DSPRelated.com
Forums

adsp-2184 strange behaviour.

Started by Sergey Vlasov May 11, 2001
Hi, group.
Just found that one of flag inputs (PF4) of ADSP-2184 machine does not work.
If it is configured as an output, it works fine, but if it's input it
always appear to be "zero", and no IRQE interrupts happen too.
Assuming the input is damaged I measured V-A characteristics of all the
flag inputs. They appear to be the same: nearly zero current at input
voltages between -0.4 and +5.5v and fast current growth out of the
interval (to some 3mA at -0.7v and +6.8v). I thought that if the device
is damaged with, say, static discharge, the bad input will show different
V-A curve than other. Any ideas about the case?
 
Another curios thing about '2184 is that when addressing DM through IDMA
port I can normally write and read 8k words of it (not 4k as in the datasheet).
I simply write different numbers to different DM locations and then read.
All the numbers are at their places at DM addresses between 0x2000 and 0x3FFF.
Yes, addresses between 0x3FE0 and 0x3FFF appear to be normal DM-memory
locations too when addressed through I-port. Is it normal?
 
Thank you in advance.
Sergey Vlasov.



At 09:52 AM 5/11/01 +0700, Sergey Vlasov wrote:
Hi, group.
Just found that one of flag inputs (PF4) of ADSP-2184 machine does not work.
If it is configured as an output, it works fine, but if it's input it
always appear to be "zero", and no IRQE interrupts happen too.
Assuming the input is damaged I measured V-A characteristics of all the
flag inputs. They appear to be the same: nearly zero current at input
voltages between -0.4 and +5.5v and fast current growth out of the
interval (to some 3mA at -0.7v and +6.8v). I thought that if the device
is damaged with, say, static discharge, the bad input will show different
V-A curve than other. Any ideas about the case?
 
Another curios thing about '2184 is that when addressing DM through IDMA
port I can normally write and read 8k words of it (not 4k as in the datasheet).
I simply write different numbers to different DM locations and then read.
All the numbers are at their places at DM addresses between 0x2000 and 0x3FFF.
Yes, addresses between 0x3FE0 and 0x3FFF appear to be normal DM-memory
locations too when addressed through I-port. Is it normal?
 
Thank you in advance.
Sergey Vlasov.

_____________________________________

I have never noticed a problem with PF4.

Historically, ADI has labeled larger memory parts with the smaller memory part numbers. For example, early ADSP-2186M were really ADSP-2185M dies. This situation is common, before they have new (smaller size) dies. This allows them to improve their time to market.

Since manufacturing cost is proportional to die size (and SRAM takes a lot of die space), I wouldn't bet that your 2184 will always be 2186.

Al Clark
Danville Signal Processing, Inc.