Raspberry PI with balanced output

As a sound system designer you will, at some point, need a source to test your endeavors.
Apart from saying 'test one two...'  for hours on an end in a torn SM58 microphone one could also use some music.
More pleasant for the people around you and you will have the added advantage of having your hands free to twist those nobs..;)

Those soundguys being around a bit longer might remember the confusion when CD arrived. Away with the crack rattle noise and buzz from these LP's (let alone cassette copies) and hey, wow, what a clear definition, but wait a sec, what is going on in the top-end?
However the method of testing and tuning your sound system with your favorite CD became a world wide spread practice.

Now fast forward: we are in the 21st century, skipping Cd's are obsolete and a lot of people will use their ipad/pod/tunes to perform their tunings.
The lot of them will have some doubt regarding the quality of mp3 and advocate the one and only use of wav (or flac for our open source friends).

Certainly, we too have been using some CD's and a 'high-end' hifi CD-player  to do our system tuning.
But, although we never did a serious investigation, we always had some suspension regarding ripped CD's playing from some OS through some USB interface.

Now enter the realm of playing music from your computer.
First you will notice that it really does matter a lot what brand of interface you will be using. In pro-audio the most common solution will be some RME interface (pricey!)
If you investigate that route further you will notice that it also does matter a lot what player (read codec) you will be using: foo-bar being one of the more popular as opposed to anything like 'windows-something-player'.
Being on this route you will also find out that you have to bypass all OS based up and down sampling, Wasapi might be a solution in a windows environment.

I decide to take an other route:
Linux.

On several instances through this blog you will find pointers, a year ago or so f.e. I started to write a (never finished) walk through in setting up a headless music player.
Now let me try to round it up:

Raspberry PI running Volumio with balanced outputs

Well, with 5 million of these boards sold, one can presume that you have at least heard of Raspberry PI.
Rpi is based on one these so called S(ystems) O(n a) C(hip) running a linux based OS.
Actually at first marketed as a entry level platform for children to get to know programming it seems that these nice toys are becoming very popular with grown ups. Ahh, well, toys for boys I guess..
On a zillion places on the internetzz you can find information and howto's on operating a Rpi.
Let us concentrate on running audio.

First of all the on board audio chip really sucks, so first thing to establish is to get your audio data of the board in a proper way.
Now you could use some sort of usb based device, maybe even the above mentioned RME DAC's, but I had a lot of trouble running bigger data streams (24bit/96k) across the combined USB/ethernet interface.
Apparently just not up for the task.

A different route only recently opened up by using the i2s protocol on gpio ports.
Several people have been making nice add-on boards that sit on top of your RPI.
Here is three of the boards I tested:

Wolfson

Duriosound

Hifi berry digital

The 4th is a board from Hifi berry and puts out analog.
All these boards perform similar. (no, this is not turning out to be a board shoot out ;)
For the story we will continue with the latter:

modded HifiBerry analog in place

As has been mentioned before and over: two things of importance in 'bit perfect' reproduction of digital audio:
1.  no interpolation by up, down or oversampling and no arithmetic functions like 'replay gain make up' or any other mockery should be happening between your audio file and your D/A converter.
2.   clock should be as stable as possible.

For the first we trustfully use MPD playing audio directly through ALSA (i2s). Volumio is a very nice Linux distribution which takes care of all these software nooks and crannies which are explained in other places. Just google the key words.
The second is a bit less documented, you will find people arguing against RPI because to derive a sample freq. from 44100hz you need a fractional division of the RPI 's clock freq..not so good..
48k / 96k should not give you that set back.
Do I feel that as a big problem?
nah.. Remember we are talking a 100 euro device here so don't expect 1K+ euro performance!

But let us try to get a clean clock signal in the first place.
The guys who really know things about digital audio like Guido Tent (Tentlabs) or Eelco Grimm (Grimm clock) have whispered PSU in my ear.
So first step in improving the basic setup was to get rid of the wallwart switched mode usb psu:

linear PSU using LM7805 with 2N3055

Very simple linear PSU with the well known LM7805. To boost the current a bit I used a parallel 2N3055 (think 70's for sound ;)
All very well known basic techniques.
I did use a mix of some fine electrolytic capacitors in a experimentally achieved witchcrafted constellation, though...


Next thing to do is to make a balanced output.
Pro-audio guys really hate unbalanced signal transport: introduction of rattle and hum being the most mentioned spoilers of the party.
But what if a asymmetric transport is anyway a fundamentally wrong concept?
Treating 'earth' like something solid might be a right concept in high voltage power distribution, but  is completely wrong in signal transport (imho).
Naturally it takes 2 conductors to transport a signal and they are both equally vulnerable to the introduction of all kind of disturbances.
Gold plating or Teflon insulation of a single conductor does not help in any way.

Here is a picture of the circuit with it's separate dual 15v PSU:

unbalanced to balanced with That1646 and TL051 driver

The TL051 preamp is selected from a range of opamps using the still 'golden ears' of my 13 year old.  :)
Being past 55 I think mine have been deteriorated  to iron/lead/wood by now.
Remarkable sensation, knowing it's there but not being completely sure, like looking through a stained glass window.
Oh well, experience replace the loss (I hope).

Lastly a picture of the finished 'product' in a vintage AKG phantom supply box :

Active X-overed HiFi ?

In the Netherlands we have this expression about plumbers always having leaky taps in their homes.

Sooo 
While business being on the slow side I decided to upgrade my personal hifi / mediacentre sound system to something with a bit more throw and detail.
Having so much stuff from prior experiments lying around I made it a design restriction to myself not to buy anything new.
This worked out to be a much longer lasting effort than I calculated for.
Three weeks in a row of precious time wasted (?)
Oh well, let's call it educating myself.
Let me try to give you a condensed report..

