5 ms·

This kind of description can be made of any hardware piece. We came from a whole room, slow computer to an iPhone that can make billions of operations per secon

This kind of description can be made of any hardware piece. We came from a whole room, slow computer to an iPhone that can make billions of operations per second. I see the previous post as complaining about this not happening on the industry. Of course making an iPhone (or a OnePlus phone if you want to get into low-scale) is not easy, but the technology has advanced and standardized enough to be considered "cheap" for what they offer, and really cheap compared with 10-20 years ago. Why hasn't that happened with ultrasounds? Certainly most of the components have been around for a while and the process of making a "simple one" (in the way the average smartphone is simple ) should be a solved problem.

crististm 10y ago
Indeed, but for the imaging industry there are only two potential clients: medical and welding. The hobby/educational/large masses market don't make the cut so to scale down the costs. The IT industry price ranges were also in the millions of dollars before people had a need to play Angry Birds on a device in their pocket
JPLeRouzic 10y ago
* Several IC makers are selling integrated devices (TI, Analog, Maxim), for example at TI: LM96530 (Beamformer), LM96511 (VGA/LNA) and LM96570 (TR Switch). Those are eight channels ICs, they cost $200 together. There are ICs for 16 or even 64 channels. * Another claim is the need for a huge number of piezos, this probably a view from a remote past: With adequate coding and enough channels to enable beamforming, there is no meaning in having this extravagant number (200) of transducers IMO.
- guitarbill 10y ago
  e.g. the LM96530 is lifebuy, so it's basically discontinued. LM96511 & LM96570: not recommended for new design. In other words, those ICs are basically obsolete.
  - JPLeRouzic 10y ago
    Thanks for pointing out right to the important thing: Those ICs are not new, they are on the market since some years.
    - temac 10y ago
      And have probably been used in some cheap systems?
      - guitarbill 10y ago
        I don't know if the $200 quoted by the OP for all ICs is bulk pricing when they were new or $200 now when they are limited and special parts. But it's moot anyway, I doubt the ICs are the major price point.
- LiesNStartupPR 10y ago
  The basics of imaging require a certain number of elements. The acoustic window must be a certain size for the application, the elements must be smaller than a wavelength to avoid grating lobes (0.5 wavelength for perfect phased arrays in that regard). Divide one into the other, and you get a number of elements that's usually between 100 and 200. On what basis do you make your statement that this is too many elements? IMO implies opinion, do you have any facts or physics to back up your statement? Laws of physics do not change with time.
  - JPLeRouzic 10y ago
    >> do you have any facts or physics to back up your statement? I do not think that beamforming needs more than a few emitting sources neither a Nobel prize. Check no further than your Wi-Fi router's antennas. The technology you are describing is entirely different and older, as in your description there is no beamforming. In the technology you describe, there is a requirement for two emitters/receivers to be separated by less than a wavelength, however this strong requirement does not exist for beamforming. What makes beamforming useful is modulating and coding the beam. Another thing important is the SNR, as it is the addition of the waves that creates "beams". When I checked Google patents I found beamforming patents in ultrasound imaging since 1986, and super-resolution (pionnered in astronomy and microscopy) in ultrasound imaging since 2005 (US 20050160817).
blackguardx 10y ago
I'm a hardware engineer and might be able to explain a bit. Industrial electronics doesn't seem to move at the same pace as consumer electronics from the outside, but there often is incredible progress at the same pace and perhaps faster than the consumer segments. It is often hard to discern to the outsider, because they lack domain knowledge to recognize these advancements. On top of this, in certain fields, costs appear to remain stable and even go up a bit over time. In these industries, there typically isn't much demand for lower specc'd (AKA outdated) hardware. Often these products last such a long time that the used old models serve as the low cost market. Apple even copies this model by keeping their older products around longer than other companies and selling them at a lower cost. In addition, the development of many industrial electronic products is extremely expensive and the markets are small (think 100s to 1000s per year). The company typically has to charge a very high price to stay in business. While it's easy to look at something and think "I can make that with an Arduino or Raspberry Pi," this is basically the hardware equivalent of telling people you can make Facebook in a weekend (which used to be a popular claim here on HN years ago).
- mping 10y ago
  For me its the same as pharma. Can you tell me if a government (ex India) would like to devote resources to make cheaper ultrassounds they wouldn't be able to? I only buy non-technical arguments, unless the bill of materials is on the same order as the price of the equipment.
  - bayesian_horse 10y ago
    With pharma I think the main "cost" involved is about the risk of investment. It takes billions of Dollars to potentially make billions of Dollars from a single product. More likely you invest hundreds of millions of Dollars with no return at all. It takes a special kind of company, almost a bank, to stomach that kind of risk. And this extraordinary risk also demands extraordinary rewards. Medical technology isn't really the same thing.
    - frozenport 10y ago
guitarbill 10y ago
I came here to make exactly the same argument. Or cars, if we pick something that also needs certifying (from a purely manufacturing perspective). I get that most people would rather have a car, so more cars get made, so they're cheaper. (Athough roughly comparable, and there's different models. To me, it seems a car is more complicated than an ultra-sound system). The author of the original article made a good point that there may be a market for cheaper, non-medical ultrasound. The author of this article totally ignores this. Of course he's going to say they ought to cost as much as they do. > but for now take my word [...] seriously, you don't want to use these things without strong safety regulations. Why? Instead of being a condescending asshat and lording his career achievements over us, this guy could have explained if and why it might be dangerous to build one yourself, again for non-medical use. Wouldn't a good article with great arguments be more convincing than a list of credentials? (edit: I'm reasonably convinced that devices in hospitals probably should cost that much, but this article didn't change that significantly. some of the details about impedance matching were interesting)
- Xylakant 10y ago
  Cars, iphones and laptops are sold in quantities that are magnitudes larger than ultrasounds. Even if every school bought one for students to experiment with, you'd still only have all schools and doctors in the world as potential market. On top of that, medical devices are supposed to work in comparatively harsh environments (strong disinfectants etc.) for a long long time. Iphone lifetimes are measured in years, not decades and don't you dare and try to get a spare part for a 10 year old macbook. All of that means that evolution is much slower in that market - more small steps rather than big disruptive changes. I'd also challenge that cars are more complicated than ultrasounds. They do have more moving parts, but allowed tolerances are magnitudes larger than a few microns in all places.
  - ChicagoBoy11 10y ago
    ... but the "magnitudes larger" that they are sold is also endogenous to the regulatory framework around them. Suppose you had to get a PhD in physics to be able to get a drivers license. All of a sudden we would see a lot fewer buying -- and therefore far fewer people selling -- cars since it is so expensive to actually be able to drive them. With these ultrasounds, there is such a heavy regulatory framework at work -- from regulations impacting the device itself to regulations preventing hospitals from being built to regulations preventing hospitals from having as many of these as they want -- that they make the actual quantity demanded of these things artificially small. Your point about evolution of the technology also falls trap to this same fallacy: Again, if all of a sudden you are artificially making something much more expensive to produce, it should be no surprise to you that the refresh rate on the device would be smaller. Here you also have the added effect that since we rarely pay for our own medical care and instead rely on insurance, many of the incentives that would exist for companies to make cheaper, faster, better equipment also disappear. Lastly, I fully concede that it is quite possible that there is some basic, inherent quality of ultrasounds that make them more complicated than cars or any other pieces of tech. But that isn't important. The importance is realizing that the derivative of the state of these machines looks very different from other technology industries where competition and the free market are much more of a factor.

