Very cool! Been watching the series a while, so satisfying to see it reach this stage. -40dB is only 1% harmonic distortion, which isn't all that bad for audio run out of a docking station. If I were evaluating this ADC, I would buy an XMOS multichannel development board, since it can hit loopback distortion levels around -110dB, then I would add the tinytapeout ADC as a parallel dummy load, ensuring that the analog capture network doesn't affect the signal quality for the loopback, then I would take the measurement via the saleae+TTOadc and compare the raw waveforms in audacity (pre-FFT)... getting them aligned manually there isn't too tough. I've had issues with python's built-in audio FFT tools giving me values ~20dB worse SINAD than what Audacity calculates, even with the same FFT algo and window size, but I can't remember the fix off the top of my head. Good luck, and I can't wait to see more!! Thank you for all the great resources and valuable education that you've shared here!
Thanks! Got it up to an SFDR of 56 dBc now. I needed the sample rate of the audio DAC to be higher than sample rate of the ADC. I think I was seeing the spectral copies (to be confirmed). My favorite source is the DS 360 from Stanford Research (www.thinksrs.com/products/ds360.html) I have at work, but it's a bit overkill for an 8-bit ADC.
44000 isn't a common rate for audio. It'd be either 44100 or more commonly now 48000 so there might be some resampling going on causing weirdness. I'm also curious if you ran the same measurement on the data I sent ? Do you see the same harmonics. I just had a quick look and I see some but they seem to be weaker ( ~ -55 dBc )
Think I fixed the problem. Indeed there was wierdness between the audio DAC and the sample rate. What I think happened is that due to the low sample rate of the DAC, and the high oversampling of the ADC, I was seeing spectral copies. I've reduced the input frequency to 1 kHz, ADC samplerate to 16 kHz (ish) and DAC rate to 48 kHz. Now I get an ENOB of 7.06, which is about what I expect from this ADC. Yes, I did have a look at your data. I'll send you a PM.
Hi professor, very exciting stuff there! I wonder if you could give a bit more insight on how you have selected the input frequency. I've seen designers selecting the input frequency the closest prime to the sampling frequency for the ADC verification, yet could not tied this piece of information to your approach..
Hi. Thanks. Sure thing. I've updated the jupyter notebook a bit since the video. In short, when you do an FFT it's sufficient to have a input signal that ends up in a odd FFT bin, since in an FFT there is always 2^N points, then any odd number is relative prime to the number of points. Some claim you need a prime number, but that's wrong in my opinion. There reason for a relative prime is to ensure that you hit all codes. If you have an even number you can end up hitting only some of the codes. In addition, you want the end of the signal (last sample in the 2^N sequence) to match up with the first sample, so you want an integer number of periods of the input sinusoid. To some extent, it's possible to fix "match the start and the end" with a window filter, I'm a fan of the Hanning window.
I have so many questions! I have been dying to learn ADC design, what is the maximum frequency that your design can resolve? Is it 50 ohm input matched? What is the sensitivity and noise figure? What is the lowest dbm power that can be detected?
The ADC input is a switch capacitor, as such, it's not matched to 50 Ohm, and it's not designed to capture RF signals. It would need an LNA, mixer, and a filter in front. Have a look at analogicus.com/aic2025/2025/03/27/Lecture-10-Low-Power-Radio.html . It should explain how to use a SAR ADC in a radio context.
*Testing the Tiny Tapeout 6 Successive Approximation Analog-to-Digital Converter (ADC)* * *0:00:16** Overview:* The presenter discusses testing their successive approximation ADC from the Tiny Tapeout 6 project. * *0:00:49** Testing the ADC:* The presenter explains the need for an input signal to test the ADC, which is connected to the board via soldered wires. * *0:01:51** Tiny Tapeout Board:* The board features a USB connector for power and communication with a computer and requires an external input for the ADC. * *0:03:36** Ease of Use:* The presenter praises the Tiny Tapeout board's interface, highlighting its ease of use for selecting and enabling the ADC. * *0:03:48** Digital Output Capture:* The Saleae logic analyzer captures the digital output, with a specific signal (da) indicating the stability of the output bits. * *0:05:19** Input Network:* The input network uses the audio output from a docking station, sent through bypass capacitors and resistors to create a pseudo-differential signal around 0.9V. * *0:06:57** Signal Generation:* A Jupiter Notebook generates audio signals at 44 kHz, which are then played through the docking station's audio output. * *0:08:11** Data Capture and Analysis:* The Saleae logic analyzer captures the ADC output, and the data is post-processed using an FFT to analyze signal quality and identify spurs. * *0:10:31** Spur Investigation:* The presenter suspects the presence of spurs in the output signal, possibly due to the input network, and plans to investigate further using INL and DNL measurements. * *0:11:48** Performance Evaluation:* The current effective number of bits is around 6, which is lower than expected, potentially due to testing issues rather than the ADC design itself. * *0:12:08** Tiny Tapeout Appreciation:* The presenter expresses enthusiasm for the Tiny Tapeout project and its accessibility for testing and developing integrated circuits. I used gemini-1.5-pro-exp-0827 on rocketrecap dot com to summarize the transcript. Cost (if I didn't use the free tier): $0.03 Input tokens: 21716 Output tokens: 460
![Lp. Сердце Вселенной #60 РОЖДЕНИЕ ЛОЛОЛОШКИ [Финал] • Майнкрафт](http://i.ytimg.com/vi/YoR0pAV9FVQ/mqdefault.jpg)
Very cool! Been watching the series a while, so satisfying to see it reach this stage.
