What is the minimum latency I can expect from a High Speed USB control transfer? I've written a little test program using libusb on OSX and WinUsb directly on Windows 10, and in both cases I get around 4 ms from submitting the transfer to receiving the result.
This seems high. (And I could have sworn I previously did a test on OSX and got sub-1 ms latency, but maybe I did something wrong...)
Does anyone know what the limitations are here? Surely there is a way to get at least close to 1 ms?
Ugh I was being stupid. I had an inadvertent printf() over serial in my control transfer handler. With it removed I get a latency of around 120-160 µs on OSX via libusb and 240-330 µs on Windows 10 via WinUsb (in Parallels).
Related
This is probably a long shot question, but I try it anyway.
I'm developing hardware using PIC Microcontrollers (MicroChip). Communication is done through a FS USB 2.0 link.
I connect the microcontrollers to a Windows 10 Home edition, version 21H1, build 19043.1826. The processor is an AMD Ryzen 5 3600 6-Core Processor.
First I used the PIC18F45K50, for which everything worked fine from day one. But due to the shortages on the market, I now am experimenting with PIC18F47J53. Both microcontrollers are working fine, as I can (for example) control a MAX7219 controlled display (3 x 7-segment) and also control a bunch of LED's using an STP08CP05TTR. Clock timings seem also ok - I measured it with an oscilloscope.
These 2 microcontrollers are pretty much the same, at least for the core functionality such as USB. The differences that are relevant for the issue I'm reporting here are:
PIC18F45K50 uses internal clock of 8MHz, and has on board correction logic to keep clock synced for HS USB - this is a 5V processor
PIC18F47J53 uses a XTAL of 16MHz, all should be within the USB 2.0 specs - this is a 3.3V processor
I'm using the MPLab X IDE v5.45 with the MCC (MPLab Code Configurator) in which I setup the System Module (to set the correct clock frequencies including the 48MHz for USB) and where I configure the USB.
In both microcontrollers, the setup of the USB is exactly the same. I even checked the 4 files that are automatically generated by MCC, and except for the descriptors (I used different names), all is exactly the same.
When I connect the USB to my PC (same port), then the PIC18F45K50 works perfect. But the PIC18F47J53 gives error code 10.
This does not happen every time. For example, if I try 10 times (connect/disconnect the cable), then I had it 7 times. 1 time the device even didn't appear, and 2 other times I read "The device is working properly.". Although, in the latter case, my software that communicates with my controller isn't working, so there is still something wrong.
Based on the above, the first I would think of is some hardware issue. Although, the strange thing is that things like vendor ID (0x4D8), Product ID (0xA), BCD Device Release (0x100), Serial Number (12345678), etc... seem always to be read out correctly. If there would be a hardware problem, shouldn't I have more random issues with this as well? Or is this data read out in a slower mode than Full Speed (because that could of course explain this)?
Below are screenshots via "Device Manager / Ports (COM & LPT) / my serial device", then selecting the property in the Details.
If I compare the properties from the working microcontroller (PIC18F45K50) with the not working one (PIC18F47J53), it looks like all are exactly the same.
I also tried to compare the D- (CH1) and D+ (CH2) signals between the 2 microcontrollers with my oscilloscope. My USB knowledge is not detailed enough to interpret the signals, but what I can tell is that both look exactly the same to me, both timing wise and voltage level wise. Be aware that the CH2 signal on the PIC18F47J53 (D+), the second screenshot, is clipping in the picture below, but I measured it later and it shows the same voltage level as for the PIC18F45K50.
Does anybody here a single clue where I should look at in the first place? The good news is that I have a working and not working version, so I can start debugging step by step and compare. But some hints as where to start would be appreciated.
EDIT 24JUL2022
I did the measurement with my oscilloscope again. Now I soldered 2 wires to the USB port to be able to easily attach my probes. This time, both D- and D+ signals have a Vpp of about 3.3V. I put some cursors which also shows a pulse-width of about 84ns, which correlates with the USB HS frequency of 12MHz (should be 83.33ns).
I found the issue. The Vusb on my PIC18F47J53 had a bad (or was even not) connected. I gave it another touch of my soldering iron, and bingo! Now the "error 10" has disappeared completely, and each time I connect/disconnect it gives "This device is working properly.", and error 10 never appears. I now also see a continues signal on my oscilloscope - not one that is disappearing after a while. And I could send/receive already some commands.
