ball-hours

[josephkk]

On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com>
wrote:

I already almost don't have any left

OK, forget about your head, then. ;-)


It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.

It isn't what bit you, but what it was/is carrying. High thee hence to
medical care.

?-/

 

[josephkk]

On Sat, 18 Oct 2014 00:39:08 -0400, rickman <gnuarm@gmail.com> wrote:

I didn't say sample rate, I said "bit rate". The I2S ports for the DAC
and ADC can be separate. The part has 4 ADC and 8 DAC, so the two ports
would run at different bit rates to accommodate the different number of
units. If you daisy chain the ADCs and not the DACs, they now both have
the same bit rate and you may be able too run both ADCs and DACs from
the same I2S port.


Well partially, the sample rate and the bit rate have relative constraints
but so long as the bit rate is sufficient for the sample rate, number of
converters, and data width it can work. That said, the interface(s) had
better help out with this or the design is sunk.

If the data gets from the ADCs to the processor in a timely fashion and
the outputs to the DACs are also on time the conversion strobes should do
the rest.

There have been a lot of posts and I'm sure you haven't read them all.

Actually i have read nearly all of them.

The CODEC in question has 4 ADCs and 8 DACs on two separate I2S ports.
The design will require the use of two of them, potentially daisy
chained to get 5 ADC and 2 DAC. So if nothing else changes the bit rate
is Fs * 32 * 8 while the design is only using a max of 24 bits from 5
ADC and 2 DACs at some bit width less than 32. Because of all the
unused data on the interface the bit rate will be *much* higher than the
actual useful data rate. That was my point.

And boy, did you ever belabor it.

I2S is nice if both ends support it in hardware in a way that minimizes
the CPU overhead. Many I2S interfaces assume there are two channels of
data which is what is used most of the time for standard stereo audio.
It would seem that I2S can be used in place of a TDM interface with
arbitrary frame sizes and channel counts. Not all MCUs, in fact, few
MCUs support that. BTW, with I2S the conversion strobe comes from the
frame sync. But then perhaps you are referring to using the SPI port.

It is not clear that frame sync is appropriately timed for conversion
strobe. In fact i doubt that it is.

Using SPI would get around all that as only useful data would be
transferred. But that would require using a different CODEC and in fact
might end up using single channel devices which I think Joerg does not
wish to do. KRW seems to feel SPI is *doomed* I tell you we're *doomed*
to failure. So far he has not explained any of his concerns.

I don't worry much about what krw says.

?-)

 

On Sat, 18 Oct 2014 18:11:53 -0700 (PDT), Bill Sloman
<bill.sloman@gmail.com> wrote:

On Sunday, 19 October 2014 11:59:51 UTC+11, k...@attt.bizz wrote:
On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com
wrote:
krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:50:25 -0700, Joerg <news@analogconsultants.com
wrote:
krw@attt.bizz wrote:
On Fri, 17 Oct 2014 07:58:10 -0700, Joerg <news@analogconsultants.com
wrote:
rickman wrote:
On 10/16/2014 4:22 PM, Lasse Langwadt Christensen wrote:

snip

OK, forget about your head, then. ;-)

It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.

You haven't been to Dallas, have you? ;-) Might want to get that
checked but it's probably too late to worry now.

Krw being as unpleasantly moronic as ever. There are lots of nasty diseases that are spread by biting and stinging insects, but Ebola doesn't seem to be one of them.

Slowman, as usual, just *has* to prove what a dumbshit he is.

 

[Jeff Liebermann]

On Sat, 18 Oct 2014 19:38:34 +0000 (UTC), mroberds@att.net wrote:

I wonder what riders in China, Japan, or the Netherlands pay - places
where lots of people commute on bikes.(...)

I buy tires according to NBC (Nothing But the Cheapest):
<http://www.nashbar.com/bikes/CategoryDisplay?storeId=10053&catalogId=10052&langId=-1&orderBy=4&searchTerm=&beginIndex=0&pageSize=32&categoryId=204729&metaData=>
The first page shows 32 tires costing $9 to $18. Cheap enough,
methinks.