crististm 10y ago

Indeed, but for the imaging industry there are only two potential clients: medical and welding. The hobby/educational/large masses market don't make the cut so to scale down the costs. The IT industry price ranges were also in the millions of dollars before people had a need to play Angry Birds on a device in their pocket

JPLeRouzic 10y ago

* Several IC makers are selling integrated devices (TI, Analog, Maxim), for example at TI: LM96530 (Beamformer), LM96511 (VGA/LNA) and LM96570 (TR Switch). Those are eight channels ICs, they cost $200 together. There are ICs for 16 or even 64 channels. * Another claim is the need for a huge number of piezos, this probably a view from a remote past: With adequate coding and enough channels to enable beamforming, there is no meaning in having this extravagant number (200) of transducers IMO.

guitarbill 10y ago
e.g. the LM96530 is lifebuy, so it's basically discontinued. LM96511 & LM96570: not recommended for new design. In other words, those ICs are basically obsolete.
- JPLeRouzic 10y ago
  Thanks for pointing out right to the important thing: Those ICs are not new, they are on the market since some years.
  - temac 10y ago
    And have probably been used in some cheap systems?
    - guitarbill 10y ago
      I don't know if the $200 quoted by the OP for all ICs is bulk pricing when they were new or $200 now when they are limited and special parts. But it's moot anyway, I doubt the ICs are the major price point.
LiesNStartupPR 10y ago
The basics of imaging require a certain number of elements. The acoustic window must be a certain size for the application, the elements must be smaller than a wavelength to avoid grating lobes (0.5 wavelength for perfect phased arrays in that regard). Divide one into the other, and you get a number of elements that's usually between 100 and 200. On what basis do you make your statement that this is too many elements? IMO implies opinion, do you have any facts or physics to back up your statement? Laws of physics do not change with time.
- JPLeRouzic 10y ago
  >> do you have any facts or physics to back up your statement? I do not think that beamforming needs more than a few emitting sources neither a Nobel prize. Check no further than your Wi-Fi router's antennas. The technology you are describing is entirely different and older, as in your description there is no beamforming. In the technology you describe, there is a requirement for two emitters/receivers to be separated by less than a wavelength, however this strong requirement does not exist for beamforming. What makes beamforming useful is modulating and coding the beam. Another thing important is the SNR, as it is the addition of the waves that creates "beams". When I checked Google patents I found beamforming patents in ultrasound imaging since 1986, and super-resolution (pionnered in astronomy and microscopy) in ultrasound imaging since 2005 (US 20050160817).

