• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Review and Measurements of Allo Katana and ApplePi Raspberry Pi DACs

johan

Active Member
Audio Company
Joined
Feb 23, 2018
Messages
225
Likes
372
I know there is a lot of controversy on what jitter level is audible and I wont make any comments on it. As a company we have a lot of hardware that cleans jitter (like Kali) and every time we think (!!) that any DAC benefits from lowest jitter possible. Of course some whitpapers argue that its not simply jitter thats important but actual phase noise thats really important to SQ. So we try our best to have the lowest jitter possible in ALL of our hardware.

Also I want to say Katana is not using asynch mode. Our DAC is in synch mode (master DAC and not slave). We are the first company that uses ess9039q2 in master mode and any dPLL added jitter is simply nil. Clocks jitter (actually buffer jitter since we dont feed clocks direct to dac IC)

Will you hear the dPLL jitter ? You can change the dPLL bandwith(lower bandwith = lower added jitter) and apparently (for some) its clearly audible. It can be a a psychological effect though.
 

Wombat

Master Contributor
Joined
Nov 5, 2017
Messages
6,722
Likes
6,459
Location
Australia
I know there is a lot of controversy on what jitter level is audible and I wont make any comments on it. As a company we have a lot of hardware that cleans jitter (like Kali) and every time we think (!!) that any DAC benefits from lowest jitter possible. Of course some whitpapers argue that its not simply jitter thats important but actual phase noise thats really important to SQ. So we try our best to have the lowest jitter possible in ALL of our hardware.

Also I want to say Katana is not using asynch mode. Our DAC is in synch mode (master DAC and not slave). We are the first company that uses ess9039q2 in master mode and any dPLL added jitter is simply nil. Clocks jitter (actually buffer jitter since we dont feed clocks direct to dac IC)

Will you hear the dPLL jitter ? You can change the dPLL bandwith(lower bandwith = lower added jitter) and apparently (for some) its clearly audible. It can be a a psychological effect though.

e336.png
 

Wombat

Master Contributor
Joined
Nov 5, 2017
Messages
6,722
Likes
6,459
Location
Australia
You take this article on face-value or can offer credible support to it?
 
Last edited:

miero

Active Member
Joined
Aug 1, 2018
Messages
241
Likes
289
EDIT: I started a new thread for generating signals using SoX -- https://www.audiosciencereview.com/...to-sox-audio-tool-as-a-signal-generator.4242/

Could do. Need to create the test file though....
Using a remote access via SSH to a RaspberryPi you can generate test signals from its command line...

First, install the SoX and bc calculator tool:
Code:
sudo apt-get install sox bc

Then you can generate & play signals. Several examples:

Sinus 1kHz at -60dBFS @48000fs playing on a default audio device:
Code:
play -V3 -r 48000 -n synth sin 1000 vol -60dB


IMD CCIF tone 19kHz/20kHz, 1:1 @44100fs playing on the fist audio device:
Code:
AUDIODEV=hw:0,0 play -V3 -r 44100 -n synth sin 19000 sin 20000 remix 1,2 channels 2


IMD SMPTE tone 60Hz/7kHz, 4:1 @96000fs playing on the second audio device:
Code:
AUDIODEV=hw:1,0 play -V3 -r 96000 -n synth sin 19000 sin 20000 remix 1v4,2v1 channels 2


Amplitude sweep at 1kHz with 60 second duration that fades in during 59 seconds and with explicit dither to 16 bit precision:
Code:
play -V3 -r 48000 -n synth 60 sin 1000 fade 59 dither -p 16


Frequency sweep from 20Hz to 20kHz with 30 second duration:
Code:
play -V3 -r 44100 -n synth 30 sin 20+20000


Twin frequency sweep from 0-19kHz and 1kHz-20kHz:
Code:
play -V3 -r 44100 -n synth 30 sin 0+19000 sin 1000+20000 remix 1,2 channels 2


J-test 24-bit @48000fs (no dithering):
Code:
JL=$(bc -l <<< "24*20*l(2)/l(10)")
play -V3 -D -r 48000 -n synth square 12000 square 250 square 0.00001 remix 1v0.5,2p-$JL,3i-$JL channels 2


J-test 16-bit @44100fs (no dithering):
Code:
JL=$(bc -l <<< "16*20*l(2)/l(10)")
play -V3 -D -r 44100 -n synth square 11025 square 229.6875 square 0.00001 remix 1v0.5,2p-$JL,3i-$JL channels 2

Each command running in a console can be interrupted by Ctrl+C key combination.