Amazingly enough, both of those numbers are pretty close to the tire
pressures you would expect for each of those vehicles.

My guess(tm) is that the equality of the tire pressure and contact
pressure is the result of design optimization, and not the result of
the physics involved. Jobst covers it nicely in:
<http://www.sheldonbrown.com/brandt/rim-support.html>
More:
<http://ask.metafilter.com/143775/Why-doesnt-tire-pressure-vary-with-vehicle-payload-much>

The higher the ground pressure, the more wear and tear on the rubber.

So if bicycles are 40 psi and cars are 30 psi, is the expectation that
bike tires would last 75% as a car tire would, or 66%, or some other
factor? There needs to be a pretty good exponential in there to make
this match reality, I think.

I don't know. There are too many variables involved. It's odd that
both automobile tire under inflation and over inflation can cause
blowouts. Too low and the shape of the contact patch changes enough
that the load is carries by the circumference of the patch, resulting
in the pressure being concentrated in a smaller area. Too high and
normal air heating causes the pressure to exceed the burst limit.
There's also heating caused by the flexing of the rubber, which is
higher for underinflated tires. Hot tires wear more quickly, so
anything that overheats the tread is going to cause more wear. When
we talk about "rolling resistance", we're actually talking about
converting forward motion (energy) into friction, which is then
converted to heat. The higher the rolling friction, the more heat
goes into the tires (ignoring the tiny wheel bearing losses). My
problem is I don't know how to conglomerate all these factors, and
probably some that I've neglected, into a model that represents
optimum tire design and operating conditions. Sorry.

Also, if you drove an automobile on what you liberally define as a
"trail", the tires would last about the same as if your mtn bike
tires had the same tread thickness as an automobile tire.

Bike tires seem to have more failure mechanisms than just the tread
wearing away. Getting a hole in the tube seems to be pretty popular,
a problem cars stopped having in the 1960s or so.

Ahem. My office is near the freeway offramp when entering the city of
Santa Cruz. As a result, drivers with flat or nearly flat tires tend
to migrate towards my office, where they know they can get assistance
on weekends and after dark. I have an air compressor, patch kit, plug
kit, bead breaker, rubber cement, etc. My guess(tm) is that I plug
about 2 to 4 tires per month. Most a caused by nails, screws, and
scrap metal penetrating the cord through the anti-hydroplane grooves,
not through the tread. In about half the junk I've removed from
tires, I believe there would not have been penetration if it had hit
the tread, instead of the spaces in between the tread pattern.

Modern tires use unusual tread patterns not to offer better hydroplane
resistance, but to use less rubber, which offers less contact area,
which offers less tire life. For a given size and construction of
tire, the contact patch outline does not change with tread pattern,
but the actual tread contact area does change. An elaborate tread
pattern could easily have half the tread contact area of a tire with
fewer anti-hydroplane grooves. The result is more tire wear.

I bought a durometer meter (Shores-A) on eBay a few years ago.
<http://www.ebay.com/itm/141250852613>
I've been testing various tires for hardness and comparing them with
the tread wear using a depth gauge. In general, the harder tires last
longer and softer tires give a better and more comfortable ride. I've
noticed that the more elaborate tread patterns seem to use a slightly
harder rubber, probably to compensate for the loss of contact area.
I'm not sure about this because I've only tested about 4 different
tire types.

If bicycle tires were built in the same manner as automobile tires,
using multiple layers of Kevlar or steel wire reinforcement, it might
be more puncture resistant. The problem is that they would weigh far
too much to be considered useful. The rotating mass of the wheels are
far more important than the total weight of the machine, so this is
not a trivial consideration. If weight is NOT a major consideration,
solid rubber foam tires work nicely. For example:
<http://www.airlesstiresnow.com/Bicycle-Tires_c_172.html>
<http://www.livestrong.com/article/222689-airless-bicycle-tires-review/>
<http://bicycleuniverse.info/eqp/flatfreetires.html>

I have mine mounted on my rear rack, which doesn't work because
anything I carry either blocks the light, or knocks it over. Same
problem with mounting it on the back of the saddle.

