Akm5578 ADC

potstip

Member
2013-10-10 8:39 pm
I am working on simple high performance ADC, the best chip i found is AKM5578, in its referance design there is an opamp NJM5534 which is not produced anymore.

What could be the best replacement. No cost objective. I am looking for best performer.

Thanks
 

Attachments

  • Screenshot_20200115-144350_Drive.jpg
    Screenshot_20200115-144350_Drive.jpg
    110.4 KB · Views: 193
  • NJM5534_E-364876.pdf
    168.4 KB · Views: 13
The first thing to do when selecting the right opamp is check the ADC's performance numbers. The aim would be for the opamp's distortion and noise to be lower than that of the ADC. The AK5578 is an 8 channel part, are you using all 8 channels or summing them together to get a lower noise floor? With all 8 summed you want better than -130dB for SNR, the THD spec is -112dB, this won't change with multiple channels being summed.

As for LM49720 its my understanding that's the same part as LM4562.
 
Do keep in mind here that the only reason why this input stage meets the noise requirements for the ADC is because the inverting buffers are used in negative gain.

This allows AKM to use a very simple input buffer but it comes with the requirement of needing a very high input drive level to actually reach maximum signal level on the ADC. You need to provide 5.28VRMS per input half to hit 0dBfs.

This is larger than standard balanced outputs provide and significantly larger than the standard 2VRMS for single ended audio outputs. This isn't a concern for AKM, all they want to do is provide an input buffer that will showcase the ADCs performance and to that end it works. With their evaluation module they can provide an input signal of the required drive level using the Audio Precision test equipment. But with standard audio equipment and their associated output levels you will not reach the ADCs maximum input level.

This may not be of any concern but it is worth pointing out. If you wish to design an analogue input buffer that provides maximum ADC input scale with a standard 2VRMS audio signal then you'll need to redesign the input stage accordingly. And while the NE5534 is an excellent and inexpensive opamp you might need something that is lower noise to hit the target specification.

Remember the ADC hits it's maximum amplitude with only 1VRMS applied at each input pin. And needs a correspondingly low noise on the inputs to hit it's specified SNR.

This is one of the reasons why using several parallel higher noise input channels is an easier way to hit high SNR than just using two higher performance inputs. The maximum input signal magnitude remains unchanged between ADCs and this is fixed by the 5V supply rail. So to reach ever higher SNRs with a single channel you need much reduced noise from the input buffer.

Very low noise and very low distortion do not go hand in hand easily as very low noise necessitates low circuit impedances. Low impedances increase the demand placed upon the opamps output stage and also increases the size of any capacitors used for low pass filtering. The capacitors adding to the opamps struggle when driving the low impedances.

Some opamps handle capacitive loads and low impedances better than others. You can see this in the AK5397 datasheet. The example input buffers have identical, low impedance, low noise circuitry. With the LME49710 used they reach their target of low distortion but sacrifice on noise as a result. With the LME49990, a lower noise part, they manage to meet the required noise specification but this opamp is clearly not as comfortable driving the input circuits characteristics as distortion increases.

Just be aware of what it is you are trying to accomplish with your input stage, expected input levels and your ADC, then design accordingly. Also keep in mind that AKMs example buffers have low input impedances too which will not suit all sources that you may wish to use.
 
Last edited:
Do keep in mind here that the only reason why this input stage meets the noise requirements for the ADC is because the inverting buffers are used in negative gain.

This allows AKM to use a very simple input buffer but it comes with the requirement of needing a very high input drive level to actually reach maximum signal level on the ADC. You need to provide 5.28VRMS per input half to hit 0dBfs.

This is larger than standard balanced outputs provide and significantly larger than the standard 2VRMS for single ended audio outputs. This isn't a concern for AKM, all they want to do is provide an input buffer that will showcase the ADCs performance and to that end it works. With their evaluation module they can provide an input signal of the required drive level using the Audio Precision test equipment. But with standard audio equipment and their associated output levels you will not reach the ADCs maximum input level.

This may not be of any concern but it is worth pointing out. If you wish to design an analogue input buffer that provides maximum ADC input scale with a standard 2VRMS audio signal then you'll need to redesign the input stage accordingly. And while the NE5534 is an excellent and inexpensive opamp you might need something that is lower noise to hit the target specification.

Remember the ADC hits it's maximum amplitude with only 1VRMS applied at each input pin. And needs a correspondingly low noise on the inputs to hit it's specified SNR.

This is one of the reasons why using several parallel higher noise input channels is an easier way to hit high SNR than just using two higher performance inputs. The maximum input signal magnitude remains unchanged between ADCs and this is fixed by the 5V supply rail. So to reach ever higher SNRs with a single channel you need much reduced noise from the input buffer.

Very low noise and very low distortion do not go hand in hand easily as very low noise necessitates low circuit impedances. Low impedances increase the demand placed upon the opamps output stage and also increases the size of any capacitors used for low pass filtering. The capacitors adding to the opamps struggle when driving the low impedances.

Some opamps handle capacitive loads and low impedances better than others. You can see this in the AK5397 datasheet. The example input buffers have identical, low impedance, low noise circuitry. With the LME49710 used they reach their target of low distortion but sacrifice on noise as a result. With the LME49990, a lower noise part, they manage to meet the required noise specification but this opamp is clearly not as comfortable driving the input circuits characteristics as distortion increases.

Just be aware of what it is you are trying to accomplish with your input stage, expected input levels and your ADC, then design accordingly. Also keep in mind that AKMs example buffers have low input impedances too which will not suit all sources that you may wish to use.

Very nice post, you've covered the main points that need to be addressed.
 
Maybe consider LME49860. About like a 4562 or LME49720, but allows higher voltage rails and larger output swings. However, all the foregoing opamps can be sensitive to environmental RF, such as from cordless home phone base stations, etc. Proper shielding design may be part of what would needed for best performance results.
 
This allows AKM to use a very simple input buffer but it comes with the requirement of needing a very high input drive level to actually reach maximum signal level on the ADC. You need to provide 5.28VRMS per input half to hit 0dBfs.

This is larger than standard balanced outputs provide and significantly larger than the standard 2VRMS for single ended audio outputs.

It depends on whether you want headroom. A professional analogue mixing deck would typically have a nominal output level of +4 dBu, or 1.2277 V RMS, and 20 dB of headroom, so a maximal level of 12.277 V RMS, 6.138 V per side. As you wrote, it all depends on the application.
 

potstip

Member
2013-10-10 8:39 pm
My idea is use ADC as input of my class-d amplifier. My class d amplifier have pwm input so basically DSD and PWM is same.

I believe that ADC pwm conversation will be superior to convensional analog - PWM conversation methods in class d amplifiers.

So it will be a regular integrated amplifier.