Visit http://sox.sourceforge.net/sox.html for a complete SoX reference manual.
 
Last edited:

Guermantes

Senior Member
Joined
Feb 19, 2018
Messages
484
Likes
561
Location
Brisbane, Australia
There are a few people making hardware solutions to reduce jitter artefacts in Raspberry Pi implementations. The RPi was never designed to be reference audio hardware and its clocks are not specced for such applications, so it has become a bit of a contest to see who can provide the best fix. Perhaps Allo has won . . .
 

audiobill

Member
Joined
Aug 18, 2018
Messages
69
Likes
30
Can you explain why that is? The threshold for audibility of jitter is way above any DAC on the market.
I thought lower everything is always better. That is why I joined and wanted to learn about testing?
 

johan

Active Member
Audio Company
Joined
Feb 23, 2018
Messages
225
Likes
372
You take article this article on face-value or can offer credible support to it?

Hi Wombat

I think its hard to corelate jitter / phase noise with sound quality improvements and I don't see how to have a definitive answer . To reply clearly to your question , this article its a personal opinion by a respected audio designer (Herbert Rutgers) . I am also aware that some papers say that anything under 50 ps is not audible .

Whom to believe ? Sound quality is very subjective and the correlation is impossible to prove.

Still , we prefer that our Katana has lower jitter rather than higher . Is there any downside to it ? I do understand that you think there is no upside though..(and you might be right )

Hope that my position is clear.
 

mindbomb

Active Member
Joined
Nov 11, 2017
Messages
284
Likes
175
I'm interested in imd vs level. Does it have the issues that some other ess dacs that use traditional opamps have?
 

miero

Active Member
Joined
Aug 1, 2018
Messages
241
Likes
289
There are a few people making hardware solutions to reduce jitter artefacts in Raspberry Pi implementations.
The right solution is a very straightforward. Use external clock(s) to generate I2S bitclock and left-right clock and configure RPi in a slave I2S mode. If you add also a galvanic isolation between RPi and DAC+clocks you will get a perfect solution with an immunity to a jitter and EMI.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Let's look at jitter and noise (at 48 kHz):
index.php
https://asia.creative.com/p/sound-blaster/sound-blaster-x-fi-titanium-hd
In fact my X-Fi is exactly using the PCM1794 DAC, with a loopback test with the PCM4220 ADC I performed the J-Test and that's what I got.

I am VERY surprised that the spectrum is very similar to yours. While it looks like my X-Fi has less spreading, but due to the limitation of ADC (which also adds noise) and different windowing/averaging parameters, I am not so sure if it is an illusion or not. Notice the pattern at around 18kHz...

x-fi j-test.PNG
 
Last edited:

audiobill

Member
Joined
Aug 18, 2018
Messages
69
Likes
30
This is a review, Measurements and comparison of "high-end" Allo Katana and ApplePi Raspberry Pi DACs. If you are not familiar with Raspberry Pi, it is an ultra-low cost small computer board that runs Linux operating system (primarily). For just $35 you get a full computer with networking, HDMI video output, multiple USB ports and expansion port that allows countless peripherals to be connected to it. These DACs are two examples of high-end audio solutions for "the Pi." The combination of the Pi and these DACs allows you to have a full networked, remote DAC that you can place next to your computer with the media server anywhere else you like. In contrast, few desktop DACs come with built-in networking and require close-in connection to your potentially noisy server connecting over USB.

The Allo Katana has been kindly loaned to me. A bit of background on this. Shortly after I created ASR Forum and started to measure audio equipment, Ioan from Allo contacted me and wanted me to test their Pi peripherals. I warned him that if sends me something and it doesn't measure well, I would be publishing it as such. To my pleasant surprise, Ioan responded with: "that is fine. If it doesn't measure well, we will work on it until it is fixed!" Alas, for a myriad of reasons despite them sending me equipment from time to time, I never got a chance to review them. I apologize to them for taking this long to do this.

The Katana is not cheap as Pi DACs go, retailing for $249. For that though, you get a pretty fancy sandwich of no less than three boards that stack up:

View attachment 14919

On the left is a microcontroller which interfaces with the Pi and the ESS Ess9038q2m DAC located on the middle board. Instead of using an integrated op-amp IC, they had a discrete one designed by Sparkos Labs which you see on the right.

I must confess, I was worried about yet another "boutique" solution in the form of the discrete output stage, thinking it would generate sub-par performance compared to dedicated ICs. As you will see later, that did not turn out to be the case.