Tow-truck style lights on a cord with magnets or ropes or something to
attach them?

Magnets won't work on an aluminum frame, may scratch the paint on a
steel frame, and generally needs a flat surface to attach. Magnet and
ty-wrap mountings are also susceptible to theft problems.

Looking at some of those slopes, I think Joerg would do good with a
rechargeable battery and regenerative brakes.

Doesn't work if he spends most of the ride airborne.

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

 

[josephkk]

On Sat, 18 Oct 2014 09:23:34 -0700, Joerg <news@analogconsultants.com>
wrote:

For $60 I can buy a tire for my car that is rated at 1100 pounds max,
supports maybe 700 pounds statically, doesn't need a tube, will run at
over 70 mph without complaint, go over concrete, asphalt, gravel, or
dirt, not need air unless the weather changes a lot, and last 50,000+
miles.


That is my gripe all along. Bicycle stuff is inferior to car stuff.
Since you can't get a car-quality tire for bicycles I so far resorted to
buying a brand from Thailand (Vee Rubber) which costs $17/tire versus
$50/tire and will probably also last 700 miles or so. Cuts that cost
down to a 3rd.

I suggest you also consider the size of the contact patch (where the
rubber meets the road) and tire thickness in life calculations.

?-)

 

[Joerg]

josephkk wrote:

On Sat, 18 Oct 2014 09:23:34 -0700, Joerg <news@analogconsultants.com
wrote:

For $60 I can buy a tire for my car that is rated at 1100 pounds max,
supports maybe 700 pounds statically, doesn't need a tube, will run at
over 70 mph without complaint, go over concrete, asphalt, gravel, or
dirt, not need air unless the weather changes a lot, and last 50,000+
miles.

That is my gripe all along. Bicycle stuff is inferior to car stuff.
Since you can't get a car-quality tire for bicycles I so far resorted to
buying a brand from Thailand (Vee Rubber) which costs $17/tire versus
$50/tire and will probably also last 700 miles or so. Cuts that cost
down to a 3rd.


I suggest you also consider the size of the contact patch (where the
rubber meets the road) and tire thickness in life calculations.

It's not much different than on a motorcycle where tires last many times
longer. My dream would be to find a dirt bike front tire that fits on
the 29" rims of my mountain bike. That should easily last 3000-5000
miles and also offer more traction in the back.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

Jeff Liebermann wrote:

On Sat, 18 Oct 2014 09:23:34 -0700, Joerg <news@analogconsultants.com
wrote:

That is my gripe all along. Bicycle stuff is inferior to car stuff.
Since you can't get a car-quality tire for bicycles I so far resorted to
buying a brand from Thailand (Vee Rubber) which costs $17/tire versus
$50/tire and will probably also last 700 miles or so. Cuts that cost
down to a 3rd.

http://en.wikipedia.org/wiki/Ground_pressure
Bicycle tires push down on the road at about 40 psi. Racing bicycles
about 90 psi. Automobiles about 30 psi. That's NOT the tire pressure
but the weight of the vehicle divided by the contact patch area
divided again by the number of tires.
http://en.wikipedia.org/wiki/Contact_patch
The higher the ground pressure, the more wear and tear on the rubber.

Mountain bike and car tires are very similar in that respect, that does
not at all explain the difference.

Also, if you drove an automobile on what you liberally define as a
"trail", the tires would last about the same as if your mtn bike tires
had the same tread thickness as an automobile tire.

My SUV now has over 75000 miles on it and it is over 17 years old. A
serious percentage of those miles were on similar "roads" except that it
had to work hard than my bike msot of te time, for example hauling half
a ton of firewood over gnarly bush roads. The ony reason I bought a new
set of tires after 12 years was age, I didn't want to risk a blowout at
freeway speeds. They still had 50% tread. Try that on a bicycle.

