Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

New Yahoo Group for Blackfin DSP/MCU

3 views
Skip to first unread message

rickman

unread,
Jul 25, 2008, 1:02:15 PM7/25/08
to
There is more than one way to skin a cat and there is getting to be a
very large number of ways to design systems that need signal
processing along with standard control processing. One of the MCU-
like devices that are good at this is the Blackfin from ADI.

I had not paid a lot of attention to it in the past, but earlier this
year I attended a seminar on it and realized that it is a pretty nice
little chip in many ways. I tried joining the Blackfin Yahoo group
only to find that it was overrun by spam (sound familiar?). So I
started another one and have taken steps to preclude it from going the
same route.

http://tech.groups.yahoo.com/group/BlackFinDSP-MCU/

If you have and interest in the Blackfin processor, you can discuss
your issues and ask questions there. Being highly targeted, it should
provide lots of good info.

Rick

Vladimir Vassilevsky

unread,
Jul 25, 2008, 1:21:15 PM7/25/08
to

rickman wrote:

> There is more than one way to skin a cat and there is getting to be a
> very large number of ways to design systems that need signal
> processing along with standard control processing. One of the MCU-
> like devices that are good at this is the Blackfin from ADI.

I've been doing the designs with Blackfin since 2003 and still do, and I
can tell you that Blackfin is same as crap as all other MCUs and CPUs. A
tremendous list of the incredible bugs in the silicon, completely
idiotic and unusable memory protection, poor set of peripherals, weak
external bus shared between SRAM and other peripherals.

> I had not paid a lot of attention to it in the past, but earlier this
> year I attended a seminar on it and realized that it is a pretty nice
> little chip in many ways.

Yes, several things are good about it, however don't get over excited.

> I tried joining the Blackfin Yahoo group
> only to find that it was overrun by spam (sound familiar?).

"Hallo freinds I am Babbadushman. I want 2 ask U a Q about 2+2=4. What
does this mean? I need this urgent and immediately."

> So I
> started another one and have taken steps to preclude it from going the
> same route.
>
> http://tech.groups.yahoo.com/group/BlackFinDSP-MCU/
>
> If you have and interest in the Blackfin processor, you can discuss
> your issues and ask questions there. Being highly targeted, it should
> provide lots of good info.

Am I already banned from there?

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

BWallace

unread,
Jul 25, 2008, 2:27:44 PM7/25/08
to

I don't know if you are banned or not but I think your signature should be
changed to Vladimir "Wet Blanket" Vassilevsky.

Vladimir Vassilevsky

unread,
Jul 25, 2008, 3:17:03 PM7/25/08
to

BWallace wrote:


>
> I don't know if you are banned or not but I think your signature should be
> changed to Vladimir "Wet Blanket" Vassilevsky.

Do you have anything else to say before being plonked?

VLV

Adrian

unread,
Jul 25, 2008, 3:33:06 PM7/25/08
to
On Jul 25, 1:02 pm, rickman <gnu...@gmail.com> wrote:
> I tried joining the Blackfin Yahoo group
> only to find that it was overrun by spam (sound familiar?). So I
> started another one and have taken steps to preclude it from going the
> same route.
>
> http://tech.groups.yahoo.com/group/BlackFinDSP-MCU/
>

What's wrong with the +4 years old forum from blackfin.org?

rickman

unread,
Jul 25, 2008, 4:41:02 PM7/25/08
to
On Jul 25, 3:17 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:

Man, you have done it now! You are threatened with being killfiled by
the big V!!!

No, actually, that might be a complement...

Rick

BWallace

unread,
Jul 25, 2008, 5:45:21 PM7/25/08
to
Considering his last two responses, I thought he would be a bit more
receptive to the humor.

Gene S. Berkowitz

unread,
Jul 26, 2008, 9:50:28 AM7/26/08
to
In article <NXnik.8868$L_....@flpi150.ffdc.sbc.com>,
antispa...@hotmail.com says...

