开云体育

Here's another try. I just put shielding around the outside and not over the top. I improved both traces a little, see the picture - top is the "old" shields and bottom is the new try on shields. CH0 average signal level levels off more after about 500 MHz. CH1 noise level is also slightly lower.

But, this is all I'm going to do about it. After all, it's not being used to design a new weather satellite it's just for ham radio, and I don't need anything over 450 MHz. I also think this general project (the NanoVNA) is just at the beginning and over the coming years we'll see more and better designs, so for now I'm actually really very pleased with both the units. Everyone should get one!

:-)

the second one I got with no shields works better

1) Component variations
2) reflections within shielding.
Hugen warned about that in an early post here.

FWIW, a metal case "upgrade" for RSP1A includes an RF absorber.
The original plastic housing had incorporated that...

On Mon, Oct 21, 2019 at 07:28 AM, <erik@...> wrote:

On Mon, Oct 21, 2019 at 07:13 AM, Jeff Anderson wrote:

if you were to short either c26 or c27 to ground, I would expect the noise
floor to increase by 6 dB

C26 and C27 connect to the output of the transmission mixer. Do not expect a
lot of sensistivity for noise going back into the output, especially because
of C32 and C33 removing all HF.

My reasoning is that, by shorting one of the caps to ground, you are effectively halving the mixer’s output level. I am assuming no new noise is injected. But, because recalibration would “renormalize” the signal level, the effect would be an increase in noise by 6 dB, because the renormalization, by bringing the signal back up by 6 dB, also amplifies the ADC’s noise floor ( and whatever other in-band noise there is) by 6 dB.

- Jeff

On Mon, Oct 21, 2019 at 07:13 AM, Jeff Anderson wrote:

if you were to short either c26 or c27 to ground, I would expect the noise
floor to increase by 6 dB

C26 and C27 connect to the output of the transmission mixer. Do not expect a lot of sensistivity for noise going back into the output, especially because of C32 and C33 removing all HF.
Most unwanted input to the transmission SA612 comes through its osc input from the SI5351 as the isolation between outputs of the SI5351 is fairly bad. The variations in leakage between SI5351 could explain all differences between products except when measuring below 50MHz. I expect stong FM transmitters to have some inpact around 100MHz but many coax cable have insufficient quality shielding to block these FM signals

I believe the main function of the shielding in this design is to improve isolation between the signal source and the SA612 receivers (so, to W5DXP’s question, I doubt there’s much improvement at HF, but this would need to be verified).

Other comments...

Mike you can make the dual shields and install them fairly easily. Make one small rectangle to cover one of the sections, and then make a similar shield, but with one of the long sides bent up, that will cover the other section.

Solder the first shield to the board per qrp rx’s Soldering recommendations, then place the second shield on the board, solder it’s 3 sides that touch the board to the board, and then solder the lifted (fourth) side to the top of the first shield.

It’s worthwhile investigating qrp rx’s suggestions. I will note:

1. Capacitive coupling to ground via the shield might be an issue, but a big question I have is how much difference in noise you are seeing. For example, if you were to short either c26 or c27 to ground, I would expect the noise floor to increase by 6 dB ( or is it 12 dB?) after recalibration. Attenuation due to capacitive coupling to ground will be less than this. So if you are seeing a large increase in noise, I would doubt that it is capacitive coupling. But do not rule it out as the cause without first verifying.

2. Similarly, I am hesitant to say that the shield itself, because only three corners are tacked to ground, is somehow injecting noise. The many ground visa should already be providing a low impedance ground path. Could there be a “slot antenna” effect due to the gap that is somehow injecting interference? Possibly, but from experience with my own designs and testing for EN61000 (and FCC) radiated and susceptibility compliance, I haven’t come across a poorly installed shield being worse than no shield. Still, do not remove it from your list of suspects until you have verified that it is NOT the problem — while I haven’t seen this as a problem, myself, QRP RX might have come across it in their designs.

3. Also, before each test, make sure you verify that the calibration is correct ( I have had some strange calibration effects). Verify that for S11 the SOL loads appear on the Smith Chart where you expect them, and verify that S21 looks correct with and without the thru jumper. (In other words, don’t rule out operator error!)

4. Finally, just a note that, although you or I might think the designs are identical, they need not be. Parts have tolerances (and might even be counterfeit). There could be manufacturing errors, etc.

- Jeff, k6jca

On Mon, Oct 21, 2019 at 03:22 PM, Mike_nano wrote:

However, how would you address the fact that the second one I got with no shields works better?

this is because shield installed incorrectly, it affects circuit performance.

Perhaps this PCB design doesn't come with shields because they aren't designed to need them?

Shield is needed. But this PCB has some design mistakes which makes proper shield installation more hard. Probably this is why they don't installing shield, they just don't wanna spend time to fix these issues caused by poor PCB layout.

