Final project chronicles

---

NB: This page is structured as a log. A R&D Chronicle.

• My FabAcademy final-project, due by June the 18th, has been simplified to "just" the thermal-management & transportation box of the spectrometer.
• Complete build files and documentation of the Gamma Cool Box, as a independant open-harware project, is expected for Jully 2018.
• Complete build files and documentaion of Gamma v8 (pre-commertial spectrometer) will be made public, as part of a crowdfunding campain, in 2019 .
  In the mean time: if one has the wish and the skills to help develope Gamma, please contact me! :)
• Build files of Gamma v6 will be shared through my FabAcademy repo in June 2018, but documentation will be very limited and version is still unstable.

---

I did a second run of measurments at the hospital. I studied the serum of patients with more or less advanced liver disease. I especially studied the drift of the absorption with respect to the serum temperature (over the range 5 to 35°C).
Results were encouraging, but I need to reduce the thermal cross-talk between the different elements of my spectrometer. Indeed while the temperature of my sample oscilate between 5 to 35°C (with a 1st peltier module) and that I thermalise my LEDs @ 20°C (with a 2nd peltier module), I need my Photodiodes and my electronics (especially the ADC) to remain at constant temperature. With deviation of less than 1°C.
I could cut the spectrometer in differents blocks, separated by 5 to 10mm each:

• The LEDs driver PCB, which needs to dissipate about 10 Watts
• The LEDs PCB & its peltier
• The cuvette holder & its peltier
• The Photodiodes PCB, which is very close from the cuvette holder

I could then attach those blocks on independant radiators (ie Aluminium base plates), press-fitted into an insulator (eg a wood plate). I would just have to be able to control the temperature of the radiator plates...
Also, since such a set-up would have to be distributed all around the spectrometer, it could be a good opportunity to integrate it as a transportation box.

---

My FabAcademy Final-Project is to make, using only FabLab materials and DIY tricks, a multi thermal control and transportation box. Connected, of course.
The box will be designed for my spectrometer, but could also be used/addapted for other kind of scientific instrumentations projects where thermal management is critical.

---

For the thermal dissipation, I will use CPU fan cooling-systems. As they are mass-produced, there are very affordable. I chose the GAMMAXX 200 of Deepcool.

Key specs:

• Easily available online for 12€/unit.
• 12V, 200mA
• 1000 to 2000 RPM, 30dB noise max
• 42 CFM (feet3/min) = 20 L/s
• 95W dissipation

Use the CPU fan 4 pins connector standard:

• GND
• Vcc: 12V DC, 13.2 max
• Speed sense: 2 pulses/rotation
• PWM control: 25kHz (+/-3kHz), 5V max, 0.8V HIGH/LOW threshold, duty-cycle tunes fan speed

In the Principle and practice assignement I documented a first sketch.

In the Embedded programming assignement I evaluated the nRF52 MCU familly. No go. Nordic is not Fablab friendly.
For now I will use Atmel AVR MCU, only. Whith USB linking through Arduino or Eclipse IDE. If time, could add an external RF IC through I2C or SPI, like this one.

---

For the peltier, Ill use the TEC1-07108HT:

• 5€
• 30x30x3.5mm
• 8.5V/8.5A max (70W)
• 0.9 Ohm => P_joules=65W at Imax
• Q = 40W at ΔT =0°C
• ΔT max = 68°C (Tmax=225°C)

For the peltier driver, I'll use a high-current DC motor driver + a low-pass LC filter in order to smooth the current. Maybe the DRV84x2 of TI:

• Dual bridge (ie: 2 channels)
• PWM controle. I would have prefer I2C, like the DRV8830, but ha.
• Fault and auto-shutdown: Over Temperature, Over/Under current
• 3A/channel for DRV8412 (6A peak) @ 3.4$/unit by 1k - 6.6€ in retail
• 7A/channel for DRV8432 (12A peak) @ 6$/unit by 1k - 12€ in retail. Few suplier have it in stock...
• CBC (Cycle by Cycle) current limit
• GND to PVDD operation! Rare for high-current bridges (usually 2 or 5V to PVDD), but versatil: allows low impedance loads.

