absorber materials

Jim Lux 2020/10/13 09:37

On 10/13/20 9:30 AM, Stephen Laurence wrote:
> Thankyou.
>
> So my best bet is Laird MCS sheet. All I need to do now is find
> somewhere I can buy some without costing ten times the cost of my vna!
>
> Easier written than done. If I can find a source of a single A4- sized
> sheet (which would last me a while), then many others might also want to
> get some.
>
> On a totally different subject, we have been off the actual topic of
> “charge led view” for sme time!
>
>
Yeah, and what frequencies are you trying to "fix" - even the MCS might
not do the job you expect.

You could get some lower frequency ferrite materials for EMI - Fairrite
77, 31, etc. as tiny cores/beads or rods and skewer probe or rubber
cement test them)

Laird may not be your best choice as a material source - each company
has their specialties.

Siegfried Jackstien 2020/10/13 21:25

i have cross checked several of the pdf files from laird ... jepp ...
mcs seems to be the best mix

maybe a "group buy" ... (or just ask in this group who wants a 6x10cm
piece and cut the a4 in pieces)

greetz sigi dg9bfc

ps what does one a4 sheet cost???

Am 13.10.2020 um 16:37 schrieb Jim Lux:

- Show quoted text -

Brian Ray 2020/10/14 00:56

This idea is not my original but have tried it on several cases where I have put an active or passive circuit in a metal box with excellent outcome. Take an old medium wave radio out of your junk box or skip. Recover the ferrite rod aerial (minus any coils or wires). If it is too long, break it (careful, it is a ceramic and can produce VERY sharp edges). Glue a piece of this to one side of the box. Do not try to be symmetric. A skewed orientation is often the best. Ferrite aerials are only good up to 3MHz or so but you are not looking for a high Q. You are looking for a very lossy device to mop up the eigenfrequencies of the box or screening.

Another idea which I have seen on very upmarket spectrum analysers is to look carefully at your PC layout. Assuming you have 50 ohm tracks which you will have calculated based on track width, the dielectric constant of the PC material (often 3 to 4 relative) and assuming a ground plane on the other side of the PC or in the middle for multi layers. But what about signals running on the top of the signal tracks? We all know that the signals on the the two sides of the outer a coaxial cable can be quite different and we use clip-on ferrite to suppress the outer signal. This is often part of a balun. The analysers that I refer to often use small ferrite “bridges” which are glued across the signal tracks. They perform the same job as the clip-on attached to coaxial cables. By suppressing the top side signals, they will make it more difficult to couple to the eigenfrequencies of the box.

Both these ideas could be retro-fitted for cents. If anyone wants a simple method to demonstrate that the density of eigenfrequencies increase with frequency, try this. Pick up a large sea shell off the beach. Put it to your ear. What do you hear? For acoustics we normally talk about eigentones. For exactly the same reason, the density of eigentones increase with frequency.

73 5B4AHW

- Show quoted text -

Jim Lux 2020/10/13 15:36

On 10/13/20 2:25 PM, Siegfried Jackstien wrote:
> i have cross checked several of the pdf files from laird ... jepp ...
> mcs seems to be the best mix
>
> maybe a "group buy" ... (or just ask in this group who wants a 6x10cm
> piece and cut the a4 in pieces)
>
> greetz sigi dg9bfc
>
> ps what does one a4 sheet cost???

a 30x30cm sheet in the US costs around $100

One might also look at 3G sheiding specialities SB032-020 is a "low
frequency microwave absorber for 0.5-3GHz".

About the same price for a 0.51mm thick sheet.

The easiest way to get small pieces would be to get a marketing sample
kit if you can convince the rep. The usual sample kit has a bunch of
10x10 or 10x5 cm pieces of all the different kinds of elastomer and foam.

But again, what frequencies are you trying to absorb. None of these
materials will work well at, say, 10-20 MHz.