>
>
> rickman wrote:
>
> > There is more than one way to skin a cat and there is getting to be a
> > very large number of ways to design systems that need signal
> > processing along with standard control processing. One of the MCU-
> > like devices that are good at this is the Blackfin from ADI.
>
> I've been doing the designs with Blackfin since 2003 and still do, and I
> can tell you that Blackfin is same as crap as all other MCUs and CPUs. A
> tremendous list of the incredible bugs in the silicon, completely
> idiotic and unusable memory protection, poor set of peripherals, weak
> external bus shared between SRAM and other peripherals.

Then why don't you design your own, get it fabbed, and enlighten the
world with the golden rays of your brilliance, captured in silicon?

--Gene

rickman

unread,
Jul 26, 2008, 11:46:35 AM7/26/08
to
On Jul 26, 9:50 am, Gene S. Berkowitz <first.l...@verizon.net> wrote:
...snip...

> Then why don't you design your own, get it fabbed, and enlighten the
> world with the golden rays of your brilliance, captured in silicon?
>
> --Gene

There are some things and some people who are better ignored. I can't
say I always do that myself, but it is a good idea anyway.

Rick

Vladimir Vassilevsky

unread,
Jul 26, 2008, 11:50:50 AM7/26/08
to

Gene S. Berkowitz wrote:

> In article <NXnik.8868$L_....@flpi150.ffdc.sbc.com>,
> antispa...@hotmail.com says...
>

>>I've been doing the designs with Blackfin since 2003 and still do, and I
>>can tell you that Blackfin is same as crap as all other MCUs and CPUs. A
>>tremendous list of the incredible bugs in the silicon, completely
>>idiotic and unusable memory protection, poor set of peripherals, weak
>>external bus shared between SRAM and other peripherals.
>
>
> Then why don't you design your own, get it fabbed, and enlighten the
> world with the golden rays of your brilliance, captured in silicon?

May be I should, why not. I don't care about the enlightenment of the
fucking world, however I have already developed an OS for BlackFin
trying to make the things more or less sensible for myself.

There are the reasons for the things to be as bad as they currently are.
Speaking of BlackFin, it was initially seen as the CPU for the
cellphones. When it became apparent that they missed the market, Intel
pulled out the plug, leaving ADI with the unripe project. ADI put the
things together as they could, and released the obviously incomplete
processor. Sadly enough, the original concept was appealing. Now ADI is
pathetic in pushing BlackFin in all directions, including those where
BlackFin doesn't fit (like the consumer video, for example).

Vladimir Vassilevsky

unread,
Jul 26, 2008, 12:05:42 PM7/26/08
to

rickman wrote:


Group Information:

* Members: 4
* Activity within 7 days: 2 New Messages

Looks like some things and some people are better ignored, huh?

Stefan Reuther

unread,
Jul 26, 2008, 2:01:24 PM7/26/08
to
Vladimir Vassilevsky wrote:
> Gene S. Berkowitz wrote:
>> In article <NXnik.8868$L_....@flpi150.ffdc.sbc.com>,
>>> I've been doing the designs with Blackfin since 2003 and still do,
>>> and I can tell you that Blackfin is same as crap as all other MCUs
>>> and CPUs. A tremendous list of the incredible bugs in the silicon,
>>> completely idiotic and unusable memory protection, poor set of
>>> peripherals, weak external bus shared between SRAM and other
>>> peripherals.

The BF54x parts are much better regarding peripherals. Lots of them, and
flash and SDRAM interface no longer shared. (Having them shared means
you cannot reliably boot the thing from flash unless you guarantee to
power down your whole board for >500ms on every reset. Duh.).

The annoying silicon bugs still remain.

>> Then why don't you design your own, get it fabbed, and enlighten the
>> world with the golden rays of your brilliance, captured in silicon?
>
> May be I should, why not.

I'll add that to my project list. Right after a C++ compiler that
doesn't suck.