Why use LC filter and not RC filter? Why a low-pass filter at all? I've read a lot of obscure explaination on the internet. Here is the simple truth:
DC motors behave like big inductances. They would smooth any PWM. Peltier behaves like a simple resitor. When the driving current of a peltier is turned off (eg in the low of a PWM), the heat flows from the hot side back to the cold side. We lose the ΔT of the peltier. It's inefficiant. Ergo the filtering. Since resistance means joules losses, especially at 8A, we don't use resistors in the filter.
Shielded Inductance with L>0.3mH and I_max>5A? 9 options on mousser...
Since cut-off frequency will be limited to ~1kHz, it would be great to have "high frequency" PWM. Meaning 20kHz or above.

---

Other parts:

• Power resistance: 10W, 5 Ohm
• DRV8830, the same one with use in our "Machine" assignement.
• ATmega32u4 (the Arduino Leonardo MCU, with integrated USB coms)
• MCP9800, and other temperature
• IMS PCB for final design
• Digiscope stock: 1206 LED, R, C.
• 220V/50Hz -> 12V DC (150 to 200W)
• Thermal paste
• Little SPI screen display?
• R_sense for monitoring consumption?

---

Raw stock:

• Aluminium: 200x500x10mm - 74€ here
• Wood (solid beech): 300x600x18mm
• Foam: 80mm thickness. It can be cut on the Shopbot using our longest bit, as documented in the Week8 assignement

---

For my wildweek assignement, I worked on a single "thermal line" of the Gamma Cool Box.
I found a simple way to assemble the wood, the aluminium, the peltier and the fan. With this sub-assembly validated, I could then sketch the complete box.

NB: 10mm aluminium may be too much. It takes a long time to cool or warm (0.02°C/s). Could be changed for 8 or 5mm.

---

Using Solidworks I adapted the design to my Gamma spectrometer. I added a 5th fan and aluminium pipe (on the bottom left). I will use it to cool the main PCB.

Source files available here.

Here is a hand sketch shematics of the PCB:

---

Track width for 35µm Cu PCB:

• 8A => 2mm (ΔT+50°C) - 1.5mm (ΔT+80°C)
• 5A => 1mm (ΔT+50°C) - 0.8mm (ΔT+80°C)

NB: Dissipation will be better if using IMS PCB. But through-hole componant would be very problematic...

---

Designing the Probe PCB with KiCAD
1st part of my Input devices assignment

---

Designing the Main PCB with KiCAD:
1st part of my Output devices assignment

The deadline is getting closer, I need to prioritize if I want to leave enough time for debugging and programming...

• The Main PCB will be simplified. No MCU only the outputs ICs (MAX31790 and DRVs). I'll use a Leonardo (which has the ATmega32U4) for now.
• Not all the 5 base plates will be fonctionnal ^^. Maybe 3.

TO DO before Assembling phase:

• Associate the footprints of the componants DONE See 1.2/ of my Output devices assignment
• Root main PCB DONE See 1.3/ of my Output devices assignment
• Engrave main PCB DONE, see below
• Solder main PCB DONE, see below
• .
• CNC the other thermal pipes (need accurate Z zero-ing): DONE
  
• .
• Test probe PCB DONE, see below
• Solder remote transistor-probe DONE, see below

Probe PCB

See 2nd section of my Output devices assignment
This PCB can measure its own temperature, and the temperatures of 3 remote transitors's diode, with +/-1°C accuracy and communicate it through I²C. For now the transistors are soldered on the PCB (top left). Later thin wires will connected the transistors to the PCB.

Above the probe-PCBs you can see the LEDs-PCB of Gamma v6, which also has a temperature sensors communicating through I²C (on the top of the PCB), accurate at +/-0.1°C.

Main PCB v0

I didn't finish the rooting yet. It is not a working PCB but it will allow me to check the footprints; and also find the motivation to finish this 1 layer-rooting challenge.

Rooting Main PCB v1

See 1.4/ of my Output devices assignment
I did have to add 18 jump resistors (0 Ohm) in order to keep the boad 1 layer. With 200µm tracks and 200µm clearance, it is possible to fit 4 tracks under a 1206 package!

Making Main PCB v1

See 2/ of my Output devices assignment

Probe PCB

See 4/ and 5/ of my Input devices assignment

---

I did a second vesion of this PCB, the KiCAD project file has been updated with the last files version. The fabrication file is here.
Then I tested it:

When there is no "adress resistance", the MCP9804 takes the address 0x1C.

Main PCB