- Show quoted text -

Siegfried Jackstien 2020/10/13 23:33

lowering s21 noise at 2-4 gig ... so low microwave area

dg9bfc sigi

Am 13.10.2020 um 22:36 schrieb Jim Lux:

- Show quoted text -

Jim Lux 2020/10/13 17:05

On 10/13/20 4:33 PM, Siegfried Jackstien wrote:
> lowering s21 noise at 2-4 gig ... so low microwave area
>
> dg9bfc sigi
>

When you say noise, are you talking broadband noise? or some discrete
spurs?

The NanoVNA doesn't have a super low noise receiver (if nothing else,
there's a resistive pad at the front end), so S21 measurements are
limited by the sensitivity of the receiver and the output power.

If we take 1 kHz detection bandwidth, then the best possible noise floor
is -134 dBm, but in reality, it's probably more like -125 dBm. What's
the stimulus power?

Jim Lux 2020/10/13 17:28

On 10/13/20 4:33 PM, Siegfried Jackstien wrote:
> lowering s21 noise at 2-4 gig ... so low microwave area
>
> dg9bfc sigi
>
>

I was looking at the wrong schematic.. The VNA 2 has slightly different
architecture

Coming in port 2, there's a resistive pad for which I calculate 14 dB
loss (120 ohm series, 75 shunt, 50 ohm system), a MXD8641 which has 0.4
dB Insertion loss, then another MXD8641 with 0.4 dB, finally into the
AD8342, which has a noise figure around 12.2 dB.

So that's 27 dB stacked up.. The mixer input noise floor is -147 dBm.
With a detection bandwidth of 1 kHz (don't know what the firmware does)
that's -117 dBm noise floor (assuming the baseband amplifier isn't
contributing any noise).

The ADF4350 ( which I've used) puts out about 0 dBm (maybe a bit more,
maybe a bit less, depending on how you program the attenuator), then
through a 3.8 dB Pi-pad, a switch (0.4 dB) and the bridge (6dB?), and
another switch (0.4dB), so let's call the stimulus coming out of Port 1
about -10 dBm.

That should give you better than 100 dB SNR for a S21 measurement.

Peter KA6Z 2020/10/14 01:26

Sigi -- Could you possibly use proper capitalization in your posts? I am tired of the stream of total lower-case. Sorry to offend, but your posts are ok but not genius enough, and you are literally the only one here who writes this dismissively regularly.

Thanks.

Siegfried Jackstien 2020/10/14 12:20

Peter

see it as my personal note ... i write all in lower case EXCEPT
HIGHLIGHTED POINTS

if YOU think i am not genius then ok ... I CAN LIVE WITH THAT!!

(all in lower case is boring you ... all in higher case is seen as
shouting ... i tRy to hAve A GooD MIx iF NEeDeD)

Have a nice day

Greetz Sigi DG9BFC

Am 14.10.2020 um 08:26 schrieb Peter KA6Z via groups.io:

- Show quoted text -

Siegfried Jackstien 2020/10/14 12:39

Hello Jim

There is some port 1 to port 2 leaking (called s21 noise) and getting
100 db isolation between the ports is the problem. If you put the v2 in
a metal case (or for testing place on a metal sheet) then there seems to
be some "reflection" inside that case (or in other words s21 noise goes
up a bit) . We can not change the pcb layout, so there is a point were
we can get no better results, but there are maybe a few db to be gained
at the higher end.

The new v2plus4 has more distance between the ports and so different
layout to get the noise down

Greetz Sigi DG9BFC

Am 14.10.2020 um 00:28 schrieb Jim Lux:

- Show quoted text -

Jim Lux 2020/10/14 06:29

On 10/14/20 5:39 AM, Siegfried Jackstien wrote:
> Hello Jim
>
> There is some port 1 to port 2 leaking (called s21 noise) and getting
> 100 db isolation between the ports is the problem.

OK.. that's not noise, which would be random, that's a coherent leaking
signal.

