Kenneth Fong wrote:

> Has anybody had any problems with the Cypress 811HS USB Host chip?  I
> am having some strange stability problems.
>
> Sometimes I reboot, the chip would work perfectly.   Other times (80%
> of times), when I reboot the chip, I can successfully write to a
> memory area on chip and read the same data back out.  However, if I
> try to read and write to the control registers, I get garbage.  The
> strange thing is that when the system works, it works perfectly for as
> long as I keep the power on, and when it doesn't work, no matter how
> long I leave it on, it won't "fix"/stabilize itself.
>
> Things I have tried:
> -adding large (470u) capacitor to power supply
> -adding 4 small (0.01u) capacitors to power supply (power supply
> currently swings 3.0 to 3.6V)
> -Tried using Epson, Citizen, EMC, Fox 48 MHz oscillators.  They all
> oscillate at 48MHz, however, I see a lot of overshoot (10V+ pk to pk)
> in original wave form.   The waveforms are also sinusoidal, or heavily
> skewered squarewave (one big overshoot that covers then entire
> plateau).  I also tried to process the signal by feeding it to 3
> diodes in series for a theoretical drop of 2.1V, which strangely,
> limits the voltage swing to about -.5V to 4V.  Note that I might be
> seeing some artifacts of the oscilloscope for mine is only rated at
> 300Mhz with a 200MHz probe.
> -Tried the 48MHz Crystal circuit as described in datasheet.  Also is
> unstable.
> -Tried to process the signal to give it sharper edges, but can't find
> anything that switches fast enough.  Each "high" signal is only 10ns
> long.  CMOS inverter didn't work. Simple BJT switch didn't work.
> Comparators didn't work.
> -Of course, I've tried all of the above with 2 Cypress Host chip.  One
> is more flaky by the other, but both are by no means stable.
> -Also tried powering the cypress chip separately from the rest of the
> circuit, although the 3.0-3.6V noise is still present.
>
> Does anybody have a solution to this?

Ken;

We used the 811HS in a USB Host product driving a printer.  As a combined
system, we had a number of problems, some of which might have bearing on
your problem and maybe some of which you might run into further down the
road.    We spend a lot of time and money getting the system to work.  In
fairness, it should be state that a good portion of that time was spent
working
with the device we attached to and not related to the 811Hs.

Open a case file at Cypress.  We found them very supportive and
worthwhile.
And feel free to email me directly at this address and we can discuss this
in
more depth offline.

Regards,

Blakely LaCroix
noone@wrenchman.com

Has anybody had any problems with the Cypress 811HS USB Host chip?  I
am having some strange stability problems.

Sometimes I reboot, the chip would work perfectly.   Other times (80%
of times), when I reboot the chip, I can successfully write to a
memory area on chip and read the same data back out.  However, if I
try to read and write to the control registers, I get garbage.  The
strange thing is that when the system works, it works perfectly for as
long as I keep the power on, and when it doesn't work, no matter how
long I leave it on, it won't "fix"/stabilize itself.

Things I have tried:
-adding large (470u) capacitor to power supply
-adding 4 small (0.01u) capacitors to power supply (power supply
currently swings 3.0 to 3.6V)
-Tried using Epson, Citizen, EMC, Fox 48 MHz oscillators.  They all
oscillate at 48MHz, however, I see a lot of overshoot (10V+ pk to pk)
in original wave form.   The waveforms are also sinusoidal, or heavily
skewered squarewave (one big overshoot that covers then entire
plateau).  I also tried to process the signal by feeding it to 3
diodes in series for a theoretical drop of 2.1V, which strangely,
limits the voltage swing to about -.5V to 4V.  Note that I might be
seeing some artifacts of the oscilloscope for mine is only rated at
300Mhz with a 200MHz probe.
-Tried the 48MHz Crystal circuit as described in datasheet.  Also is
unstable.
-Tried to process the signal to give it sharper edges, but can't find
anything that switches fast enough.  Each "high" signal is only 10ns
long.  CMOS inverter didn't work. Simple BJT switch didn't work. 
Comparators didn't work.
-Of course, I've tried all of the above with 2 Cypress Host chip.  One
is more flaky by the other, but both are by no means stable.
-Also tried powering the cypress chip separately from the rest of the
circuit, although the 3.0-3.6V noise is still present.

Does anybody have a solution to this?