Are these USB series resistors important?

[zapta]

This diagram is from the datasheet of the LPC11U35  (USB full speed)



I tried it with and without the series 33 ohm resistors and the signals looks just the same and so is the functionality. How important is it to have those series resistors?

I am not interested in formal USB compliance, just want to thing to work reliably.

[paulie]

In stripping down your Arm-Mini design I came to pretty much the same conclusion: excess baggage. In many cases I found 90% of circuitry superfluous. A lot of so called engineering these days would best be described as religious ritual. For example the impedance of those resistors is inconsequential compared to the relative drive characteristics of the MCU. Almost exactly the same situation with some AVR designs like USBASP. For super high speed USB with special low impedance dedicated devices it's a different story.

There are some who argue the zeners are required even for 3.3v circuits thereby demonstrating almost complete lack of understanding of how these really work.

[langwadt]

This diagram is from the datasheet of the LPC11U35  (USB full speed)



I tried it with and without the series 33 ohm resistors and the signals looks just the same and so is the functionality. How important is it to have those series resistors?

I am not interested in formal USB compliance, just want to thing to work reliably.

together with the resistance of the output drivers it forms a series termination of the ~100R differential twisted pair
probably not terrible critical, but is it worth the risk to save two resistors?

[DanielS]

I am not interested in formal USB compliance, just want to thing to work reliably.

If you want stuff to work reliably, you usually have to stick with reference designs. If the reference design includes 33 ohms resistors, it most likely means the chip's IO driver requires them for proper cable impedance termination. Without these resistors, you might be compromising signal integrity on the USB cable and this may translate into fussy behavior based on specific cable build quality and length - it might work fine using a cheap 2' cable but be unusable on a quality 6' cable.

If you want to test your luck, you could put the resistors on your PCB design and then use a solder bridge to see if you can get away without one. Or you could put both of the SMD resistors' pads on the same net to have a trace between pads and cut that trace if you find out you really need the resistors.

[free_electron]

There are multiple reasons for those two resistors

1) it is bloody hard to meet the impedance on the board in terms of track width and clearance to meet the usb spec. putting these resistor directly at the pin allows for an easier way to do the pcb layout. if your stackup allows you to easily create the 90 ohms differential you don't need them. fact is most stackups with standard pre-preg material on a 4 or 6 layer board are in the 60 to 75 ohm...
tack on 22 or 33 ohm and you are there.

2) these resistor limit current for an esd strike. integrating these resistors is a no-go becasue of 1 above and because they cannot handle a lot of power nor standoff voltage in integrated form.

go read this if you want to know why and how to do it properly ( as opposed to believing in ritual )
 http://www.usb.org/developers/docs/hs_usb_pdg_r1_0.pdf

[free_electron]

I am not interested in formal USB compliance, just want to thing to work reliably.

If you want stuff to work reliably, you usually have to stick with reference designs. If the reference design includes 33 ohms resistors, it most likely means the chip's IO driver requires them for proper cable impedance termination. Without these resistors, you might be compromising signal integrity on the USB cable and this may translate into fussy behavior based on specific cable build quality and length - it might work fine using a cheap 2' cable but be unusable on a quality 6' cable.

If you want to test your luck, you could put the resistors on your PCB design and then use a solder bridge to see if you can get away without one. Or you could put both of the SMD resistors' pads on the same net to have a trace between pads and cut that trace if you find out you really need the resistors.

the problem is that this 'test' involves you pulling eye-patterns... most hobbyists don't have  a scope fast enough to pull that off nor do they have access to the test criteria.

[paulie]

If you want to test your luck, you could put the resistors on your PCB design and then use a solder bridge to see if you can get away without

That's exactly what I did on one of the boards zapta sent me. Then ran it over the coals with a variety of PCs and cables. For one particularly tricky ASUS and my longest and shortest unshielded cables it worked fine so the resistors were left out of the remaining builds.

These "rituals" do serve a purpose in military, medical, etc applications but in many cases more for insurance and liability reasons than religious. This is HOBBY world guys, much as you might like to pretend otherwise.

[langwadt]

There are multiple reasons for those two resistors

1) it is bloody hard to meet the impedance on the board in terms of track width and clearance to meet the usb spec. putting these resistor directly at the pin allows for an easier way to do the pcb layout. if your stackup allows you to easily create the 90 ohms differential you don't need them. fact is most stackups with standard pre-preg material on a 4 or 6 layer board are in the 60 to 75 ohm...
tack on 22 or 33 ohm and you are there.

sticking a series resistor at the end of a transmission line does not change the impedance

but 2x33R + 2x impedance of an output  makes a roughly  ~100R termination

[zapta]

together with the resistance of the output drivers it forms a series termination of the ~100R differential twisted pair
probably not terrible critical, but is it worth the risk to save two resistors?

It's an exercise in keeping things simple.

[miguelvp]

Find a way to induce noise on the USB cable, without those resistors I will think your differential information would be too degraded in the other end depending on the distance and noise picked up by the cable.

For example, if the cable runs in parallel near an AC power cable, without the proper termination your communication distance might degrade.

At least that's my understanding but I'm a noob at this.

[tggzzz]

In stripping down your Arm-Mini design I came to pretty much the same conclusion: excess baggage. In many cases I found 90% of circuitry superfluous. A lot of so called engineering these days would best be described as religious ritual.

Translation: I don't understand why X is there, therefore there is no reason to have X there.

A very unimpressive example of the Dunning-Krueger effect.

[senso]

But also using the "its a HOBBY" argument, a pair of resistors will cost you next to nothing, and you are paying PnP time to place them, soldering those is a 15 seconds job, and the space occupied in a pcb is irrelevant.

