inputs and impedances

pythagorean-comma

giuliomoro Hi Giulio, a follow-up to your post about the Multi's inputs being DC-coupled with a range of about 1.5 V ±1 V. I'm feeding IN2–IN7 audio from my own board (wings off, inputs taken from J6). Each input goes through a series 10 µF capacitor, and each IN− goes to ground through its own 10 µF. That's the arrangement datasheet §2.2 recommends for AC-coupled single-ended inputs, with no bias resistor.

With the inputs DC-coupled in the shipping software, does the TLV320ADC3140 hold each INxP and INxM pin at its common-mode voltage itself, so that a series capacitor with no external bias works, as on the wings? Or does the Gem side of the capacitor need biasing externally?

What is that DC level? The datasheet says 1.375 V; this thread said 1.5 V and 1.65 V.

Which input impedance does the software set (10k or 20k), and can a project change it?

Thanks!

giuliomoro

By default they are AC-coupled, 10k. The easiest way to change that right now is to edit these lines in ~/Bela/core/Tlv320Adc3140_Codec.cpp

    setInputSource(-1, AnalogDifferential, false);
    setInputDcCoupling(-1, false, false);
    if(setInputImpedance(-1, Impedance10k, true))

I think the ADC takes care of the biasing, so you don't need to add additional external biasing. The diagram in the datasheet works without additional biasing.

pythagorean-comma What is that DC level? The datasheet says 1.375 V; this thread said 1.5 V and 1.65 V.

I think it's 1.375V. Trust the datasheet over me.

pythagorean-comma

giuliomoro I want to run an expression pedal (a 10k–100k pot) from J3.1, the 4.096 V reference, into analog0.

In the Gem's netlist, J3.1 goes straight to the ADS8168's REFP pins (IC3.5/6), with only C5 and C8 (10 µF each) on it and no series resistor. TI's datasheet (SBAS817D) gives the reference buffer no DC load or short-circuit rating; the only figure is the REFby2 buffer's 2 mA. So I'd like to know what you designed J3.1 for:

Load: what's the most current, or the smallest pot, you intend J3.1 to supply? Would 0.5 mA continuous be fine?
Short to ground: a mono (TS) plug in a TRS expression jack shorts the ring, and so the reference, to ground. I plan to put about 10k in series so a short draws about 0.4 mA. Is that safe, and does REFP recover by itself afterwards? Do you know what it does with no series resistor?
Sinking: if something outside drives J3.1 above 4.096 V through a resistor, can the buffer sink current, or does REFP get pulled up? I'm planning a series diode so it never has to; I'd like to know if that's needed.
Effect on the other channels: REFP is the ADC's reference for all eight inputs. Have you measured how far it moves under load, say at 0.1 mA and 1 mA?
Future revisions: in post 10 you mentioned possibly changing the analog input range. Is J3.1 likely to change too?

Thanks again for an amazing ecosystem to build on!

pythagorean-comma

pythagorean-comma Following up on my post above. I found TI's answer on their E2E forum "The internal buffer is only intended to drive the ADC inputs, and is not specified for external loads." He points people to REFIO (which has 1 kΩ in series) or REFby2 (2 mA). As far as I can read the netlist, neither reaches a socket on the Gem.

So I've changed the plan: J3.1 won't be loaded at all. It will go into the input of a rail-to-rail op-amp follower on my carrier board. The follower drives the pot through a series resistor and clamps, so a TS plug, a reversed cable or a stray voltage on the ring lands on my op-amp, never on the Gem. The op-amp runs from the 5 V that feeds P1.1, which is always at or above the Gem's +5VA. Readings stay ratiometric to the ADC's own reference.

That makes my first four questions above moot (though I'd still be curious whether you ever tested REFP with a load). What I'd value your view on now:

Is an op-amp input on J3.1 fine? Is there any reason not to put one there, perhaps with about 10k in series for protection? Does anything else on the Gem rely on REFP being undisturbed?
Is a high source impedance on analog0 OK? The wiper reaches J3.10 through about 120k in series, with 10 nF at the pin and a 2.2M pull-down so an empty jack reads 0. As I read the netlist, analog0 goes straight into a TSV524 follower, so that should do no harm. Is there filtering or protection on the Gem side I've missed?
Which ground should the pot return to? J3.2 is GND, not GNDA. I'm returning the pot to GNDA (at the jack, which bonds to the chassis), so its current, about 0.1 mA and constant, reaches the ADC's ground through NT1. Would you rather it returned on J3.2?
Will J3 change? Is J3's pinout or REFP likely to change in later revisions? (My question 5 above.)

Thanks!

giuliomoro

Do not hijack threads, I moved you to a new conversation.

pythagorean-comma "The internal buffer is only intended to drive the ADC inputs, and is not specified for external loads

Then maybe they should have not labelled it like this in the datasheet:

pythagorean-comma Is an op-amp input on J3.1 fine?

Well, you say that TI says no. But in practice an opamp with large-enough input impedance should be fine. Not sure what 10k in series would do, considering you should have >> Mohm input impedance in your opamp.

pythagorean-comma Is a high source impedance on analog0 OK?

TSV524's input impedance is probably very high (>> Mohm), although the datasheet doesn't specify any value.

pythagorean-comma Which ground should the pot return to?

IC3 uses GND

pythagorean-comma Will J3 change?

We are not planning changes at this stage.

pythagorean-comma Future revisions: in post 10 you mentioned possibly changing the analog input range

That might have been about changing REFIO via JP7, e.g.: when powered via lipo or anything below 4.5V.

In general, I wouldn't overthink this. You want to connect an expression pedal, just connect it between 3.3V and GND and use the reduced range (you still get more than 15-bit precision). It's sampled at 22kHz, so you can filter away any additional noise that may be coming from 3.3V being more noisy than VREF.

pythagorean-comma As far as I can read the netlist,
pythagorean-comma As I read the netlist,

Humans rarely read netlists. Please do not post AI-generated questions, because I respond by hand. The schematic is there and it is complete. Ask your AI to bring up all the datasheets, the electronics textbooks and all the relevant forum entries and guess the answer to your questions.

pythagorean-comma

Thanks for the guidance and candour. I appreciate all the work being done by yourself/bela. I want to be a good community member and will be sure to be mindful of all of the feedback you have provided. Hope you have a great weekend 🙏