Raspberry Pi I2S to SPDIF Hat

Nothing more than shameless promotion! I'll dial it back.

Cheers,
Michael

 

Well... a good cd transport is still way over the raspberry league. This evening I had a listen to a teac vrds 25. Unbealivable clarity all over the room. Back in the days its msrp was about 3000USD IIRC, It may sound unfair but it is like that.
The chinese digipi is no screechy. I have it at home for a couple of days of testing and it's more or less very close to the digi+ but less meaty in the bottom.

I'm waiting to receive the 503dac1, unfortunately there's an ocean between me and the US.
I'm curious to find out if the 503dac1 can compare positively with the other digital hats.

No reason to dial it back imho. Maybe we could talk about it in another dedicated thread but since most of us are focused on its digital performance I think it's pertinent to discuss it here.

 

Hi Michael, agree with Vastx-- don't dial anything back! This is just for my own focus identifying the digi+/D2 changes... want to be concise/methodical on testing and describing the differences, and was just picking your brain in case was missing something in the thread.

 

Stock digi+?

this config:

USB Switching Power - piSomething - digi+

 

Yes, stock digi+.
Not sure if you are asking me if I refer to that configuration when I make comparison, but if that is the case I never use a switching ps.

 

Edit, deleted-- will post this later.

 

I used to own the Esoteric P3 and D3 transport and dac combo. To be fair, you can't say a transport is in a different league than the Pi because there is a lot that goes into the system prior to the Pi that will impact the sound. Now I am not saying the best pc based transport system including the Pi as a renderer or player or server is better than my Esoteric combo. But I am not sure it won't be close assuming you have a top flight pc transport with optimized software, hardware, power, cabling, etc.

 

To be honest I don't think it is fair telling me what I can or cannot say. Not sure also why you put the PC into the equation... The raspberry\digi+ setup doesn't necessarily needs a PC to work. So far IME I feel pretty comfortable saying that a good cd spinner is way superior to my Pi2\digi+\lpsu in my setup.
I'll be happy to change my mind if new hats prove to be improving on the situation though.

 

Let's hope the SPD1/2 can make you happy! When you get the DAC1, what you will be testing is the same as the DIGI+ with one major exception, no use of the Pi 3V rail. This plus a high PSRR LDO (ADP150) should go a long way to improving what you now hear with the DIGI+. But, until we have the SPD1/2 built (sometime this week), get the low noise clock option working under SW and run it with your linear supply, we won't know if we can give your CD transport a run for it's money. I certainly hope so, but I won't be surprised if we fall a bit short.

Cheers,
Michael

PS Off topic AND shameless promotion, but you all said I could! Our 503HTA Hybrid Tube DAC/AMP Kickstarter campaign is down to under two weeks. We still have some discounted units so hop on over and get one!

 

Hello Michael,
Would love to support your 503HTA Hybrid Tube DAC/AMP project - but I was wondering if we could use the Tube Amp part of the 503HTA Hat together with your 503SPD1/2 Hats?

I've also been looking at the Rpi Isolator HAT from iancanada, and wondering, how things might fit together with your HATs, whether things are duplicated or might compliment each other. http://www.diyaudio.com/forums/pc-b...-i2s-dsd-isolator-hat-native-dsd-decoder.html (I'm not sure if linking to other forums is OK, will edit this post if appears to go off topic. With the whole HAT lego building blocks approach, it is easy for 'non super techie minding people' to get confused as to how HATs from different developers might fit together).

My 'New setup' is most likely not to be a RasPi, but an Up-Board with one of your 503SPD1/2 boards; powered by a HDPLEX 100W LPS https://github.com/rootscript/audio-dump/blob/master/HDPLEX-100w-LPS.md.

There is so much development happening in this area, so I'm trying to piece together all the bits on my shopping list. It looked like the Up-Board was ready to ship last week, but those guys have gone a little quiet, as they noticed some last minute improvements could be made to the GPIO.

