（评论）

（评论）
(comments)

原始链接: https://news.ycombinator.com/item?id=39930195

这篇有趣的文章讨论了一个名叫 Henrik 的人，他使用现成的零件构建了一个自制雷达系统。尽管 ITAR 限制带来了监管挑战，特别是在进口零部件方面，他的目标是创造一种专业雷达系统的低成本替代品。中国制造商生产更易于使用、可能更便宜的夜视摄像机、热成像仪和包含 ITAR 保护技术的无人机。然而，购买它们有违反规定的风险，导致法律复杂化，特别是在军事设施附近。雷达由发射器和接收器组成，可通过频率、脉冲重复频率和波形等特定特征进行识别。法规控制某些频段的使用，使得民用建设此类设备存在风险，尤其是在雷达设施周围。禁止民间参与创建雷达数据库或指纹。尽管复杂且存在潜在问题，作者还是成功构建了雷达项目的一部分，展现了令人印象深刻的工程技能。他还指出，由于元件布局的限制，在多层 PCB 上实现接地层通常不是标准的。建议使用第二层，以最大限度地减少信号干扰并提高建模的简易性。不寻常的布线选择可能是由于 FPGA 制造商提供的封装延迟特性造成的。此过程可以优化组件定位，同时保持噪声源分开。人们对已完成的项目在同时处理多个角度时是否能够有效运行仍然感到好奇。

yes, it's safe. It's not that different from regular WiFi or cellular signals.

It's non ionizing (aka it doesn't have enough energy to instantly destroy cells unlike uv radiation) but it can heat up tissue, which is linked to cancer and worse (think microwave ovens).

I did some reading and it seems you are right.

No Idea why phones come with a SAR rating then and why there are safe limits defined, when no one found unsafe limits.

Tiny amounts of cell heating/cooling have been found to change the cell differentiation mechanism. That's the reason ultrasound energies are limited during pregnancy and it's recommended not to do too many ultrasounds.

I don't see why microwave heating wouldn't cause the same effect.

Afaik the transmission energy of a cell phone is limited to a maximum of 1 or 2 Watts (depending on which frequency it operates on). This power is a maximum that will only be used if the reception is bad.

In the wavie scene (people who are paranoid about radio waves controlling them) shielding those devices is a typical approach — this is also where the meme of the tinfoil hat comes from. Ironically that shielding makes your cellphone operate with the maximum transmission power at all times.

> It's as awful as it sounds.

Well to me it sounds like you'd get heat stroke and pass out before anything reaches particularly painful temperatures, because it's heating you pretty evenly and not via contact, but maybe that's not the right way to think about it?

At higher frequencies I would guess it gets closer to normal burning?

This isn't really correct I'm afraid. The cause of SAR is predominantly dielectric losses and the loss tangent is a strong function of tissue type -- CSF, fat and bone are really quite different in terms of epsilon r and sigma, and one has to solve Maxwell with a human voxel model to work out SAR effectively which computationally is a pain.

Once tissue heating has occurred what happens next is well described by the bioheat equation, which is basically the thermal diffusion equation with a massive percussion term. The blood supply is very different to different tissue types and the depth at which peak heating occurs is a very strong function of wavelength.

For frequencies, the combination of these effects means that your eyes are most at risk -- water like and terrible blood supply. This gives rise to the first piece of advice I was given when a graduate student playing with electron paramagnetic resonance -- never look down a waveguide and treat them like a loaded gun!

This is great.

It’s the last piece I needed for my suburban missile guidance system! This will be the last year the Joneses survive the annual block party.

Can you do phased array radars next? I need the extra precision. There’s a few neighbors who don’t clean up after their dogs…

Interesting article. I wonder if any progress on the ITAR issue has been made since 2022? If Brad Parkinson can't steer ITAR in the direction of common sense, nobody can. (For those who don't know, he was the principal architect of the original Navstar GPS system.)

Sure, it's getting increasingly easier to just buy China made for ITAR-protected products. Night vision cameras, thermals, drones. They're fine, way more user friendly, often cheaper too.

Wow. I haven't looked at the ITAR list in decades. There are lots of electronics items on there now that have useful civilian applications and can be built at low cost. You can build such things, but can't ship them cross-border to or from an ITAR country.

