Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter

[iancanada]

Originally Posted by 2A3SET

9uV, is that ADP-151? A shunt reg (like TPA trident or salas) sounds better with XO.


2. Powering two oscillators at same time with 9uV RMS low noise high PSRR LDO and independent EMI suppression filters.

Hi 2A3SET,
Good point. I'd like to give a try. Do you konw who has the PCB of salas shunt regulator, 3.3 or 5V version? Comparing with linear amp, the XO is more like a RF circuit, it even sensitive to the pin inductance of a decoupling capacitor. Ian

[2A3SET]

Hi Ian,

Quanghao should have the 3.3V version,
Salas low voltage shunt regulator (V 1.0 ) for DAC and more

Or you can try the TPA trident without the CCS,
Trident

So far I have two setup using shunt for XO with good result,
1. swap ADP-151 with Salas shunt on audio-widget for Silab Si532 dual rate XO.
2. bypass "The flea" reg on AckoDAC with Audio-GD shunt (sounds even better than LifePO4 battery).

[iancanada]

Originally Posted by 2A3SET

Hi Ian,

Quanghao should have the 3.3V version,
Salas low voltage shunt regulator (V 1.0 ) for DAC and more

Or you can try the TPA trident without the CCS,
Trident

So far I have two setup using shunt for XO with good result,
1. swap ADP-151 with Salas shunt on audio-widget for Silab Si532 dual rate XO.
2. bypass "The flea" reg on AckoDAC with Audio-GD shunt (sounds even better than LifePO4 battery).

Hi 2A3SET,
Thank you for the links. I'll try to get one. Reagrds,Ian

[iancanada]

Originally Posted by ANADIGIT

Hi, Ian
Now I'm on the hard work of ESS9018 parallel 8DACs per channel and when it completed I shall send you asap :-)

Thanks
Anadigit

Thanks Anadigit, How's going? I'm looking forward to that!
Ian

[1audio]

Ian asked me how I measure jitter. It can get pretty involved but here are the basics.

The most useful is probably measuring at the main analog output. This process was documented pretty thoroughly by J Dunn, better than I can do, so I'll concentrate on the essentials of doing it.

You will need the following:
1) Test files (I have pulled some from the web and created some).
2) A way to play them (obvious)
3) A good analog to digital converter that won't limit the measurements. I use an ESI Juli@ card. The host PC can affect things and you will need to do some calibration to be sure of what you are seeing with the system.
4) An FFT capture package. There are free ones but I have not looked too deep. I use Praxis, primarily a speaker package, because it has very versatile and useful tools in it. Its pricey at close to $1,000 but I have got really good value from it.
5) lots of care in the setup since you will be looking a low level noises it helps to have a pretty clean setup.

I make a long file, 5-10 minutes, and do a hi resolution, up to 16,777,216 points and usually 5-6 averages, which can take almost 10 minutes to capture. I will play the test track at 44.1 for example and capture at 192, partly to make sure the two are not synchronous, which could hide things.
The files can be made with various tools, including Audacity. The test tone should be an integer sub-multiple of the sample rate. At 1X sample rates (44.1K and 48K) the tone will be 11,025 or 12,000. At the 2X or 4X rates they can be higher but then they are beyond the audio range and even some dacs that handle those rates will attenuate or even remove them.

I have attached three representative captures to illustrate what you will see. First is a Bryston BDA-1, which is really very good. You can see the textbook Jtest pattern. The second is the Apogee Minidac, which shows much higher jitter and especially the broadband jitter close to the "carrier". That is similar to the noise plot of a fair crystal oscillator. Finally is a real close in measurement of the Bryston. It has 10Hz span so each division is about 1 Hz. You can see two symmetrical blips at 2.5 Hz from the carrier. At this resolution I can not be sure if they are from the dac or the measurement system. The last two plots are an Auraliti PK100 analog output which looks a little different. The first plot is the same 8 KHz span as stereophile uses for these tests, the second is a zoomed version showing the close in phase noise, what my homework has been for the last year.

I'll be happy to post the test files if they will fit here.

[iancanada]

Originally Posted by 1audio

Ian asked me how I measure jitter. It can get pretty involved but here are the basics.

The most useful is probably measuring at the main analog I'll be happy to post the test files if they will fit here.

Thanks Demian,
You did great job! Very smart idea, I like it. I'm reading your post right now. It seems I need set up same system to follow up your test. Ian

[iancanada]

Originally Posted by 1audio

3) A good analog to digital converter that won't limit the measurements. I use an ESI Juli@ card. The host PC can affect things and you will need to do some calibration to be sure of what you are seeing with the system.

Hi Demian,
I have a couple of questions about the sound card:

1. How do you deal with the A/D sampling jitter from the clock of the sound card?

2. Usually the EMI noise from the PC switching PS is terrible. Is that the USB sound card with external PS any better than the PCI card doing this kind of measurment job?

Thanks, Ian

[1audio]

What you get the the additive composite of the source and the sound card. I lucked out using the Juli card since it has a pretty low residual jitter. I have tried other cards. The USB capture cards I have tried are poor for some reason. They are the EMU 0202 and the Maudio transit. Lots of noise and capture issues that may be resolvable but more work with no certain end. I have also tried an RME card that seems to have its own plan for capture and prevented me from getting good data even on a digital loopback. Again, possibly resolvable, but I don't have the time to wade into it.

For jitter measurements the important part is to look for the symmetrical sidebands. Everything else is simply noise in the pickup. Calibrating the system is important so you can separate the signals. You need several clean sources since none will be as clean as you would like. An analog oscillator may have a pretty low noise floor but a higher residual phase noise and less frequency stability. A crystal controlled source may have other junk but a pretty clean tone. My Boonton 1121 has a crystal stabilized analog oscillator, which should be really good, but shows the PLL effects of wandering pretty easily on these tests. No easy answers. The system and its grounding are also a big issues but careful setup and possibly isolation in the right places will reduce the issue a lot. This is more like FCC testing in that you will need to know the sources of what you see and can ignore those you know to not be from your DUT.

The host PC has changed since the first captures, with a new motherboard (MSI AMD AT MB) and power supply.

I just did a "quick" test with a Minirator crystal controlled battery powered generator and a Krohn Hite KH4400 low distortion oscillator (ac powered) to show the basic cal process. neither of these are really low enough as a source to be without fault. For that I use a Wenzel 10 MHZ oscillator divided down to 10 KHz. That division yields a really low noise floor, much lower than nay ADC or DAC but its a PITA to setup. I need to build it into a standard test fixture when time permits.

The Minirator in the first two captures was done with the 16M point fft. The KH4400 in the following two could not be measured that way. Its frequency stability is not good enough so the center frequency moves bins and is averaged out. The measurement is a 500K point FFT.

If time permits I will set up the original cal stuff but I'm scrambling to get ready for Burning Amp tomorrow.

[iancanada]

Originally Posted by 1audio

If time permits I will set up the original cal stuff but I'm scrambling to get ready for Burning Amp tomorrow.

I'm gonna try to get one of that sound card. Is that Juli@, right? Everything connected to jitter is not easy.

Enjoy the Burning Amp, have a good time.
Ian

[HpW]

Hi Demian,

may look ant my home page some RME BabyFace Loopback measurement I did lately... a big improvement against the HDSP

What really counts is the used ADC and the clock part.

Cheers

Hp