How good is bad? (Raspberry Pi 4B headphone output DAC review)

[Berwhale]

Oh, the distortions and noise are so tube-like and vinyl-like

There was a HAT for that...

https://www.kickstarter.com/projects/1815752970/hybrid-tube-amp-for-the-raspberry-pi

 

[antcollinet]

Great work.

Though I have a suspicion that some of the distortion might be still coming from clipping, even at 80% volume.

A plot of distortion against level would be interesting to check this. Does the "knee" of clipping distortion start below 80% volume?

 

[AudioSceptic]

For the life of me I didn’t realise the Pi had a headphone output.

I think they've always had a basic audio output, but it's there just to check that there's *any* sound, not for serious listening, a bit like the speakers on a notebook/laptop computer.

 

[GeorgeWalk]

There was a HAT for that...

https://www.kickstarter.com/projects/1815752970/hybrid-tube-amp-for-the-raspberry-pi

I have that HAT. I bought it because I thought it was an interesting dichotomy of modern and old tech. I haven't used it much, it doesn't sound good. But it looks good on my shelf.

 

[Multicore]

That's the ideal performance for use in a DIY emulator of classic 8-bit computer games. Put it in a wood and pelxiglass case with a nasty but efficient little loudspeaker to compliment the period-style game control buttons...

 

[JakeK]

I used this as a line out briefly before the DAC HAT arrived for my Pi based streamer and the distortion was definitely audible. This gives us a benchmark for a Sinad of 56 as so bad you can easily hear it. Somewhere a little north of this the noise and distortion become inaudible.

 

[formdissolve]

I used the analogue out with both video and audio running to an old CRT for a project one time.. let's just say the quality for both is pretty poor! I ended up using an HDMI to VGA converter, then another converter for VGA to Composite and it was much better. Thanks for doing this!

 

[Steven Holt]

Finally, The Flanker has some competition!

 

[Barrelhouse Solly]

The Pi5 doesn't have the headphone jack. HAT and USB are the only options. I always wondered how bad the 1/8" output was. Thanks for the measurements. I'll bet that there are USB DAC dongles in the same class.

 

[Timcognito]

Thanks. Enjoyed your writing style as well as the tech. 70/1 likes/posts ratio. Please keep posting.

 

[little-endian]

@Linards

Great overview and measurements! However, a final summary would have been nice. As bad as the measurements are by today's digital-to-analog-reconstruction-standards, probably even that will objectively still be beyond what even decent vinyl records provide, those admittedly doing so with more style (risking to stir up a hornet's nest even here).

You can expect around 9 to 10 bits of quite clean dynamic range[...]

Actually overall "signal to noise - ratio", as this is what any bit depth effectively limits, given proper dithering. The dynamic range may be decoupled from the SNR in general (non-linear PCM) or in terms of particular frequency ranges (dithered LPCM with noise shaping).

[...]which, I guess, would translate to roughly to 400kbps of bitrate at 44.1 kHz.[...]

If you're referring to lossy codecs, one can't translate any dynamic range into such bit rates as it is what remains unaffected even at very low ones. In fact, due to the use of the floating point, lossy formats may exhibit a tremendous amount of dynamic range which may by far exceed even what is possible with 24 bit PCM (in theory as the practical limitations kick in way earlier). Of course that also will come at the cost of a lower SNR.

Something that could be enough for lossy audio? Perhaps.

Maybe I missed the irony, but otherwise I'm astonished that @amirm left that uncommented. After all, this forum is supposed to be mostly free of the usual audio voodoo and debunked stereotypes.

Given halfway decent bitrates (which in practice are way lower than those suggested 400 kbps) in conjunction with proper codecs and psycho-acoustic models such as (Apple) AAC, (Lame) MP3 or Vorbis, partially no chance to distinguish that from a PCM original as countless blind tests have shown.

Looking forward for your reasonable critique and suggestions.

You got it.

 

[Linards]

@little-endian , Thank You! Your reply is genuinely appreciated! ^_^

I probably should have also included disclaimer, that I have no formal or informal experience in audio engineering, R&D etc. All my "expertise" stems from listening to (mostly streamed) music, DIY-ing and reading ASR time to time. I don't feel comfortable nor competent enough to draw really precise conclusions (or stir the nest, yet) given my limited knowledge in this are, thus the writing style. Also I might have used incorrect terms here and there or interpretation is off. Sorry about that... Hopefully, your (and other members) input will lead to improvements. ^_^

I have a question about dynamic range measurement and (useful) interpretation.

