Musiland 01US USB DAC I2S into Sabre

[jkeny]

I started my recording of mods to the Musiland 01US USB DAC board here: http://www.diyaudio.com/forums/digi...usb-spdif-24-192khz-asynch-4.html#post1955840

I first focused on getting the most from the internal PCM1793 DAC that's on this board so my mods involved:
- external USB power
- changed digital decoupling caps for Oscons
- changed 24MHz crystal for low jitter 24MHz Crystek clock
- bypassed op-amp output stage & went direct from PCM1793 DAc Vouts to transformer


But somebody told me I should open a separate thread. Seeing as I have just done another round of mods, I decided now would be a good time.

I decided to experiment with the power supplies to the board so I cut some traces to isolate sections that I could then supply separately:
- First I wanted to isolate the PS to the PCM1793 DAC
- this has two PS a 3.3V digital & 5V analog supply
- The PCM1793 DAC is at the end of a 3.3V supply that feeds the Cypress USB chip, the Xilinx FPGA - so I cut the trace just before the decoupling cap to the DAC
- The 5V supply is provided through a circuit that has a dc-dc converter chip (NCP3063) - I cut the trace feeding the USB power to this circuit
- I also lifted the + & - PS pins on the OP275 output op-amp as I was using output transformers anyway.

So now I have the complete out stages isolated & they can be powered off also.

So now my total supplies to the board are:
- external 5V through USB cable
- 3.3V to clock
- 3.3V to digital supplies of DAC
- 5V to analog supplies of DAC
This improved the sound some but I finally tried the cheap Sabre 2 channel Vout DAC (ES9022) board which I had tried before but found then that I had a lot of noise on it. Because I was now able to power off the output stage I was happier feeding I2S to the Sabre DAC. So I only need a 3.3V supply for the Sabre - I disconnected the 3.3V & 5V analog supply to PCM DAC. I don't believe I need an external USB supply anymore as it's only supplying the Cypress & FPGA chips - not as critically noise sensitive as the PCM DAC.

This sounds far superior to the PCM1793 (not surprisingly, I guess) & there's no noise on it. It brings a much more natural, rich sound & there seems to be more resolution of detail which might have to do with the ESS jitter handling - hyperstream?. Sound now more divorced from the speakers. The output level is far higher than the Vout from the PCM1793 & The bass is now noticeably superior.

So the Sabre is a good cheap DAC (I think it was $2-3?), I would recommend (not sure if you need a NDA though - I had one already for the ES9018). I'll be listening some more & will post some pics.

In terms of bang for buck, I reckon this combo is dam good.

 

[jkeny]

One thing I've noticed that bugs/confuses me & I wonder about
- when I bypassed the Op-amp & went with an output transformer, I noticed that If I select the HP button the Musiland control panel, I can hear the relay click & the vol went down slightly on my output. The same happens with this external Sabre DAC set-up being fed from I2S. Any ideas what is the cause of this - some change inside the FPGA, I really don't know?

 

[glt]

jkeny,

Very nice!
Concerns about the multiple options of the US01 was one of the reasons that led me to get a US Mini to do the I2S mod since it is a single function device (still waiting for it).

That bug/feature is very strange. Why would it affect the digital lines? Maybe you can post in the musiland forum. I find using the google bar the best way to get instant translation.

 

[jkeny]

Thanks glt,
I may well post it there but I can't always understand the answer in translation

 

[olev]

Well my Musiland 02 US performs just like this - it has a memory for LINE and HP settings. Set LINE vol to 99, click HP set volume to 89. Now when you click LINE the volume turns to 99 and again when you click HP it goes to 89. You have to set both to 100 for digital volume control bypass. BTW I call this a REALLY good feature - LINE should be always at 100 but when switching to HP and volume is 100 you probably damage your ears/headphones (and this is why HP volume defaults to a much lover value).

 

[jkeny]

Well my Musiland 02 US performs just like this - it has a memory for LINE and HP settings. Set LINE vol to 99, click HP set volume to 89. Now when you click LINE the volume turns to 99 and again when you click HP it goes to 89. You have to set both to 100 for digital volume control bypass. BTW I call this a REALLY good feature - LINE should be always at 100 but when switching to HP and volume is 100 you probably damage your ears/headphones (and this is why HP volume defaults to a much lover value).

Thanks Olev, & I agree with you about Line & HP on analog outs but this is running I2S into Sabre DAC so analog controls shouldn't effect it.

But, I'm even more confused now as I turned it on again tonight & the analog CH. left vol slider is now controlling the volume as well as the ASIO vol slider Last night the Analog CH slider had no control over the volume (as I would have expected) only switching between Line & HP there was a noticeable volume difference. I reckon it's a glitch/feature in the software - probably some memory storage issue.