I'm amazed at what people are building as hobbyist RF gear. I wonder what test equipment they have. The test equipment for GHz RF is very expensive. If you build it and it's not working, you may need test gear.

(I tried to build a LIDAR in 1990s. It didn't work, and I didn't have access to gear that would let me see what was happening up there.)

I’d be interested in knowing if you’ve got that. I was going to buy one but the pull of the code put that on ice. I may pull the trigger if I can get the code!

If you are a foreigner, you're asking the OP to commit a federal crime.

If you're an American national, you're likely asking the OP to commit s federal crime.

ITAR's a mother.

Incredible work. Crazy that this could be done by a single person. Stick it on a rotating pedestal and you've got a planar radar detector. Add some tilt and then it's not much different than aircraft/weather radar I suppose.

The cost (in $LOCAL_CURRENCY, or $570 according to another comment) isn't great but I can only imagine how many hours this took.

Given a proper budget, the sky is the limit.

Anyone know how .mil aircraft would interpret being tracked by a 6 GHz radar build by a civilian(yes I am aware that his estimated max distance is 1200m, assume he increased that by a factor of 10 with larger antennas or something)?

> Anyone know how .mil aircraft would interpret being tracked by a 6 GHz radar build by a civilian(yes I am aware that his estimated max distance is 1200m, assume he increased that by a factor of 10 with larger antennas or something)?

Regulations for signal strength are ERP, so a more directional antenna could make it no longer legal to use the 6GHz band.

I doubt anyone intentionally tracking military aircraft is gonna be too worried by ISM radio band regulations.

(They're either the kind of idiot who aims laser pointers at police helicopters, or they're doing it with intent to get in much more serious trouble than just ignoring radio regulations...)

There are different ways

First and foremost, aircraft radars have receivers, which will also receive the transmitted pulse from the other radar. In fact, that is how radar jamming works - you deluge the other radar with power on the same band / wavelength, and it will fill the display with clutter.

With that said, radars have a bunch of characteristics that makes it somewhat easy to identify. Frequency, PRF (pulse-repetition frequency), waveform, etc.

The bands at which radars operates are regulated, and various agencies - military and otherwise - will pick up radar transmissions. These tiny DIY projects aren't a huge "threat" in that sense, but if you're active anywhere near a radar installation, and especially military ones, it could land you in hot water, real fast.

I'm not sure about Finland, but where I live, civilians for example are not allowed to fingerprint or build fingerprint databases of military radars. In fact, the military is the only ones allowed to do so.

The power-scaling makes these things go from DIY to non-DIY real quick.

Would be cool to see this guy build a phased-array (radar)

Source: Worked with radars

>Anyone know how .mil aircraft would interpret being tracked by a 6 GHz radar build by a civilian(yes I am aware that his estimated max distance is 1200m, assume he increased that by a factor of 10 with larger antennas or something)?

No specific knowledge, of course, but I'd imagine it wouldn't trigger a serious threat warning. Military TWRs are highly tuned systems dedicated to the threat environment they're expected to operate in.

Depends on the system, any pulsed emission of sufficient strength should start triggering things as you can't know for sure what bands and systems are going to be used against you. I believe they usually have 'unknown' IDs for stuff that isn't in the threat database.

There is a story of a police speed camera catching a fast jet going around. The snopes debunking has some interesting technical details of why this is really unlikely https://www.snopes.com/fact-check/police-radar-missile/

But of course people have built nuclear reactors at home, so they could probably assemble a sufficiently powerful radar too.

Edit: didn't mean that people need to power homemade radars using homemade nuclear reactors! I'm sure that's not necessary :)

14dBm EIRP is extremely low power for a radar however. Off the bat the author is using 1W (30dBm) PA, and you could readily get another 20dB from a moderate sized antenna (e.g. repurpose a scrap DirecTV dish).

Unless you're next to a base, it's the FCC that will come knocking long before the military.

Regarding this "The suspiciously cheap $15 FPGA had equally suspiciously date and lot codes covered (white rectangles on the FPGA chip in the picture above). I have a development board of the same series chip with markings intact, so it definitely shouldn't look like this. It did end up working, but I wonder what the origin of these chips is."