The dynamic range may be decoupled from the SNR in general (non-linear PCM) or in terms of particular frequency ranges (dithered LPCM with noise shaping).

When I look at 1kHz sine (80% volume, lets assume it is max output level) Signal-to-Noise Ratio (SNR) = 72dB (12 bits of depth at 44.1 kHz sampling rate). Something that is (?) still within the ballpark of RPI PWM driver and could be anticipated. There is Spurious-Free Dynamic Range (SFDR) =56dB. When I look at Multitone test (192 kHz sampling rate) SNR = 116dB (seems suspicions) and SFDR = 50dB.

The question is - Which is the best, useful metric for real-life, hear-able dynamic range when playing music (if any of mentioned)? And if none - how to estimate it (decouple it from SNR) or what other test I should run?

Perhaps I simply don't understand (fine?) difference between SNR and DR...

Maybe I missed the irony, but otherwise I'm astonished that @amirm left that uncommented. After all, this forum is supposed to be mostly free of the usual audio voodoo and debunked stereotypes.

Probably bad wording from my side. What I meant here, is if RPI headphone output is capable of delivering compressed-lossy (think Spotify) signal without further major degradation of quality (noise, distortion etc). Honestly, I don't know is it true. Perhaps there is a good way to test it?

Given halfway decent bitrates (which in practice are way lower than those suggested 400 kbps) in conjunction with proper codecs and psycho-acoustic models such as (Apple) AAC, (Lame) MP3 or Vorbis, partially no chance to distinguish that from a PCM original as countless blind tests have shown.

Agree. ^_^

 

[fordiebianco]

I have a variety of Pis distributed around the house, all with a Hifiberry dac. They feed the rca or optical input of various REVO sound systems and are working well for that purpose. My main music room at present has a PI as a streamer going into a SMSL dac via USB, feeding two Adam Audio active monitors. Sounds amazing.
The Raspberry Pis have been a godsend for the tinker/audio fraternit and long may they live. Never used the headphone out before.

...great writeup, btw.

 

[melomane13]

I vote yes , because it is "near" audible cd quality.
THD is only 2nd and 3td harmonics. the audibility level is 14% and 4% with pure tones, so it is ok.
Noise is also ok, if you dont use with horn tweeter.
linearity of frequency response can be increased with some eq, so not a problem.
output level is low, but not relied with audio quality
IMD is high , yes
using rpi4 audio output with some basic equipments should make no difference compared to perfect dacs.
Will someone want make ABX?

 

[GWolfman]

Great work! Thanks for sharing. Looking forward to see what’s coming up next from you.

 

[RandomEar]

I used this as a line out briefly before the DAC HAT arrived for my Pi based streamer and the distortion was definitely audible. This gives us a benchmark for a Sinad of 56 as so bad you can easily hear it. Somewhere a little north of this the noise and distortion become inaudible.

Are you sure you were not operating above 80 to 85% of volume? Otherwise, you would have heard clipping, which - to be fair - is a form of distortion, but at a much higher level in this case.

 

[JakeK]

Are you sure you were not operating above 80 to 85% of volume? Otherwise, you would have heard clipping, which - to be fair - is a form of distortion, but at a much higher level in this case.

I didn't make any notes so I can't remember exactly but I think I did try it at different volume levels. Possibly if I had known how it measures I could have managed the volume level better to get the best from it. I think I did adjust the player volume level down and turn up the speakers to have a sound that was tolerable for a couple of days until the DAC HAT arrived but it still didn't sound good. Of course if you turn down the digital source you lose resolution that way down to maybe 14 bits from the full 16 bits which isn't CD quality any more anyway.

I also tried it with some headphones just to find out and that did not sound good at all either!

So I made a mental note that I remember clearly: Pi headphone out does not sound good. Don't use it for music.

 

[JakeK]

I vote yes , because it is "near" audible cd quality.
THD is only 2nd and 3td harmonics. the audibility level is 14% and 4% with pure tones, so it is ok.
Noise is also ok, if you dont use with horn tweeter.
linearity of frequency response can be increased with some eq, so not a problem.
output level is low, but not relied with audio quality
IMD is high , yes
using rpi4 audio output with some basic equipments should make no difference compared to perfect dacs.
Will someone want make ABX?

Fairly confident I can hear the difference in this case!

 

[aagstn]

PS Audio designing a DAC near the quality of the audio jack on a PI is somehow not surprising. I'll stick with USB DACs and digital out hats for my Pi setups.

 

[aagstn]

For the life of me I didn’t realise the Pi had a headphone output.

They dropped it on the Pi 5. No big loss apparently.