Design imperative was to build a small system (needed to fit in the bookshelf) with enough throw and definition to listen music at very low levels at night while the family would be asleep.
For that we would need FIR filters and so a Soundweb 9088ii will be at the hart of it all, see my prior posts on the upgrade and repair  of these nice tools.

 

Next thing would be the speaker cabinets themselves.
Top-end wasn't going to be that time consuming where as I was to recycle my priorly build tractrix waveguides.(also see previous posts)
Loaded with some DH450 drivers from my preferred, and highly appreciated supplier A&D-audio

rearview

frontview

Now why is that resistor in series?

Three reasons: 

First of all, evidently this horn/driver combination will have a lot more gain than the accompanied (hifi-) mid speaker.
So a big attenuation in the digital domain would be needed to overcome a magnitude mismatch.
Rather than choking on the associated loss of digital headroom/resolution, I decided to make a passive attenuator.

Secondly the amplifier will have to make some amplitude swing so distortion introduced by the amp (class AB crossover distortion) would be relatively small.
For some reason (presumably distortion associated) I also like the sound of warmer electronics (warm not hot!).
Ok, ok writing this I detect mumbo-jumbo alert, but believe me there really is a difference.
While I don't have the engineering skills to explain this difference, this doesn't mean I am lacking in, on firm science based, perceptional skills!

Third and not the least:
The driver amplifier combination will now operate in a (partly) current drive mode (again also written about in a prior post).
Naturally we now have to use some (min. phase) EQ to compensate the magnitude differences introduced by the impedance curve of the driver in series with this resistor, but the overall reduction of distortion is quite convincing.


Now for the mid speaker:

left is hifi      -      right is PA

From the maximum total cabinet hight of 28 cm (had to fit in the bookshelf, remember) I now only had 17cm left.

In pro-audio industry more and more smaller speakers are being used in those line arrays you see at every bigger event.
So (being a pro-audio designer) one would say I have plenty of sample speakers lying around to be used as this mid speaker.
Quite disappointing!
From all the samples I tested  (yup, a lot of 'big' names) the only one remotely optional for this purpose is the one shown in the picture.
(This experiment alone actually took a whole working week, so fast forward here ;)
Does this imply that all PA speakers are horrific?
Yes and no, for this purpose they will not be suited in a same way as a hifi speaker will not rock you.

Although this statement seems quite obvious one really has to add in a lot of question marks here..why?..

Actually this is a pledge for a paradigm shift in our industry.
We neither have the tools nor the (scientific) language to quantify the above described statement.
Surely this has something to do with distortion but how, what, why?

Anyway, for the purpose of the story let's continue with this Seas mid speaker.
In the pictures above you can still see an open back.
Actually being a bit too lazy to do a proper cabinet tuning (either closed or reflex) I thought this dipole configuration would be a good idea.
Stupid!
Horn loaded, zero phase shift, directed high's don't combine with room loaded low's..
So in the final cabinet this is a closed box (with luckily the correct volume for a .7 tuning)

Same thing for the subs which I made from some no brand speakers I had lying around.
Nothing special here, only highly over dimensioned for the purpose, in that they all ways will be operated in their linear region only.

soon follow up on electronics and (FIR) filtering

for now here's a preview of the final result:

Bringing Soundweb greens back to live

We are still using quite a lot of  old green Soundwebs 9088 and 9088ii in our soundprojects. The main reason being  that these ancient machines were allready capable of some FIR filtering.
Mind you: there is a bug in the software which makes you have to reverse the filter kernels. Presumably nobody ever noticed using linear phase filters (symmetrical as these will be) ;)

Anyway I had some broken ones lying around so I thought it would be time to do some repairing.
One of the apparent vulnerabilities is the power supply cap marked as C216.
So after replacing these I was able to get some off them back to live.
For all you out there plagued by the dreaded searching.... (at infinitum) modus I will repeat some 'rules':

1. replace C216 cap, while you are at it maybe it's a good idea to upgrade cap C265 with something beafier as well.
2. use the spare jumper at the option A position to set your machine in debug modus.
3. switch on power, you now should get the 'nightrider' row of flashing leds.
4. use the backup loader.exe to reload the firmware. make shure you have the correct *.a21 file. I learned the hard way do not to make a mistake!
5. when done switch of power, replace jumper and reboot.

As allways use correct settings of com ports and bitrate (38400 front, 115200 back). Also the FTDI driver / USBtoRS232 works fine on windowsXP (and 7) but on windows 8 it appears to be troublesome.

Now for the fun (and utterly time consuming and by no means commercially viable) part.
Let see if we can do some 'audiophile' modifications.
In the issue 1 model (9088) there are some VCA IC's build in to facilitade volume controle directly after the DAC's.
These suck! ;)
I forgot the type number but in a earlier experiment some years ago I removed these and  shortened the signal path. So no more volme controle on the output and maybe (with high gain power amps) a bit excessive noise.
But hey, who cares for noise?
It's the universe coming through our speakers!
Big plus!

Next thing I thought would be whorthwhile is replacing all these (non-polar) caps in the signal path.
Well this wasn't going to happen.
All my regular dope dealers like Farnell and RS and the like seem to have stopped selling these, hmmm, bummer.
Now what if we replaced them with regular caps back to back with a DC bias.
(This is an idea I actually stole from a very old Midas PRO4 console I once owned).

Big improvement?
Argh euuhh, yes with all my skills I can notice the difference (some more clarity) but I 'm not even shure if this is desirable.
So, nah not going to do that on all our machines...

 ...edit...edit...edit...
Not completily convinced by the dispointing results of the above modification I decided to give it a second try.
Same mod but this time I tested a stereo a/b comparison..wow.. this is some serious shit.. much broader and deeper image and a very convincing musical impact..now why the hell is this???