The M68HC812A4 at 5Vdd needs Vih (input high) to be 3.5v (0.7xVdd) for external reads. Most memories have a minimum Voh (output high) of 2.4v. This is a problem.

Started by "Bump, Jerome" February 15, 2007
Hi,

The M68HC812A4 at 5Vdd needs Vih (input high) to be 3.5v (0.7xVdd) for external reads. Most memories have a minimum Voh (output high) of 2.4v. This is a problem.

We have been using this design since 1998 and it has worked great. We recently connected a 1 Meg NVRAM ( from Benchmark & ST Micro) so we could store large amounts of data and the 812A4 would not read the data correctly. After verify all of the timing issues, we realized that the 1 Meg memory had a Voh of only 3.2v. Even though our other memories (Dallas Semi, Chiplus, Hynix, Samsung) also state a Voh of 2.4v, they actually have a Voh at 3.5v or greater and work fine. That is why we have not seen problems during the last 9 years.

So, it looks like we need to add a level shifter/buffer on our data bus if we want to insure that our design meets the external memory spec? I guess I can understand this for adding oddball peripherals, but it doesn't seem to make sense for the 812A4's design to require a level converter to communicate with external RAM & ROM.

If you are using external memory in your design, what have you done to address this issue? Even the newer parts in the MC9S12X families show Vih of 3.2v (.65xVdd).

Freescale Support has not offered any good suggestions, but to continue using the memories that work?

Regards

Jerome
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 3:38 PM
> To: 6...
> Subject: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
>

Maybe pullup resistors will fix this?

Regards,
Anders
We did try that. Adding 3.3K pull up resistors helped lift the signals up from 3.4v to 3.8v -AND- the RAM's started working. But we don't know why the signals didn't go higher, closer to 5v. The only connections to the data bus are the 6812, the RAM and a tri-stated FLASH part.

We thought about dropping the resistance down to a smaller value (2K) to pull it higher so we would have more breathing room above the 3.5v input, but we are wondering if we are just asking for other problems. A 3.3K should have done the trick. Since we are not sure why we should have to pull the dataline up so hard with a resistor, we started looking at using a level converter.

Do you have any ideas why the signals would not have gone closer to 5v with 3.3K pullups?

Regards,

Jerome
-----Original Message-----
From: 6... [mailto:6...] On Behalf Of Anders Friberg
Sent: Thursday, February 15, 2007 10:04 AM
To: 6...
Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input high) to be 3.5v (0.7xVdd) for external reads. Most memories have a minimum Voh (output high) of 2.4v. This is a problem.
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 3:38 PM
> To: 6...
> Subject: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
>

Maybe pullup resistors will fix this?

Regards,
Anders
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 6:09 PM
> To: 6...
> Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
>
> We did try that. Adding 3.3K pull up resistors helped lift
> the signals up from 3.4v to 3.8v -AND- the RAM's started
> working. But we don't know why the signals didn't go higher,
> closer to 5v. The only connections to the data bus are the
> 6812, the RAM and a tri-stated FLASH part.
>
> We thought about dropping the resistance down to a smaller
> value (2K) to pull it higher so we would have more breathing
> room above the 3.5v input, but we are wondering if we are
> just asking for other problems. A 3.3K should have done the
> trick. Since we are not sure why we should have to pull the
> dataline up so hard with a resistor, we started looking at
> using a level converter.
>
> Do you have any ideas why the signals would not have gone
> closer to 5v with 3.3K pullups?
>

Can you give some info about the NVRAM part number?

Regards,
Anders
Here's what we tested.

These NVRAM's all work:
Dallas RAM Size Output
DS1744-070 32K x 8 with RTC 3.6 Vdc
DS1747-070 512K x 8 with RTC 3.6
DS1249Y-70 256K x 8 5.0

These other 32K RAM's also work:
Samsung
K6T0808C1D-DL70 32K x 8 5.0
Hynix - HY62CT081ED70C 32Kx8 --
These NVRAM's all FAILED:
Dallas RAM Size Output
DS1250Y-100+ 512K x 8 (100 nsec) 3.4
DS1265Y-70 1 Meg x 8 3.4

ST Electronics
M48Z512AY-70PM1 512K x 8 3.4

Benchmark
bq4015YMC-70 512K x 8 3.4
bq4016YMC-70 1 Meg x 8 3.4

Regards,

Jerome

-----Original Message-----
From: 6... [mailto:6...] On Behalf Of Anders Friberg
Sent: Thursday, February 15, 2007 1:02 PM
To: 6...
Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input high) to be 3.5v (0.7xVdd) for external reads. Most memories have a minimum Voh (output high) of 2.4v. This is a problem.
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 6:09 PM
> To: 6...
> Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
>
> We did try that. Adding 3.3K pull up resistors helped lift
> the signals up from 3.4v to 3.8v -AND- the RAM's started
> working. But we don't know why the signals didn't go higher,
> closer to 5v. The only connections to the data bus are the
> 6812, the RAM and a tri-stated FLASH part.
>
> We thought about dropping the resistance down to a smaller
> value (2K) to pull it higher so we would have more breathing
> room above the 3.5v input, but we are wondering if we are
> just asking for other problems. A 3.3K should have done the
> trick. Since we are not sure why we should have to pull the
> dataline up so hard with a resistor, we started looking at
> using a level converter.
>
> Do you have any ideas why the signals would not have gone
> closer to 5v with 3.3K pullups?
>

Can you give some info about the NVRAM part number?

