Teensy 3.1 Power supply clarity.

Status
Not open for further replies.
I doubt a 3.3V zener would do much other than eliminate overshoot. It would not eliminate undershoot, since the breakdown voltage of the zener would be above the ambient voltage. Thus, I would go after a zener with a lower voltage than 3.3V.
 
So I got my hands on a 1N4370A zener diode, which is supposed to be 2.4V, however I am only seeing 1.9V on the Oscilloscope. My connections are, diode to ground, then AREF to diode (cathode). This part was recovered from an older piece of equipment, not sure if that is an issue. I do remember a discussion about this before, I just need to find it. I think it had something to do with adding another resistor.
 
I have been looking further into your problem of reading the values from your load cell which has a wheatstone bridge output ( see diag in #19 reply ).
You need to take the two voltages at D and B into the ADC as a differential pair. If you are powering from lets say 5v D will be 2.5v or there about ( depends upon R1=exactly R2 ). The input from B will be 2.5v plus your mV of strain gauge input which represents your LOAD value.
When ADCs are used in a differential mode then TWO reference voltages are used, VrefH which will be set to be just higher than your maximum input voltage and VrefL which will be just lower than your lowest input voltage i.e. VrefH will be just greater than voltage at B and VrefL will be just less than the voltage at D. The difference between VrefH and VrefL represent the scale of the reading that will be converted by the ADC. If it's a 16bit ADC then VrefH-VrefL will be divided down by 65,536. I.e. lots of resolution.
Unfortunately in the case of Teensy 3.1 the VrefL is tied to 0V. Immediately you have lost 2.5v of precision as everything is referenced to 0v.
I will come back with some suggestions for solutions when I have had a little while to think further.
 
Unfortunately in the case of Teensy 3.1 the VrefL is tied to 0V. Immediately you have lost 2.5v of precision as everything is referenced to 0v. I will come back with some suggestions for solutions when I have had a little while to think further.

One thing I have done in the past when I had this sort of issue is to lift a pin and solder a wire to it.

You could run a three-resistor voltage divider from the wheatstone excitation input to AREF (aka VREFH on the schematic) through the first resistor, then connect AREF to to pin 15 (aka VREFL) through a second resistor, and then connect pin 15 to AGND via a third resistor. In the process, you can create a very narrow 'window' of voltages that the ADC can sense by choosing large resistors for VREF-AREF and Pin 15-AGND while using a small resistor for AREF to pin 15.

Lifting the pin takes a very fine soldering iron, patience, etc. Then, you can attach a single strand of wire to it that you can then lead to your voltage divider. Again, takes time and patience. Certainly suggests rolling your own board if this were ever to go into any level of production. Lifting pins is not something you want to do en-mass with fine-pitched chips like the K20.
 
I'm back...If anyone disagrees with my suggestions then please jump in, this is a discussion board after all.
A few possible solutions come to mind.
First of all we need to get the voltage at V down to or near 0Va. W1.png
The drawing above shows one solution. It does require an isolated (from the Teensy) supply for the Load Cell. Any variation in it's supply voltage will give a change in the ADCs reading. Likewise any drift in the Zenner will change the ADCs reading.

Second thought

W2.png

Please excuse the rubbish sketch.
Connect the voltage at D to Teensy 0Va. Derive a Teensy reference voltage from the Load Cell's isolated supply.
Read the SINGLE ended voltage at B with ADC. No need for a differential input since the I/p is now ground referenced.

Lastly, and probably the better but more expensive solution is to use an external Differential I/P ADC connected to Teensy via I2C or SPI.
 
I think that using multiple power supplies is probably more complicated than it really needs to be.

You guys seem to be suggesting that you narrow the ADC references to get better absolute resolution on the ADC. That will work to a point, but I'd suggest instead using an instrumentation amp IC on the two outputs of the wheatstone bridge to convert the differential signal to single ended, and give you lots of gain so that you can use the full 3.3V input range of the ADC from the small range of the wheatstone bridge.

This has the added advantage of presenting a high impedance input to the wheatstone bridge and a low output impedance to the ADC input. If you look at the load cell website, it mentions that the output impedance of the load cell (basically the nominal value of the resistors in the wheatstone bridge) is 1k Ohm. The sampling cap on the ADC needs to charge through the output impedance of the ADC input, which is slower when the output impedance is high, and can affect the measurements if you don't allow for a long sampling time.

I would still suggest using a separate 3.3V supply for the wheatstone bridge and the instrumentation amp, but you may see good results even connecting directly to the Teensy's 3.3V supply because the wheatstone bridge and instrumentation amp both inherently provide high rejection of power supply noise.

You can find instrumentation amps in DIP packages, which should be very easy to work with (such as http://www.digikey.com/product-detail/en/INA126PA/INA126PA-ND/300992)
 
Good point. The ADC does not like high impedances and the op-amp is just the ticket to ensure a clean, buffered signal. The only downside is added cost and complexity…

Also, IIRC, the VREFL has to be attached to VSSA. See p. 36 in the K20 data sheet. So an op-amp is a better option than a multi-resistor divider.
 
Last edited:
The PGA in the 3.1 would almost be perfect, but you need a gain of about 1000 to bring the load cell max output swing to the full range of the ADC and the PGA only does 64x max.
 
I've used the INA125 in the past, but trying to keep this device simple and small and cheap. I already have to add a coin cell and crystal, since we have to log the time the files are recorded. I will try the 3.3V to the load cell and see how it goes. Since I am already using an LDO to get 5V and 3.3V, can I bypass the 3.3 regulator in the teensy. Meaning, can I feed 3.3V into the Vin pin, or is that pin always going to go through the regulator. Thanks again for all the info.
 
Status
Not open for further replies.
Back
Top