Would forum members chip in $$ for a USB analyzer

[Wavelength]

Guys,

 

There are 3 types of USB testing equipment available. The most basic of test is the protocol analyzer. These are available from TotalPhase, ITIC and others (I have those 2) $500-$1500. The second is a USB compliance analyzer. This is what companies use to determine the basics of USB. These go between a computer and a device. My Tektronix version of this cost about $18K. It's not really good for testing how well things are working. But you can set this up and see more information than a protocol analyzer can. You can go a step further with Tektronix and I have been working with them on this so that you can test real time between devices. Look at EYE patterns and stuff and calculate USB jitter (not to be confused with digital audio jitter). The big problem here is on HS/FS devices you really have to sit down and take note of who's talking. Remember it's a bi-directional bus system. Anyway this adds another $10K on top of the compliance testing for HS and don't ask category for USB3 stuff.

 

But really what's it going to tell you? You bought the wrong product? Why not spend the money wisely on something else?

 

Thanks,

Gordon

[Wavelength]

Ok guys a few things...

 

First Audio Jitter/Phase Noise is measured from 10Hz down. Most jitter analyzers you see like the microsemi is used for communications or other stuff.

 

Good high speed scopes with EYE pattern and differential probes will tell you a lot about USB, Ethernet and other protocols. Still need a 5x scope for that with a lot of storage.

 

Ethernet is a good protocol, not great for streaming but non the less it is isolated and capable at 100Mhz speeds. The controllers are good at 10/100 and some of them Microchip MZ, i.MX6, Renesas RZA1 etc.. are good choices as they have I2S capable outputs. A number of the ARM processors do not have good I2S interfaces. The processors listed above have very good I2S. The other ARM processors do not have good I2S and need to have FPGA's or other logic to lower the I2S jitter. The problem with more powerful processors such as say an A53 which has 10/100/1000 Ethernet is now you have processor noise to deal with. Isolation becomes and issue and expense. Ethernet does require a protocol and even tough there is AVB, in our case the Roon code is probably best. UPnP/DLNA is hit or miss in most setups.

 

I have a Thunderbolt license and designed a few things. I have the v1, v2 development system here chips and so forth. Thunderbolt is basically a wired extension of PCIe. Some companies in the pro industry are basically doing USB and Thunderbolt. They have a USB multiplexer on board and a USB Host controller hooked up the the Thunderbolt bridge and can select one or the other interface and then that goes to usually an ARM processor and or DSP (can be FPGA) to do all their stuff. Is this better? Well you could argue that Thunderbolt could be a more controlled setup and I bet it would be.

 

USB, USB3, USB-C we made USB easy to work with. We created platforms that allowed digital and analog designers easy access to build their products. Full Speed sure is ceiling is pretty low at 32/96 being the max stereo sample rate. High Speed sure is limited to sixteen channels @ 32/192. USB SS is pretty much wide open. The big problem here is isosynchronous is not error correcting. You could create a custom protocol that is error correcting but really with the error rates on a well constructed system being pretty low there really is no need.

