Am I unreasonable to complain about USB 3 performance?

[Kenwstr]

Just a thought:

I've had a lot of reliability issues with USB3 as well. While there can be processing issues. It seem to me to often relate to the electrical connection of the USB3 plugs themselves. A quick spray with Deoxit F5 often does the trick. I suspect there may be small particles getting into the plugs making for poor and intermittent electrical contact. The contact area is smaller with USB3 after all. The F5 product is a quality contact cleaner lub and conditioner that improves and protects electrical contact and performance of mechanical components like pots, faders, trimmers, switches and plugs etc. It's expensive but with over 30 years messing about with audio and electronics, it's safest and most reliable one I have found. It's amazing how many computing hardware issues just come down to electrical continuity, disappearing files on drives etc interrupted registry/index update for example. I've even had missing files reappear after restoring connections in this way. As mentioned by others, alternate high capacity connection type wouldn't hurt.

 

[shjacks45]

Summary:: I have bought a Cmos camera for astronomy. It is sold as USB3 but it will not talk to my computer reliably. Astro 'users' just say it's a known problem and are prepared to work around but I don't think that's good enough.

USB 3 was heralded as the next best thing to sliced bread and, imo, has been marketed in a not - too - honest way. It's not always as useful as the bare numbers suggest. I know the same could be said about 1GB Ethernet but Ethernet can handle all sorts of rag tag and bobtail traffic streams and STILL WORK. I guess the Engineering of Ethernet is a lot more mature and the Internet just has to work.

Astro cameras are often required to transfer image data pretty fast and the movies that are shot to get good planetary images need a good data rate. USB3 claims 5GB/s but that figure is only when things are 'just right'. I am well aware of that and my expectations are realistic.I bought a low-end ZWO ASI290 MC camera with a 3MB sensor. Sometimes I get half decent images. With the short, supplied lead I have had 100 frames per second at times but mostly the capture programs hang up. I can accept speed restrictions but is there any excuse for hang-ups? If I have a box that produces data for transmission to another box, I would expect the comms channel to know about the channel capacity and to regulate the data transfer rate to fit. Using the camera (and others) with a USB2 lead, the frame rate goes up and down, according to cable lengths and other traffic. I can't think there's any excuse for the USB3 shortcomings.Afaics, USB3 is very much a bolt-on system. The connectors have two data wires which give USB2 compatibility and then, in another part of the connector, two twisted pairs provide fast duplex comms. I know it's busting its little braces to achieve 5GB/s but when a system fails, it really should fail gracefully

I have a question about the "USB Traffic" control that some capture progs have. ZWO have told me to fiddle with that control and it does have some effect. It could allow some control of priorities amongst several data streams - potentially very useful but that still doesn't allow it to work unless things are just right (and I don't know what that actually means). The poor computer has frozen completely on occasions and works as soon as I pull out the camera cable. Crazy.

It's like trying to get a racehorse to pull a milk cart. And it shouldn't be like that. Are the chips just not up to requirements~?

Actually you are right in assuming that vendors claiming USB 3.0 compatibility under the umbrella of USB 2.0 subset of USB 3.0 operation. Or their device is USB 3.0 if it uses more than USB 2.0 0.5 Amp power limit. When you see wifi, ethernet, fiber, or usb: claims are "Up to" 6 Gbits per second is actually a guarantee that that is max speed possible. If you know a little physics, you will recognize the term "bit jitter". It is the fact in the physical transmission of data, that higher frequencies (rapid data changes) travel more slowly along a cable than slower frequencies. All fast data transmission is serial since the bit jitter issue affect sync of parallel data transmission worse. Any deviation from straight or proximity to other conductors slow transmission speed and muddle those data bits. Equivalent speed copper ethernet has similar distance limitations. Optical fiber allows long distance transmission of high speed data but alas the maker of your sensor array didn't consider that. USB-over-Wifi does bypass USB cable length issues. USB 1.1 max was 18 ft., USB 2.0 high speed 6 ft, USB High Speed (480 Mbits/sec) max length 18 inches. For computers built with internal USB 3.0 DVD or hard drive, data rate (burst data; physical devices don't read data that fast) on USB 3 is faster than SATA or PCIe.
CRITICAL to speed is USB 2 uses polling (cpu wastes its time checking port) vs USB 3 interrupt mode (optional) which automates data transfer signaling when data is freshly available. The faster mode is not a given. Pardon my predjudice, but the first Rasperry Pi had a "synthetic"(software) USB port and even the Pi 3 had underwhelming tranfer rate with external devices, HOWEVER addon USB/network hats for the Pi (with real usb controller chips) will boost the Pi's USB speed considerably.

 

[harborsparrow]

I didn't read all replies here so sorry if duplicating. But I would not diagnose any performance issue, including IO devices, without asking about RAM and cache capacity of the system, and whether SSD or HDD. This is because of caching and swapping to disk. Sometimes simply beefing up RAM can make a difference. Surprisingly often in my experience.

 

[shjacks45]

I didn't read all replies here so sorry if duplicating. But I would not diagnose any performance issue, including IO devices, without asking about RAM and cache capacity of the system, and whether SSD or HDD. This is because of caching and swapping to disk. Sometimes simply beefing up RAM can make a difference. Surprisingly often in my experience.

