TUSB2036: EEPROM and hub reset timing

Wayne:

EEPROM read should be start after power on reset. TUSB2036 performs a one-time access read operation from the EEPROM if the EXTMEM pin is pulled low and the chip select(s) of the EEPROM is connected to the system power-on reset, btu EEPROM read its self took some time.

 One more critical is CLK should be start at the end of reset, did you look at crystal signal and the relation between reset?

Best

Brian

I couldn't find a timing specification for the crystal relative to the reset within the datasheet other than "Clock signal has to be active during the last 60 μs of the reset window".  But what voltage level at XTAL1 is "active", since I'm using a passive crystal with a start-up envelope?  Does "clock active" mean that XTAL1 has VIL(TTL) <= 0.8V and VIH(TTL) of 2V min, during which time RESET is <= VIL(TTL) min? 

Wayne:

      the only timing for clock is  "Clock signal has to be active during the last 60 μs of the reset window".

      Clock active means here that crystal clock reaches Vih (2v) and stable . which is 60us before RESET is done.

      Did you look at power supply and clock waveform yet?

Regards

Brian

Brian:

You stated that  "Clock active means here that crystal clock reaches Vih (2v) and stable which is 60us before RESET is done."  However, my crystal takes about  6-7ms to reach Vih (2V) and stable (see link below).  But a TI document SLLA314 March 2011 stated "Texas Instruments recommends a minimum of 100 µs to a maximum of 1 ms of reset timing. If the hub is held in reset for a longer period of time, it can fail to respond promptly to USB host signaling and not complete enumeration. This is typically an issue with embedded system applications."  So apparently the hub switches-in it's D+ pull-up when the power supply comes up rather than after the reset pin is Vih (2V)?

So the time it takes for my crystal to start up and stabilize is about 7x longer than the maximum suggested 1ms reset time.  

When I changed R2 from 10K to 100K to increase my RC reset time, I observed EEPROM communications about 6.7ms after the "brick wall" / fast rise time 3.3V power supply comes up.

3.3V to XTAL1 start-up: https://www.dropbox.com/scl/fi/uhemmocptg0l0zj1hnsdt/Vcc-to-XTAL1.bmp?rlkey=4b0gcw1bovyxe37b2fivmynfw&dl=0

Thanks,

Wayne

P.S.  The above 3.3V to XTAL1 start-up link above was with a 680 Ohm resistor instead of the recommended 1.5K XTAL2 series resistor which didn't give me a high enough amplitude with my crystal.  I'm now testing lower values to pull the low voltage closer to 0V.

Wayne:

       First issue we need to resolve is that crystal startup time is too long. TUSB2036 will not get reset correctly  without stable clock when RESET signal goes high.

      We may need to find crystals with fast startup time.

Regars

Brian

Brian:

Here's the crystal specifications from the TUSB2036 datasheet: "Fox Electronics – part number HC49U-6.00MHz 30\50\0±70\20, which means ±30 ppm at 25°C and ±50 ppm from 0°C to 70°C. The characteristics for the crystal include a load capacitance (CL) of 20 pF, maximum shunt capacitance (Co) of 7 pF, and the maximum ESR of 50 Ω. In order to insure enough negative resistance, use C1 = C2 = 27 pF. The resistor Rd is used to trim the gain, and Rd = 1.5 kΩ is recommended."

The Fox Electronics part number is no longer available.  The closest crystal I could find was ECS Inc: ECS-60-20-5PX-TR which met all of the above specifications, except that the equivalent ESR was 70 Ω instead of 50 Ω which was the lowest I could find compatible with the other specs.  I can't change the drive level and am using the recommended crystal capacitance.  So I'm looking for suggestions to shorten the crystal start-up.

The below screen capture shows the reset voltage to crystal start-up.  I'm assuming that the hub does start driving the crystal until Reset Vih = 2V?

Reset to XTAL1

Wayne:

     Let me discuss with our designer if ESR is critical for crystal startup time.

Regards

Brian

Brian:

I know that the ESR does affect the start-up time.  Please recall that I couldn't find an ESR lower than 70 Ohms while still meeting the other electrical specifications.  The crystal I chose was a ECS Inc: ECS-60-20-5PX-TR with a HC-49/US footprint.

Thanks,

Wayne

 One customer used ABC2-6.000MHZ-4-T from Abracon. and startup time is around 2ms, it looks much better than 5ms you saw.

Regards

Brian

Brian: 

While this has 60 Ohms ESR (better than the 70 Ohms of my ECS crystal), it's still higher than the TI spec'd 50 Ohms.  The problem with this part is that it has a 4-SMT footprint and my PCB has already been fabricated with a HC49/US footprint.

Can you confirm that the hub does start driving the crystal until Reset Vih = 2V?  Or does the hub start driving the crystal as soon as Vcc=3.3V is applied?  When does the hub switch-in the D+ pull-up, at which time USB discovers the hub is connected?

Wayne:

      Design has not respond the question about how hub driving the crystal today. I will find it out by Monday.

Best

Brian

Brian:

It's been a week from last "Monday" since your last correspondence.  Think that means that the design group isn't going to respond?

Thanks,
Wayne

Wayne:

   Sorry for the delay, since this device was designed 20 years ago,it took some time for design to find the database.

 This is  the block diagram for crystal oscillator. The GZ port controls the oscillator's functionality. GZ can disables the oscillator and applies high impedance to the output. Design is still checking if XI/XO is off during reset.

Regadrs

Brian