50 unit.    

https://siglentna.com/spectrum-analyzers/sva1000x-spectrum-vector-analyzer/

You are an optimistic man if you think the Nano will get to 3.2 GHz
The nano is a nice 300MHz device.

Also you might have missed this:
"Vector Network Analysis from 10 MHz – 1.5 GHz / 100 kHz – 3.2 GHz (Now included as standard)"

I didn't notice that they now include the VNA software.   

I'm not sure where they are heading with the Nano or next generation.  My friend bought these just for learning.  From my own home use, I would more than likely pull the trigger on a used HP to replace my 1970s one.   3GHz would not be a problem.   

However wider spans with the same number of points take larger frequency steps. I wonder if the Si chip needs a longer time to settle when making larger frequency jumps? Maybe simply increasing the wait time prior to sampling will improve things in wide spans?

I wonder if this is part of the problem I saw with that last test.  I would assume most noticed that that data was always corrupt in the area leading up to 900MHz.  Once it reached 900, the data was stable.   Hopefully the people putting the firmware together are aware of these problems and are able to solve them.       

[joeqsmith]

For my hobby, I would most likely buy a used 8753. Siglent isn't on the radar yet.     

Oh.. Fine choice indeed. I believe that you have to get extremely lucky to find such in working condition for a "hobbyist-friendly price".

I would have expected the people making reviews for the Siglent products to compare them against these low cost analyzers. Instead, all the videos I saw for Siglent were fairly basic.

Considering that Siglent is new kid in the VNA block, perhaps their reviewers are such as well

For my personal use,  the $5K USD spent on a Siglent would cover the cost of a used HP that would more than fit my needs.  But I am used to buying used TE and restoring it.  Actually, my antique HP still fits most of my needs which is why I have not replaced it. 

I would say that your last statement is spot on.   IMO, this is not how you want to market your new line of products. 

[ogden]

However wider spans with the same number of points take larger frequency steps. I wonder if the Si chip needs a longer time to settle when making larger frequency jumps? Maybe simply increasing the wait time prior to sampling will improve things in wide spans?

I wonder if this is part of the problem I saw with that last test.  I would assume most noticed that that data was always corrupt in the area leading up to 900MHz.  Once it reached 900, the data was stable.   Hopefully the people putting the firmware together are aware of these problems and are able to solve them.     

Possible explanation in nanovna-users@groups.io post:

hugen@...Aug 5   #719 
The si5351 manual shows that the internal VCO operates at a maximum of 900MHz and a 4-divide-frequency output with a maximum frequency of 225MHz. In order to output a frequency of 300MHz, the internal VCO needs to be overclocked to 1200MHz. Not every si5351 can be stably overclocked to 1200MHz. As the temperature increases, the internal VCO operating limit frequency of the si5351 will decrease. If you notice a significant spike(>0dB) in your nanoVNA at 300 MHz or 900 MHz, I recommend that you use the 800MHz firmware.

[OwO]

This change will reduce ADC dynamic range from 100 dB to 70 dB. So, it seems that the new NanoVNA will be worse...

No, the VGA will extend dynamic range to 100dB, plus ADC dynamic range is nowhere near being the bottleneck anyway.

Where exactly you can get (genuine & new) ADF4350 for 0.4$? AD web: $6.05 @1000+

Not my decision. I just heard this from the internal chat group.

Right, MiniVNA Tiny. Existence of such design does not mean it is good. Problem with such approach - leakage through switches. Search this forum to see how bad this VNA actually is.

Please do not jump to conclusions before you've seen the design. There are 3 stages of switches between the reflection path and the receiver, for a total switch isolation of 90dB.

The bottleneck of the dynamic range is actually the common mode inductance of the two edge mount SMA connectors, and this can't be improved much other than separating the two ports as far as possible. All low cost USB VNAs on the market currently suffers from this problem and is why they are all limited to a dynamic range of around 70dB < 1GHz and 50dB at 3GHz.

Again - do not agree to such design decision. ADC of stm32 have barely 11 bit ENOB and 69dB SNR. Such VGA-augmented ADC will have worse linearity and temperature stability comparing to literally any generic audio ADC. Not to mention that VGA acting as part of ADC may slow sampling speed down because some/many points needs to be sampled at least 2 times while correct VGA gain is found.

See above - ADC dynamic range is far from being the bottleneck. If the ADC is sampling at 1Msps and the VBW is 1kHz, a 60dB ADC dynamic range leads to a 90dB measurement dynamic range, which is more than sufficient. What makes you think autoranging is slow? A typical measured transfer function is fairly smooth and in practice you likely won't ever see more than 5 to 10 autoranging events per sweep, which is equivalent to adding 10 points to a 100 point sweep.

[OwO]

And before you jump in and say RF switches are expensive - https://item.szlcsc.com/270817.html

[radiolistener]

If the ADC is sampling at 1Msps and the VBW is 1kHz, a 60dB ADC dynamic range leads to a 90dB measurement dynamic range, which is more than sufficient.

Oversampling cannot eliminate non-linear distortions of a cheap ADC. If you will be able to get good results with integrated ADC it will be nice win, none can do it. This is why all using external codec.

[radiolistener]

Where exactly you can get (genuine & new) ADF4350 for 0.4$? AD web: $6.05 @1000+

Chinese clones are much cheaper

[OwO]

I have a prototype of a NanoVNA variant using the built in ADC on the stm32. The measurements using it are nearly identical to the audio codec variant. The reason all NanoVNAs on the market do not use this version is because of software compatibility.

[OwO]

I just got more info about the baseband VGA design; a RFIC switch is used to switch the shunt resistor in the feedback path. The switch is basically "transparent" because the off state capacitance is in the femtofarad range (it is an RF switch) which is negligible at the IF frequency. The on state resistance is small compared to the resistors being switched in. Since the amplifier gain is mainly dictated by the feedback network, and the switch is "transparent", there is nothing other than the tempco of the physical resistors that can cause a temperature dependence. The RFIC used is the same as for the receiver RF switch, and it turns out all the maxscend switches do not have the shunt diode problem (most RF switch ICs have parasitic diodes from RF input to ground which will start to conduct at lower frequencies), so it has no theoretical lower frequency limit and can be applied at the IF frequency. This is a big improvement over using normal analog switch ICs which have capacitance in the pF range.

EDIT: also asked about linearity. The answer is that the code will perform a calibration of each VGA step on boot up. Since there is no temperature dependence the calibration only needs to happen once.

[OwO]

Here is a preliminary block diagram of the design: