PDA

View Full Version : High Frequency Boost?



JohnSwenson
2006-06-13, 21:31
I'm new to the forums here so a little bit of brief background info.

I've been building my own DACs for quite a while and wanted to try a SqueezeBox direct into one of my DACs bypassing the S/PDIF interface. I recently purchased an SB3 and was quite impressed, no its analog out is not as good as my own DAC, but still quite good.

Anyway I added a cable connected to the I2S signals going to the DAC chip, disconnected the 11.2896 crystal and fed the output from a Tent XO. That Tent clock also feeds a reclocker in the DAC (Synchronous, its just some flip flops), this should give very low jitter without having to worry about any asynchronous issues.

First off I noticed that it was distorting on peaks, that turned out to be the data format, its left justified, not I2S. This was not a problem, I just reprogramed the CPLD that I use to do the I2S splitting (Its real nice having programmable logic in there!).

Then comes the issue at hand, my ears hurt when listening to this combination. This usually means there is some extra very high frequency noise or oscillations, but I looked at the output and its VERY clean. Tonight I was running the spectrun analyzer screensaver and noticed that there is high output in the very upper frequencies on almost all my songs. I've run these files through several programs that have high resolution analyzers and have never seen this high frequency emphasis.

Is this something the squeezebox or slimserver is doing? If it is can I turn it off? Is it some form of EQ that is sent to the DAC chip but not to the S/PDIF out? Is it just a fluke of the analyzer or is it really there?

I'd like to get this resolved because this configuration is by FAR the best sound I have ever heard from digital, it just hurts my ears!

I'm running 6.2.2 with firmware 48.

John S.

PS this is pretty amazing piece of gear, especially for the price!

ezkcdude
2006-06-14, 06:58
Suppose you can post some pics of your mod? If you don't mind, could you share some details of your DAC as well? I'm also wanting to build a DAC for myself, although it will be a first for me. I've been working on a schematic, if you care to take a look. As for your HF problem, haven't got a clue. I'm sure Sean or someone else will chime in about it.

http://www.cellandtissue.com/ezdac/schematic.pdf

fuzzyT
2006-06-14, 08:43
JohnSwenson wrote:

> Tonight I was running the spectrun analyzer screensaver and noticed
> that there is high output in the very upper frequencies on almost all
> my songs. I've run these files through several programs that have high
> resolution analyzers and have never seen this high frequency emphasis.

Could this be the result of the display values of these freqs are being
boosted to give a "better looking" spectrum analyzer screensaver
graphic, but not actually altered in the audio data?

You might want to track down the author of that code to discuss the
accuracy of that graphic.

--rt

ezkcdude
2006-06-14, 09:02
I think it's been said before that the spectrum viewer is not extremely accurate.

seanadams
2006-06-14, 10:07
We do not do any equalization on the audio. However, the spectrum analyzer does emphasize the higher bands for more visual impact. Although it is a real spectrum analyzer in the sense that it comes from a high resolution FFT of the stream, it is NOT intended to be to scale. If you search the forums there are some comments from Vidur describing exactly how it works.

mfieger
2006-06-14, 10:14
the explanation is here:

http://forums.slimdevices.com/showthread.php?t=13515

and here:

http://forums.slimdevices.com/showthread.php?t=14935

JohnSwenson
2006-06-14, 16:48
Thanks for the replies. I'm going to run some frequency response measurements tonight just to make sure.

The other option is that some EMI from the SB is getting into the analog part of my DAC, the SB IS sitting right on top of the DAC because there is a very short cable between them. I didn't use any special line drivers or terminated transmission lines so I had to keep the cable very short.

As to specifics of what is inside, I'll try and get a close up picture of the connection, I'll have to see how good the macro mode on my camera is.

On the DAC specifics, this is a colaberation with a high end manufacturer so I'm not at liberty to give out specific details at this point. (I've been doing this stuff for fun and someone wanted to PAY me to do it!)

Up to this point it has been a USB input using a 2706 with the I2S signals going into a CPLD which splits the stream to a pair of 1704s, no digital filter, which feed an I/V stage which is essentially a current mirror made out of MOSFETS running at high current and voltages. This is followed by a very low distortion MOSFET buffer stage that can drive just about anything (including efficient speakers). Somewhere on the Audio Asylum I posted a schematic of an early version of the output stage designed to work with a 1543.

John S.

JohnSwenson
2006-06-16, 13:08
I found the problem. I was using some DIY very good sounding but unshielded interconnects, they were picking up EMI from the SB3 which was doing strange things in the preamp.