I'm really looking forward to the Audiophile/Audiophool direction this thread was going towards, and would love to hear your thoughts on how you think a 'general version', would compare with Audiophile version.
My feelings are that at the prices your suggested earlier would be fair for a 'general version', and I was wondering what you & the other guys thought were as to the upper ceiling in price of what the 'Audiophile/Audiophool' version would be?

I guess you can design something without cost being a factor, or is there still a 'cost factor' in mind when designing/engineering a 'Audiophile/Audiophool' version? I hope you understand what I'm trying to say (Do you design the product with no limits & just add up the costs at the end, or do you get an idea of the upper cost limits people would be prepared to pay, and design within that upper price point in mind)?

There has been quite a lot os honesty & openness here, which I find to be VERY refreshing. I'm sure that all of the guys here feel quite a lot of loyalty towards this type of approach. It looks like Vastx, tboooe, Abartels, myself & others are all looking at things from different angles/perspectives, would be COOL to get a summary of the options after your testers have given their feedback.

 

My apologies...no offense meant.

The reason why I brought up PC and other peripherals is because a transport is completely self contained and designed to work optimally with all the internal components. The designers probably had a chance to do listening tests and select components to get the best sound in their opinion. The Pi needs components supplied by the user which could and does impact the sound. For example, the user has to provide the power supply, power cable, music source (USB, NAS, etc.)...Perhaps the user selects components that do not have synergy so if this system including the Pi does not sound as good as a transport, how can we be sure that it is the Pi causing this?

That being said, just my thoughts only.

 

The 503HTA supports analog in from external sources. The I2S bus is available on the GPIO connector and I have finished an XMOS based USB to I2S that takes any of our Audio hats and truns them into USB DAC's or DAC/AMP's. But, there is no S/PDIF in option at this time.

As for "cost no object" design, I do not start with a fixed dollar amount. However, I do a sort of self filtering where I weigh the merits of each item against it's cost. Ultimately I do have to put it onto the market and compete with other, perhaps lesser, perhaps better, products at their price points. For the SPD1/2 I honestly don't see what else would make sense. We have high PSRR LDO's, low phase noise clocks, isolated BNC and XLR plus external 5V input. I am always open to suggestion, as you have noted from the back and forth in this forum. So please do not hold back! What would make this better? Blinky lights? Chrome fins? Bedazzling?

Cheers,
Michael

 

Thanks for your reply Michael, I guess I'm just trying to assess what has been discussed already in this thread in order to look for areas that might be available for improvement.

So just to recap (based on my very VERY limited & lacking understanding (Google is my best friend)):

- high PSRR LDO's (Power Supply Rejection Ratio in Low-dropout regulators)
so these DC linear voltage regulators will keep external noise down, and will be effected by

1. the internal reference voltage
2. the error amplifier

and some of the methods to reduce LDO noise are:

1. Filtering the reference (if an external capacitor is used, what quality options are there?)
2. reducing the noise gain of the error amplifier

- low phase noise clocks
So there are 3 clocks on the 503SPD1/2, what do each these do?,
so one of these clocks bypasses the use of the jittery/noisey clock on the Rasberry Pi?
In terms of the clocks used, do these need to be soldered to the board, or are there sockets available so that they could be swapped out ? (if someone wanted to try something else, a bit like 'tube rolling' or 'op amp rolling')
(please remember that I'm an idiot - i guess that it isn't just about the quality of the components used, but how they are implented)

- isolated BNC and XLR
If people require RCA, then a BNC to RCA cable/adaptor could be used
I think that 'WM8004 has a SPDIF input' was also mentioned earlier, so could we input a CD player (or another source) via TosLink?
Also can you explain a little (for simple people like myself) what Transformer coupled outputs are, does this option require us to do/have something else?

- external 5V input
So we can feed the 503SPD1/2 from a Linear Power Supply (LPS), which will in turn feed the Rasberry Pi, or we can switch the Jumper to isolate PI 5V from Hat 5V, and feed them both separately.

I hope you don't mind recapping a few details for the dumb ass users like myself?
I do hope that something useful might come from my comments, particularly because my perspective is a little naive; sometimes it is useful having an idiot in the room !

