Interesting new project

How hard can you bang the source?

You could almost do time resolution of depth. Use a gain-switched laser and a fast receiver. Maybe average one photon per shot and use an avalanche-type detector. That might be worth thinking about. I guess the velocity of light drops in a scattering medium like a baby.

Does pulse ox use multiple wavelengths?

It's nice to do stuff that helps people.

Hi Phil,

that sounds as if effect modulation (like in "lock-in amplifier" ...), in addition to sensing the baby's pulse frequency, could be useful.

Would that help? And be feasible using a low-cost sensor?

Good luck for that project!

Dieter Michel

About a watt, peak. Don't want to fry the baby's eyeballs if it's presenting face forward, or burn the mum's skin.

You need a lot of detector area, which makes fast receivers hard unless you use something like a microchannel plate PMT. The light is diffuse, so you can't usefully play concentrator tricks to reduce the area.

It's very much analogous to carriers propagating by diffusion in semiconductors--on the order of c*(mean free path/propagation distance)**2, with a factor of order unity to take account of 3D propagation.

Usually two, often 600- and 900-nm-ish, depending on available LED wavelengths. You can see the Hb/Hb02 absorption spectra at

. It also helps a lot with rejecting room lights and stuff. You typically phase-lock to the pulse rate for noise rejection purposes. (This requires some smarts because the jitter is very large.)

The blood spot detector we did some years ago also sensed haemoglobin, but needed to reject a huge overlapping absorption from the brown-egg pigment protoporphyrin--like 10,000x stronger than the detection threshold, which is 80 dB electrical. (That's looking for a 1% absorption signal in the presence of a two-AU background, i.e. 99% peak absorption from the protoporphyrin.)

That one worked by picking another wavelength on the opposite shoulder of the protoporphyrin peak, so that the two went up and down together as a function of egg colour, and then take one minus the ratio. We didn't invent the scheme, but we did come up with some new calibration and post-processing wrinkles that allowed the wavelength tolerance of the filters to be relaxed quite a lot, which helped reduce costs. (Just getting rid of the usual flashlamps and PMTs was the main cost-reduction freight.)

The challenge of the present project is more in the SNR and background rejection areas.

Yup. Especially mums and babies. My younger daughter is planning to get married in the fall, so grandchildren are a strong possibility in the next few years.

Cheers

Phil Hobbs

Chopping the source can help reject additive background, for sure. We'll almost certainly do that, maybe with direct-sequence spread spectrum to make it robust against fluorescent and LED lights, PWMed displays, variable-frequency motor drives, and suchlike vaguely-periodic interference.

What it can't reject is the shot noise of the background, so we'll need a black elastomer sheet as part of the probe. The modulation also has to be slowish, because big photodiodes have a lot of capacitance. Capacitance differentiates the voltage noise of the front end, causing a noise current that rises linearly with frequency. In a low-light measurement that gets ugly in a hurry.

And be feasible using a low-cost sensor?

Illumination LEDs are pretty cheap per watt, IR ones less so. There'll be a tradeoff of source cost vs. light transport cost, for sure. We might wind up with a disposable skin rather than a disposable probe, the way they do with fever thermometers.

Thanks! Praise God for mums and babies!

Cheers

Phil Hobbs

Modulating the source is always helpful. It doesn't have to be tens of kHz but it would be best to avoid the usual suspect frequencies such as the first five or ten odd harmonics of 60Hz (such as 180Hz) and for international markets also those for 50Hz. This could be made adaptive so that the system first ascertains which kinds of optical noise is present and then automatically selects an operating frequency for the LED and the photodiode RX. Icing on the cake would a wee bit of spread spectrum behavior in there.

Yes! Especially those who then stay home with them when they grow up. I am forever grateful to my mom that she did. Can also be dad who stays home like it is with neighbors.

I think we'll probably wind up with a moderate-bandwidth direct-sequence SS system, ideally designed to have holes where the well-known interference sources live. I've never gone through an FDA certification process, but I expect that making the system's behaviour vary with interference conditions could make that a lot more complicated. Is that likely to be so?

Yup. Maybe harder to maintain spiritual leadership that way, but there are lots of ways to succeed. Having children raised by strangers is usually very sub-optimal, unless the parents are hopeless.

Cheers

Phil Hobbs

Generally not, as long as you can describe the behavior and show that the energy stays within a certain band. In this case it's all just optical spectrum anyhow when it comes to tissue exposure. Modulation would, from a regulatory POV, come into play if if was highly pulsed. High peak energy versus low average, et cetera, but yours doesn't sound like that. The FDA is mostly concerned about stuff such as maximum permissible energy into tissue and with ultrasound the max levels were very different between adult and fetal exposure. That has to be very well documented. Also, if you are around the limit and must maintain a max exposure time that timing cannot be done solely by software or firmware. It has to have a hardware limiter.

It's been almost two decades for me but the golden path was always a

510(k) FDA process which basically claims that your device is very similar to existing and already approved technology. "It's like a mouse trap but ours is better". If there are no predicate devices then its the full PMA process and at least in my time that used to be super-onerous and almost insurmountable by a start-up, financially.