I only use the car when I have to and the reason is mostly hauling a lot
of heavy stuff.

I am still looking for a good rear light. Ideally hardwired to a central
battery. Or at least not with that flimsy clip-in mount. The clip
thingies won't even last one ride in the mountains, that whole concept
is junk.

https://www.google.com/search?q=led+road+flasher&tbm=isch
Don't forget the turn signal indicators.

The big problem seems to be where to mount the rear lights. I have
mine mounted on my rear rack, which doesn't work because anything I
carry either blocks the light, or knocks it over. Same problem with
mounting it on the back of the saddle. Clipping it to my back pocket
worked, until it fell off. The best compromise I've seen are two
arms, one for each side of the bicycle, and attached near the rear
axle. That might to work for your style of busy riding, but it seems
practical for street riding.

I have them all on the seat tubes. On a FS MTB that is the only option
because anywhere on the rear wheel mount it would have to be fully
mil-spec or won't survive the first few miles.

Once I can semi-retire and have more free time (some day ...) I want to
build a small box bolted to the seat tube that contains a Li-Ion
battery, some electronics, a uC, a switch and an angle-adjustable very
bright rear light. The bicycle industry is not capable of such ideas :-(

I also want to try a light on the down tube or seat tube:
http://rockthebike.com/the-down-low-glow/

Neat. I like the red bar.

Another idea for the rear rack:
http://rockthebike.com/sites/rockthebike.com/files/imagecache/gallery_image-1_and_one_third_ratio-450px_width/IMG_0019.preview.jpg

That's the one/ But can't use it on MTB. One could use a shorter one
mounted vertically to the seat tube though.

For your style of riding, I suggest a Gorilla glass tube, armored
cable, and an explosion proof battery.

Yeah, has to be almost mil-spec. All those lights with flimsy clip-in
holders are just junk, useless. Even on my road bike because that isn't
always enjoying smooth blacktop either.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sat, 18 Oct 2014 09:31:47 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/17/2014 8:27 PM, Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:

The Blackfin has to work because Lasse showed that they sell a board for

it with their audio codecs on there. That would be like hauling the

garbage pail with a 5-ton truck but maybe it has to be that way.
BTW, I wouldn't put 100% faith in the idea that the I2S *has* to work.
Certainly this approach has the best chance of working the way you want.
Just make sure the eval board will work in daisy chained mode with the
two devices. TI isn't the only company to ship product with errata. I
had a design scrapped once because the chip maker had a fatal flaw in
the device and had no intention of fixing it.

They claim in the manual that they can address 16 channels in TDM mode
if you throttle the clock to 48kHz which is all we need. At 96kHz it's
still eight channels.

That does make the bit clock 24.576MHz (assuming 32b samples). Point
to point, that's not too difficult but it can be problematic in large
systems. The bit clocks tend to have high edge rates so can cause
problems with EMI.

It's not a problem if you ac-terminate properly and/or bury the traces.
The trick is to always know and design towards the trace impedance. I've
run north of 200MHz clear across teh large backplane and family-pizza
size boards in ultrasound machines which all had to pass class B.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sat, 18 Oct 2014 09:26:58 -0700, Joerg <news@analogconsultants.com
wrote:

Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:
rickman wrote:

On 10/17/2014 4:03 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 2:40 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 10:20 AM, Joerg wrote:
rickman wrote:
On 10/16/2014 8:56 PM, Joerg wrote:
krw@attt.bizz wrote:

[...]

but I don't know that you can get the blackfin in anything with pins,
and it is expensive, you could probably get an M4 and an FPGA for the
same

They come in QFP which is what I'd prefer.

I prefer QFNs (space). The ones with the wettable flanks make for
easy hand soldering, too.

I am a striong believer in compliance. A completely rigid structure on
top of FR4 which by nature is flexible is IMHO just not a good thing.

