Cheap DAC chips

[André RT]

Came across a dilema... Got lucky still can't believe this happened. Maybe my references are still very low and I haven't afforded decently good proper audio hardware... Anyway. What happened was...

Spotify is sounding better playing on a cheap smartphone than on the cheap IQ audio DAC pro on a raspberry pi.

Went to buy a set of cheap Mackie CR4x and brought along the pi with the IQ audio but the pi wasn't being able to access the wifi hotspot provided by the phone that well and the sound wasn't that impressive so i ditched it and connected the phone directly to the speakers in the store shelves by feeding it the output of the headphones with a cheap RCA to 3.5mm jack. Hold and behold. The sound was amazing and with that 2 pieces of hardware I was filling the whole store with an bright and colored sound. The store was like a 200 square meter by 6 meters high full of keyboards guitars all sort of musical instruments and the little 4 inch monitors we're filling it completely with a loud bright defined sound, even the low end was punchy and present. People that were shopping in the other corner of the store said that they felt and liked the sound... All of this coming from the DAC of a cheap Alcatel 1S 2019 smartphone.

This phone is known for having and extremely loud internal speakers but no definition like for example the Alcatel idol 4s (aka Vodafone Smart Platinum 7)with stereo speakers and Waves™ technology (software DSP I guess "inside"), which made it sound as good as a portable Bluetooth speaker back in 2019 , but still don't believe that it's ( Alcatel 1S 2019) DAC has such a flat and clear live sound coming out of the headphones jack.
At the i use this phone to record surveillance video or eventually control Spotify running on the raspberry pi, every other ocasion I indulge in connecting it to the amplifier were i have the Mackie CR4x connected and play Spotify from it because the IQ audio DAC on the pi is no match for it in making the sound come alive while streaming Spotify

I will eventually buy a proper DAC and subwoofer and will then share how the new sound compares with this cheaper hardware.

 

[DVDdoug]

As always, the first question is, are you comparing with the volumes matched?

Personally, I've never heard any difference with any DAC, starting with the DAC in my 1st CD player. ...I also don't have golden ears and I've never bothered doing any careful listening tests because I've never heard anything wrong, unless it was the music itself.

If you hear any difference it will probably be noise (hiss, hum, or whine in the background). I HAVE heard noise from a soundcard, but not from my current/recent computers. That's a DAC, but I don't blame the DAC chip. It's probably the noisy computer power (which is usually noisy) or just digital switching-data noise getting into the analog electronics.

I will eventually buy a proper DAC and subwoofer and will then share how the new sound compares with this cheaper hardware.

Headphones (almost any headphones) will usually reveal any defects or "differences" in the audio signal.

At least 90% of audio limitations/problems are with the speakers/headphones (and room acoustics with speakers). And of course you need enough amplifier power so you don't over-drive your amplifier into clipping (distortion). You can also sometimes clip your headphone amp. And of course your speakers need to handle the amplifier power.

And there can be issues with the recording and low bitrate MP3s can sound bad etc.

Digital defects are very rare and usually very obvious, like crackling or audio dropouts, etc.

 

[André RT]

The tests in the store were made with the volume maxed out both the smartphone/Spotify and the Mackie CR4x. No distortion noted. Again impressive. The higher the volume the better the pair sounded. The Mackie were oob so not even broken in, (if that is still a thing as it were 20 years ago with HiFi.)

The IQ audio dac seems to have more power. Perhaps it reaches near 2 volt RCA line out. But if falls apart, loses the definition it has in lower volumes. The Alcatel no. It makes the sound come alive... can't believe this.

Anh ps : fun "fact"... from my calculations the phone is able to run line out and wireless more than 16 hours with its 3060 mAh battery.

 

[André RT]

Theres that thing, in the acr reviews, that is the resampling works best at full volume. Now there is some sort of software and volume control between raspotify linux alsa and the IQ audio dac hardware. At least it raspberry pi os.
I tested distros like Volumio/ moode to see it I would give any control or at least perception of how the sound decoding and resampling happened but nothing seems to be able to be tweaked to make sound better. I set Spotify to 320 kbps and thats it's. I don't know how to tweak alsa inside Linux to optimize further. I hoped Volumio did that by itself but nothing noted a
t least with this IQ audio dac pro through RCA.
To force 32 bit depth or 192 khz sampling doesn't make the sound better. Volumio sounded worst than normal raspberry pi os.
I think that balenasound docker IOT device sounded better.

 

[Rednaxela]

Without measurements you have absolutely no idea what you have been comparing.

 

[JasonLG1979]

I know I'm a little late to the game but Raspotify maintainer and librespot contributor here. I read this and I felt it may be helpful to clear up some misconceptions.