QRP RX,

There is no easy way (impossible) to make two sections under the larger shield and have any sort of continuous solder line to the board, so perhaps I'll just take it off. I'll take your other comments into account on the other shield but ...

... maybe it's not worth the trouble. I'm not "repairing" a broken PCB, I thought it would be an improvement. However, how would you address the fact that the second one I got with no shields works better? Are we really sure this PCB design actually requires shields at all? This PCB design (there is a different one) appears to never come with shields. Perhaps this PCB design doesn't come with shields because they aren't designed to need them? I need to find the response plot of the two traces (CH0 return loss with calibrated load and CH1 noise level with load) shown from another NaNo with shields and see if I'm wasting my time. Do you know where I can get those plots to compare mine to? It does look pretty with shields though.

Question: If the NanoVNA is used only for HF, are the shields necessary?

Also it seems that this PCB layout is not good. Because it don't allow to place shield edge on safe distance from c26, c27. Also it looks that there is impedance problem with wire to SMA connector. The wire thickness looks like too thin for two layers pcb.

But these things cannot be fixed. This is poor PCB design.

ideald shield should be soldered with no gap. Because any gap leads to ground loop. Longer gap leads to higher noise. Did you see these through holes on the pcb ground? These holes are named "vias", they plays the same role as shield soldering. As you can see, there are a lot of vias with a short distance. This is needed to avoid ground loops between front and back ground layers of pcb.

On Mon, Oct 21, 2019 at 01:04 AM, Mike_nano wrote:

What am I missing? Did I do something wrong?

yes, you did it wrong. There are two problems with your shields:
1) You're forgot to solder left top corner of shield. The distance to the ground is long, so the shield works as loop antenna and receiving noise.

2) the distance between shield and c26, c27 is too close. Short distance leads to capacitive coupling. You're needs to cut a small piece of left top corner of your shield in order to increase distance between shield and c26, c27.

Also, you didn't shown what is under the hood of the top shield. There is also may be problem with capacitive coupling due to short distance and missing soldering at some point. Also, there are needs two sections under shield for CH0. It's not clear from your photo if CH0 shield has two sections or one section.

Mike, looking at the photo of the shields and how they mount, there is nothing that is visible to me that is obviously wrong.

I say this because the ch1 shield is mounted on pads that seem to be tied to ground with plenty of vias, so if there were a ground noise issue, I’d expect noise to be improved, not worsened, when the shield is in place.

One thing I do notice, though, is that ch0 and ch1 both seem to have their own ground islands. And I’m wondering if the back plate of the instrument was screwed in place when you noticed the improvement in ch1 noise when it’s shield was removed. If the back plate is conductive (and from memory I think it’s made of PCB material), then it’s possible that the back plate is shorting all the grounds together through the standoffs and thus making a possible path for supply noise via rf ground. Compare the back plates of the two versions. Is one conductive and the other not?

Anyway...it’s difficult to troubleshoot a problem long distance! Experiment, see what improves performance and what worsens it, and always compare these effects to how you think the changes ought to be affecting the performance.

Jeff, k6jca

Hmmm ... thanks for that. Let me play some more tomorrow and see what I can do.

Hi Mike_nano, Jeff...

One difference I note from prior photo's and shielding is the connection of the shield itself TO THE SMA CONNECTOR BODY. Each shield was returned to ground via the SAM. This connection could reveal an improvement and hence the presence of ground currents without that connection(s).

I am afraid that without the appropriate tools such as signal integrity simulation and EM, fooling around with shielding and cause and effect relations is like poking around in a dark closet with a sharp stick.

Alan

Jeff,,

Thanks for taking an interest. The answer is no. I'm very careful and took into account all the SMD parts locations. No, I don't see how I could have done that. I made sure that the edges of the copper were clearly on the ground plane outline.

Mike, any chance that the shield is shorting out C26 or C27?

Jeff, k6jca

That's a great case, I'm going to print one today :)

To the first nano I got (the one with the white 3D printed case) I added my own carefully-made copper shields. I recently got another Nano (the one with the orange case). It also came with no shields. However, I noticed something - the second one has a better return loss with load for CH0 and a better noise floor for CH1. Better than the Nano I had added shields to. I did a lot of experimenting and calibrating and swapping of cables and loads, and it's clear - the second one with no shields seems to behave better in the two aspects noted before. Also, the two PCBs are identical - no differences in design or components.

In the picture you see one of the shields removed because I wanted to see what was going on. When I removed it, it also improved the CH1 response (blue trace) more like the "orange" nano. I have not yet taken off the other shield yet. I first wanted to ask the community to help me understand this. It sure seems to me that the shields are not helping things. What am I missing? Did I do something wrong?



.