I actually prefer it this way as it gives me finer control over the volume - last night a slight increase in the ASIO vol slider & it was too loud - now I have finer control over the vol. Just wish it was consistent.

The software also sometimes shows me a control panel is without any settings - it seems to lose it's settings stored in memory. This seems to only happen if I unplug it a couple of times - music still plays & if I can click on all the settings again

 

[jkeny]

I'm listening to this again tonight & it is fantastic sound - I'm thinking of boxing it up into the best USB portable amp I've heard. The configuration I have is
- a Musiland USB 01 US feeding I2S ($80) into
- Sescom MI-97 transformers ($50) into
- a Sabre ES9022 2 channel Vout DAC ($3) into
- A modded Lepai Tripath TA2020 running off batteries (recovered from "dead" laptop battery pack)
- into DIY Jordan JX92S speakers

All this (except for the speakers!) could be boxed up into a neat enclosure with batteries for a really portable & great sounding USB music player.

 

[jkeny]

Some pics:
- Musiland with all the output stage disconnected & red wire at the top = 3.3V dig supply for PCM1793 DAC. Red wire at bottom = 5V analog supply to PCM1793
- Sabre ES9022 connected via ribbon cable to Musiland using I2S interleaved with ground so 8 wires needed. The two black wires facing down are the previous connections from the PCM1793 DAC to the transformer. The black wire curving up to the top of the picture is the connection from the Sabre to the transformers.
- IS2 connection to Musiland is by plug into a socket installed underneath board. Fine copper varnish insulated wires connect to I2S tracks leading from Xilinx to PCM1793 & run to bottom of board through vias.
- Bottom of ES9022 Sabre on protoboard
- Top of protoboard

It's amazing this sounds so good with such lash-up wiring but it does with dead quiet background & fine sonics!

 

[glt]

How to install caps

jkenny,

I got the MINI and wanted to upgrade the 10 uF caps around the regulators. However, those caps are TINY! The leads of typical sized caps I'm afraid will not fit in those holes. How did you solder the caps?

Here is a picture of the caps I plan on using. Right now I think it would be best to solder them in the backside of the board in parallel with the existing caps.

 

[jkeny]

jkenny,

I got the MINI and wanted to upgrade the 10 uF caps around the regulators. However, those caps are TINY! The leads of typical sized caps I'm afraid will not fit in those holes. How did you solder the caps?

Here is a picture of the caps I plan on using. Right now I think it would be best to solder them in the backside of the board in parallel with the existing caps.

Hi Guillermo,
I think I used something like 22uF/16V Oscons to replace any digital decoupling caps. The 01US has more clearance above the board for these caps but is tight underneath so I'd imagine the bottom of the "Mini" will be similarly restrictive if you still want to fit it back in it's case? Your caps may have to lie down! I also didn't always try to fit the new cap leads in the vacated cap holes but found attachment places nearby.

 

[wackyterbacky]

Can you provide the Xilinx pinout for I2S?

Hi jkeny:

Thanks for your pioneering work. I just received a Musiland 01 USD. I see you have modded the 01 US to provide I2S outputs by using the DAC datasheet pinout.

I had posted the original Juli@ I2S output information on head-fi back in 2007.
Easy I2S from Juli@ PCI sound card - Head-Fi: Covering Headphones, Earphones and Portable Audio

I think the I2S interface is the way to go, but the PCI card in my machine is too subject to PCI latency issues despite lots of effort. That is why I am intrigued by this Musiland async USB implementation.

I am currently using foobar2000 + Sox resampling ( 4X upsampling with max BW and Linear phase) - > stock Musiland 01 USD ( 1.0.3.2 driver, did not get license yet ) -> BNC -> DIY BNC coax RG179 -> Buffalo32S DAC -> DIY Balanced AKG 701 headphones. Wonderful engaging detailed sound.

Can you tell me which are the I2S pins on the Xilinx part? Not sure if the I2S interface will be active, but since these devices all share the same driver I think its a safe bet.

Thanks.

 

[jkeny]

Ah, Wackyterbacky, you should have checked first - the 01USD does not output I2S, AFAIK, only SPDIF - that's why I bought the 01US - it sends I2S to the internal PCM1793 DAC from pins 3,4,5,6 of the Xilinx. On the 01USD only pin 5 is connected & this outputs SPDIF which ends up as BNC ouput. Sorry about that!

As a consolation, it may be the case that I2S gets resampled to 88.2KHz or 96KHz and isn't output at it's native speed - I'll be checking this today!