I work with an embedded device that has a USB host port. I would like to connect an iPhone to it and communicate via USB. I have done development on this and ported the functionality to connect to usbmux on the iPhone and have successful communication, however there is another problem.
All development was done with the iPhone connected to a powered USB hub that was connected to my device, as soon as I connected it directly, after enumeration it starts to drain the battery of my embedded device and causes a tension (voltage) drop that causes my device to turn off.
I know that after enumeration usb devices can draw up to 500 mA from the usb port, but I was wondering if there was a way to limit that to 100 mA (while still having the iPhone registered).
I found various questions regarding controlling voltage on the data pins or vcc from the usb port and I understand that's not possible, I'm looking for a software solution (although hardware solutions are welcome).
tl;dr: Is there a way to supply the iPhone with less than 500 mA after enumeration? Could I do this in software? Or do I need a hardware solution? I don't want to turn the port on/off, just limit the power draw of the iPhone.
NOTE: I am using Windows CE 6.0, if it is something that can only be done by modifying the drivers, or having direct access, there is no problem.
P.S. also, if there is a way to do this in *nix (or some other open source OS) that I could look at the source code and port it to Windows CE please let me know.
When a device shares its available configurations (see USB chapter 9), it specifies how much power it requires for each configuration. The host should look at all the available configurations and choose which one it wants.
In practice, however, these things don't work so smoothly.
The last time I looked at this, Windows always chose the first configuration. MacOS always chose the lowest power configuration (or highest, I can't remember). I never looked at WinCE or Linux.
If you're writing/modifying the driver, you can set your own rules for which configuration to choose, including looking for one that's 'self powered'. The iPhone, however, might only have one descriptor that always requests 500mA, bus powered. If so, then you're pretty much screwed since there's no way to let the iPhone know it's not OK to draw power.
That being said, I believe all the iPhone accessories are actually USB host (as opposed to USB device), and given that they don't always supply power, the iPhone must be capable of enumerating self powered.
I like the answer by Russ Schultz but I want to add another one:
No.
The descriptor of the peripheral device, iPhone in this case contains bMaxPower. If you enumerate this device, you also accept the power demand. It is not possible to only supply less, lets say 300 mA, if you already enumerated the device with the 500 mA desriptor. If this is what you wanted.
If the device provides multiple configurations, you are as mentioned by Russ free to write a driver which selects the configuration with less power. Hopefully, the device will then only consume the granted power.
Many peripheral devices just don't care. Most devices only provide one configuration with 500 mA. And there are a lot of devices which just consume more than they say ...
I've been given the task of getting ADC samples onto an embedded linux computer at the highest rate I can (up to about 300kSPS). I am playing with several different platforms (odroid, edison) but easrly on I realized the limitations of using the build in ADCs from within linux and timing (I am relativly new to this).
Right now I am reliably getting 150kSPS using a teensy 3.2 with a very basic swapping buffer, a PDB, and the USB connection. USB writes take 2.5usec no matter my buffer size so any faster and the ADC read interrupt collides with the USB and I get nothing.
My question is: Would using an external ADC chip enable faster speeds? I see chips on Digikey and Mouser advertising 600kSPS and higher with SPI and even parallel outputs... but I fell like the bottleneck is the teensy with USB writes. Even if it could (and I am sure it could) read values 600k times a second how do you get it onto the computer without falling behind?
also, it is for long term collection so I can't just store everything and write it once the collection is over. The edison has a built in microcontroller, but no SPI implemented yet.
Edit:
To clarify, my question is weather there is any way to get large amounts of data very fast into my embedded linux device programmatically or is there some layer between a fast SPI device and the comptuer that I don't know about. So far my mentors have suggested I 1) learn to write a device driver for the SPI device or 2) recompile an image with RT_PREEMPT.
I am a physicist, and I had a revelation a few weeks ago about how I might be able to use my personal computer to get much finer control over laboratory experiments than is typically the case. Before I ran off to try this out though, I wanted to check the feasibility with people who have more expertise than myself in such matters.
The idea is to use the i/o ports---VGA, ethernet, speaker jacks, etc.---on the computer to talk directly to the sensors and actuators in the experimental setup. E.g. cut open one side of an ethernet cable (with the other end attached to the computer) and send each line to a different device. I knew a postdoc who did something very similar using a BeagleBone. He wrote some assembly code that let him sync everything with the internal clock and used the GPIO pins to effectively give him a hybrid signal generator/scope that was completely programmable. It seems like the same thing should be possible with a laptop, and this would have the additional benefit that you can do data analysis from the same device.