I tried some old shielded ICs I had laying around and the problem went away but they didn't sound nearly as good as my DIY ones. Last night I tried some much better shieled ones (that cost twice as much as the SB3!) and its back to sounding fantastic but without the headache.

Sean, in another thread you mention having some test equipment that can measure low levels of jitter, it would be really nice to test out what I'm actually getting with this synchronous reclocking scheme. I'm in Fremont so I'm not too far away. Also if you'd be interested you're welcome to come over sometime and hear the difference between the raw squeezbox and using it as a front end to my way more expensive DAC.

John S.

Triode
2006-06-18, 04:11
Hey John,

Out of interest, what sort of psu are you using for the Tent XO? I've been playing with one based on the AD797 based psu discussed on Pink Fish http://www.pinkfishmedia.net/forum/showthread.php?t=18014
Which seems to work well, definately better than a simple 5V reg.

I'm still in the land of spdif, so am playing with Tent VCXOs to get a low jitter clock next to the dac. SB sounds great with this.

Adrian

JohnSwenson
2006-06-18, 23:44
Hi Adrian,

I'm using a modified version of what Guido has on his website. I start with the normal very good "raw" supply which comes from a bridge/CLC filter. This goes into a panasonic regulator which is pretty low noise as is, then into a voltage divider which is filter by a 220uf solid polymer cap, which then feeds a darlington emiter follower. The noise is quite low, I can't measure it with any of my test instruments, its below the equipments noise floor.

Yes going with a VCXO is probably the best thing you can do if you insist on having S/PDIF in the neighborhood. I've tried the VCXO approach as well as the buffer approach with a local Tent clock and I still wind up getting S/PDIF nasties leaking through, just having it on the same board seems to infect things. Thats why I went with ANYTHING that would let me get rid of S/PDIF.

I think I'm still getting some EMI from the SB3 although its much lower now than it was before. I think I'm going to build a shielded box to put it in with some LVDS tranceivers at both ends and send the data over shielded CAT5 or something equivalent.

John S.

parrydave
2006-06-20, 03:39
Hi John,

I read in your other post (Audiophiles forum) that the analogue output from the SB3 were greatly improved by using the Tent XO. Thats got me thinking about this as an analogue only mod. How simple a job would this be? And what would be your recommended route?

Presumably I'd have to put the Tent XO in an external box with suitable power supply. There's no way the Tent could live inside the SB3 and use existing power feeds? Or would this negate benefits?

I am already planning a few other analog mods, including upgraded internal supply caps etc.

Interesting stuff!

JohnSwenson
2006-06-21, 14:49
It MIGHT be possible to squeeze a Tent clock into the SB3, they are not very big, it depends on how much space there is between the board and case. Putting a board with a low noise regulator and a Tent clock into the SB3 would be a VERY good thing to do. I think some of the modders already do this. Remember there are two crystals, a 12.288 and a 11.2896. If you just want the improvement on ripped CDs its fine to just replace one and leave the other.

My experience is that doing just this without any other changes will improve the sound quite a bit. Of course upgrading the analog section will be even better.

John S.

tingtong5
2007-03-15, 05:16
I'm also considering slaving my DAC clock to the SB3 (removing the onboard 11,28 MHz crystal). However the clock output of the Tent XO in my DAC is 5V. Can I simply use a resistor voltage divider or would thsi screw up the low jitter output of the XO ?

Regards,

Ronald

Jitterbug
2007-03-15, 07:14
John mentions elsewhere that he uses low impedance resistors in a divider to do just this 50R & 25R. I'm in the process of doing the same thing - just waiting for my X03 to arrive

tingtong5
2007-03-15, 07:43
John mentions elsewhere that he uses low impedance resistors in a divider to do just this 50R & 25R. I'm in the process of doing the same thing - just waiting for my X03 to arrive
Very interesting :-)
So how will you get the master clock to the SB3 then. Using 75 Ohm coax? If so the resistor divider would terminate this coax in an incorrect way (not 75 Ohm) at the DAC.

And by the way , I am also waiting for my 11,28 MHz XO to arrive :P

Jitterbug
2007-03-15, 07:59
I use one of Guido's XO-DACs on my DAC so will just run the tent link to that, to keep the clocks in sync. The two clocks may well be overkill but I already had the X0-DAC before I thought of reclocking the SB.

tingtong5
2007-03-15, 08:16
I use one of Guido's XO-DACs on my DAC so will just run the tent link to that, to keep the clocks in sync. The two clocks may well be overkill but I already had the X0-DAC before I thought of reclocking the SB.
I see.. I guess this is indeed the easiest and best method of doing it! Still I like to try first using only one clock (saving me quite some money :P).

Let us know how it sounds ! :-)