[zapta]

But also using the "its a HOBBY" argument, a pair of resistors will cost you next to nothing, and you are paying PnP time to place them, soldering those is a 15 seconds job, and the space occupied in a pcb is irrelevant.

It adds another SKU. Trying to keep the design simple. Only 2 resistor values.

[zapta]

1) it is bloody hard to meet the impedance on the board in terms of track width and clearance to meet the usb spec. putting these resistor directly at the pin allows for an easier way to do the pcb layout. if your stackup allows you to easily create the 90 ohms differential you don't need them. fact is most stackups with standard pre-preg material on a 4 or 6 layer board are in the 60 to 75 ohm...
tack on 22 or 33 ohm and you are there.

2) these resistor limit current for an esd strike. integrating these resistors is a no-go becasue of 1 above and because they cannot handle a lot of power nor standoff voltage in integrated form.

go read this if you want to know why and how to do it properly ( as opposed to believing in ritual )
 http://www.usb.org/developers/docs/hs_usb_pdg_r1_0.pdf

That intel document seems to focus on high speed USB (480mpbs), my board is full speed only (12mpbs). Does it matter than much in the lower speeds?

[Pjotr]

Anyway, a proper design is by default 90% crap by the new proposed definition of "proper" 

So also leave out proper ESD protection and rely on the small on-chip diodes only.

[zapta]

So also leave out proper ESD protection and rely on the small on-chip diodes only.

Thanks for the suggestion.

[tggzzz]

Anyway, a proper design is by default 90% crap by the new proposed definition of "proper" 

So also leave out proper ESD protection and rely on the small on-chip diodes only.

Good joke!

[zapta]

Anyway, a proper design is by default 90% crap by the new proposed definition of "proper" 

So also leave out proper ESD protection and rely on the small on-chip diodes only.

Good joke!

I expect mature communication here.

[Pjotr]

Hi Zapta,

Leaving out the series resistors will probably not corrupt USB 1.1 but they are there also for EMC reasons (slew limiting) sometimes. A QAD design that works in a particular occasion is not always a good design.

There is a load of good application notes about USB implementation available on the www. So why not doing it the standard right way and as advised by the manufacturer, instead of making short cuts?

[donotdespisethesnake]

I am not interested in formal USB compliance, just want to thing to work reliably.

Definitely leave them in.

I and many others have been bitten by "hobby" designs that take shortcuts, and then spend a lot of time and effort tracking down why my board resets when the fridge turns on (or other common equipment). Turns out for the cost of a few SMDs, the problem is easily fixed, but the amount of frustration caused was ridiculous. The designer says similar thing "wanted to save money/keep it simple/didn't think it was needed".

Typically, the Chinese take these shortcuts, most of the time they get away with it, many times users get bitten.

Beginners, or even professionals, trying to develop embedded code can find it tricky at the best of times, having the CPU reset randomly does not help!

I would therefore argue for more protection on USB, rather than less. Best practice is not to connect shield directly to ground, but use an RC pair.

I also use a buffer chip like USBUF01W6, which has series resistors and 1.5k pull up built in and claims to have ample ESD protection , so replaces at least 3 components with 1. There are several other similar chips, although downside is that they are usually a quite small component if you are doing DIY.

In general, negative experience due to unreliability tends to create more of a bad rep, than simplicity can create a good rep.

[free_electron]

Oh it's one of them 'hobby' things eh ?
Why don't you just forget about all of the rules and simy wire it up with a bunch of recycled power cords. After all, those work reliably with 10 to 15 ampere at 220 volt so surely that measly few milliamps at 3 volt of usb will be no problem

  hopeless. We are all doomed. Nobody takes pride in what they do anymore.

[zapta]

I also use a buffer chip like USBUF01W6, which has series resistors and 1.5k pull up built in and claims to have ample ESD protection , so replaces at least 3 components with 1. There are several other similar chips, although downside is that they are usually a quite small component if you are doing DIY.

That's USBUF01W6 looks interesting except that I need to switch the 1.5k resistor. Do you know of a similar device with an open ended 1.5k resistor or even better a switched 1.5k resistor?

[T3sl4co1l]

Hi Zapta,

Leaving out the series resistors will probably not corrupt USB 1.1 but they are there also for EMC reasons (slew limiting) sometimes. A QAD design that works in a particular occasion is not always a good design.

There is a load of good application notes about USB implementation available on the www. So why not doing it the standard right way and as advised by the manufacturer, instead of making short cuts?

Actually very little to do with EMC; that's why the shield is there, and why it's supposed to be grounded as directly as possible to the circuit ground plane.

Although the bulk of a USB packet (at 1.1 Full Speed) is balanced, the signals are essentially equivalent to 74HC CMOS drivers, and the line status / negotiation signals are explicitly unbalanced (both lines high/low, one open, etc.).  Which is a good reason not to route it as full differential -- it isn't!  Likewise, common mode chokes on the data pair won't save you much trouble.

I'm still waiting to hear what Zapta's oscilloscope bandwidth is.

Tim

[senso]

I also use a buffer chip like USBUF01W6, which has series resistors and 1.5k pull up built in and claims to have ample ESD protection , so replaces at least 3 components with 1. There are several other similar chips, although downside is that they are usually a quite small component if you are doing DIY.

That's USBUF01W6 looks interesting except that I need to switch the 1.5k resistor. Do you know of a similar device with an open ended 1.5k resistor or even better a switched 1.5k resistor?

If the only thing really using the 3.3v then you can just use and IO pin from the micro to the 3.3v pin in the package, needs to be tested, but it might work.

[zapta]

I'm still waiting to hear what Zapta's oscilloscope bandwidth is.

It's an old 60Mhz tektronix DSO, similar to that batch of Military DSO's Dave scored a few months ago. I used one probe with standard earth crocodile, and look at D+ and at D- while communicating with an Apple Macbook Air notebook.