So SPDIF may be a better interface if done properly! Particularly when interfacing with the Sabre DAC. One thing to change would be the resistor divider before the BNC output (schematic attached but I think R4 should be 07ohm? - reading from the pcb component)

Taken from another site:

Assuming: C5 is very low Z at S/PDIF frequencies, and paralleled U5 ouputs are very low Z source, and that T1 is 1:1
Then: effective Z (ac) s/b R2||R4 or ~89R (using the schematic values of 470 & 110)
with R2=330R and R4=100R, then Zout=76R

I'm not sure what the formula for this derivation is - can anybody help?
You need to bring the Zout to 50ohms for BNC, I believe!

 

[wackyterbacky]

Check out Thevenin Equivalent Circuit

Heres a link on wikipedia: Thévenin's theorem - Wikipedia, the free encyclopedia

So the output impedance is computed by shorting all voltage sources to ground and calculating the resulting impedance, which ends up being 470ohm in parallel with 110ohm. Equation = (470 *110)/(470+110) or 89 ohms.

In this case he assumes the capacitor is very low impedance at the spdif frequency, and that the output impedance of the parallel inverters is extremely low, and that the output transformer is 1:1.

So, for true 75ohm BNC, the suggested resistor changes yield 76ohms using standard values.

Thats a pity about the 01 USD, but I did not find this information on any of the threads. I should have asked first.

 

[jkeny]

Thanks Wacky, for the link. so it's just the parallel resistor calc, doh!
Oh yes 75ohm for BNC not 50ohm.
OK, so with the component values on the pcb 470 & 78 (that must be it rather than 07R) it gives 67ohm - jeez why didn't they just use 470 & 90ohm for 75ohm impedance?

Another thread you may have seen is the DiyHiFi supply thread on modding the Musiland http://www.diyaudio.com/forums/diy-hifi-supply/154446-music-server-computer-transport-hop.html where he took out the transformer, changed the resistors & reported a much better sound. He has since put in new transformers which he says sound better again but hasn't spilled the beans on what make/model - guess he wants to sell them through his shop

I posted on Headfi, that I was choosing the 01US specifically because it gave access to I2S.

 

[danny_66]

Hello Wackyterbacky,
it's possible that you also can tap I2S out of the 01USD,
allthough the I2S pins are not connected it's very likely that you'll find I2S,
because the 01USD uses the same Xilinx IC and software driver as the 01US or 02US, just there's no pcm1793 connected to it.
Regards,
Danny

 

[jkeny]

Hello Wackyterbacky,
it's possible that you also can tap I2S out of the 01USD,
allthough the I2S pins are not connected it's very likely that you'll find I2S,
because the 01USD uses the same Xilinx IC and software driver as the 01US or 02US, just there's no pcm1793 connected to it.
Regards,
Danny

It's possible & worth checking out but I very much doubt it - one of the 4 pins used for I2S signals to the PCM1793 DAC is used to carry SPDIF in the 01USD - so this suggests to me that there is different firmware delivered to the Xilinx depending on the model. The model is probably identified by the ID held on the ROM chip. However it's done the software in the Xilinx seems to be different for different models.

 

[glt]

Danny,

It seems the driver tells the fpga what to do, so it depends. I got a MINI and even though the MINI only has audio out, the control panel lists MINI-speaker and MINI-spdif. So the function is there, but not used.

jkenny,

If you are going to measure the data rate of the i2s lines, post the raw data too. Also it would be interesting to see how that changes with different material: 44.1, 96, 192...

 

[jkeny]

If you are going to measure the data rate of the i2s lines, post the raw data too. Also it would be interesting to see how that changes with different material: 44.1, 96, 192...

How do you mean "post the raw data too"? I was intending to send a file of each sample rate that I have (44.1, 88.2 96, 176 & 192) through the I2S lines to the DAC & measure the I2S clock while doing so?

 

[glt]

Yeah, the measurement results (frequency, period...) not just the conclusion.

Also did you trace the spdif output of the US-01 to the fpga pin? which pin is it?. Maybe the MINI can also output spdif because there is a driver for it.

 

[jkeny]

Yeah, the measurement results (frequency, period...) not just the conclusion.

Also did you trace the spdif output of the US-01 to the fpga pin? which pin is it?. Maybe the MINI can also output spdif because there is a driver for it.

Oh, yeh that's what I intended!

I've attached a pic of the Mini 01US which shows the same pins being used for I2S to the DAC as is used in the Monitor 01US. The pins used are 3,4,5,6. On the Monitor 01USD (blue pcb) pin 5 is used for SPDIF, so I'm presuming the Mini 01USD is the same.