I set Spotify to 320 kbps and thats it's

320 kbps is the bitrate of the Ogg Vorbis stream. It has nothing to do with the sample rate. The sample rate is 44.1kHz no matter the bitrate of the stream.

Now there is some sort of software and volume control between raspotify

All processing internally in librespot is done with 64 bit floating-point precision. The software volume in librespot is applied as a last step before format conversion from Float64 PCM to whatever output format you choose. Digital volume control implemented in that way is objectively superior to any and all forms of analog volume control and very more than likely better then the digital volume control of your DAC for formats above 16bit. For 32bit (linear) dithering is not even necessary during format conversion since the quantization noise is well, well below the noise floor of any DAC by almost an order of magnitude or more. 32bit (linear) has a SNR of just slightly more than the dynamic range of air at sea level (194.42 dB vs 194 dB).

In short if you care about sound quality use the digital volume control provided by librespot.

To force 32 bit depth or 192 khz sampling doesn't make the sound better.

Well of course it doesn't...
The 1st rule of resampling is; don't resample unless you have to.

Resampling is basically making an educated guess about what should be there, how educated depends on the interpolation algorithm, but none the less, it's still a guess. It will NEVER be completely accurate. It will NEVER objectively improve the sound quality. It's a matter of how much will it degrade the sound quality.

In short if you care about sound quality don't resample if you don't have to.

 

[JasonLG1979]

As far as DAC HATs are concerned, don't waste your money on one unless you absolutely need super cheap and/or an all-in-one box setup. The cost to performance ratio just isn't there compared to even a mediocre stand alone USB DAC. You can get a nice USB DAC (like a Topping D10s) for $100 or a nice USB Dongle DAC (like a TempoTec Sonata HD PRO) for like $50 that will smoke any DAC HAT. For my money if I wanted a stand alone dedicated Spotify Connect device I'd combine a Pi Zero v2 with a TempoTec Sonata HD PRO or something comparable. All in you'd end up spending about $100.

 

[notsodeadlizard]

Raspi is a good small machine but it never was designed for audio applications.
It has a lot of issues with the powering (historically), the power subsystem (if it can be called a subsystem at all) has always worked to the limit, PCB was traced for non-audio specific requirements (only God knows how noisy it is and how high-frequency harmonics are distributed by return currents on the groundplain etc.).
All this obvious does not mean that Raspi is not suitable for audio applications, it means that it must be used wisely.
Any DAC chip, for example, has several different power domains and the requirements for them are clearly defined and indicated in the datasheet (and if they are not followed, the declared parameters will not be obtained). Somebody can call it snake oil from illiteracy (it’s generally fashionable to do this), in fact, this is just a competent solution for powering analog-digital systems. Therefore, there is no need to invent anything, this is pure engineering, in which it is customary to power digital and analog parts from different sources and use galvanic isolation of these parts in especially difficult cases so that the return currents of the digital part do not "roam" the ground plane of the sensitive analog part (in reality, we are talking about the ppm level of distortion, and this is already something like the input sensitivity of receivers).
The really good Raspi hat designers know about this stuff.
That's why Ian Canada (who is really competent designer) has galvanic isolator hat in his portfolio, and provides separate powering for the galvanically isolated DAC and Raspi.

Why your old phone is better than a simple combination of Raspi + cheap DAC hat?
Because the internal "audio" part of phone schematics and PCB is designed for what it's meant to be.
In the case of Raspi, which comes to you ready-made, you cannot optimize the circuitry and topology of the board (you can, of course, with the help of a complete redesign, but there is no point in this).

I have no more secrets for you.

 

[JasonLG1979]

The power is isolated enough on a Pi that even a remotely competently designed USB DAC isn't going to have any issues. DAC HATs are the only things are are going to suffer from performance issues from dirty power on a PI, namely because they're mostly designed to be cheap and based on DAC chip reference designs. I stand by what I said. Based on my experience and objective measurements on this very site you get more for your money from a USB DAC. Looking at the prices of the Ian Canada boards you're still better off with something like a Topping D10s.

 

[JasonLG1979]

You can go the 75 stacked HATs route if you like but you'll never even get to good USB dongle level performance and you'll be out a couple hundred bucks.

 

[notsodeadlizard]

The power is isolated enough on a Pi that even a remotely competently designed USB DAC isn't going to have any issues. DAC HATs are the only things are are going to suffer from performance issues from dirty power on a PI, namely because they're mostly designed to be cheap and based on DAC chip reference designs. I stand by what I said. Based on my experience and objective measurements on this very site you get more for your money from a USB DAC. Looking at the prices of the Ian Canada boards you're still better off with something like a Topping D10s.

