Sunday, November 29, 2015

 

Soekris R2R Dam Dac - Modding

Mods I made

Summary
Thought it would be good to rate the mods below
Biggest one is changing the reference voltage to the shift registers.  That is one of the more complicated changes, but it makes a big difference.

Next was using battery power for the front end, either the amanero or exD card and for isolated 3.3.  For some reason, this made a big improvement.

3.3V is next, and 1.2V is last.

At least that is my current rating of the main changes I made.  As always YMMV.

Links
tirnahifi dac thread
Most of the mods are from this thread.  Its where I got the ideas for what is presented below.

NOTE on Batteries
For the mods described below, I'm using these batteries
batteryspace 26650 lifepo4 clone
This is a clone of the 26650 lifepo4 cell.  They have worked well for me.
They are easy to use, and easy to charge.  To charge, you just need to provide them with around 3.3VDC, regulated.  I've charged as high as 3.6V, and as low as 2.5V.  2.5 volt is using two in series to generate 5vdc.  I did this in my PC.

For some of the cells, I add a small value, 1W resister, to current limit between the charger and the battery.  I used 1.5 ohm resistors, will describe where later.

These types of batteries can draw a lot of charging current, and you do not want to provide too much current.  But if you keep them charged all the time, or at least close to full charge, this really is not an issue.  I haven't decided yet if it matters if I apply the battery charging voltage when I listen to the DAC.

MODs
Remove output opamps
Removed the 4 output opamps on middle of photo below.  U28, 29, 31 and 32.  As you can see U29 and U32 have thermal pads underneath.  You may have to heat up the body of the parts to remove them.

EDIT:  Need to mention another part to remove, two really, C135 and C143.  In the picture below, they were already removed, the purple wires pass right by them.  They caps are on the output, as part of a high frequency filter, but I think they hurt the sound.  If you think you need high frequency filtering, I would add a high quality, small value film cap from each output to ground.  I didn't find a need to add a cap here.


I only use the raw output, so I don't need these opamps.  And these are the only parts that use "PWR A+" and "PWR A-".  This makes modding the power section of the DAC easier, and you can run the DAC from a lower voltage.


New Reference Voltage to Shift Registers
The reference voltage used to power the 4 rows of shift registers is crucial to getting good sound from this DAC.  There are many mods out there associated with improving the opamp circuit used to generate +/-4VDC to power the shift registers.
But I went a different route, I supplied +3.3V and -3.3V from LIFEPO4 cells to power the shift registers.  They run fine on this voltage, and the dac sounds really good using lifepo4 batteries here.

The first and third row of shift registers get negative power.  In my pictures, purple is negative 3.3V, and orange is positive 3.3V.  For the shift register chip, pin 8 is ground (last pin on the pin 1 side) and pin 16 is VCC.  Pin 16 is opposite pin 1.
So, on the first row, the +3.3 ends up on pin 16 for all the shift registers on rows 1 and 3.
Row 2 and 4 get -3.3VDC.  For these shift registers, ground is on pin 16 (VCC), and -3.3 is on pin 8 (gnd).  So these devices see +3.3 from ground to VCC, but they are really running at -3.3V.

I ended up removing a few passives around each op amp and the opamp itself.  You can see this in the picture above.  With the opamp, I heard some distortion that went away when the opamp is removed.

Initially, I wired the + and -3.3VDC to a cap next to the op amp with a small wire.  Wire goes directly to a switch.  Purple/grey wires are -3.3VDC, and orange are +3.3VDC.  You can see this in the pictures below.


In the picture below, the opamps are still installed.  Compare to the picture above, where they have been removed.

Later, I changed the wiring, to wire +/- 3.3V to each shift register, so I have 16 wires, 8 for positive and 8 for negative.  I wired a couple directly to the shift register, but I was making solder bridges, so the rest I soldered the wire to a cap next to the shift register, which is much easier.




And it does sound better with the 16 wires versus 4, more detail and better instrument separation.

To charge the batteires, I used an old, Twisted Pear dual regulator laying around.  I set both sides for 3.4VDC.  It needs AC, I used two walwarts that supply about 12VAC each to the dual regulator.  Charging power is applied through another switch.

Figure of battery and switch connections.

FIGURE Batteries Charger and switch connections

Here is a picture of the two batteries that supply the voltage to the shift registers.  I used some copper tape to connect the battery return to ground on the dac.  Copper tape creates a good, low impedance path because it has lots of surface area.   The twisted pear dual reg is at the top of the picture.
The two switches control the charging voltage for the lifepo4 batteries.  The switch on the left turns on charging voltage to the shift register batteries.  The one on the right turns on the charging voltage to the 3.3V and 1.2V batteries described below.
At some point in the future, I may redo this section, and make the battery ground shorter by moving the batteries over a bit.



