Fluctuating USB jitter from CPU activity

[jkeny]

I'm sure Gordon has seen this post already - from John Westlake, a well respected E'ee & audio designer who is now measuring the Uptone USB Regen device - a device that improves the SQ through USB - see here http://www.pinkfishmedia.net/forum/show ... tcount=639

Basically he sees fluctuating USB eye pattern & jitter based on the PC's CPU activity

"I've been "getting to grips" with the USB interface and Regen - as the Regen will be a competitors product I'll have to be careful so as not to appear critical - I'll have to claim the 5th amendment if some question might force me into being too harsh about its design.

As a USB source I'm using a small Sony Pocket PC (VGN-UX380) and the USB connected via its docking station - running WinXP SP3.

Early days, so I'm just performing simple tests - but I can already see on a fast Analogue scope that the USB Packet it jitter visually effected by what the PC is processing - with eye pattern Jitter increasing from 320pS to 400ps simply by running a Youtube video.

The Screen is a mess and I'm just getting the feel of whats going on - while its easy to see the jitter modulation on the Analogue scope screen, I'm not sure how to post it here.... the screen is a mess....

You can see via the jitter modulation on the scope when the PC is performing other tasks - with big "apparently" random jumps in jitter, I suspect that these jumps are caused by other "hidden" operations being performed by the Windows OS.

I'm not in anyway endorsing the idea behind "optimized" media players (as the are trying to resolve an issue that should not effect the analogue domain in the first instance - IMO they are an indirect poor and dubious bodge fix) - but seeing the OS / CPU "Process" modulation comes as no surprise and opens a whole bag of hurt...

This observations are with the Upstream USB input to the Regen, it will be interesting to see if the USB hub is able to reformat the data packets and thus attenuate the OS / CPU process related jitter modulation."

[jkeny]

It will be interesting to keep an eye on this thread as JohnW is designing his own version of the Regen, he's calling it Detox - but first investigating & measuring the Regen, itself. As he states, he has to be careful about what he says regarding the Regen which is a competing product (or will be in competition with his Detox)

These measurements will/should include how such USB jitter affects the output of a USB audio device attached to it i.e a DAC's analogue output.

For anybody not following the thread I bring you the highlights here (hopefully not skewing the coverage to any particular slant in my editorialising?)

"You can clearly see discrete jumps in time as various windows are open and closed on the PC (and then some LF activity while I presume the OS is doing its house keeping tasks) - also the USB is very heavily modulated by the audio signal, as the frequency content of the audio signal changes, so does the Jitter eye pattern, Bass Frequency content is clearly viable - its a real mess....

There is a surprising amount of LF "Stuff" going on..."

What can I say - "interesting" but too soon to jump to any conclusions

[jkeny]

Just keeping you all abreast of posts of interest:

OK, as I expected the USB HUB does "repackage" the data and remove huge amount of the Host PC's USB Packet jitter - I can no longer see the vast amount of Process related jitter on the scope, its all much cleaner.

The edges of the waveforms still display a "Fuzzyness" which is indicative of jitter - but atleast its not an unholy mess of the direct PC connection.

I see the Regens biggest advantage is removing a lot of the "differences is sound quality" between software players and computers (With Bit Accurate Data).

I'll now have to use a more advance test step up to analyse the jitter as its now the jitter is at a more acceptable level - I was worried with my first tests (USB Direct to the PC) as it was such a mess.

[mick]

Does he say what USB Hub is used?
Does it matter?

[Sligolad]

Glad you have an inbuilt immunity to all the dung that gets slung over on PFM John and thanks for posting the relevant bits as I do not have the patience to tolerate the closed minds there.

Interesting that someone as apparently capable as John Westlake comes out with the following:

I'm not in anyway endorsing the idea behind "optimized" media players (as the are trying to resolve an issue that should not effect the analogue domain in the first instance - IMO they are an indirect poor and dubious bodge fix)

Surely he understands that optimized medial players are working towards unloading the hardware and processes that he rightly identifies as contributing to all the mess he is seeing on the USB interface!!

[jkeny]

Does he say what USB Hub is used?
Does it matter?

Good question. The particular device that's being investigated is called Regen from a company Uptone Audio. It's specifically designed to address issues with USB audio. It uses a specific USB hub chip but any such chip can be used. Can any USB hub be used - I don't know but my experiments have shown that the PS to the hub chip is critical.

[jkeny]

The first really interesting measurements of the Regen - a before & after shot of the USB signal.
How USB signals are viewed/measured is by overlaying multiple passes of the USB signals over one another - each signal should be a squarewave waveform. USB data is composed of two datalines D+ & D- in other words the positive signal is transmitted on D+ & it's exact reverse is transmitted on D-. This is called a differential signalling system i.e where there's a 1 on D+ there will be a -1 on D-. It makes for a more robust transmission of the overall signal because by design it should cancel any noise that is on both datalines.

Anyway when you measure both D+ & D- signals on an oscilloscope you see both square waves but one is 90 degrees out of synch with the other so it looks like an eye & is called an eye pattern. When you overlay multiple square waves, if there's any fluctuations in timing between each squarewave, the lines making up the squarewave plot will spread out & instead of a single well-defined line you will see a more blurred looking line.

So the eye pattern plots are what is seen in these diagrams. The without plot has two major characteristics - blurry lines but also a background which is not black - this shows that some major excursions of the waveform have happened to cause this bluish background but they are random so you don't see any individual lines. Both of these characteristics indicate timing issues - the blurry lines show small mis-timings (jitter) but the bluish background show large mistiming.

When the Regen is inserted, the background is black & the lines are less blurry

The Analogue scope measurements of the USB direct and ReGen USB revealed the truly "horrid" nature of the USB traffic from a PC with heavy "CPU Process" jitter modulation - and I've been trying to think how to graphically capture it so I can post it here on PFM.

So I dug up a Digital scope - rearranged the lab to make space for it... I think its the first time I've really used it - I hate them, but it proved useful to capture the data.

Without Regen:


Above is a colour graded eye pattern of the Direct PC USB connection to an XMOS based DAC (streaming 1KHz, 0dB FS) High speed USB.

Note the multitude of lines, but more importantly that the background is not black (Black means no data points fell in this area) - the background its dark blue because during the acquisition low frequency "Runt" waveforms where sampled in this area - these "Runt"waveforms occurred as the PC / operating system processed other applications / Operating system house keeping tasks etc. On the Analogue scope it was a horrific mess...

With Regen:


Now with the ReGen USB Data, notice not only the much cleaner waveform but far more importantly that there is very little "Runt" data - the background is Black. The USB Hub IC in the Regen has for the most part cleanly repackaged the data.

The colour grade data sample hit size is the same for both plots (20K Hits on the Red).

[jkeny]

Can someone extract the pics from the dropbox & post them here - I can't?

[sebna]

I can do it no problem but the links you are quoting are truncated and do not work. It is a common problem with quoting links. Can you send me via pm full link?

Cheers

[Diapason]

Put the links between

Code: Select all

[url] [/url]

tags and you'll be grand.