We often got certified in Europe or Japan first because that was less difficult and faster. For that reason we did clinical trials a lot over there as well.

+1

For spiritual leadership that's actually the best upbringing. By far.

A factor of two is available if you polarize the light source; that can also diminish multiple-scattered light from non-target parts of the intended light path.

How about using a pulsed source, and delay-line-amplification to allow the receiver to have a slow pulse response, but still get roughly one-pulse-worth of received signal while rejecting the long tail? EMI crosstalk might be easier to design around than electron recombination lifetime, at high pulse frequency. So, you let the receiver be slow.

Aren't fetuses liquid cooled ?

Sure. So are you!

Cheers

Phil Hobbs

Nah, highly-scattering media destroy polarization information. The mean free path is tens of microns.

It wouldn't even be that complicated--we could make the pulse as short as we like, and just keep the rep rate down and detect slowly. Thing is, the allowable dose doesn't go quite as fast as 1/(duty cycle), so you wind up losing SNR instead of gaining it.

It's the e_N*C noise that's the main issue.

Cheers

Phil Hobbs

I have a technique that I invented long ago. I call it binary sampling. It is similar to delta modulation, invented in France in the late 1940's. I have searched for the patent but was unable to find it.

Delta modulation samples the waveform multiple times per cycle, and generates a single bit to tell if the waveform has increased or decreased since the last sample.

It is very efficient since it uses a single bit per sample instead of one or more bytes. It was used in some old Sony compact discs, in some telemetry applications, and on the space shuttle for voice communications. It has the problem of establishing and maintaining the voltage reference.

Binary sampling is very similar, except instead of sampling multiple times per cycle, it allows one or more cycles between samples.

The amazing thing is this rejects noise riding on the signal. It can follow a signal that is buried in -30 dB of noise. Since you control the sample time, you can sweep through the entire waveform and recover signals that would be impossible any other way.

The system works. I have already built and tested it. I am having trouble getting google drive and sabercat to work, so I can't post the schematic and waveforms. However, if you are interested, I could email them to you.

I am too old, and after a number of strokes, I am too weak to pursue starting a company to commercialize this technique. So I am making it available to anyone who can use it.

If it can help you, I would be very happy.

BTW, it also has application in GPSDO's where it can help remove the jitter caused by sawtooth error in the 1 pps timing. It operates in real time, so you don't need the time delay and granulation noise of current methods.

Sounds very interesting. I'd like to have a look at it.

Cheers

Phil Hobbs

[...]

Except our liquids often have hops in there :-)

Mine mainly ice cubes.

Cheers

Phil Hobbs

[...]

OK, I'll go through the documentation and clean it up a bit. A lot of the links have gone 404 - I'll see if I can find newer versions. I also have some new information to add. I hope to be able to send it in the next day or so.

Depends on your IR wavelength; about three thousand microns of my skin is pretty transparent to a security IR camera (lookit all those veins!) and skin is pigmented to stop light. Amniotic fluid and fetuses are less opaque.

I'm thinking there's scattering, but mainly forward scattering.

Try it at 30 mm. (It has to go through some significant amount of subcutaneous fat, for a start.)

And the obvious veins in your hand are large compared with their depth, so they're still visible despite the blurring due to diffusion. You sure aren't seeing your biceps in that NIR image.

Well, you'd be wrong then. Forward vs. reverse depends on the size of the scatterer--Rayleigh scatter by air molecules has a dipole pattern, i.e. a single molecule scatters into 2 pi steradians.

Multiple scattering destroys polarization for various reasons, but one inescapable one is topological phase. If you make light traverse a path such that the k vector encloses a solid angle Omega, the polarization rotates by 2 Omega. See e.g.

and the Crabtree reference at the bottom of the second page. This works regardless of how it's done, including mirrors and optical fibres. Doing that over an ensemble of 3D paths through a scattering medium homogenizes the polarization pretty thoroughly.

Cheers

Phil Hobbs

Did you get my emails? I sent them on Monday to your electrooptical.net address. There were two. The second is more important.

Do you have a spam filter that rejects attachments or large files? I notice you have no protection on your email address. You must get swamped with offers from Nigerian princes:)

There are a number of ways to obfuscate your email address to block spambots. Here's one that uses javascript, but leaves the "mailto" prefix exposed:

The problem with this is spambots also use javascript, and they search for the "mailto" prefix.

I wrote a bayesian filter to block spam, but it never worked very well. Spambots have an infinite variety of ways to get past these filters.

I like this approach much better since it also encodes the mailto:

After switching to email encoding, I never got a single piece of spam from my online address in 20 years. Of course, that didn't help block spam from vendors I purchased items from. So I had to use a different email address.

For those, I use disposable email addresses at

It never fails. If a vendor persists in sending you junk mail, just delete the address you assigned to him.

Other email forwarders turned out to be unreliable, or do not allow you to assign the email address yourself. E4ward. com has never lost an email, but some sites reject the forwarding address and want you to give a real one.

Then it is a choice if I want to do business with them or not, and how to persuade them to stop sending me junk advertisements. I usually regret giving them my real address.

Let me know if you got my emails:)