Looking to Hire a Veteran Engineer to Do a Design Review of Schematic + PCB

[Potomac]

Quick intro - I'm a hobbyist/maker developing a lab sensor that shines a beam of IR light across a distance of about 12 mm to get a density measurement for a given liquid full of yeast culture. I'm focusing at the moment on building the best version (high accuracy/high repeatability/low noise) of the light transmission sensor possible using off-the-shelf parts.

I'm attacking the problem in two ways.

1). First method was to use  IR slot sensor from Omron to take analog readings.  You can read a little about my first attempts in this thread.   https://www.eevblog.com/forum/projects/should-i-be-using-a-constant-current-or-constant-voltage-device-for-my-project/msg948298/#msg948298

I have since hired an independent engineer to design me a schematic and PCB in Diptrace that addresses some of my concerns about noise levels.  I want to hire someone from the EEVBlog community to do a critique of that design.  I hope this will serve to find any flaws before I send off the files to Gold Phoenix in China to get manufactured, and teach me something about the design review process.  This article recommends strongly that novices hire independent designers to do design reviews.   http://predictabledesigns.com/4-ways-for-hardware-startups-to-reduce-risk/


2). Second method I'm currently working on is to use an IR light-to-frequency converter that outputs a digital square wave that is proportionate to light levels. There is a lot of external circuitry needed to combat the noise with the Omron sensor. I'm hoping that the onboard light-to-frequency IC on this sensor will bypass the need for all that, and maybe even produce lower noise data.  http://www.digikey.com/product-detail/en/ams-taos-usa-inc/TSL245R-LF/TSL245-R-LF-ND/3095042

I have built a custom case mount that looks similar to the U-shaped Omron sensor described in bullet point 1.  I am having my first designer work on a schematic and PCB for that right now, and would in the future like to do a design review.  Again, the point here is to have a devil's advocate who catches any flaws, and also it will be a good learning experience for me. 

Does this sound like a good way of going about things to produce a good product and to teach myself a little bit about design reviews?   

I have a rough idea of what a design review would cost --- $100, $200, $300 - whatever it costs I'm happy to have a serious crunch of the numbers.

Please feel free to respond here or via PM if you think you can help me out. 

It would be great if you have Diptrace experience too.  I chose that program since it is more user friendly than Eagle. Makes it easier for me to follow along.   (http://predictabledesigns.com/pcb-design-software-which-one-is-best/)

[ebclr]

A free design point

For the purpose of match sensor's a god idea is to have "standard sensor" and a bridge circuit who will compare the tested device against the samples, The good point of this approach is that a lot of the problems like noise instability will occur in both sides and will be seldf cancelling,

After that you will have several groups of sensors that will match with other

[tautech]

Hardly a veteran but Tim specialises in contract design.
https://www.eevblog.com/forum/profile/?u=92330

[T3sl4co1l]

Hardly a veteran but Tim specialises in contract design.
https://www.eevblog.com/forum/profile/?u=92330

Thanks   

As with any optical project, the central questions will be:
- What are you doing about interference?
- Repeatability?
- Linearity?
- Is the wavelength appropriate for the sample?  Holder/cuvette?
- Does your setup give good specificity, i.e., does it respond well to turbidity and scattering, while rejecting transmitted light, ambient light, specular reflections, internal reflections between any parts, and etc.?

If the answers are "no" enough to be a problem, you'll have to figure out ways to deal with that.  For example, using a second ambient light sensor to subtract background.  Even better, using the same sensor, but chopping the illumination, so you measure ambient and illuminated levels in the same place.  Measure the two, subtract, repeat ad nauseum, and you can get quite excellent sensitivity.  If you need transmission (or absorption) versus scattering, two sensors could give you the ratio.  A rainbow of LEDs could be used to provide more color/wavelength options, or you can use an incandescent lamp (black body radiation -- white LEDs are actually blue and yellow, not truly "white"), maybe with an adjustable filter or prism to do it for a range of wavelengths.

Just some ideas.

As for just schematic review, that could be a couple hours for a simple polish, or it could be several weeks of involved work, if it turns out it needs serious redesign.

I wouldn't see a small project taking more than a week or two at the worst.  Complicated design projects can run into the months, and that's about it.

Tim

[AndyC_772]

I agree with Tim, it's important at this stage to have a clear idea of how good the end result needs to be, before it's possible to make any real assessment of whether or not the design you have will be fit for purpose.

A couple of hours' work might be enough to catch any actual bugs in the design and suggest changes. By 'bugs' in this case, I mean errors that will prevent it from working as designed, not features of the design that will mean it ends up not working as well as it needs to.

