Public Shielded Room Work
mirimir at riseup.net
Mon Oct 15 18:48:06 PDT 2018
On 10/15/2018 07:20 AM, Karl wrote:
> On 10/14/18, Mirimir <mirimir at riseup.net> wrote:
>> I worked on this for a while. I was thinking about compartmentalizing
>> services in multiple Raspberry Pi, connected via opto-isolators. Like
>> Markus Ottela's Tinfoil Chat, but expanded to work more like Qubes.
>> And indeed, the Qubes team have announced work on Qubes Air:
>> | This approach even allows us to host each qube (or groups of them)
>> | on a physically distinct computer, such as a Raspberry PI or USB
>> | Armory. Despite the fact that these are physically separate devices,
>> | the Admin API calls, qrexec services, and even GUI virtualization
>> | should all work seamlessly across these qubes!
>> But, you know, I wondered about EMF cross-talk between qubes. So I
>> decided to learn how to measure that :)
> How far did you end up getting on this?
I gave up :(
Basically because professional signal/spectrum analyzers cost so much.
For example, used Tektronix RSA306 (9kHz to 6.2GHz) go for ~$3000. And
you really want something that detects higher frequencies, such as an
Anritsu MS2850A (9 kHz to 44.5 GHz). I can't even find used ones, but
I'm guessing several $1000.
Maybe less-expensive alternatives that are good enough. I find a lot on
<https://www.used-line.com/list-signal-analyzer>. For example, Anritsu
But ignoring stuff over ~6GHz might be problematic.
>>> I don't have any formal training and have developed some cognitive
>>> issues, resulting in slow progress, but this is all I am spending my
>>> free time on. I do not work a job, being supported for now by a
>> Yeah, me neither. But no trust :(
> Maybe we can help each other. Do you have any experience with funding
> platforms like opencollective.com or something?
> My trust is small. I'm happy to share as much as it will allow me to,
> but maybe if we could use something general, donations would
> eventually come in.
Thanks :) But I'm not qualified, so hey.
>>> I'm currently residing in Green Bank, WV, where emissions are
>>> regulated for a radio observatory. I am trying to develop some
>>> relatively simple software using the rtl-sdr to measure the power of a
>>> noise source independent of background traffic, so as to quickly and
>>> repeatedly measure shielding effectiveness.
>> I tried that approach. And it was a nightmare. The setup -- SDR stick,
>> upconverter and laptop -- generated far too much EMF noise. I played
>> with testing stuff in a Faraday cage. But I didn't manage a signal feed
>> to the SDR etc that didn't introduce unacceptable noise.
> My current setup is an oscillating noise source powered by a
> single-board computer that toggles a relay, turning it on and off at a
> consistent rate.
> By averaging the noise level when the source is powered, and averaging
> the noise level when the source is unpowered, over many thousands of
> samples, I believe I can determine the power of the emitter without
> regard to background noise by comparing the statistical distributions
> of the two sets of samples.
> I make the assumption that the foreground signal is the arithmetic sum
> of the generated noise and the background noise.
> Any thoughts?
That sounds reasonable. But again, I did some RF stuff in physics lab,
and that was a _long_ time ago ;)
>> I believe that you'd need professional equipment, which is properly
>> shielded, and doesn't bleed noise into the testing environment.
> Have you tried or researched any professional equipment to report
> back? I haven't, at this time.
As noted above, it seems that https://www.used-line.com is a good
source. And eBay, of course ;)
>>> I then plan to try to measure a variety of setups ranging from
>>> homemade aluminum foil & iron paint to soldered copper and welded or
>>> bolted stainless steel, to identify ways for everyday people to
>>> cheaply create shielded environments that are actually effective. I
>>> would like to find a way people can use off-the-shelf supplies to make
>>> environments that are isolated from DC to light, if desired.
>> That is also harder that it might seem. For high frequencies, with very
>> small wavelengths, even tiny cracks are enough to leak horribly. You
>> need joints with elastic seals, to mitigate against misalignment and
>> wear. Such as beryllium copper finger strips, elastic beryllium copper
>> tubular braid, etc.
> I've looked into that a little. After skimming through some shielding
> books, I've got the following thoughts:
> Permanent Seams:
> - Alu foil can be stapled or tightly taped (see David Weston's paper
> on aluminum foil rooms) to increase its frequency range. I expect
> using a wire brush to remove oxidation, and tightly flattening it,
> would help too. Testing is needed to see if this is worth the effort.
> - Metal filings can be mixed with paint. This allows for tight
> sealing, but the conductivity is likely poor. Advantage is that
> sweepings are available for free. Testing needed.
> - Metal can be tightly bolted, as is done for modular rooms. The
> bolts must be frequent and close, to pull microvariations of the metal
> into each other.
> - Steel can be welded. This is the gold standard. Welding is not that hard.
> - Copper can be soldered. This is easier than welding !
> Temporary Seams:
> - Fingerstock is purchasable and not that expensive, but complicated
> and needs cleaning.
> - Bolts can be temporary, to bolt a door on as modular walls are
> bolted together. It's laborious, but it's workable and cheap and
> doesn't require mail-order. A robot could tighten and loosen them.
> Testing needed.
> - A door could perhaps be given pressure that does not penetrate it,
> to keep a tight seal, perhaps via an automatic mechanism. Cleaning
> will be needed. Testing too.
> - A copper door could be actually soldered closed, and then desoldered
> to open it. A robotic door could automatically do this. Very tight.
> Research needed.
> My understanding is that high frequencies are attenuated mostly by
> reflection. Hence I'd expect these tight seals to be needed mostly
> for very thin, highly conductive material, which could keep costs down
> if true.
That's consistent with what I've read.
More information about the cypherpunks