Allo was kind enough to send me a full stack in acrylic casing including the Raspberry Pi running Volumio OS. Volumio is a shrunk-wrap version of Linux with a nice web interface so you can remotely control and access it without knowing much about Linux. It essentially turns the Pi and the DAC into a networked DAC appliance. Alas, it is easy to screw things up as I did when attempting to shut off the wireless networking on the Pi. In doing so, somehow I changed the network settings and I could no longer connect to the unit. I had to hook up a monitor and keyboard to the Pi and manually configure the Ethernet settings to get it back online. I also found that sometimes Volumio would not play right requiring a reboot to get things going again. Overall it "works" but is a bit flaky and you need to be comfortable with messing with what is essentially a DIY solution.

I like to use Roon with my remote DACs but alas, Volumio does NOT support Roon's native RAAT protocol. Instead if speaks Apple's airplay which is fine for playing 16 bit content but anything above that, gets truncated down. Sample rate is likewise is limited to 44.1 and 48 kHz. They really need to step up and support RAAT seeing how Roon itself gives away a Pi solution.

For comparison purposes, I ordered the ApplePi DAC which started as a kickstarter by now Orchard Audio. I purchased it through Amazon at $200 plus $5 shipping. Alas, that is an incomplete solution to get this DAC going. If you look at this picture you see that it requires stand-offs and an extra long "stacking header" to mate with the Pi:

View attachment 14920

Otherwise the board interferes with the Pi's components. This is a $1 connector and 50 cents worth of stand-offs that is left off. The company sells these pieces through ebay for something like $4.99 but then charges another $8.45 for shipping! And you still don't get the stand-offs you need to mount it. So here I am sitting with a $200 DAC and no way of getting it even connected. I went on Amazon and found a seller that offered four of those connectors and another who sold the stand-offs. All in all I think I spent over $20 on parts and had to wait for them to arrive to plug this thing in.

And oh, there are no manuals or instructions of any kind with the unit. I had to go hunt around online to finally find a Youtube video that shows how to put the thing together. Yes, this is an enthusiast that has put this together but come on, not including essential connectors and stand-off to make a $200 DAC work is just wrong. As a minimum there should be a strong warning in the listing that you need to buy these extra parts.

The ApplePi uses Ti/BurBrown PCM1794 DAC chips. And uses integrated OP-amps for the output stage. So this makes for a interesting battle of different approaches to building a DAC.

I am sure you all are anxious to see the measurements. Per above note on Volumio, I could not use Roon to play my test tracks. Instead, I ran them from a USB thumb drive. Seeing how my analyzer cannot control the output signal, the number of measurements I can run is highly limited. It is even less than what I could run on my older Audio Precision analyzer (where I had after many hours of work, created some asynchronous tests for such things as linearity). Over time, I should be able to augment these results. I think we can get a good picture of them regardless though.

Another note is about power. Katana DAC can be powered using one, two or three power supplies! Jumper settings determine this. They recommend two or three supplies. For my testing, I used a switching supply Allo sent me for the Pi and an Sbooster Linear Power supply for the DAC portion. For ApplePi DAC, I just used the Sbooster linear supply. FYI in limited testing, switching power supplies generate similar results but required more attention to grounding. Otherwise mains leakage caused some degradation of SNR.

Measurements
Let's start with out dashboard view as usual. First, Allo Katana:
View attachment 14921

Wow, this is pretty good performance! We have a bit more output than our 2.0 volt nominal requirement for unbalanced/RCA output. SINAD is 109 dB which is quite respectable.

Here are the results from ApplePi DAC:
View attachment 14922

So another respectable performance with SINAD of 107 or so.

Allo advertises an a-weighted THD+N (Sinad) of 112 dB which I had no trouble replicating (above measurements as always have no weighting).

ApplePi's rated SINAD is 106 dB and per above, it nicely matches or slightly exceeds that number.

This is how they stack against recently reviewed products:

View attachment 14923

Not bad at all!

Here is the frequency response:
View attachment 14924

The Katana DAC starts its roll off a bit early at 18 kHz. The ApplePi is flat to our 22 kHz limit.

Here is what they do when we feed them white noise which has much larger bandwidth than our sampling rate allows, enabling us to see how the DAC filters behave:

View attachment 14925

So first we see the wider bandwidth of ApplePi. But strangely, one channel is much less attenuation than the other (ideal case requires infinite attenuation).

Katana doesn't have this problem but then has an ultrasonic idle tone at 88 kHz or so. I understand Allo is working on better filtering of ultrasonics in the current revision of the DAC. Hopefully they will be addressing this even though it is inaudible and your system likely can't reproduce it anyway.

