AFE5801: Digital Gain on the AFE5801

Brian,
I moving this to the medical forum where this device is supported.
Thanks,
Vishy

Hello Brian,

There are at least 3 different questions here and we will try to address them one by one.

(i) In comparing the AFE5801 vs. the ADS5282, note that the AFE is full signal chain with an analog front-end (AFE) followed by an ADC, whereas the ADS5282 is only an ADC. The input referred noise of the signal chain is dependent on both the noise contribution of the AFE as well as the quantization noise of the ADC. The purpose of the VGA before the ADC in the AFE5801 is to amplify the incoming signal so that the voltage swing seen at the ADC input is maximized. The maximum gain of the VGA (31dB (typ) in the AFE5801) is selected such that gives you the gain range necessary to ensure that the overall signal chain is limited by the analog amplifier and not the A/D converter. Digital Gain after A/D conversion allows varying the digitized data without any impact to the analog portion of the signal. This allows an extra degree of freedom for system design and helps to optimize SNR and distortion of the entire signal path. For a better generic understanding of the benefits of digital gain, please refer to the EDN article by one of our colleagues:
www.edn.com/.../ADC-programmable-digital-gain-allows-tradeoffs-in-SNR-and-distortion-performance

(ii) Question: If there are extra internal bits keeping track of quantization information, then why aren’t these bits also used in the case when no digital gain is applied?

Response: Typically in any part (AFE5801 or ADS5282) the internal digital data-path widths are much wider than what is available at the output to account for loss of precision during digital math. If you can give us some back-ground of the exact application, and/or a diagram of your signal chain, we can help better clarify what you need to do and how to use the various knobs in the AFE5801.

(iii) Question: It seems that the best thing to do, for a small voltage signal, would be to adjust the scale of the ADC. What’s wrong with doing this?

Response: There are two reasons why adjusting the scale of the ADC is a very complex way to solve a straight forward problem:
(i) The “scale” of an ADC is typically set by the ADC reference voltage. The ADC reference voltage is set by some kind of a fixed low-noise reference. The choice of reference is critical because any noise on the reference directly impacts the quality of the A/D conversion. (see: www.ti.com/.../slyt331.pdf). To change the ADC scale implies manipulating the ADC reference voltage. Based on the type of ADC used, there can be various noise contributors that set the input referred noise of the ADC itself. There are fundamental limitations (such as kT/C, flicker noise, etc.) that set the input referred noise floor of the ADC independent of the reference voltage.
(ii) Secondly, the most straightforward way to address small input signals is to gain them up before the A/D conversion. That way the ADC input referred noise floor is much smaller than the incoming signal. If you are using a stand-alone ADC part like the ADS5282, you will need an external variable gain amplifier to achieve the same functionality. This is what the paragraph quoted from the ADS5282 is trying to say. This is why having an integrated flexible/ VGA before the ADC as part of the signal chain – such as what is done in the AFE5801 – is very helpful.

Again, we can help better clarify what you need to do and how to use the various knobs in the AFE5801 if you can give us some back-ground of the exact application, and/or a diagram of your signal chain. A block diagram of your signal path with input voltage levels will help greatly to help you optimize your solution.

Hi Praveen,

Thank you for your detailed response. For clarification on what "digital gain" is exactly, I have the following questions:

1) I assume that a "digital gain" of 2 is not simply a bit shift to the left (multiplication by 2). In such a case, resolution is lost and no SNR would be gained. Indeed, quantization noise also increases by 2, since the quantization step is now larger. This is not what you do, correct?
2) Does the AFE5801 have "extra internal bits" that are used to reduce quantization noise when the digital signal is amplified?
3) Does the ratio of signal strength to quantization noise actually improve with increasing "digital gain"?

In regards to my application, I have signals of different signal strength in the 8 ADC paths. The amplitude difference is as high as 8dB. I would like the SNR performance to be the same (or close to the same) for each signal path, but the AFE5801 sets the analog gain in each path to be the same. However, the digital gain can differ. So, that's why I'm trying to figure out what the "digital gain" does.

- Brian

Brian,

AFE5801's ADC is one generation ahead of ADS5282. your undestanding is correct. AFE5801 has internal bits as ADS5282 as well. so when you set gain of 6dB, the SNR slightly improved. for your applicaiton, digital gain seems to a good approach. this is assuming the channel to channel gain difference is fixed or can be predicted.

Thanks!

Thank you Xiaochen,

Do you know how much the SNR will improve for a digital gain of 6dB?

- Brian

we don't have exact number. my esitmation will be around 1~1.5dB. typically 12bit ADC SNR is 70dB, 14bit ADC SNR is about 72~73dB. so 6dB means shifting by one bit. so close to 13bit performance.

Yeah, but those SNR numbers are for a full-scale input.

Consider an input that is 6dB below full-scale (after the analog gain). Assuming a perfect analog amplifier, if I boost the gain of the analog amplifier by 6dB, I should get 6dB improvement in SNR, because my quantization noise is the same but my signal strength has increased by 6dB. However, if I leave the analog gain the same and instead use a "digital gain" of 6dB, what is my SNR improvement? In this case, I am guessing that the quantization noise may increase slightly.

- Brian

Brian, we can't assume the digital gain as a ideal amplifier. i.e. signal +6dB and Noise is the same. actually noise will be amplifier by several dB as well. a better way to think is still the 13bit ADC. for 13bit ADC, the SNR dBFS is slightly better than 12bit one, let us say 1.5dB better. when you have digital gain, you through away the MSB. in the new 12bit data, signal is doubled and noise is increased not by 6dB, maybe 4.5dB. So overall SNR in dBc improvement is about 1.5dB.

Thanks!

It would be nice to have measured, quantitative information about this in the datasheets. How else can the designer know how useful this "digital gain" is? There is also no mention of how it works in the datasheet. You're saying that it is like having a 13bit ADC. So does that mean there is literally 1 extra "internal bit" that is being used? Considering only quantization noise, shouldn't this mean that you would indeed get 6dB improvement in SNR gain, according to the formula in electronicdesign.com/.../signal-noise-ratio-snr-equation That would be great. Otherwise, only 1.5 dB gain is kind of disappointing.

The datasheet should discuss these internal bits.

Is there any TI ADC that has individually-programmable analog gain on each channel?