"they're mostly designed to be cheap and based on DAC chip reference designs"

Are you saying that external DACs are not made based on reference designs of DAC chips?
I'm very interested.

 

[notsodeadlizard]

I forgot one "terrible secret".
I'm sorry.
Any really good designed DAC based on the high quality DAC IC or chipset is required to have galvanically isolated USB input.
Will it improve the sound?
In general, no.
Will this avoid power problems, such as noise on the ground plane due to return currents, such as "clicks" in the speakers, etc.?
Absolutely yes.
And because a good DAC or chipset can cost about $100 in batches (ROHM and AKM for example), just one digital isolator chip does not change the picture.
Digital TOSLINK is generally made galvanically isolated by design because it's the right design.

So you can consider a top-end DAC based on top-end ROHM, AKM or ESS DAC chips/chipsets as a rattle if it does not provide for a simple, absolutely "transparent" device protection for electronics and firmware from possible power problems in the system.
It simply speaks of the relation of the producer to the consumer, and nothing else.

 

[JasonLG1979]

Are you saying that external DACs are not made based on reference designs of DAC chips?
I'm very interested.

You tell me?

Any really good designed DAC based on the high quality DAC IC or chipset is required to have galvanically isolated USB input.

You've really latched onto that term, "galvanically isolated".

And because a good DAC or chipset can cost about $100 in batches

You can get 5 ES9038Q2M chips for $82.80 retail. I'm not sure where you're paying $100 a chip but you're getting ripped off.

I can't tell if it's a language barrier issue or that you really have no idea what you're talking about or a combination of both, but very rarely have I seen someone use so many words to actually say so little.

 

[JasonLG1979]

Back on topic.

A bit of clarification about formats in librespot and formats in general.

1st a little bit of lossy audio basics.

Most (if not all?) lossy audio does not have a bit depth. The source did but upon encoding the audio is transformed to exist only in the time domain, sorta like DSD.

When the audio is then decoded back to PCM a bit depth has to be calculated. Generally decoders output Float32 or Float64 PCM and leave the format conversion to the app in question.


In the case of choosing an output format bigger is better. The larger the format the less quantization noise during conversion and the more usable bits for digital volume control before you actually start to throw away audio information.

What does all that mean?

Say the source was 16bit. That means it has a digital SNR of 98.09 dB. It then gets encoded and decoded. The decoded Float PCM still contains nothing below -98.09 dB except for maybe artifacts. By converting it to a higher format you're essentially padding it with silence.

So lets say you choose 32bit (linear). As I said before 32bit (linear) has a SNR of 194.42 dB. That means you end up with 96.33 dB of usable digital volume control before you actually start to throw away audio data. Meaning you retain all information until you set the volume to below -96.33 dB. For 24bit (linear) it works out to 48.17 dB of usable volume control.

Considering that the background noise in an extremely quite room is going to be well over 30 dB and prolonged exposure to sounds over 80 dB can cause hearing damage, I would consider a usable digital volume control of > 50 dB to be transparent in normal, safe conditions.

In short, when playing lossy audio and/or using (properly implemented) software volume control choose the highest bit depth that your hardware will support. If your device supports 32bit (linear) use that.

 

[JasonLG1979]

I missed this gem...

Digital TOSLINK is generally made galvanically isolated by design because it's the right design.

TOSLINK is electrically isolated because it sends signals via a non-electrically conductive fiber optic cable...

 

[antcollinet]

I missed this gem...


TOSLINK is electrically isolated because it sends signals via a non-electrically conductive fiber optic cable...

A new member who knows exactly what he is talking about and making absolute sense. What a rare and welcome sight

Welcome to ASR @JasonLG1979.

 

[JasonLG1979]

A new member who knows exactly what he is talking about and making absolute sense. What a rare and welcome sight

Welcome to ASR @JasonLG1979.

Thanks. I'm far from an expert in hardware, but I would consider myself almost competent on the software side. You learn pretty quick in software. It's not forgiving at all. When you screw something up it's pretty obvious because it mangles the audio,lol!!!

 

[notsodeadlizard]

I missed this gem...


TOSLINK is electrically isolated because it sends signals via a non-electrically conductive fiber optic cable...

You are absolutely right.
It is necessary to increase in every possible way the already highest scientific level.

 

[JasonLG1979]

You are absolutely right.
It is necessary to increase in every possible way the already highest scientific level.

You keep posting without saying anything? Do you have anything of value to add to the conversation or are you just trying to bump your comment numbers? So far you're just noise distracting from the topic.

Times like this I am reminded of a quote that goes something along the lines of:

"Better to remain silent and be thought a fool than to speak and to remove all doubt"

 

[JasonLG1979]

Because you are definitely more strong in hardware than in software.

Talk to me when you write a resampler from scratch with no external libs...