I spoke with a GoPro operations guy once, and asked how you can get the chinese ripoffs at $40, that looked the same as their $400 "real" deal. He said that they literally get chips from scrapyards, like they get batches of memory chips that didn't pass the tests etc. They might work only in a very limited temperature range for example. Also many other components are picked and speced to work 6 months on avg not 6 years.. They know that most of those "gadget" products are bought and used for a weekend and then put in a cupboard anyway.

It would be fine for an R&D project like this but I would be very careful if I was going to make a commercial project (then you also have the politics side, probably fuelled by some paranoia - there are legislations coming up which prevents you from using chinese silicon components in products for certain markets)

>GoPro operations guy once, and asked how you can get the chinese ripoffs at $40, that looked the same as their $400 "real" deal. He said that they literally get chips from scrapyards

Chinese consumer electronics companies simply do not put huge margins on stuff. Here a great example, a $45 AlienTek DP100 100W USB-C micro lab power supply https://www.youtube.com/watch?v=Pd6LG7iP2GQ something like this would cost $500 with Digilent logo.

>speced to work 6 months on avg not 6 years

while original gopros suffer above average defect rates and cant record at 4K without baking itself to death/random shutdowns.

The Finns strike again. Just the depth and number of different areas of expertise is insane to me. It seems he planned it all out, had it printed in China, and then 1/2 of the boards actually worked. That's like building a whole backend+frontend app and hoping it works after a month of coding on startup.

Badass work. Rarely do you see such technical depth being demonstrated across the entire stack from RF to hardware to firmware to software. The article just gets better and better as you read on.

Speaking of radars... Did you know a useful radar is now just 34 US cents?[1]

And that chip? Not even designed for this application - it is a PIR sensor chip repurposed.

The core of the radar is just a single multi-Ghz transistor (in the middle of the left half of the PCB), and a super clever feedback circuit made out of fancy shapes of copper traces that probably took someone years to design and perfect.

[1]: https://www.aliexpress.com/item/1005006240906322.html

Why the ground planes are on layer 2 and 6 instead of 1 and 6 ?

Naively, as someone who doesn't have high frequency PCB design experience, I would have placed my grounds to form a shield and put my voltage plane on layer 3 or 4. I am sure that there is a good reason behind that choice but I don't see it.

Putting ground planes on top and bottom layers isn't usually done with high speed PCBs because components are there. There would need to be a cutout on the ground plane near every chip. High speed signals really need continuous ground plane and ICs, especially RF ICs, need short access to ground. Second layer is the best layer to minimize the distance from ICs to ground plane.

It is common to route high-speed signals on the top and bottom layers to avoid vias that cause impedance mismatches even when back drilled ($$$).

To maintain a specific characteristic impedance, you need a plane (GND) some distance from the traces which themselves have a specific width. Without a plane under/above the signals, you can't get a specific impedance value.

You can additionally fill the top and bottom layers, which marginally affects the impedance.

> Above is the final DDR3 routing on all the PCB layers. Layers 2 and 6 are ground, 5 is supply voltage, and others are reserved for signals. Two grounds are needed for correct impedances on the top, middle, and bottom traces of the PCB. With only one ground plane, the distance from the signal to ground would be too large on either the top or bottom layer.

Is the textual description in the article correct? To me the images make it look like signals are on 1 (red), 3 (orange) and 6 (blue), with ground on 2 (green) and 5 (pink) and supply voltage on 4 (teal). If you match some vias, you will find that 2 and 5 are definitely connected.

thanks, layer 2 and 5 would fit the rest of the description and the reason would then be in the text :

  Two grounds are needed for correct impedances on the top, middle, and bottom traces of the PCB

Obviously not the designer, but my guess is to lower capacitive loading on the matched high frequency signals (those with squiggles on top and flood underneath) and make modeling and IC mounting easier without flood. Note there's not flood near those either. I'm not quite sure on the ordering in the picture but it looks like 123/654? (edit ahh looks like the text was wrong and it is 123/456 with 2&4 Gnd). I'll note it looks like the internal "low speed" digital signals are squeezed between the gnd/pwr (edit: between gnd/gnd) planes, which is probably good since they're usually the biggest source of "noise" if you keep it away from other PCBs. You'd definitely want power/gnd planes immediately next to each other since bypass caps don't work at anything close to this frequency.