I would think it would cancel out in the calibration, but I'd have to
think about that. Port 2 is seeing a combination of the signal from the
DUT and the leakage path, so very low signals from the DUT could wind up
being less than the leakage, and I'm not sure the cal used in the Nano
would fix it.

If you put the v2 in
> a metal case (or for testing place on a metal sheet) then there seems to
> be some "reflection" inside that case (or in other words s21 noise goes
> up a bit) . We can not change the pcb layout, so there is a point were
> we can get no better results, but there are maybe a few db to be gained
> at the higher end.
>

Yes - but it's going to be serious cut and try. Pieces of absorber on
sticks and move it around to see what happens.

It *might* be easier to figure out to remove the leakage mathematically.
The leakage signal should be constant (varying with frequency of
course), so you should be able to subtract it out from S21 (that would
be the naive approach, and is probably not the best way)

- Show quoted text -

Stephen Laurence 2020/10/14 06:32

On Tue, Oct 13, 2020 at 10:28 PM, Siegfried Jackstien wrote:

>
> i have cross checked several of the pdf files from laird ... jepp ...
> mcs seems to be the best mix
>
> maybe a "group buy" ... (or just ask in this group who wants a 6x10cm
> piece and cut the a4 in pieces)

Well, as I remember, a single A4 sheet was about £140 plus local tax plus carriage, so getting on for £200 (250 euro) delivered. It may need more than one layer of it to improve reflective absorption - it depends if some emf gets through and is reflected by the metal case behind it.

It would be cheaper to buy a V4, as I would not want the work involved in “marketing” pieces to others and putting up with recipients who expected a greater improvement than the stuff actually produces.

When I get home, I will poke around with a small piece if ferrite ( eg a small tuning slug) on the end of a cocktail stick. The vna may have to be viewed in a mirror, or with the screen on an extension cable.

Steve L

n2msqrp 2020/10/14 11:23

Sigi,

Bravo. You are another E.E.Cummings.

<">https://en.wikipedia.org/wiki/E._E._Cummings> https://en.wikipedia.org/wiki/E._E._Cummings

I had to study his poems in high school and college.

Mike N2MS

- Show quoted text -

Stephen Laurence 2020/10/14 08:32

Dear Jim,

You have given an excellent breakdown of the signal path through port 2 (channel 1) and suggest the dynamic range should be 100db, but it is 85db at low frequencies reducing to 40db at 4.4ghz (60 db at ghz). Some of this at the high frequencies presumably is increased noise within devices at the higher frequencies, but the rest.....?

What I would like to ask is why the 14db attenuator at the input of port 2? I understand some attenuation can help to normalise the input imedance, but as much as 14db? What would happen if the this attenuator pad was reduced to 4db; would this increase the dynamic range at all?

Steve L

Jim Lux 2020/10/14 08:50

On 10/14/20 8:32 AM, Stephen Laurence wrote:
> Dear Jim,
>
> You have given an excellent breakdown of the signal path through port 2
> (channel 1) and suggest the dynamic range should be 100db, but it is
> 85db at low frequencies reducing to 40db at 4.4ghz (60 db at ghz). Some
> of this at the high frequencies presumably is increased noise within
> devices at the higher frequencies, but the rest.....?

Well, that's an interesting question - I don't know how the 8nV/sqrt(Hz)
noise on the baseband amp adds in (should be negligible, one usually
designs so that the later stages contribute less noise than the front end)

It could be the ADC - a 16 bit adc typically will give you around 80-85
dB SNR, but the actual dynamic range/noise floor will be set also by the
downstream signal processing, which I've not looked at. A 24 bit ADC
would be better (but it's *hard* to get more than 100dB in practice).
And, it also depends on where the nominal "0 dB" is - good practice
would be to keep it well below (10 dB) the P1dB level, but, then, this
is a VNA, not a receiver and it's a bit different. You do want to keep
it in the linear range, though. So if the designer chose the gains
such that max input is 80 dB above the noise floor, there you are.