> There are the reasons for the things to be as bad as they currently are.
> Speaking of BlackFin, it was initially seen as the CPU for the
> cellphones. When it became apparent that they missed the market, Intel
> pulled out the plug, leaving ADI with the unripe project.

Did Intel ever plan to make a "standalone" MSA core? The only part I've
seen (PXA400F? don't remember the exact number) used the MSA core as a
DSP co-processor to an Xscale core.

> ADI put the things together as they could, and released the obviously
> incomplete processor. Sadly enough, the original concept was appealing.
> Now ADI is pathetic in pushing BlackFin in all directions, including
> those where BlackFin doesn't fit (like the consumer video, for example).

If you just talk about decoding (playing) stuff, it isn't too bad. Hey,
people did full-screen animation on 66 MHz 486's, so a 533 MHz Blackfin
should be more than enough :-)


Stefan

problems@gmail

unread,
Jul 28, 2008, 12:29:07 AM7/28/08
to
Vladimir Vassilevsky wrote:
> I've been doing the designs with Blackfin since 2003 and still do,
> and I can tell you that Blackfin is same as crap as all other MCUs
> and CPUs.

] Speaking of BlackFin, it was initially seen as the CPU for the


]cellphones. When it became apparent that they missed the market,
]Intel pulled out the plug, leaving ADI with the unripe project.

]ADI put the things together as they could, and released the

]obviously incomplete processor. Sadly enough, the original
]concept was appealing. Now ADI is pathetic in pushing
] BlackFin in all directions, including those where BlackFin
] doesn't fit (like the consumer video, for example).

So it's got low power consumption ?
How would it be for a 2..4 AA-cell driven Ebook ?
Or is the CPU power consumption irrelevant compared
to a 80 * 20 char display ?

How's the uClinux project/capability ?

== TIA.


rickman

unread,
Jul 28, 2008, 9:53:09 AM7/28/08
to

I don't recall for sure if uCliniux is ported, but I want to say yes.
The power consumption is not especially low. The units I considered
were about the same as one of the high end ARM CPUs, both in
performance and power. They may have come out with some versions that
are better on static power.

You might try asking this in the Yahoo group...

Rick

Vladimir Vassilevsky

unread,
Jul 28, 2008, 10:17:02 AM7/28/08
to

problems@gmail wrote:
> Vladimir Vassilevsky wrote:
>
>>I've been doing the designs with Blackfin since 2003 and still do,
>>and I can tell you that Blackfin is same as crap as all other MCUs
>>and CPUs.
>
>
> ] Speaking of BlackFin, it was initially seen as the CPU for the
> ]cellphones. When it became apparent that they missed the market,
> ]Intel pulled out the plug, leaving ADI with the unripe project.
> ]ADI put the things together as they could, and released the
> ]obviously incomplete processor. Sadly enough, the original
> ]concept was appealing. Now ADI is pathetic in pushing
> ] BlackFin in all directions, including those where BlackFin
> ] doesn't fit (like the consumer video, for example).
>
> So it's got low power consumption ?

Although Blackfin consumes somewhat less power then the other DSPs, I
wouldn't call it the low power processor.

> How would it be for a 2..4 AA-cell driven Ebook ?
> Or is the CPU power consumption irrelevant compared
> to a 80 * 20 char display ?

The very minimum power consumption of Blackfin is at the order of 80mA
at 3.3V. This is if the core is running at 24MHz and idling most of the
time. At full speed, BlackFin burns about a couple of watts.

> How's the uClinux project/capability ?

Kinda working. As good as the illustration of the possibility.
If you want BlackFin to run with the speed of AVR, use ucLinux.
uClinux is linux without MMU, which is nonsense indeed.

Adrian

unread,
Jul 28, 2008, 10:22:41 AM7/28/08
to
>
> > How's the uClinux project/capability ?
>
> I don't recall for sure if uCliniux is ported, but I want to say yes.
> You might try asking this in the Yahoo group...
>
> Rick

Again, www.blackfin.org has all information you need + active
discussion groups (including a uClinux - subforum).
I don't think there is any need for another discussion group.