Relays for the power switching
I also decided to use relays to control power to the dac.  Went to my local Frys, and picked up two of these 4 pole relays.
4 pole dt relay

I made a simple, unregulated, 24 VDC power supply to turn on the relays.  It uses a 16VAC walwart I had from a previous project, a bridge, cap and a switch.  Threw into a little box I had that has used to be a battery charger.




The relays are "mounted" with earthquake putty, which is like blue tak.  BTW, since the relays stick up, and have pretty large contacts to solder to, they are easier to solder the 16 wires to, for shift register power.  

1st relay switches, 
    +3.3VDC to shift registers
    -3.3VDC to shift registers
    +3.3VDC to power logic on the board

2nd relay switches
    +1.2VDC to power FPGA
    +3.3VDC to power Amanero board (this uses two poles, will describe later)


Removing some parts
Once the shift registers have battery power, you don't need some parts.  You don't need anything that generates a negative voltage.  I had already removed the opamps, but I also removed the +/- 5VDC regs, and their output caps.  You can see the removed parts on the right of the picture below.  I also disconnected the walwart I had connected to -V in.
+V in is only used, now it supplies the 3.3V linear reg, and the 1.2V switcher.



3.3VDC upgrade
The dac comes with a 3.3V linear reg.  This is the 3 pin device with the tab, on the right of the clock chip in the picture above.  I removed it by cutting the three pins, and then heating up the tab until it came loose.

I added another lifepo4 battery, and wired it through the relay.  +3.3 from the battery/relay go to the middle pin of the regulator.  I soldered the - tab from the battery directly to a ground pad on the board, a pad that had been used by a cap I removed.


I also added a separate wire to VCC and ground of the clock, you can see in the the picture above.  I also removed a resistor to the above, and to the left of the clock.  This isolates the clock from it's other +3.3V path.  This clock change had a small positive effect, but still recommended because its easy and just takes a couple of wires.

Made another 3.3 VDC regulator to charge this battery, and the one that powers the 1.2V linear reg.
It is a simple reg made from a LM317.  I feed it power from an unregulated walwart.  


The output of this regulator goes to a 4 pole switch.  Two of the poles switch the positive output of the reg, and two poles switch the regulator ground.  From the switch, one set of positive and ground wires go to the 3.3v battery.  

Here is a schematic of my battery charger circuit from the picture above.  The LM317 is set to 3.3VDC output.



R3 and R4 are 1.5ohm, 1W resistors intended to current limit the charging current, and to provide a little isolation between the charger and the battery while the dac is on.

1.2VDC Switcher replacement

An LT1764 linear regulator generates the 1.2VDC.  This is the minimum voltage that this part can generate.  I used it because I had a few on hand.  As shown in the schematic above, pins 1 and 2, and 4 and 5 are shorted.

I put a short wire between these pins, close to the body of the part, and then cut off most of pin 2 and 4.
Pins 3 and 5 solder directly to the board, to J2-1 and J2-2.  Soldering these to pins to the board "mounts" the LT1764 to the board.
Pin 1 in bent up, and then a wire is soldered to it from the relay.  When the relay is on, 3.3V is fed to the LT1764, and it generates 1.2V to the DAC.


I tested the LT1764 stand alone first, after shorting pins 1 to 2, and 4 to 5, I connected it to 5VDC, and measured the output.  The output was high, around 1.5V.  I decided it needed an output cap, so I found a small cap (I think it's 47uf, but 10uf or larger would work), and soldered the cap from pin 3 (ground) to pin 5 (output).  Tried again, and I now have 1.2V out of the 1764.

After checking it, I installed it on the board.  With this installed, I would no longer use the 1.2V switcher, and I had already removed the 3.3V linear from the board, so I don't need to provide 12VDC to the A+ input anymore, and I removed the two wires I had connected to J1 to provide that voltage.

Powered it up, and all was good :)

Then I removed the 1.2V switcher, most of the large input caps, and the two inductors by the switcher.  The inductors are difficult to removed, I broke them with pliers (ferrite is brittle) and then removed all the pieces.

It sounded slightly better after going to the 1.2V linear, and a little better after removing the old components.  Not a huge change, I honestly expected more, but still worthwhile to do.

Amanero Power and isolated 3.3V power
Initially, I powered the Amanero and the isolated 3.3V with a simple LM317 reg and a 9V walwart.  Sounded fine, and I didn't think this power was critical since it is isolated from the dac.  I was wrong.

On advice from nige2000, I added another lifepo4 battery here, and it was a major improvement.  More so than the 1.2V regulator mod, and also I think more than the 3.3V mod.




Battery grounding
As used in this mod, each battery is like a little power supply.  So for grounding the battery, I tried to connect the battery ground as close as possible to the parts that the battery was powering.

So the ladder batteries are grounded relatively close to the shift registers.  The 3.3V battery is connected to ground by the 3.3V parts, etc.