>
> What I would like to ask is why the 14db attenuator at the input of port
> 2? I understand some attenuation can help to normalise the input
> imedance, but as much as 14db? What would happen if the this attenuator
> pad was reduced to 4db; would this increase the dynamic range at all?
>
I was surprised by that. I wonder if that's basically to set the levels
to a "safe" (i.e. not pushing linearity) level for the down stream
components (switches, mixer). In theory, of course, that gives you a
"good match" into port 2.

The AD8342 has a P1dB of 8.4 dBm but only if configured for 100 ohm
shunt input. If it's matched, it's 2.3 dBm (and it improves the NF to
10.5 dB). With 75 ohms, it kind of looks like around 9dBm.

The v2.2 schematic shows a 75 ohm shunt fed through a 10 microfarad DC
block (which is quite large.. C306 - Is that right? A "microwave
appropriate" capacitor at 10 microfarad is pretty unusual)

So 14dB pad plus the switches, makes the level at "max output from PLL
(+5dBm) " around -10dBm...

More than that, you need to talk to the designer.

- Show quoted text -

OwO 2020/10/15 00:58

1. The leakage seen at port 2 at a particular frequency is dependent on
DUT S11 and S22. The former can be and is corrected for in the firmware
and in nanovna-qt (it is a linear dependence), but not yet in
nanovna-saver. The latter can not be corrected for, so in practice the
achievable dynamic range after calibration/leakage removal is
bottlenecked by port 2 shielding. Otherwise, there is no reason you
can't remove almost all leakage.

2. The noise floor is not bottlenecked by mixer noise figure. It is (1)
power supply noise getting into the mixer and (2) mixer 1/f noise. This
is only a factor in S21 measurements. For S11 the signal level is high
and synthesizer phase noise is the limiting factor. All are solved by
increasing IF frequency.

3. In a VNA you absolutely need to keep mixer input well below 1dB
compression, in fact it must be at least 10dB below P1dB and 20dB below
IIP3 to achieve 1% error magnitude which corresponds to 0.1dB S11
nonlinearity error/trace noise. I've seen VNAs being marketed that have
far greater than 0.1dB S11 nonlinearity, and these can not be called
network analyzers.

4. There is no such thing as a "microwave appropriate" MLCC capacitor,
unless we are talking high power levels. Look at the characterization
data for any 10uF 0603 capacitor (X7R or X5R), and you will see the
impedance is < 1ohm all the way to many GHz, with no kinks or internal
resonances to be seen. In fact the larger capacitors have better
characteristics at several GHz. The S-A-A-2 BOM does not call for any
capacitor manufacturer and everything on the BOM are generic. Any
shenzhen variety component will work well. This is an art that many RF
designers are yet to master.

Jim Lux 2020/10/14 10:44

On 10/14/20 9:58 AM, OwO wrote:
> 1. The leakage seen at port 2 at a particular frequency is dependent on
> DUT S11 and S22. The former can be and is corrected for in the firmware
> and in nanovna-qt (it is a linear dependence), but not yet in
> nanovna-saver. The latter can not be corrected for, so in practice the
> achievable dynamic range after calibration/leakage removal is
> bottlenecked by port 2 shielding. Otherwise, there is no reason you
> can't remove almost all leakage.

Got it, that's sort of what I figured..

>
> 2. The noise floor is not bottlenecked by mixer noise figure. It is (1)
> power supply noise getting into the mixer and (2) mixer 1/f noise. This
> is only a factor in S21 measurements. For S11 the signal level is high
> and synthesizer phase noise is the limiting factor. All are solved by
> increasing IF frequency.

interesting.. We've used LT3042s to help against the Power Supply
rejection - but maybe that's not a good solution in this context. And of
course, mixer 1/f (both in the mixer and the following baseband amp)..
I don't know that AD publishes useful noise data (the data sheet is sort
of vague and doesn't cover down low)..