I'd be tempted at the FPGA solution because I know that it would be
quick and easy to do and is guaranteed to work.

Right now I am looking for the solution that require the leat in
engineering effort. Price is not very important as long as it isn't
outrageous. So even $100 more is fine.

Holy crap! I thought I'd never hear *you* say that! $100 is a
fortune for this stuff. You're in high-end SHARC territory, with the
ADCs thrown in, too.

This is a feasibility project but there must be a path to later turn it
into a marketable product where cost begins to matter. So even a fancy
BGA is (grudgingly ...) acceptable now, just not long term.

I think you mentioned something about the DAC being driven with a DDS
that would also be very easy to off-load to the FPGA

One of the DACs could be replaced by a DDS which is addressed via SPI.

Why?

Only in case we'd run out of I2S resources. Which now with the Blackfin
or the ADSP-CM40x series doesn't look like a problem anymore.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sat, 18 Oct 2014 12:16:10 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:27:55 -0700, Joerg <news@analogconsultants.com
wrote:

Lasse Langwadt Christensen wrote:
Den fredag den 17. oktober 2014 22.12.44 UTC+2 skrev Joerg:
Lasse Langwadt Christensen wrote:

Den fredag den 17. oktober 2014 20.45.44 UTC+2 skrev rickman:
On 10/17/2014 10:43 AM, Joerg wrote:
krw@attt.bizz wrote:
Shouldn't be a problem. Look at a Blackfin.
I will, especially since they come in QFP packages. They also
do not have I2S/TDM but at 400-600MHz they should be able to
bang it all out on the SPI.
http://www.analog.com/static/imported-files/data_sheets/ADSP-BF531_BF532_BF533.pdf

You need to read the data sheet more carefully. SERIAL PORTS
(SPORTs) The ADSP-BF531/ADSP-BF532/ADSP-BF533 processors
incorporate two dual-channel synchronous serial ports (SPORT0
and SPORT1) for serial and multiprocessor commu- nications. The
SPORTs suppo rt the following features: ?I2S capable operation.
This may or may not support your TDM mode of operation, but it
needs to be looked at.
We've used BF537 and the sports can do almost everything if you
know how to decipher the datasheet you can get a plugin board for
the BF537 development board with two AD1939s
http://www.mouser.com/ds/2/609/Blackfin_audio_extender_man_rev.2.2-256212.pdf

Aha! Thanks, I sure will check that out. If there is a demo or
plug-in

board that means it can do that and the doc says 16 channels at
48kHz.

Which is just fine for us.

The blackfin is ok, I don't know how well GCC etc. works with it.
I've only used VisualDSP, I think it is ~$4K for visualDSP with a
debugger

A bit painful but ok as long as they don't force us into a strangehold
contract where you must pay a regular maintenance tax.
I don't know the extent of your requirements but the signal processing
portion (including control of the CODECs) should be able to be done in
Sigma Studio (all GUI).

That's where I'd leave it to our SW guy. I am not at all experienced in
compiler suites although I've used some, and only because I had to.

Sigma Studio is drag-n-drop. If you need to do some uC sorts of
control stuff, the compiler is likely needed. Sigma Studio will do
the DSP stuff and control the CODECs, easily enough.

That sounds enticing. There will always be control stuff, in our case a
serial port accessed display and, of course, control buttons. Plus
calibration routines and whatnot.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:50:25 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 07:58:10 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/16/2014 4:22 PM, Lasse Langwadt Christensen wrote:
The STM32f429 will do, it has two dual Serial audio interfaces so
it'll run
two ad1939 with no trickery.

It'll do the 256 bits, 8x32bit slots (in/out) shown in the AD1939
datasheet
and each slot can optionally be set as unused so it'll output zero for
that
slot
I don't think this will do it. The ADC and DAC interfaces run with
different bit rates so it will need three I2S interfaces. Maybe the ADC
interface can be daisy chained to give 8 channels and the DAC interface
not chained which could ideally work off of one I2S.