You understand that SATA, m.2, USB, PCIe, are all peak burst transfer rate and the speed of the speed of the media is typically orders of magnitude slower. Like gamers getting "primo" video card that's using 4x of their 16x PCIe card slot. Speed? Depends on storage device and your controllers. Generally SATA gets request from program, moves the data, then tell cpu its done. USB 2.0 (allowable in usb3) requires your CPU to check for every data byte (polling) which is slower than the transfer rate even on a fast computer doing nothing else. True USB 3 allows true device to memory data movement with minimal cpu intervention. The m.2 interface is a modified PCIe 1x channel. The PCIe 1x interface found on USB3, ESATA, Ethernet, RAID controller, is slower than the interface standard.
Note on SSDs. How come the SSD makers say the mean time before failure is 150 years, but the memory chip makers say you data only good for 10 years max? However SSDs, the flash chips, have a relatively low maximum writes compared to magnetic and the flash controller uses wear leveling to make that less catastrophic. Flash memory has transient bad bits (not an issue with stored images as artifacts aren't noticed) and requires error correction bits to prevent data errors. A digression to point out that using SSD as swap drive is a bad idea. Think Microsoft "Ready Boost".

 

[russ_watters]

Astro cameras are often required to transfer image data pretty fast and the movies that are shot to get good planetary images need a good data rate. USB3 claims 5GB/s but that figure is only when things are 'just right'. I am well aware of that and my expectations are realistic.

I bought a low-end ZWO ASI290 MC camera with a 3MB sensor. Sometimes I get half decent images. With the short, supplied lead I have had 100 frames per second at times but mostly the capture programs hang up. I can accept speed restrictions but is there any excuse for hang-ups? If I have a box that produces data for transmission to another box, I would expect the comms channel to know about the channel capacity and to regulate the data transfer rate to fit. Using the camera (and others) with a USB2 lead, the frame rate goes up and down, according to cable lengths and other traffic. I can't think there's any excuse for the USB3 shortcomings.

I am not sure what your complain is. Is it with USB3 as a protocol? Is it that USB3 is slower than advertised? Is it that your (your words) "low end" camera doesn't implement it well?

[sigh] I have a ZWO ASI1600 that I bough largely for the 2017 solar eclipse. I don't advertise it, but I didn't get much out of it -- almost all of the finished photos I got were essentially from my back-up rig. In addition to being hard to use (weird settings, thin documentation, no native software), it also had an un-documented bug that ruined a significant fraction of the images, that I didn't get rectified until months later when tech support emailed me a firmware update that fixed it.

That saga's not over yet. A year and a half ago the mount for my scope loosened I guess due to the temperature drop, and pile-drove the camera into my deck. The USB cable was shredded and shoved the USB port up into the camera body. I waited more than a year to try and fix it and it turns out that after removing a loose body and straightening-out the USB port (and replacing the *USB3 cable), it seems to work. But I haven't tried yet to see if it is really usable...

I have a question about the "USB Traffic" control that some capture progs have. ZWO have told me to fiddle with that control and it does have some effect. It could allow some control of priorities amongst several data streams - potentially very useful but that still doesn't allow it to work unless things are just right (and I don't know what that actually means). The poor computer has frozen completely on occasions and works as soon as I pull out the camera cable. Crazy.

So I just bought a new camera for planetary imaging, a QHY290. It's another Chinese manufacturer, with a lot of the same shortcomings. But if it is good enough for Chris Go, it should be good enough for me. Nope. The first night with it was yesterday; balky connections, and when it did work, I was getting like 4 fps. It's advertised as capable of 200+fps at 1920x1080.

Whelp, after a little troubleshooting I found that the unreasonably short, included USB3 cable was the culprit. Fortunately I had *a spare USB3 cable lying-around to try. Now it gets a spinning-hard drive-limited 58fps, transferring data at 235 MB/sec.

Anyway, USB traffic: Near as I can tell it is some sort of throttling function, reducing the framerate and transfer rate to avoid overloading the bus.
https://astrojolo.com/gears/cmos-setting-challenge/

The manufacturers seem to be comparing notes, because it is a rare feature that exists on both my ZWO and QHY cameras -- but with opposite settings.

 

[sophiecentaur]

The manufacturers seem to be comparing notes, because it is a rare feature that exists on both my ZWO and QHY cameras -- but with opposite settings.

Commiserations on all that. The cameras used on telescopes in major observatories just have to be better than that but I guess that involves ten times the cost as least. Zwo and the like are getting away with the sort of behaviour that DSLR manufacturers wouldn't. When you think of what an amateur level DSLR has inside it (mechanical and optical), the so called high end amateur astro cameras must be only 'average' value.
I was staggered to learn that the (usb2) ZWO ASI120MC was actknowledged by ZWO to 'give problems' with Apple systems. What possible excuse could there be for that if they are sticking to the spec rules in their cameras? It's only DATA fgs.

 

[shjacks45]

A Mac is theoretically BSD so it should catch hardware interrupts. I thought you might be a science type with a Raspberry Pi. Note that camera connector on the motherboard is not USB but I believe DCI, for direct sensor connect. High speed USB 2.1 is 480 Mega bits per second, assuming a minimal overhead and 10 bits per byte gives 48 Mega Bytes per second. If the camera is truecolor (4 bytes per pixel color info), 20 Megapixels x 4 bytes is 80 Megabytes. 80 Megabytes x 30 frames per second is 2.4 Giga Bytes per second. Serialized (x10) is 24 Giga bits pers second compares to SATA, USB3, PCIe, et al MAXIMUM 6 Gigabits/sec.
Q.E.D.