The main potential difficulty that I foresee is that the hardware on a BeagleBone is designed with this sort of i/o in mind, whereas I expect the hardware on a laptop will probably be harder to control directly. I know for example (from some preliminary investigation, http://ask.metafilter.com/125812/Simple-USB-control-how-to-blink-an-LED-via-code) that USB ports will be difficult to access this way, and VGA is (according to VGA 15 pin port data read and write using Matlab) impossible. I haven't found anything about using other ports like ethernet or speaker jacks, though.
So the main question is: will this idea be feasible (without investing many months for each new variation of the hardware), and if so what type of i/o (ethernet, speaker jacks, etc.) is likely to be the best bet?
Auxiliary questions are:
Where can I find material to learn how I might go about executing this plan? I'm not even sure what keywords to plug in on Google.
Will the ease with which I can do this depend strongly on operating system or hardware brand?
The only cable I can think of for a pc that can get close to this would be a parallel printer cable which is pretty much gone away. It's a 25 wire cable that data is spread across so that it can send more data at the same time. I'm just not sure if you can target a specific line or if it's more of a left to right fill as data is sent.
To use one on a laptop today would definitely be difficult. You won't find any laptops with parallel ports. There are usb to parallel cables and serial to parallel cables but I would guess that the only control you would have it to the usb or serial interface and not the parallel.
As for Ethernet, you have 4 twisted pair with only 2 pair in use and 2 pair that are extra.
There's some hardware that available called Zwave that you might want to look into. Zwave will allow you to build a network of devices that communicate in a mesh. I'm not sure what kind of response time you need.
I actually just thought of something that might be a good solution. Check out security equipment. There's a lot of equipment available for pc's that monitor doors, windows, sensors, etc. That industry might what your looking for.
I think the easiest way would be to use the USB port as a Human Interface Device (HID) and using a custom built PIC program and a PIC that includes the USB functionality to encode the data to be sent to the computer and in that way be able to program it independently from the OS due to the fact that all mayor OS have the HID USB functionality.
Anyways if you used your MIC/VGA/HDMI whatever other port you still need a device to encode the data or transmit it, and another program inside the computer to decode that data being sent.
And remember that different hardware has different software (drivers) that might decode the raw data in other odd ways rendering your IO hardware dependent.
Hope this helps, but thats why the USB was invented in the first place to make it hardware and os independent.
In the world of embedded software (firmware) it is fairly common to observe the order of events, take timings and optimise a program by getting it to waggle PIO lines and capturing their behavior on an oscilloscope.
In days gone by it was possible to toggle pins on the serial and parallel ports to achieve much the same thing on PC-based software. This made it possible to capture host PC-based software events and firmware events on the same trace and examine host software/firmware interactions.
Now, my new laptop ... no serial or parallel ports! This is increasingly the case.
So, does anyone have any suggestions as to go about emitting accurate timing signals off a "modern" PC? It strikes me that we don't have any immediately programmable, lag-free output pins left.
The solution needs to run off a laptop, so using add-on cards that only plug into desktops are not permitted.
Laptop with a docking station and old-skool parallel port. An alternate is to use a "smart" box connected via USB that handles the timing for you and simply reports the results over USB.
You may want to look into some of the USB logic analyzers like the Saleae Logic kit.
You can get USB dongles that create serial and parallel ports. Make sure you do your homework though, as you want to be sure that you can get all the Pins and all their data, some of the cheaper units don't do everything you need.
I've got a PCMCIA parallel port from Quatech - model SPP-100. It is a "real" parallel port - not a USB-Parallel port one. It is "real" enough to allow me to use a parallel port dongle with Windows 7 x64 on my laptop which doesn't have a parallel port.
How about using the audio line-out port? That should at least have consistent latency since audio applications care about that. That's the only modern computer output I can think of that isn't packet-based or dedicated to other purposes. It only has two channels (though external USB audio devices could expand that, and hopefully also have consistent latency through whatever the protocol does), and there might be a DC filter, but you could at least produce reliably timed pulses.
USB logic analyzers exist, some are even chip. But keep in minds #xtofl's comment about timing.
sump.org
The Bus Pirate