Keith

unread,
Jul 31, 2008, 12:01:11 AM7/31/08
to
Vladimir Vassilevsky wrote:
>
> The very minimum power consumption of Blackfin is at the order of 80mA
> at 3.3V. This is if the core is running at 24MHz and idling most of the
> time. At full speed, BlackFin burns about a couple of watts.

One of my Blackfin projects (based on a '533) uses less power than
that:

http://www.afibalert.com/

I am not at liberty to divulge the actual numbers, but it appears that
one's
mileage varies on power consumption.

>> How's the uClinux project/capability ?
>
> Kinda working. As good as the illustration of the possibility.

Hmmm, I have delivered a couple of solutions using uClinux for the
Blackfin.
I am not sure if it's ready for prime-time (i.e. ready to
power something you could sell at Best Buy), but it is definitely
good enough to deliver single-digit prototypes to customers or perhaps
if you are making some sort of test equipment that is for the lab and
not for consumer use. So I guess I agree with you there.


> If you want BlackFin to run with the speed of AVR, use ucLinux.

Hahaha! Quite a bit of an exaggeration, but the point is correct.
gcc for the Blackfin is about 40% slower than ccblkfin (the ADI
compiler), and linux adds some overhead that isn't there if you
use your own OS (like I usually do).


> uClinux is linux without MMU, which is nonsense indeed.

I am not involved in the usual linux or OS religious wars, so
I don't know if linux w/o an MMU is nonsense. I mean, is VDK
without an MMU(*) retarded? Is any pre-rolled OS running on a chip
without an MMU silly?

The reason I used uClinux was that the client wanted TCP/IP running
over gigabit ethernet from their blackfin, and they needed it
pronto. My analysis (given the time and $$ involved) was that the
only way to achieve all the goals was to port the existing gigabit
linux driver to uClinux for the Blackfin, and it worked great. All
I had to do was to get the ethernet chip working, and from there
TCP/IP was a done deal.

Plus since I had to follow the GPL, you all can benefit from my work
on the gigabit driver (it's rolled into the base distribution now).
Since then, ADI has taken it over and increased the performance
of the driver since I last touched it. Sure wish they had been
interested in the driver before I spent all those late nights
in kernel panic :-)

--Keith

(*) BTW the Blackfin does have an MMU, just not one that gives the
granularity needed for a port of the real linux kernel.

Vladimir Vassilevsky

unread,
Jul 31, 2008, 10:28:41 AM7/31/08
to

Keith wrote:

> Vladimir Vassilevsky wrote:
>
>>The very minimum power consumption of Blackfin is at the order of 80mA
>>at 3.3V. This is if the core is running at 24MHz and idling most of the
>>time. At full speed, BlackFin burns about a couple of watts.
>
>
> One of my Blackfin projects (based on a '533) uses less power than
> that:
>
> http://www.afibalert.com/
>
> I am not at liberty to divulge the actual numbers, but it appears that
> one's mileage varies on power consumption.

I was planning the BF-537 for the low power project, and I measured the
actual power consumption that can be attained. There could be some +/-
difference, however the ballpark is going to be as I cited.


>>uClinux is linux without MMU, which is nonsense indeed.
>
> I am not involved in the usual linux or OS religious wars, so
> I don't know if linux w/o an MMU is nonsense.

By design, linux was meant for the 386 protected mode. A lot of the
functionality of the linux core relies on the MMU. Linux without MMU is
the perversion of the base concepts.

> I mean, is VDK
> without an MMU(*) retarded?

VDK is retarded regardless.

> Is any pre-rolled OS running on a chip
> without an MMU silly?

The dynamic memory management without MMU is unsafe and involves many
tradeoffs. The memory protection is the other highly desirable feature
of MMU. Blackfin has a rudimentary MMU which doesn't provide neither
address translation nor a sensible mechanism for the memory protection.
The exception handling of BlackFin is simply a nightmare: there is no
sense in the way it is implemented, and there are the ugly bugs in the
silicon in the addition to that.