AFE7225 is my goto for applications that require both DAC/ADC. ~$40 in single qty. Just limited to half the sample rate achieved here.

I personally wouldn't have bothered with Zynq here, so much easier to interface any FPGA with any COTS SBC over Ethernet. Or just use the $160 ZUBoard 1CG, perfect for quick prototyping.

Can someone explain why those differential pairs are routed with so many curves instead of straight paths? (E.g. see the photo under the “ADC and DAC rotting” section)

"The traces are length-matched with squiggly lines[...]. The trace matching requirement is ±10 ps according to the Zynq PCB design guide, which is approximately ±2mm in trace length. [...] There is also some delay difference inside the FPGA package which should be considered in the length matching."

This is a shortened excerpt from the article. It can be found below the image with six colorful images of PCB layers. I'm curious how the delays inside the FPGA package are known. Is there a table which pin adds how much delay to a signal or something like that?

The FPGA manufacturer has characterized all their package pin delays. It's possible to export a csv file with internal delays of the package for each pin from the FPGA design tool. With Xilinx Vivado it's just File -> Export I/O Ports.

Does KiCad allow to take these delays for each pin into account automatically or do you need to do all that manually?

Edit: it's using the pad property of "pad-to-die-length" if doing it manually, right?

No, it needs to be done manually. It wasn't as tedious as it sounds though. Most of the pins are very close in delay already and there were just few traces that I had to adjust a little.

How is kicad getting along with RF work in 2024, would you say?

I did some prototyping but never got the stage if actually fumbling around with PCBs a few years ago, things seemed to progressing quite well.

Could this be adapted into a voxel system? Seems like it could be cheaper than LIDAR which are a huge cost for why the HW for self-driving systems are so expensive & work in far more environments that LIDARs struggle with. I suspect getting multiple directions simultaneously is hard?

Not without significantly complicating your antenna setup (and the data processing setup too). You guessed it, getting multiple directions simultaneously is hard. Note how the current system only detects distance and speed in 1 dimension.

Here's an analysis from someone smarter than me:

> To enable the new features, radar systems now use multiple input/multiple output (MIMO) antenna arrays for high-resolution mapping. Traditional radar systems usually contain two to three transmitting antennas and three to four receiving antennas, which lead to a beam providing limited short-range coverage and a narrow field of view unable to generate images. The limited angular resolution is insufficient to differentiate among vehicles, pedestrians, or objects that are close. The MIMO approach increases the underlying channels from only nine to anywhere between 128 and 2,000. Given radar’s significantly lower costs — even with all the enhanced technology — it’s easy to see how the two technologies will increasingly be on more equal footing.

https://www.oliverwyman.com/our-expertise/insights/2023/jul/...

Cutting-edge AESA radar like on the F-35 is incredible. It actually looks like a black and white photograph. I think your guess on antennas is roughly correct based on what we know about modern AESA.

Isn't that Synthetic Aperture Radar though? You can get similar results (black and white aerial pictures) by strapping a pretty basic siso radar system on a drone.

Look at the forward looking automotive 4D (distance, speed, azimuth angle, elevation angle) radar systems. The new ones work at around 80GHz, and the entire thing comes in one integrated, tiny package, 16x16 phased array antenna already included with the MCs and FPGAs on the same board.

To go from those 4D radar maps to a voxel system requires a whole lot of software, of course.

The end goal seems to be to beat LIDAR on price and reliability (turns out moving mirrors don't like years of constant vibrations), while delivering enough resolution for self-driving.

Beautiful piece of work. Henrik has a long history of interesting radar and other RF data-acquisition projects of the sort that you don't see publicly documented much, at least not at this level of quality.

> Cost was 330 USD for PCB manufacturing and assembly of two PCBs and additional 225 EUR (240 USD) for components from Digikey that I soldered myself. This is including 24% VAT and shipping costs.

I actually didn't use any expensive test equipment, only oscilloscope and multimeter. Even design and simulation software was all open source.

Expensive signal analyzer or spectrum analyzer would have been useful, but they aren't absolutely necessary. It's possible to use the radar itself for many tests and debugging.

I have tried to limit the projects I do on my own to only the equipment that I have home and open source software.

（评论） (comments)

（评论）
(comments)