 

I included these quotes from awhile ago, where you were talking about LT3042 noise... why filter the audio band, or just the audio band, when we're still digital at this point. Since SPDIF operates far higher at 2.8MHz--6MHz etc. and higher, why not worry about filtering noise at these frequencies? Trying to learn a little, Thanks!

 

I will take these in order:

High PSRR LDO - The LT3042 has a PSRR of 80db+ up to 100Khz. This insure 5V power supply noise does not get to the 3.3V rail that powers the WM8805 S/PDIF Transmitter. For example if you had 100mv of ripple/noise on the 5V supply (fairly high, but not uncommon for a switching supply) it would be reduced to 6 uV on the 3.3V side! That's micro-volts! At this point it simply can't affect the digital chain. If this was powering a DAC I would add filtering to the LDO feedback which might reduce it by another 20-30%.

Low Phase Noise Clocks - There are three clocks. One is 27Mhz feeding the WM8804 internal PLL. This one is used by default when you run any SW that supports the Digi+. This is not a particularly low noise clock, but still higher quality than the Digi+. The other clocks are NDK NZ25420SD and are used to drive the I2S clocks directly, without the PLL. This greatly reduces noise, but requires that we modify the software to both select the right one (22.5792 for 44.1Khz derived audio and 24.576Mhz for 48Khz based audio, as well as set the PLL in bypass mode and set the right divider for the actual sample rate (48/96/192Khz or 44.1/88.2/176.4Khz).

An important distinction between power supply noise and clock phase noise. PS noise rides on the high and lows of our ideal square wave that is our clock. Phase noise or jitter, affects the edges where the clock transitions from high to low and vice-versa. When the clock is essentially moving in time, the down stream S/PDIF devices have trouble locking onto the data stream. The nature of S/PDIF makes this worse because it does not actually supply a separate clock. It is embedded in the data stream and has to be "decoded" by the receiving device. If it locks to the stream and then the edges shift, it can incorrectly decode data, or even lose lock altogether.

Isolated BNC/XLR - The primary use of isolation is to insure that each side of your audio system (transmitter and receiver) accurate communicate those important bits. Isolation insures there is no ground voltage difference that could cause the receiver to "misinterpret a high as a low or vice versa. Ground differences between systems can easily be 100's of millivolts. That might represent more than 20% of the actual signal strength the receive gets.Copper, as opposed to fiber optical, can also have external noise "imposed" imposed on it due to high strength magnetic fields, such as those around speakers or power amps. Balanced XLR insures this noise, which is imposed on both the plus and minus (AKA common mode noise) is easily filtered out by the receiver that only counts a bit when it sees the plus and minus are opposite.

External 5V Input - As clean as our 3.3V can be using the LT3042, providing the cleanest 5V possible certainly won't hurt. Honestly, this one is the most questionable to me. But I am prepared to be wrong!

Bottom, line is that getting the data to the receiver with 100% of the bits and an accurate clock are the only job we have. Nothing else matters!

Cheers,
Michael

 

Good question. The PSRR we care about is from 5V PSU and it's switching noise. A typical switching PSU runs at anywhere from 100Khz to 2Mhz. Across that band, the LT3042 with 5V in and 3.3V out, maintains that 80db+ PSRR.

Cheers,
Michael

 

None taken, No need to apologize, mate, I'm sorry if I can sound harsh sometimes. It's not my native language afterall
Being a little offtopic I sent you a pm.

 

I will also get an UP Board and load it up with my software optimization packages I have. I think this will be a closer comparison to my current PC.


I believe the SPD has the capability of being powered from a separate 5V source than the Pi. So the options are:
1. supply Pi with 5v which would also power the SPD
2. supply the Pi and SPD with separate 5V

Mike, did I get this right?

 

Eager to see the combo that does compare, your Esoteric is pretty up there, and I do want to get to all the files I have in Roon.

I am comparing it right now with a CD player/spinner through a Gungnir Multibit, and its not really close. The player alone is even better, but I got the Mutec 3+ today, and will compare it that way to see if it helps.

 

Yup!