> The reason I used uClinux was that the client wanted TCP/IP running
> over gigabit ethernet from their blackfin, and they needed it
> pronto. My analysis (given the time and $$ involved) was that the
> only way to achieve all the goals was to port the existing gigabit
> linux driver to uClinux for the Blackfin, and it worked great. All
> I had to do was to get the ethernet chip working, and from there
> TCP/IP was a done deal.

We developed the proprietary RTOS for Blackfin, including the posix
compliant FAT32 file system and TCP/IP stack over the BlackFin native
MAC. At minimum, it can run from the bare 537-class chip using only the
internal memories.

> Plus since I had to follow the GPL, you all can benefit from my work
> on the gigabit driver (it's rolled into the base distribution now).

I could never understand the joys of the GPL.

> Since then, ADI has taken it over and increased the performance
> of the driver since I last touched it. Sure wish they had been
> interested in the driver before I spent all those late nights
> in kernel panic :-)
>
> --Keith
>
> (*) BTW the Blackfin does have an MMU, just not one that gives the
> granularity needed for a port of the real linux kernel.

Obviously they wanted to make the real MMU (look at all of those empty
registers), however it didn't happen.

nos...@isp.com

unread,
Aug 3, 2008, 12:48:24 AM8/3/08
to

So then what device/configuration would you advise for a
black & white 80 char wide, probably 20 lines, reader from
eg. CF, running from 2 to 4 rechargable AA cells & no-frills
miminal cost ?

TIA,

== Chris Glur.

Vladimir Vassilevsky

unread,
Aug 3, 2008, 11:06:24 AM8/3/08
to

nos...@isp.com wrote:


> So then what device/configuration would you advise for a
> black & white 80 char wide, probably 20 lines, reader from
> eg. CF, running from 2 to 4 rechargable AA cells & no-frills
> miminal cost ?

The task is stated very vaguely; it can be anything from as simple as
PIC/AVR to as complex as iPod/TabletPC depending on what the device
should be capable of.

nos...@isp.com

unread,
Aug 7, 2008, 12:53:43 PM8/7/08
to
In article <JQjlk.6204$np7....@flpi149.ffdc.sbc.com>, Vladimir Vassilevsky
<antispam_bo...@hotmail.com> wrote:

>
> nos...@isp.com wrote:
>
> > So then what device/configuration would you advise for a
> > black & white 80 char wide, probably 20 lines, reader from
> > eg. CF, running from 2 to 4 rechargable AA cells & no-frills
> > miminal cost ?
>
> The task is stated very vaguely; it can be anything from as simple
> as PIC/AVR to as complex as iPod/TabletPC depending on what the
> device should be capable of.
>

It starts from the belief that the 'one laptop per child'
hyped-idea from MIT can't handle the lack of mains electricity
in 3rd world locations.

So the 2 to 4 rechargable AA's need to give about 8 hours
reading. I didn't yet find out about display power consumption,
and a 'hi-volume used for other devices' display would be used.
Of course if the display's consumption is significant then the
CPU is less relevant. I guess a 8-bit CPU would do, except for
the lack of ability to leaverage existing software, eg. uclinux.

Here's some ideas of required capabilities for the student/reader:
* hyper link to random sections,
* bookmark where I was last time,
* bookmark the stuff/s that I want to show Joe & Jill..
* cutNpaste ..........
* search: where-else/if XYZ was mentioned........
* embed(?sp) [simple] sketches in the text...posibly animated
* ... etc. ..your ideas/requirements...

So, I realise that the half second delay of low-power Epaper
is a problem for eg. cutNpaste.

I'm thinking of a 5 or 10 finger-capacitive-pad input
mouse-like/pointer.

Probably the starting decision must be the display ?

Thanks for any feedback,

==TIA.


0 new messages