>
> 3. In a VNA you absolutely need to keep mixer input well below 1dB
> compression, in fact it must be at least 10dB below P1dB and 20dB below
> IIP3 to achieve 1% error magnitude which corresponds to 0.1dB S11
> nonlinearity error/trace noise. I've seen VNAs being marketed that have
> far greater than 0.1dB S11 nonlinearity, and these can not be called
> network analyzers.

I've always wondered if one could "calibrate" this out (the usual
problem with P1dB is harmonic growth) - in theory, you could know the
Voltage in vs voltage out curve and work backwards. (not as part of the
standard VNA cal - that assumes a linear system)..

In other forms of VNA that use only amplitude sensors (like a 6 port)
where diode detectors are popular, it's standard to calibrate out the
nonlinearity of the detector.

Probably not worth it here - just design, as you did, for enough headroom.

>
> 4. There is no such thing as a "microwave appropriate" MLCC capacitor,
> unless we are talking high power levels. Look at the characterization
> data for any 10uF 0603 capacitor (X7R or X5R), and you will see the
> impedance is < 1ohm all the way to many GHz, with no kinks or internal
> resonances to be seen. In fact the larger capacitors have better
> characteristics at several GHz. The S-A-A-2 BOM does not call for any
> capacitor manufacturer and everything on the BOM are generic. Any
> shenzhen variety component will work well. This is an art that many RF
> designers are yet to master.
>

I suppose - I'm coming from an environment (space) where we like big
margins on voltage and temperature, and there's not a huge number of
"big" MLCCs, so there's been a tendency to not use them. (i.e. there's
no 10 uF from AVX on the NASA approved list or as a QPL device, but
there is on the MIL and ESA lists) - Clearly for your design you don't
need QPL parts :)

Useful to know that they have decent RF properties.

>
>

Awesome answers.. thanks

Stephen Laurence 2020/10/14 10:46

Dear Gabrielle,

Thankyou for your comprehensive reply. I am learning all the time.

Is the leakage between port issue the reason why all the old HP and other boatanchor vnas had the two (N type) port connectors miles apart?

Steve L

Jim Lux 2020/10/14 10:59

On 10/14/20 10:46 AM, Stephen Laurence wrote:
> Dear Gabrielle,
>
> Thankyou for your comprehensive reply. I am learning all the time.
>
> Is the leakage between port issue the reason why all the old HP and
> other boatanchor vnas had the two (N type) port connectors miles apart?
>
> Steve L
> _._,_._,_

ANd that's why the inside of the box was full of other metal boxes -
Look at an 8663 signal generator for an example boxes within boxes
within boxes and lots of hardline interconnects.

Making that happen in a single board design is a challenge -

Peter KA6Z 2020/10/14 22:54

Whatever. You're obviously problematic.
e.e. cummings was at least interesting.

Will ignore you. Great contributions from others I will follow.

Peter

Stephen Laurence 2020/10/15 08:18

Dear Peter,

I am not quite sure who your comment was intended for.

If it was me, then I would reply in the manner of a saying by the late Tom Lehrer (look him up):-
“ .......if you cannot communicate, the least you can do is shut up.”

Not everyone on these forums are high-powered professional techs. some of us are amateurs, hobbyists and are learning. Those of us who are learning benefit greatly from those who are professionals.

ps, in the United Kingdom there is a dickhead called Dominic Cummins (look him up too).

Steve L

Peter KA6Z 2020/10/15 12:47

Not at all meant for you, Stephen, and apologize for any misunderstanding. I was referring to the habit of never using capitalization, or often punctuation, in these communications, that's all. As you say, the forum is meant for communication. I find it tedious and unenjoyable to read tracts that dispense with any form of punctuation and capitalization. It's difficult to read or follow. That's all. Sorry again for sounding pompous or singling anyone out (not you). I'll just skip the posts that are shoddy (again, not yours).

With respect,
Peter KA6Z

Beware of cheap underperforming clones