FPGA Cap mod
There are a bunch of caps, 36 to be exact, between the FPGA and shift registers.  The caps are required because some of the shift registers are running at -3.3V,  So, without the caps, the either the fpga or the shift registers would be damaged.

I'm changing the caps to 100pf panasonic film caps.  Mouser and digikey both carry.
mouser link to caps
I bought a hundred because of the quantity discount it was just a little more than 50.  I would order some extras because they are tiny, and easy to lose a few.

To remove the existing caps, I use two soldering irons. I recently picked up a 2nd iron from hobby king, a Hakko clone that was really cheap there
hobbyking soldering iron

If you heat up each end of a cap with an iron, it just pops out really easily.

I replaced one cap, made sure the DAC still worked, then replaced 4 (first little group), and checked again.
Then I did 5 more, for a total of 9.  That is for one set of shift registers, so 1/4 of the way done, and checked again.  Checking this often slows it down, but gives me piece of mind :)

To install, I add a little dab of solder to one pad, and then tack the cap down to that pad.  Then I solder the other pad with a little more solder.  I use tweezers to hold the cap.  With this method for removal and replacement, it actually goes pretty quickly, easier than I thought it would be.

After replacing 27 of the caps, I heard a tiny bit of static in one channel.  I tried to touch up some of the pads, and the tiny bit of static became a lot of noise.  Pulled out the scope, and started looking at the shift register side of the caps.  On a few of the caps, the signal was very low, as compared to the others.  On these same caps, the signal was fine on the FPGA side.

I replaced those caps, carefully using as little heat as possible.  I checked the signals again, and they were fine.  Then I hooked it up, and static/noise was gone.

These film surface mount caps are very delicate, and you should use as little heat as possible to install them.  You should also check them after you install them.

But, I also heard an improvement in detail, realism and low end presence after making this change, so I still recommend the change, but its a borderline recommendation, because the caps are so delicate and easy to break.


exD USB to I2S card
Got a used exD USB to I2S card to try instead of the Amanero, to see how it would sound.
I had to email exD to get drivers for Windows 2012R2, but he sent me a new driver, and with that I installed it fine.

exd audio

This card needs 9V to run, it does not use USB power.  It has a few onboard regs that run off of the 9V.
There is a 5V reg, and a 3.3V reg that runs from 5V, and one around 3.1V that runs from 9V.


EDIT on March, 2018
First update in a long time lol.

Changed USB to I2S card.
Bought this one from DIYINHK
Its the

XMOS 768kHz DXD DSD512(DSD1024) high-quality USB to I2S/DSD PCB

Also wired in a Potato Semi 74G374 octal flip flop between the DIYINHK card and the DAC.


Pin 20 to 3.3V,
Pin 10 and pin 1 to ground
pin 11 to i2s master clock
pin 12 to dac data
pin 13 to xmos data
pin 14 to xmos lr clk
pin 15 to dac lr clk
pin 16 to dac bit clk
pin 17 to xmos bit clk

Pretty sure the above is correct, but I can't see the bottom silkscreen on the xmos board since it's wired up, and the xmos signal names are there.


I soldered FF into a 24 pin adapter, but its only a 20 pin device so the pin numbers are off.
The top of the adapter board, in the pic below, has pin 11 (which has a white wire in it) on the left, and pin 20 (blue wire) on the right.  So, pins 1-10 are on the bottom.





I also removed two isolators, and jumpered across to make the connections, so pic below







Comments:
Hi,
I have a few questions.
1. For the "FPGA Cap mod", could you show us a photo about how it looks like after replacing the capacitors?
2. For the Amanero USB card, is it needed to remove the 3.3v regulator onboard? As when there is usb cable plugged in, there will be 5V power from the PC side.
Thanks
 
Hi,
About charging, I am new to the battery thing and I found that it requires some circuit to make current and voltage limiter at different stage of the battery power level, or simply use some IC specific for charging lithium battery. So is is ok in long term if you use a fixed 3.3v for charging? I am thinking to implement this and I aware of the safety of battery in long run as I am thinking to charge it whenever the DAC is off.
Also, if I fix the batteries in parallel and not relocate them when charging, that means the batteries are charging in parallel, is it ok?
Thanks a lot.
 
Hi Obee
Sorry about the long time to answer, I never remember to look for comments/questions to my blog pages.
1. OK, I can add a picture, but it just looks like a cap, although slightly different from the original.
2. Yes, I did remove the 3.3 regulator for the Amanero, did that a long time ago.

LIFEPO4 batteries are easy to charge. Yes, you can buy a charger, or build a special circuit, but you really don't need to. Some batteries do need a special circuit to charge, but LIFEPO4 batteries are really easy to charge, but give them a constant voltage like I did and you're good to go

 
Thanks Randytsuch
 
Randy, great stuff...will you try the dam1941 as well ? I got one of those and would love to apply LIpO to ots shift registers....
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?