A very simple circuit might just apply a constant current to a LED and measure the resulting current through an opto-sensor of some kind. This might be perfectly OK, but it will drift with temperature and over time, it won't be repeatable from one unit to the next, and will likely suffer all manner of other performance limitations which mean it's not suitable for the task in hand.

Or, you could have a much more sophisticated set-up which includes the features Tim has described, and which will undoubtedly work much better but will cost more to develop, debug and produce. You need to be clear about what level of performance is needed, and to discuss it in detail with whoever designs your board.

I deal frequently with new customers who are unfamiliar with the process of having an electronic device designed, and by far and away the most difficult thing is establishing the design requirements on day one. The one thing I try and emphasize above all else is the need to specify clearly what the product is required to do, and how it will be used. Design changes are costly, and best avoided wherever possible.

It's also worth mentioning that if you have two different engineers involved on a project, you'll likely get three different opinions on it. Everyone likes to have their own input, has their own experiences and preferences, and there's no definitively 'right' answer to most questions. I guarantee that, if I were to review your design, I'd pick holes in it and recommend parts of it be done differently. And some other engineer would say exactly the same about any redesign I carried out.

[nctnico]

Or, you could have a much more sophisticated set-up which includes the features Tim has described, and which will undoubtedly work much better but will cost more to develop, debug and produce. You need to be clear about what level of performance is needed, and to discuss it in detail with whoever designs your board.

I agree. It might even be necessary to divide the project into a research phase and a product development phase. The research phase should be used to test the measurement method thouroughly and include part variations.

To the OP: why aren't you using a light to digital sensor with I2C interface? Intersil has several (including highly sensitive ones) and they have a specified accuracy over temperature. I'm about to use one in a project I'm currently working on to measure low light conditions. When I look at the specs from the sensor you proposed it seems the temperature has a big influence on the output frequency (see Output frequency and dark frequency versus temperature).

[Potomac]

Thanks Ebclr, tauntech, T3sl, Andy and nctnico

I'll comment on some of the technical questions you asked

T3sl4co1l, I responded in bold blue text below

As with any optical project, the central questions will be:

- What are you doing about interference?   In regard to interference, I plan on taking two measurements--one with the LED on and one with the LED off. Then "subtracting" as you say. However, my initial tests, don't have that code in place yet.

- Repeatability? Repeatability is important.  For the new PCB design, I will have a low ripple current AC-DC wall wart hooked up to the board.  Then, there is a constant current device supplying power to the 16 LED's in series.  There is also a noise filter for the phototransistors, among other things.

- Linearity?  I think it's safe to say that nothing here will be linear.  Will have to rely on curves, especially if I'm using anything involving a BJT, like that Omron sensor.  The Taos light-to-frequency sensor also has curvy data.

- Is the wavelength appropriate for the sample?  Holder/cuvette?  Infrared around 950 nm will be good I think. You can also use red around 600 nm, but it's not of big importance. Just have to pick something a stick with it.

The holder and cuvette width/distance is good. Most measurements are done 12 mm including the plastic walls


- Does your setup give good specificity, i.e., does it respond well to turbidity and scattering, while rejecting transmitted light, ambient light, specular reflections, internal reflections between any parts, and etc.? Accounting for ambient sunlight is the important thing.  Like I mentioned above, I think this can be compensated for by taking a measuring with the LED on and one with it off, then do subtraction in your Arduino code. 

Not sure about the scattering and reflections yet

If the answers are "no" enough to be a problem, you'll have to figure out ways to deal with that.  For example, using a second ambient light sensor to subtract background.  Even better, using the same sensor, but chopping the illumination, so you measure ambient and illuminated levels in the same place.  Measure the two, subtract, repeat ad nauseum, and you can get quite excellent sensitivity.  If you need transmission (or absorption) versus scattering, two sensors could give you the ratio.  A rainbow of LEDs could be used to provide more color/wavelength options, or you can use an incandescent lamp (black body radiation -- white LEDs are actually blue and yellow, not truly "white"), maybe with an adjustable filter or prism to do it for a range of wavelengths.

Just some ideas.

As for just schematic review, that could be a couple hours for a simple polish, or it could be several weeks of involved work, if it turns out it needs serious redesign.

I wouldn't see a small project taking more than a week or two at the worst.  Complicated design projects can run into the months, and that's about it.