blackguardx 10y ago

I'm a hardware engineer and might be able to explain a bit. Industrial electronics doesn't seem to move at the same pace as consumer electronics from the outside, but there often is incredible progress at the same pace and perhaps faster than the consumer segments. It is often hard to discern to the outsider, because they lack domain knowledge to recognize these advancements. On top of this, in certain fields, costs appear to remain stable and even go up a bit over time. In these industries, there typically isn't much demand for lower specc'd (AKA outdated) hardware. Often these products last such a long time that the used old models serve as the low cost market. Apple even copies this model by keeping their older products around longer than other companies and selling them at a lower cost. In addition, the development of many industrial electronic products is extremely expensive and the markets are small (think 100s to 1000s per year). The company typically has to charge a very high price to stay in business. While it's easy to look at something and think "I can make that with an Arduino or Raspberry Pi," this is basically the hardware equivalent of telling people you can make Facebook in a weekend (which used to be a popular claim here on HN years ago).

mping 10y ago
For me its the same as pharma. Can you tell me if a government (ex India) would like to devote resources to make cheaper ultrassounds they wouldn't be able to? I only buy non-technical arguments, unless the bill of materials is on the same order as the price of the equipment.
- bayesian_horse 10y ago
  With pharma I think the main "cost" involved is about the risk of investment. It takes billions of Dollars to potentially make billions of Dollars from a single product. More likely you invest hundreds of millions of Dollars with no return at all. It takes a special kind of company, almost a bank, to stomach that kind of risk. And this extraordinary risk also demands extraordinary rewards. Medical technology isn't really the same thing.
  - frozenport 10y ago

guitarbill 10y ago

I came here to make exactly the same argument. Or cars, if we pick something that also needs certifying (from a purely manufacturing perspective). I get that most people would rather have a car, so more cars get made, so they're cheaper. (Athough roughly comparable, and there's different models. To me, it seems a car is more complicated than an ultra-sound system). The author of the original article made a good point that there may be a market for cheaper, non-medical ultrasound. The author of this article totally ignores this. Of course he's going to say they ought to cost as much as they do. > but for now take my word [...] seriously, you don't want to use these things without strong safety regulations. Why? Instead of being a condescending asshat and lording his career achievements over us, this guy could have explained if and why it might be dangerous to build one yourself, again for non-medical use. Wouldn't a good article with great arguments be more convincing than a list of credentials? (edit: I'm reasonably convinced that devices in hospitals probably should cost that much, but this article didn't change that significantly. some of the details about impedance matching were interesting)

Xylakant 10y ago
Cars, iphones and laptops are sold in quantities that are magnitudes larger than ultrasounds. Even if every school bought one for students to experiment with, you'd still only have all schools and doctors in the world as potential market. On top of that, medical devices are supposed to work in comparatively harsh environments (strong disinfectants etc.) for a long long time. Iphone lifetimes are measured in years, not decades and don't you dare and try to get a spare part for a 10 year old macbook. All of that means that evolution is much slower in that market - more small steps rather than big disruptive changes. I'd also challenge that cars are more complicated than ultrasounds. They do have more moving parts, but allowed tolerances are magnitudes larger than a few microns in all places.
- ChicagoBoy11 10y ago
  ... but the "magnitudes larger" that they are sold is also endogenous to the regulatory framework around them. Suppose you had to get a PhD in physics to be able to get a drivers license. All of a sudden we would see a lot fewer buying -- and therefore far fewer people selling -- cars since it is so expensive to actually be able to drive them. With these ultrasounds, there is such a heavy regulatory framework at work -- from regulations impacting the device itself to regulations preventing hospitals from being built to regulations preventing hospitals from having as many of these as they want -- that they make the actual quantity demanded of these things artificially small. Your point about evolution of the technology also falls trap to this same fallacy: Again, if all of a sudden you are artificially making something much more expensive to produce, it should be no surprise to you that the refresh rate on the device would be smaller. Here you also have the added effect that since we rarely pay for our own medical care and instead rely on insurance, many of the incentives that would exist for companies to make cheaper, faster, better equipment also disappear. Lastly, I fully concede that it is quite possible that there is some basic, inherent quality of ultrasounds that make them more complicated than cars or any other pieces of tech. But that isn't important. The importance is realizing that the derivative of the state of these machines looks very different from other technology industries where competition and the free market are much more of a factor.