ADC -> uController +DAC or MCU -> MCU?

PWM for 2¹⁴ different values would require 16 Gbps.

If you transmit the samples with a simple protocol like I²S, you do not require a microcontroller. And specialized ADCs/DACs will be able to get the desired quality at lower cost than some built-in MCU module.

That was our initial hope. However, to the very best of our knowledge it is not possible to get an ADC/DAC pair that 
can work without a microcontroller. Maybe we are missing something. Have you ever seen such reference design or
article that talks about it?

Also, I just realized that I2C would pose a data throughput bottleneck and not allow 1MS/s as the highest transfer rates
are 3Mbps. Can we have a microcontroller-less solution using SPI?

Sorry, I was thinking about audio converters, which run continuously with only a clock. (And I²S is not I²C.) But that would make it hard to go above 192 kSPS.

As far as I can see, you need an ADC, a single microcontroller that reads samples through one SPI port and sends them out through another SPI port, and a DAC. The second SPI goes through a three-channel isolator. (The microcontroller might have an integrated ADC; you have to decide if that one is good enough for you.)

Yes, now you understand the challenge.

The main issue is now to determine if a microcontroller plus ADC would be better than going with MCUs to simplify design and reduce number
of components (and potentially cost)

Do you, or anybody else have some experience trying to figure these tradeoffs out?

Also, the microcontroller probably would be on the other side of the isolator together with the DAC to reduce the power consumption on the isolated
power side. Unless, again, we were to go with 2 MCUs.

Thoughts?

Hi Sheilon,

Thank you for posting to E2E! What are the isolation requirements of your system? Using less components is typically desired, and a common approach to isolation is to input analog signals to an ADC, isolate the digital output to maintain signal integrity, and interface with an MCU, as shown in the ISO7741 datasheet image below:



TI also offers ADCs and amplifiers with integrated isolators, if those are of interest. What will the isolated signal be used for? Our ISO77xx digital isolators can isolate PWM signals anywhere from DC to 100Mbps with any duty cycle provided the pulse width of signal is > 10ns.

I will await your response.


Thank you!
Manuel Chavez

Dear Manuel,

We already use some isolator chips that can isolate to 5kV RMS. However, for our applicaiton we need isolations higher than 10kV and
therefore need to come up with some other methodology. We are considering using some optical fiber as isolation. However, we need to
figure out a way to translate the analog signal into digital (PWM or ADC) then, reconstruct after the isolation barrier.

The parameters hold. We'd like to have signal bandwidth of 1MHz (if possible) with 14 to 16 bits resolution.

The solution you proposed seems neat, however, does not meet our isolation requirements. 
Thus, we are looking for some expert advice on the tradeoffs of having some ADC and microcontroller or full MCUs.

Can you help?

Clemens,

As a potential alternative for yet another project with lower sampling rate we'd be very interested in knowing more about your suggestion of using purely I2C. Do you have
any reference material on how to achieve that? I see a challenge as both the ADC and DAC are slaves.

Hi Sheilon,

Multiple of our digital isolators can be connected in series to achieve greater isolation capabilities. Regarding the signal transmission, our team is looking for potential options using our devices. I will follow up in 1 - 2 days.


Thank you,
Manuel Chavez

Hi Manuel,

Thank you for your suggestion, however, we really do not think that placing multiple units in series is an option. Given their extremely high impedance, 
fluctuations are expected and they could cause imbalances in the voltage drops. One of the chips could potentially be subject to a much higher voltage than
the other and break prematurely.

Thus, we hope to achieve isolation via fiber optical barrier.

The question still remains on if that would be possible via some smart placement of your components to reduce number of components while preserving performance.

Thank you

Hi Sheilon,

Although there could be a voltage discrepancy between two isolators placed in series, our tests have found high voltages to divide between them about equally; however, we understand this solution is not acceptable and TI does not have an appropriate digital isolation solution at this time. Is the 10kV rating you are designing for defined by a standard spec like Viso per UL or a surge rating?

To best assist with the conversion of analog signal to digital for fiber optical isolation, the Analog to Digital Converters team is joining this conversation.


Thank you,
Manuel Chavez

I²S is not I²C.

To build a software-less analog→digital→analog system, select an ADC with I²S output and hardware control and a DAC with I²S input and hardware control.

Configure the ADC for master mode, which means that it derives bit clock and sample clock from the system clock, which is the only clock signal that needs to be supplied. Typically, the system clock is 256 times the sample rate. (Some ADCs like the PCM1861 also have a built-in crystal oscillator to generate the system clock.)

The four output signals (three clocks and one data) go directly (or through an isolator) to the DAC.

"Hardware control" means that the devices can be configured by tying certain pins to V_CC/GND; for an ADC like the PCM1861, these are the MD0…MD6 pins; for a DAC like the PCM1754, these are the DEMP/MUTE/FMT pins.

Hi Sheilon,

From the isolated ADC side of things, a measurement of 1MHz and 14-16bits of resolution is not possible. 

Our isolated amplifiers have bandwidth in the ~300kHz range and resolution around 12-14bits. 

Our isolated modulators can be used for a higher accuracy or higher bandwidth measurement, but at the expense of the other. For example, if you take a look at table 2. of the document below, you will see that a bandwidth of 1.25MHz can be achieved, but you will only have 4.65 ENOB. Similarly, if you want high resolution, an ENOB of 14.39 can be achieved at a bandwidth of 19.53kHz.  

http://www.ti.com/lit/an/sbaa359a/sbaa359a.pdf