Teensy 4.1 AGND (Retired?)


Teensy 3.6 calls out AGND.
Teensy 4.1 calls the same pin GND.

Can I please confirm that on Teensy 4.1 that the pin previously designated as AGND is no longer separated by inductor, and can be considered a standard digital ground?

... I ask only because the SnapEDA footprint for Teensy 4.1 drops with all the grounds on different nets (arbitrary names, which is fine)... but I am unclear as to whether AGND is still inductively separated in hardware or if . . . it is just now a netname difference.


Anyhow -
This is my first post. My experience with the Teensy 3.6 has been quite good and we have experimented thru 25pcs of them. Keep up the good work.

Thanks for responding.

I had reviewed the schematics and seen as much, but wanted to be sure. Things move so fast these days (I work startups. . .) that the most current documentation is often "hardware in hand".

Feedback on AGND

On a 3.6 I ran some basic ADC inputs.

* 0-3V
* LPF of arbitrary value

Very stable, rock solid running 10bit.

If you are reading this and trying to get better performance out of your ADC in a somewhat noisy environment, you MUST have a LPF (Low Pass Filter) on ANY of the microcontrollers. That is a universal truth and one not often enough spoken of.

Almost any value will work... and you dont have to go and calculate knee frequencies or any of that.

Resistor inline
Cap to Ground.


The LPF blocks changes that are in violation of Nyquist's Limit, so anything that is faster than half your sampling rate. Folks get mixed up with that (and calculating cutoff and roll off) so I like to explain things in a much simpler way

* Cap collects a charge and stabilizes
* Cap has only 2 places to spill out
* Into the ADC (very high impedance)
* Back out thru the series resistance (relatively high impedance)

Of course.. there is some self discharge and discharge thru paths like grease, but for what we are talking about... if you "fill the cap up" it will stay full. The key to understanding this is to understand that electronics work WAY faster than humans

* Seconds
* mSeconds
* uSeconds
* nSeconds
* pSeconds

Most signals... they change on the order of SECONDS. Like... Temperature. We DO NOT NEED (EVER EVER EVER!!!) Temperature readings at 1Khz. That is asinine and bordering on mental retardation. For something like temperature, once a second is PLENTY GOOD.

That being the case,
We are free to filter out all of this higher speed data... all of these swings and changes... all of these ups and downs (noise)

* Integrate
* Average
* Collect

The average value of the voltage coming in with the capacitor. We hold it there with the resistor. The resistor allows for SLOW CHANGE on the order of the RC time constant. Fill slow, changes slow... and that is EXACTLY what we want if we are trying to filter out periodic noise.

Make sense?

Capacitor on the ADC input is like a little safe zone. A little buffer. A bucket...

Resistor slowly fills this bucket and resists fast changes back out

ADC input is practically an Open Circuit, so what happens is... it fills up... and you can read it. IF you go too large with the values you can actually SEE this slope in your data. You can see "lag". Lag is your friend, but only to a point.

Something like... hey... 1K and 4.7uF
Try that

Something like 200 ohms and 10uF
Try that

Something like 2k and 10uF
Try that

All of them work and it really does not matter. Just... Get a LPF (single pole) on your ADC input if you want to see clean and steady readings.

Final Point
This will have little to no impact on the circuit being measured. The resistance (be it 20ohms, 200 ohms, 2000ohms) is effectively hanging out on a wire with no current flowing thru it. Those who have completed their EE program of study know that this means there will be no DC impact to the circuit under test. OF COURSE there will be AC impact, but... then... that is what we are trying to do - right?

Pull off only the DC
Leave the AC
Because... the AC is the Noise part :cool:

(that is my Knuckle-head way of thinking about it and it has gotten me a very long way with taking good measurements in bad environments using only an Arduino, a resistor, and a capacitor)

I'm in the process of making an Eagle Library device of the 4.1. I did not find one in SnapEDA as noted above.
I'm modifying the 3.6 footprint available in a zip file on the forum.
I noted (as above) the same change of the pin adjacent to VIN, now labeled GND instead of AGND, and I cannot find an AGND on the Teensy 4.1 ref card.
- Is the 4.1 GND clean enough for analog measurement? I have been using AGND for this on Teensy 3.2?
- Is there a typical circuit I should add when using GND as AGND on the 4.1?

Thanks, though I did not grok an answer for Teensy 4.1 from #4.
(I have degrees, but none in EE)

My analog signal measures breath pressure, which varies over a few milliseconds (and I sample it that often).
I end up rounding the 10-bit input to 7-bit (MIDI expression control).

I can experiment to see if it's unreliable on the Teensy 4.1, but I guess I'm looking for a quick answer if possible:

- will an analog input comparing to GND on the Teensy 4.1 behave similarly to an analog input comparing to AGND on the Teensy 3.2?
Simple answer, NXP has no accessible AGND on these chips so Teensy 4.x can’t have one. While the analog inputs aren’t as ‘good’ as prior models they should more than suffice for 7-bit MIDI with little to no issue.
- will an analog input comparing to GND on the Teensy 4.1 behave similarly to an analog input comparing to AGND on the Teensy 3.2?

Yes, more or less. Mostly less. Teensy 4.1 doesn't have quite the analog performance of Teensy 3.2. As a general rule for analog features embedded inside mostly digital chips, improvements in silicon process that allow for faster digital circuitry tend to trade off analog performance. It's not just some arbitrary decision NXP or PJRC made, is the physics of how the transistors and signal coupling between analog and digital happens inside the chip that impacts analog performance.

Still, treating 1 GND pin as "analog ground" and the other GND pin as "normal ground" will help to avoid ground loop currents which cause noise on the measurements.

Keep the analog signal's impedance low. If using pots, 1K to 10K is the usable range. Lower is better, but consumes more power. A 1K pot wired across 3.3V power will draw constant 3.3mA current.

If you connect a capacitor at the ADC input, the type of capacitor matters more than the amount of capacitance. Ceramic capacitors rated NP0 or C0G are best. More than 1nF (1000pF) is probably diminishing returns for most applications. Aluminum electrolytic capacitors are a poor choice, even if large capacitance, due to their higher impedance at high frequency. Even ones claiming to be low ESR or described as high frequency aren't a good choice. Likewise for tantalum. Use a ceramic capacitor, ideally one NP0 rated.