-40dB is only 1% harmonic distortion, which isn't all that bad for audio run out of a docking station.
If I were evaluating this ADC, I would buy an XMOS multichannel development board, since it can hit loopback distortion levels around -110dB, then I would add the tinytapeout ADC as a parallel dummy load, ensuring that the analog capture network doesn't affect the signal quality for the loopback, then I would take the measurement via the saleae+TTOadc and compare the raw waveforms in audacity (pre-FFT)... getting them aligned manually there isn't too tough. I've had issues with python's built-in audio FFT tools giving me values ~20dB worse SINAD than what Audacity calculates, even with the same FFT algo and window size, but I can't remember the fix off the top of my head.
Good luck, and I can't wait to see more!! Thank you for all the great resources and valuable education that you've shared here!
Thanks!
Got it up to an SFDR of 56 dBc now. I needed the sample rate of the audio DAC to be higher than sample rate of the ADC. I think I was seeing the spectral copies (to be confirmed).
My favorite source is the DS 360 from Stanford Research (www.thinksrs.com/products/ds360.html) I have at work, but it's a bit overkill for an 8-bit ADC.
44000 isn't a common rate for audio. It'd be either 44100 or more commonly now 48000 so there might be some resampling going on causing weirdness.
I'm also curious if you ran the same measurement on the data I sent ? Do you see the same harmonics. I just had a quick look and I see some but they seem to be weaker ( ~ -55 dBc )
Think I fixed the problem. Indeed there was wierdness between the audio DAC and the sample rate. What I think happened is that due to the low sample rate of the DAC, and the high oversampling of the ADC, I was seeing spectral copies.
I've reduced the input frequency to 1 kHz, ADC samplerate to 16 kHz (ish) and DAC rate to 48 kHz. Now I get an ENOB of 7.06, which is about what I expect from this ADC.
Yes, I did have a look at your data. I'll send you a PM.
Hi professor, very exciting stuff there! I wonder if you could give a bit more insight on how you have selected the input frequency. I've seen designers selecting the input frequency the closest prime to the sampling frequency for the ADC verification, yet could not tied this piece of information to your approach..
Hi. Thanks.
Sure thing. I've updated the jupyter notebook a bit since the video. In short, when you do an FFT it's sufficient to have a input signal that ends up in a odd FFT bin, since in an FFT there is always 2^N points, then any odd number is relative prime to the number of points. Some claim you need a prime number, but that's wrong in my opinion.
There reason for a relative prime is to ensure that you hit all codes. If you have an even number you can end up hitting only some of the codes.
In addition, you want the end of the signal (last sample in the 2^N sequence) to match up with the first sample, so you want an integer number of periods of the input sinusoid. To some extent, it's possible to fix "match the start and the end" with a window filter, I'm a fan of the Hanning window.
@analogicus oh that makes a lot sense now.. thanks a lot for the explanation!
I have so many questions! I have been dying to learn ADC design, what is the maximum frequency that your design can resolve? Is it 50 ohm input matched? What is the sensitivity and noise figure? What is the lowest dbm power that can be detected?
The ADC input is a switch capacitor, as such, it's not matched to 50 Ohm, and it's not designed to capture RF signals. It would need an LNA, mixer, and a filter in front. Have a look at analogicus.com/aic2025/2025/03/27/Lecture-10-Low-Power-Radio.html . It should explain how to use a SAR ADC in a radio context.
*Testing the Tiny Tapeout 6 Successive Approximation Analog-to-Digital Converter (ADC)*
* *0:00:16** Overview:* The presenter discusses testing their successive approximation ADC from the Tiny Tapeout 6 project.
* *0:00:49** Testing the ADC:* The presenter explains the need for an input signal to test the ADC, which is connected to the board via soldered wires.
* *0:01:51** Tiny Tapeout Board:* The board features a USB connector for power and communication with a computer and requires an external input for the ADC.
* *0:03:36** Ease of Use:* The presenter praises the Tiny Tapeout board's interface, highlighting its ease of use for selecting and enabling the ADC.
* *0:03:48** Digital Output Capture:* The Saleae logic analyzer captures the digital output, with a specific signal (da) indicating the stability of the output bits.
* *0:05:19** Input Network:* The input network uses the audio output from a docking station, sent through bypass capacitors and resistors to create a pseudo-differential signal around 0.9V.
* *0:06:57** Signal Generation:* A Jupiter Notebook generates audio signals at 44 kHz, which are then played through the docking station's audio output.
* *0:08:11** Data Capture and Analysis:* The Saleae logic analyzer captures the ADC output, and the data is post-processed using an FFT to analyze signal quality and identify spurs.
* *0:10:31** Spur Investigation:* The presenter suspects the presence of spurs in the output signal, possibly due to the input network, and plans to investigate further using INL and DNL measurements.
* *0:11:48** Performance Evaluation:* The current effective number of bits is around 6, which is lower than expected, potentially due to testing issues rather than the ADC design itself.
* *0:12:08** Tiny Tapeout Appreciation:* The presenter expresses enthusiasm for the Tiny Tapeout project and its accessibility for testing and developing integrated circuits.
I used gemini-1.5-pro-exp-0827 on rocketrecap dot com to summarize the transcript.
Cost (if I didn't use the free tier): $0.03
Input tokens: 21716
Output tokens: 460