Regards,
Anders
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 10:38 PM
> To: 6...
> Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
> Here's what we tested.
>
> These NVRAM's all work:
> Dallas RAM Size
> Output
> DS1744-070 32K x 8 with RTC 3.6 Vdc
> DS1747-070 512K x 8 with RTC 3.6
> DS1249Y-70 256K x 8 5.0
>
> These other 32K RAM's also work:
> Samsung
> K6T0808C1D-DL70 32K x 8 5.0
> Hynix - HY62CT081ED70C 32Kx8 --
> These NVRAM's all FAILED:
> Dallas RAM Size
> Output
> DS1250Y-100+ 512K x 8 (100 nsec) 3.4
> DS1265Y-70 1 Meg x 8 3.4
>
> ST Electronics
> M48Z512AY-70PM1 512K x 8 3.4
>
> Benchmark
> bq4015YMC-70 512K x 8 3.4
> bq4016YMC-70 1 Meg x 8 3.4
>
> Regards,
>
> Jerome
>

Just two ideas (hard to say without the schematics or source code):

- The 3.5 V seems to be close to the internal battery voltage (plus
parasitic diode drop), is it guaranteed that no voltage is present on the
NVRAM lines when the Vcc is missing? All devices on the bus including the
pullup resistors must be powered by the same 5V, if not, the battery
switchover may be disturbed.

- Some NVRAMS seems to require a startup time after powerup to become
functional (a couple of hundred ms), is this accounted for?

Regards,
Anders
Anders

Thanks for all of the ideas and comments.

- Power Supply. All devices run off the same 5v. When the system powers down, everything is off. Also, we have been using the Dallas 32K & 128K NVRAM's with RTC's for 9 years without problems, so I don't feel we have a battery switchover issues.

- Startup Time. Yes, we designed our system to be held in reset long enough to give the NVRAM's ample time to switch over from battery to Vcc. We also have a software delay at startup using only internal RAM just to insure that the NVRAM is up & running correctly. Then, on top of that, we perform write/read test on the NVRAM before we use it to again verify that everything is ready to go.

Regards,

Jerome

-----Original Message-----
From: 6... [mailto:6...] On Behalf Of Anders Friberg
Sent: Thursday, February 15, 2007 8:32 PM
To: 6...
Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input high) to be 3.5v (0.7xVdd) for external reads. Most memories have a minimum Voh (output high) of 2.4v. This is a problem.
> -----Original Message-----
> From: 6... [mailto:6...]
> On Behalf Of Bump, Jerome
> Sent: Thursday, February 15, 2007 10:38 PM
> To: 6...
> Subject: RE: [68HC12] The M68HC812A4 at 5Vdd needs Vih (input
> high) to be 3.5v (0.7xVdd) for external reads. Most memories
> have a minimum Voh (output high) of 2.4v. This is a problem.
> Here's what we tested.
>
> These NVRAM's all work:
> Dallas RAM Size
> Output
> DS1744-070 32K x 8 with RTC 3.6 Vdc
> DS1747-070 512K x 8 with RTC 3.6
> DS1249Y-70 256K x 8 5.0
>
> These other 32K RAM's also work:
> Samsung
> K6T0808C1D-DL70 32K x 8 5.0
> Hynix - HY62CT081ED70C 32Kx8 --
> These NVRAM's all FAILED:
> Dallas RAM Size
> Output
> DS1250Y-100+ 512K x 8 (100 nsec) 3.4
> DS1265Y-70 1 Meg x 8 3.4
>
> ST Electronics
> M48Z512AY-70PM1 512K x 8 3.4
>
> Benchmark
> bq4015YMC-70 512K x 8 3.4
> bq4016YMC-70 1 Meg x 8 3.4
>
> Regards,
>
> Jerome
>

Just two ideas (hard to say without the schematics or source code):

- The 3.5 V seems to be close to the internal battery voltage (plus
parasitic diode drop), is it guaranteed that no voltage is present on the
NVRAM lines when the Vcc is missing? All devices on the bus including the
pullup resistors must be powered by the same 5V, if not, the battery
switchover may be disturbed.

- Some NVRAMS seems to require a startup time after powerup to become
functional (a couple of hundred ms), is this accounted for?

Regards,
Anders