Wow that's really awesome, I'm just skimming the site at the moment, are you planning on selling the pulser boards etc. out of interest, if so roughly how much do they cost?
Hey! Theres a link for tindie in the doc. It's quote expensive, high voltage being the main reason, but I'm planning on reducing the costs and have a cheaper on-board extension for the pi once this hardware stabilizes
Very cool. There was something on HN about a year ago about why ultrasound machines are so expensive. This might provide some background on why something low cost makes sense.
Before trolling a bit, I remember the main costs don't come from hardware (though it definitely costs something) - rather for r&d, patents, and most of all proper certification, since, in the end, it's a matter of life and death.
That being said, this experiment is mostly a dev kit to help curious tinkerers explore the topic :)
Yes. Just come up with a non-medical/safety use for it and you are good. Check your car's engine to see if it will last through the next race - a very useful thing to do and nobody will think twice. Check a boiler to be sure it won't explode - you better get this right or people will die so don't talk about potential uses here.
The most enlightening comment there was the one that contained this:
"Still I ended up north of $3k for hardware, and several hundred projected work-hours. A lot of the cost comes from needing 50-100 transducers to get a useable image."
Ah, that makes a lot of sense - I was trying to find information on how ultrasound machines worked awhile ago, and most of the information I found just said that a ceramic piezo is used to emit the sound and measure the time until it returned. I could not fathom at the time what kind of math must be required to get a 2D image out of a single data point like that, and I couldn't find much elaboration on the subject.
But if it's just a few hundred piezos in an array, now THAT makes a lot more sense.
Yep. Array is the keyword that makes it easier to Google...phased arrays, curved arrays, etc. And of course, with a bunch of piezos, your signal processing problem is larger.
The point being that on these experiments, the sensor isn't an array of piezos, rather a single-element piezo. Cuts the costs (a fab invoices around 100-200e for such a sensor), but lowers the overall image quality.
It's not an array of piezos (or at least, not a one-per-column array of transducers). It's a phased array which they use to rapidly scan the volume, iirc something like 32 emitters and a single receiver in a commercial unit.
Source: I looked into building one last time we had a baby but didn't get around to it.
I have no idea how hackernews always reads my mind about this kind of things but I started thinking about this topic due to a personal interest a few days ago.
And my question is. Can the ultrasound probe be made of off-the shelf parts. I think I understand why it's not possible with electromagnetic parts alone (kind of a speaker) they just can't vibrate fast enough to reach the Mhz frequency range.
But would it be possible to use a regular crystal oscillator they should be cheap, can easily be found for Mhz frequencies and are basically the same technology as the PZTs used in ultrasound probes. Just remove the casing and excite them, would that work?
I'd been thinking about this technology recently as well, but hadn't gotten any further than perusing data sheets a little.
My impression is that quartz crystals as found in oscillators are so brittle that it's hard to use them for things like this; I guess PZT (lead zirconium titanate) is more robust somehow? There are also polymer-based piezo materials, like PVDF.
Another potential issue is that ultrasound machines send out an impulse instead of a single frequency. So maybe the quartz crystal oscillators aren't good at producing a short high-bandwidth signal?
That being said, there's a type of radar system called "stepped frequency continuous wave" radar that uses a bunch of single frequency transmissions instead of an impulse transmission. The basic idea is that instead of using a high-bandwidth transceiver to send and receive the impulse, you can use a low-bandwidth transceiver to send and receive tones, and then hop this transceiver over the large bandwidth to get a high-resolution image (the tradeoff being that it takes longer to acquire an image). Since ultrasound is basically radar, I'd imagine this technique could be used for ultrasound too.
The spatial resolution depends on the bandwidth of your transmission. An infinite-bandwidth signal is a delta in time, and gives infinite precision. Any finite bandwidth impulse will be a sinc in time, with temporal width proportional to the inverse of the bandwidth. So a higher bandwidth impulse will be shorter in time, which would give better temporal resolution.
Typical radar works by sending and receiving a high-bandwidth impulse, which requires high-bandwidth transceivers. Let's say that transmission occupies frequencies between f0 and f1. SFCW radar works by sending a bunch of individual low-bandwidth transmissions between f0 and f1. So each transmission is small, but together they occupy the same f0 to f1 bandwidth. Assuming the environment didn't change in the time it took to send all of those low-bandwidth transmissions, you've effectively simulated a high-bandwidth impulse using a bunch of low-bandwidth impulses. So the spatial resolution will be the same.
Could be indeed!
Augmented sensing, there's definitely something one can do. A ping (radar-like) every 5s, and in time you listen to echoes. The more intense, the higher the frequency.
Ping me through the contact in the doc if you're keen on discussing about this!
32 comments
[ 1449 ms ] story [ 1697 ms ] threadhttps://news.ycombinator.com/item?id=13230741
But also if you can't afford one.
"Still I ended up north of $3k for hardware, and several hundred projected work-hours. A lot of the cost comes from needing 50-100 transducers to get a useable image."
https://news.ycombinator.com/item?id=13231831
So you can make a very low cost one, but you trade off what use cases it's good for.
But if it's just a few hundred piezos in an array, now THAT makes a lot more sense.
Source: I looked into building one last time we had a baby but didn't get around to it.
As a side note, does anyone reading this have MAX2082EVKIT# lying around in a trash pile somewhere? I'm looking for one.
And my question is. Can the ultrasound probe be made of off-the shelf parts. I think I understand why it's not possible with electromagnetic parts alone (kind of a speaker) they just can't vibrate fast enough to reach the Mhz frequency range.
But would it be possible to use a regular crystal oscillator they should be cheap, can easily be found for Mhz frequencies and are basically the same technology as the PZTs used in ultrasound probes. Just remove the casing and excite them, would that work?
My impression is that quartz crystals as found in oscillators are so brittle that it's hard to use them for things like this; I guess PZT (lead zirconium titanate) is more robust somehow? There are also polymer-based piezo materials, like PVDF.
And if it can't withstand the vibrations it generates, how far off would it be from the required energy (many orders of magnitude?)
That being said, there's a type of radar system called "stepped frequency continuous wave" radar that uses a bunch of single frequency transmissions instead of an impulse transmission. The basic idea is that instead of using a high-bandwidth transceiver to send and receive the impulse, you can use a low-bandwidth transceiver to send and receive tones, and then hop this transceiver over the large bandwidth to get a high-resolution image (the tradeoff being that it takes longer to acquire an image). Since ultrasound is basically radar, I'd imagine this technique could be used for ultrasound too.
Typical radar works by sending and receiving a high-bandwidth impulse, which requires high-bandwidth transceivers. Let's say that transmission occupies frequencies between f0 and f1. SFCW radar works by sending a bunch of individual low-bandwidth transmissions between f0 and f1. So each transmission is small, but together they occupy the same f0 to f1 bandwidth. Assuming the environment didn't change in the time it took to send all of those low-bandwidth transmissions, you've effectively simulated a high-bandwidth impulse using a bunch of low-bandwidth impulses. So the spatial resolution will be the same.
How small can the probe head get? Can I tape it to my ringfinger? Any health hazzards?
My first try would be taping it to my finger and listening to the signals enveloped mapped to my audible range.
But thank you for the offer to discuss more :)