Not really. My plan is to at least initially run them all at the same
standard 48ksps.
That's a perfect environment for I2S. Forget SPI. You won't have any
hair left.
I already almost don't have any left
OK, forget about your head, then. ;-)

It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.

You haven't been to Dallas, have you? ;-) Might want to get that
checked but it's probably too late to worry now.

Haven't been there in years but I like listening to West African music.

The fever is gone but it's still all red down there and when I get out
of bed the first 10-20 steps on that leg hurt like heck, then the pain
always vanishes completely until I get up from a horizontal position the
next time.

Meaning no mountain biking today. Hurumph, grumble.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

josephkk wrote:

On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com
wrote:

I already almost don't have any left
OK, forget about your head, then. ;-)

It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.

It isn't what bit you, but what it was/is carrying. High thee hence to
medical care.

If it doesn't get better I will. But all they usually do is prescribe
some pain pills and maybe antibiotics.

--
Regards, Joerg

http://www.analogconsultants.com/

 

On Sun, 19 Oct 2014 08:00:21 -0700, Joerg <news@analogconsultants.com>
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 09:31:47 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/17/2014 8:27 PM, Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:

The Blackfin has to work because Lasse showed that they sell a board for

it with their audio codecs on there. That would be like hauling the

garbage pail with a 5-ton truck but maybe it has to be that way.
BTW, I wouldn't put 100% faith in the idea that the I2S *has* to work.
Certainly this approach has the best chance of working the way you want.
Just make sure the eval board will work in daisy chained mode with the
two devices. TI isn't the only company to ship product with errata. I
had a design scrapped once because the chip maker had a fatal flaw in
the device and had no intention of fixing it.

They claim in the manual that they can address 16 channels in TDM mode
if you throttle the clock to 48kHz which is all we need. At 96kHz it's
still eight channels.

That does make the bit clock 24.576MHz (assuming 32b samples). Point
to point, that's not too difficult but it can be problematic in large
systems. The bit clocks tend to have high edge rates so can cause
problems with EMI.


It's not a problem if you ac-terminate properly and/or bury the traces.
The trick is to always know and design towards the trace impedance. I've
run north of 200MHz clear across teh large backplane and family-pizza
size boards in ultrasound machines which all had to pass class B.

Sure, you can easily do Gb across boards, too, but that's a
predictable system with dynamic compensation at both ends. The edge
rates on many of these parts are much faster than you might expect,
they're single ended, and there is no dynamic impedance control.
tolerances can eat you alive. Just watch it, high speed I2S can get
you into trouble.

 

On Sun, 19 Oct 2014 08:04:38 -0700, Joerg <news@analogconsultants.com>
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 09:26:58 -0700, Joerg <news@analogconsultants.com
wrote:

Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:
rickman wrote:

On 10/17/2014 4:03 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 2:40 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 10:20 AM, Joerg wrote:
rickman wrote:
On 10/16/2014 8:56 PM, Joerg wrote:
krw@attt.bizz wrote:

[...]


but I don't know that you can get the blackfin in anything with pins,
and it is expensive, you could probably get an M4 and an FPGA for the
same

They come in QFP which is what I'd prefer.

I prefer QFNs (space). The ones with the wettable flanks make for
easy hand soldering, too.


I am a striong believer in compliance. A completely rigid structure on
top of FR4 which by nature is flexible is IMHO just not a good thing.

I've never seen problems with BGAs. We did have problems with QFNs
for a while but it was self-inflicted. Wetable flanks would have
solved those problems.

I'd be tempted at the FPGA solution because I know that it would be
quick and easy to do and is guaranteed to work.

Right now I am looking for the solution that require the leat in
engineering effort. Price is not very important as long as it isn't
outrageous. So even $100 more is fine.

Holy crap! I thought I'd never hear *you* say that! $100 is a
fortune for this stuff. You're in high-end SHARC territory, with the
ADCs thrown in, too.