 

~~~

Were at a point in audio were we are moving sideways. The DAC chips really are having enough problems at 384 going to 768 is just silly. So we are giving the customer base different connection options.

 

There will always be ways and products that will better the listening experience. Really I have boxes of shit all the way back to college. Who remembers the sharpie product that you would apply to the outside rim of a CD. Basically the green was suppose to inhibit stray laser bouncing that would effect the receiver and cause less errors. This is a niche market area of high end audio that will never go away. No company can be sure that they get everything perfect in computer audio. They can't because there are too many variables. Some do better than others, that's the way it is.

 

Buying and knowing how to use test equipment is something you should leave to Stereophile. John has spent decades working with Ap, Prism and other companies creating test suites. Buying inside a group to do testing would first require that you totally understand what it is you are testing.

 

Just think of what would happen if you guys did testing and the results were XYZ. Do you think that this would be argued over and over again and really what would that accomplish? Probably nothing...

 

Thanks,

Gordon

[Wavelength]

Ok,

Quote

Not widely integrated into motherboards so forget about SoCs (cpus with integrated I/O etc

 

Some of the top end SOC's have PCIe and SATA. I have worked with the ARMADA 7040 which has both.

 

Quote

Zynq can do Ethernet and I2S both in FPGA side and logic/drivers in ARM side -- so equal facility with USB and SGMii (SFP -> fiber) ... better phase error than eg Marvel Armada

 

The larger ARM chips like the Armada, double digit ARM like the A53 the 64 bit units and so forth are great for network bridges, NAS drives and stuff like that. They are not very good at Audio. Doing Ethernet in an FPGA is not something I would try. I bet if they are using an ARM processor on the board the Ethernet is probably going there.

 

Quote

Could you define "low" and "limited" in terms of audio? Most "high end" audio rigs are 2ch.

 

As I said low in terms of sample rate. Full Speed is limited to 32/96 stereo or 2ch. High Speed USB is limited to 16 channels at 32/192. Or 8 stereo channels at 32/192 or higher rates with less channels. 4 stereo channels at 32/384 or  two stereo channels (4 ports) at 32/786. But remember with higher sample rates, means more data being pushed through which does equate to more USB errors.

 

Thanks,

Gordon

[Wavelength]

1 hour ago, jabbr said:

Regarding 1,2 and 3 (scope , spectrum analyzer. vector analyzer) there is overlap among the 3 and each tool is used for different applications or may have specs that are tuned for a specific measurement.

 

(short answer : lots of stuff)

 

a high resolution scope can do eye-patterns for example. Maybe you'd do 4 gsps for 1-2 Ghz resolution but at 8 - 12 bits. : realtime info

 

spectrum analyzer: FFT plots so my old HP 3561a 125 microHz to 100 kHz at 640 micro Gz intervals -- thus also forms part if a really really close in phase noise measurement (using a really good phase comparitor) -- generally great for looking at harmonics : non realtime info

 

vector analyzer: eg HP 89441a -- all in one phase noise much faster & quicker & easier

 

newer packages are digital & have special software modules to automate measurements -- this is less than scratching the surface  

 

I have a mint 3561A yea with the bubble memory. It's a great piece for testing low noise analog stuff. I also have a Stanford SR760 which is a little better. The problem is getting the phase noise into these.

 

I have a Symmetricom phase noise analyzer that we use for testing crystals and oscillators. It's got a custom downloadable FPGA in it and a 10MHZ super low phase noise reference in it. We get good plots to 1Hz with this. Before we had that it was kind of hit and miss using a boat load of reference stuff.

 

Thanks,
Gordon

[Wavelength]

1 hour ago, jabbr said:

 

You are probably on one of the best positions to answer some of the questions many of us have that have prompted this thread -- there is a seemingly endless proliferation of gadgets and exotic cables designed to "dramatically " improve the sound of DACs connected by USB.

 

What then are the issues with USB that require correction? What are these devices (external USB widgets) doing to the DAC and how can the effects be measured? Can they be measured? Why would it be so hard for DAC manufacturers to isolate their DACs?

Jonathan,

 

Trying to fix a USB problem with one solution is never going to work. This is going to sound like a freaken marketing answer, but... You have basically 5 variables that can interact with each other:

1. Power
2. USB Cable Data integrity
3. USB Cable power
4. Device
5. Host

#5 Ok so everyone knows that not all USB host ports are not created equal. Some are off the main host controller, others off an internal hub with other BS on it. Second not all brands are the same. I always double back on my days designing motherboards and the fact that there were 2 accountants that shared my work space. Not to mention that these things are made for a price point, but also you have to look at the engineering. Were is the motherboard, were are the ports, are these cabled?

 

#4 How well is the downstream port designed? I designed a product for a company one time a rather large company who had their contract manufacturing make some prototypes. The company called me up &%&^%$%& me out. I said send over the schematics and a sample. Crap I am not sure if my reference design was even understood by the company. USB is a differential pair, they had the + signal go like 2" longer than the - leg. It would hardly work Full Speed.

 

3# Power can effect #2, but it can also be effected from outside noise. We have all seen the split cables and the thought that this fixes the effect on data, but some of these missed the other points.

 

2# Look the cable is a huge variable. Even I have learned a bunch in the last 5 years of my 14 years in USB. Well lets face it Full Speed is pretty easy to pull off. Most of these companies took their SPDIF cable and added VBUS & Ground. We even saw companies trying to sell 50-60 foot cables. They hadn't even tried these with DACS, let alone Asynchronous ones which require bi-directional data. On the Tektronix analyzer and a TDR you can test these pretty easily with known Host and Device.

 

1# Power... part of the problem with USB is the common ground. So in the basic sense the computer ground is shared with the USB and the other devices connected to that USB cable. Isolation will fix that problem. But what effects does it have on the rest of the system and were do you isolate? Do you isolate at the I2S or at the USB? The problem with the Silanna and Analog Devices isolators is that they do not reclock the data and therefore any USB Jitter on the data stream coming in will be increased on the other side. This is why putting a HUB on the device side is really necessary. That will reduce the data errors, but can't fix the data errors. At least with I2S isolation you can reclock the data with the known Master Clock which is synchronous with the data and hopefully low jitter so the dac sees exactly what you want it to see.

 

For me I run my power for my system (WWWS speakers, WA Corona Silver VT52 amps, WA Sine6 pre, WA Crimson or Cosecant, MacBook Pro Retina 15, Drobo Quad SSD raid Thunderbolt) as a spit system using CG T filter. One to the mains, one filter to the digital MBP+Drobo+isolated ethernet switch, the other side amps, dac, pre. This way the digital stuff has to go through two filters to infect the other side.

 

I plug my DACS through a Sonnett Ethernet/USB3 Thunderbolt dongle. It works better than the MBP ports. Cables USB AudioQuest Diamond or Curious.

 

Thanks,

Gordon