AndyC:

I agree with Tim, it's important at this stage to have a clear idea of how good the end result needs to be, before it's possible to make any real assessment of whether or not the design you have will be fit for purpose.  Understood, will focus on setting specs from the beginning

A couple of hours' work might be enough to catch any actual bugs in the design and suggest changes. By 'bugs' in this case, I mean errors that will prevent it from working as designed, not features of the design that will mean it ends up not working as well as it needs to.

A very simple circuit might just apply a constant current to a LED and measure the resulting current through an opto-sensor of some kind. This might be perfectly OK, but it will drift with temperature and over time, it won't be repeatable from one unit to the next, and will likely suffer all manner of other performance limitations which mean it's not suitable for the task in hand. That's exactly what I experienced with my first perfboard setup. I will post an Excel with data in my next post below so you can see. I think there were a whole bunch of factors causing this, from the mistake of using a higher ripple current switching supply, to putting the LEDs in parallel, to not having a noise filter, to not taking two measurements and doing subtraction to account for ambient light, etc

Or, you could have a much more sophisticated set-up which includes the features Tim has described, and which will undoubtedly work much better but will cost more to develop, debug and produce. You need to be clear about what level of performance is needed, and to discuss it in detail with whoever designs your board.

I deal frequently with new customers who are unfamiliar with the process of having an electronic device designed, and by far and away the most difficult thing is establishing the design requirements on day one. The one thing I try and emphasize above all else is the need to specify clearly what the product is required to do, and how it will be used. Design changes are costly, and best avoided wherever possible.

It's also worth mentioning that if you have two different engineers involved on a project, you'll likely get three different opinions on it. Everyone likes to have their own input, has their own experiences and preferences, and there's no definitively 'right' answer to most questions. I guarantee that, if I were to review your design, I'd pick holes in it and recommend parts of it be done differently. And some other engineer would say exactly the same about any redesign I carried out. Understood about lots of different ways to accomplish the same thing

nctnico

I agree. It might even be necessary to divide the project into a research phase and a product development phase. The research phase should be used to test the measurement method thouroughly and include part variations.

To the OP: why aren't you using a light to digital sensor with I2C interface? Intersil has several (including highly sensitive ones) and they have a specified accuracy over temperature. I'm about to use one in a project I'm currently working on to measure low light conditions. When I look at the specs from the sensor you proposed it seems the temperature has a big influence on the output frequency (see Output frequency and dark frequency versus temperature).

Is this the Intersil sensor you're talking about?  On first glance, it looks like there are no configurations in infrared, just visible light. http://www.intersil.com/en/parametricsearch.html?g=optoelectronics&sg=ambient-light-sensors&f=light-to-digital-sensors#g=optoelectronics&sg=ambient-light-sensors&f=light-to-digital-sensors

In regard to the temp curve on the Taos TSL245 light-to-frequency sensor that I posted, I plan on putting this device in a temperature controlled space (incubator), so I think that should solve things.

If I were to put it out by itself with the window open, I think there would be more of a problem. Does that sound reasonable?

[nctnico]

Look at the ISL29033. It offers both IR and visible light sensing and has a specified error over range and temperature.

Edit: Another important thing to consider is that the amount of lights the LEDs emit isn't constant. This also varies with age and temperature so you need a way to measure the amount of light emitted as well. IMHO using several LEDs in series makes this harder. Did you think about using an infrared laser? AFAIK these usually have a build-in detector to measure the light output.

[mrpackethead]

US $250,000 and i'll have a look at it.
 

[Potomac]

Look at the ISL29033. It offers both IR and visible light sensing and has a specified error over range and temperature.

How do you get an infrared reading off of that one? The datasheet says there are two photodiode arrays.  Does one pick up visible light and the other picks up IR?  http://www.intersil.com/content/dam/Intersil/documents/isl2/isl29033.pdf

If so, what is the peak wavelength for the IR photodiode?  It appears to be around 830 nm on the chart on page 12 but that's all I can find

It looks like there are 6 pins,  with one providing data out  (6)

Attached a screenshot from page 12

[nctnico]

The ISL29033 has an I2C interface which allows changing settings (like selecting visible light or IR) and gain. The picture you show is the solder side of the ISL29033. Yes the peak IR wavelength seems to be 830nm. Also look at what I added above about using an IR laser with a built-in photodiode to measure and control the light output.

[brumbarchris]

I guarantee that, if I were to review your design, I'd pick holes in it and recommend parts of it be done differently

I guarantee that if I review my own designs after some time elapses (years) I'd pick holes in them and recommend different parts.

Regards,
Cristian

[seeigecannon]

Let me know if you have any questions on how to take a measurement like this as I have experience with this kind of thing. I can also throw together a circuit for you upon request if you want a second option.