This is a feasibility project but there must be a path to later turn it
into a marketable product where cost begins to matter. So even a fancy
BGA is (grudgingly ...) acceptable now, just not long term.

If it's just a feasibility project, why all the hand-wringing about
BGAs? I don't understand your problem anyway, but...

I think you mentioned something about the DAC being driven with a DDS
that would also be very easy to off-load to the FPGA

One of the DACs could be replaced by a DDS which is addressed via SPI.

Why?


Only in case we'd run out of I2S resources. Which now with the Blackfin
or the ADSP-CM40x series doesn't look like a problem anymore.

You can get CODECs with single I2S ports. The ADC would then be free.

 

On Sun, 19 Oct 2014 08:06:26 -0700, Joerg <news@analogconsultants.com>
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 12:16:10 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:27:55 -0700, Joerg <news@analogconsultants.com
wrote:

Lasse Langwadt Christensen wrote:
Den fredag den 17. oktober 2014 22.12.44 UTC+2 skrev Joerg:
Lasse Langwadt Christensen wrote:

Den fredag den 17. oktober 2014 20.45.44 UTC+2 skrev rickman:
On 10/17/2014 10:43 AM, Joerg wrote:
krw@attt.bizz wrote:
Shouldn't be a problem. Look at a Blackfin.
I will, especially since they come in QFP packages. They also
do not have I2S/TDM but at 400-600MHz they should be able to
bang it all out on the SPI.
http://www.analog.com/static/imported-files/data_sheets/ADSP-BF531_BF532_BF533.pdf

You need to read the data sheet more carefully. SERIAL PORTS
(SPORTs) The ADSP-BF531/ADSP-BF532/ADSP-BF533 processors
incorporate two dual-channel synchronous serial ports (SPORT0
and SPORT1) for serial and multiprocessor commu- nications. The
SPORTs suppo rt the following features: ?I2S capable operation.
This may or may not support your TDM mode of operation, but it
needs to be looked at.
We've used BF537 and the sports can do almost everything if you
know how to decipher the datasheet you can get a plugin board for
the BF537 development board with two AD1939s
http://www.mouser.com/ds/2/609/Blackfin_audio_extender_man_rev.2.2-256212.pdf

Aha! Thanks, I sure will check that out. If there is a demo or
plug-in

board that means it can do that and the doc says 16 channels at
48kHz.

Which is just fine for us.

The blackfin is ok, I don't know how well GCC etc. works with it.
I've only used VisualDSP, I think it is ~$4K for visualDSP with a
debugger

A bit painful but ok as long as they don't force us into a strangehold
contract where you must pay a regular maintenance tax.
I don't know the extent of your requirements but the signal processing
portion (including control of the CODECs) should be able to be done in
Sigma Studio (all GUI).

That's where I'd leave it to our SW guy. I am not at all experienced in
compiler suites although I've used some, and only because I had to.

Sigma Studio is drag-n-drop. If you need to do some uC sorts of
control stuff, the compiler is likely needed. Sigma Studio will do
the DSP stuff and control the CODECs, easily enough.


That sounds enticing. There will always be control stuff, in our case a
serial port accessed display and, of course, control buttons. Plus
calibration routines and whatnot.

A lot of that can be done in Sigma Studio, too. I'm certainly not an
expert, though. If you limit the parts to ADI, SS is a piece of cake.
Other parts are certainly usable, just need more setup.

 

On Sat, 18 Oct 2014 21:51:58 -0700, josephkk
<joseph_barrett@sbcglobal.net> wrote:

On Sat, 18 Oct 2014 00:39:08 -0400, rickman <gnuarm@gmail.com> wrote:


I didn't say sample rate, I said "bit rate". The I2S ports for the DAC
and ADC can be separate. The part has 4 ADC and 8 DAC, so the two ports
would run at different bit rates to accommodate the different number of
units. If you daisy chain the ADCs and not the DACs, they now both have
the same bit rate and you may be able too run both ADCs and DACs from
the same I2S port.