Switching to dynamic range, we get this for Katana:
View attachment 14926

Interestingly ApplePi does a bit better:
View attachment 14927

That is well shy of advertised 128 dB however.

Let's look at jitter and noise (at 48 kHz):
View attachment 14928

Here, Katana (in red) is textbook perfect other than a tiny 4 kHz jitter component below -140 dB.

ApplePi on the other hand (in blue) has a lot of spurious noise and jitter sources. Broadening of the "skirt" around our main 12 kHz tone indicates low frequency random jitter (could be a clock issue or noise from the Pi). The higher frequency peaks are likely different timers and clocks running around in the Pi, bleeding into the DAC. Fortunately all of this is either very low in amplitude or masked so not an audible concern. But when you pay $200 for a DAC, you like to see such mistakes avoided.

An interesting chart is THD, harmonic distortion WITHOUT noise:

View attachment 14929

We see that the Katana pulls ahead as frequencies increase. With noise (not shown) the two are neck and neck as the noise is dominant in this measurement.

Conclusions
I must say, I am pleased with the performance of both of these DACs. They rival good desktop DACs. Especially interesting is the inclusion of a discrete output stage in Katana and still managing to keep up with integrated solutions. This is no doubt to hand trimming of the pots in the factory (those blue boxes on the board).

I am a bit biased given my virtual relationship with Allo folks but if I had to give the nod to one, Katana would be the one I would pick. It has far cleaner jitter and noise and doesn't have the channel imbalance in ultrasonics that ApplePi had. As a bonus, the Allo folks are very active in forums and that means they would be listening to comments and suggestions.

That said, the ApplePi is also a very good DAC too. So if you go that route and get the necessary bits to go with it, you can have another good solution.

The combination of the Pi allows you to have an under $300 networked DAC. Alas, as with all things DIY and networking, reliability is a bit compromised so this is not for everyone.

As always, questions, comments, corrections, etc. are welcome.

-----
If you like this review, please consider donating funds for these types of hardware purchases using Patreon (https://www.patreon.com/audiosciencereview), or upgrading your membership here though Paypal (https://audiosciencereview.com/foru...eview-and-measurements.2164/page-3#post-59054).
I am not sure I understand when you say high price for pi dac. Is a pi dac different from a regular dac? The system is so cheap, why would anyone want to purchase anything other than this?
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Well it's still too DIY for many. Lots of people just want a box that works and anything more than pl ugging it in is intimidating.

Secondly some want a piece of jewellery. It's got to look good.

Thirdly some will never accept that this will perform as well as some seriously expensive dacs
 
Last edited:

johan

Active Member
Audio Company
Joined
Feb 23, 2018
Messages
225
Likes
372
Well it's still too DIY for many. Lots of people just want a box that works and anything more than ugging it in is intimidating.

Secondly some want a piece of jewellery. It's got to look good.

Thirdly some will never accept that this will perform as well as some seriously expensive dacs

Good answer. Still we think that lots of people (for example anyone reading those forums) understand the value. Of course , if you want CNC casing and such there are many companies that can give you exactly that...usually for an inflated price.

Katana complete player with an acrylic box , RPI will cost 329$ and comes with an SD card flashed and ready to use. Volumio , DietPI , Moodeaudio are tremendous OSes for music (dietpi has roon endpoint) and you can switch anytime between them just by flashing the card. No Linux knowledge required.
 

TimW

Major Contributor
Joined
Aug 15, 2018
Messages
1,065
Likes
1,406
Location
Seattle, WA
Speaking of CNC casing do you plan on releasing an aluminum case for the Katana DAC like what is available for some of your other products?
 

Olli

Senior Member
Forum Donor
Joined
Mar 6, 2018
Messages
331
Likes
242
Amir, you said you couldn't use Roon with it.. but in this review
it looks like you can run Roon bridge on it? Around 6:30...
 

audiobill

Member
Joined
Aug 18, 2018
Messages
69
Likes
30
Well it's still too DIY for many. Lots of people just want a box that works and anything more than pl ugging it in is intimidating.

Secondly some want a piece of jewellery. It's got to look good.

Thirdly some will never accept that this will perform as well as some seriously expensive dacs
But none of these things are objective, why should they even be discussed. The Alio is better, end of discussion?
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
Amir, you said you couldn't use Roon with it..
Oh I can. It is just that airplay in volumio can't handle 24 bit content which messes up the measurements.
 
Top Bottom