Well partially, the sample rate and the bit rate have relative constraints
but so long as the bit rate is sufficient for the sample rate, number of
converters, and data width it can work. That said, the interface(s) had
better help out with this or the design is sunk.

If the data gets from the ADCs to the processor in a timely fashion and
the outputs to the DACs are also on time the conversion strobes should do
the rest.

There have been a lot of posts and I'm sure you haven't read them all.

Actually i have read nearly all of them.

The CODEC in question has 4 ADCs and 8 DACs on two separate I2S ports.
The design will require the use of two of them, potentially daisy
chained to get 5 ADC and 2 DAC. So if nothing else changes the bit rate
is Fs * 32 * 8 while the design is only using a max of 24 bits from 5
ADC and 2 DACs at some bit width less than 32. Because of all the
unused data on the interface the bit rate will be *much* higher than the
actual useful data rate. That was my point.

And boy, did you ever belabor it.

I2S is nice if both ends support it in hardware in a way that minimizes
the CPU overhead. Many I2S interfaces assume there are two channels of
data which is what is used most of the time for standard stereo audio.
It would seem that I2S can be used in place of a TDM interface with
arbitrary frame sizes and channel counts. Not all MCUs, in fact, few
MCUs support that. BTW, with I2S the conversion strobe comes from the
frame sync. But then perhaps you are referring to using the SPI port.

It is not clear that frame sync is appropriately timed for conversion
strobe. In fact i doubt that it is.

Using SPI would get around all that as only useful data would be
transferred. But that would require using a different CODEC and in fact
might end up using single channel devices which I think Joerg does not
wish to do. KRW seems to feel SPI is *doomed* I tell you we're *doomed*
to failure. So far he has not explained any of his concerns.

I don't worry much about what krw says.

Oh, good grief. Ricky is trying to force a pig to be a cow. Your
application screams I2S but Ricky is screaming SPI anyhow. You've
even "settled" on the CODEC, which is I2S. Thinking that SPI will
somehow be easier is just nuts but that's SED's Ricky.

 

On Sun, 19 Oct 2014 08:12:27 -0700, Joerg <news@analogconsultants.com>
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:50:25 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 07:58:10 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/16/2014 4:22 PM, Lasse Langwadt Christensen wrote:
The STM32f429 will do, it has two dual Serial audio interfaces so
it'll run
two ad1939 with no trickery.

It'll do the 256 bits, 8x32bit slots (in/out) shown in the AD1939
datasheet
and each slot can optionally be set as unused so it'll output zero for
that
slot
I don't think this will do it. The ADC and DAC interfaces run with
different bit rates so it will need three I2S interfaces. Maybe the ADC
interface can be daisy chained to give 8 channels and the DAC interface
not chained which could ideally work off of one I2S.

Not really. My plan is to at least initially run them all at the same
standard 48ksps.
That's a perfect environment for I2S. Forget SPI. You won't have any
hair left.
I already almost don't have any left
OK, forget about your head, then. ;-)

It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.

You haven't been to Dallas, have you? ;-) Might want to get that
checked but it's probably too late to worry now.


Haven't been there in years but I like listening to West African music.

The fever is gone but it's still all red down there and when I get out
of bed the first 10-20 steps on that leg hurt like heck, then the pain
always vanishes completely until I get up from a horizontal position the
next time.

If it's still bothering you that much, it might be a good idea to seek
help. The time to do it was when the fever started, though. A day
later, it's not now likely to kill you. ;-) I know people who have
had nasty encounters with spiders and lost good chunks out of their
limbs. Muscle doesn't grow back.

>Meaning no mountain biking today. Hurumph, grumble.

 

[Joerg]

krw@attt.bizz wrote:

On Sun, 19 Oct 2014 08:00:21 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 09:31:47 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/17/2014 8:27 PM, Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:

The Blackfin has to work because Lasse showed that they sell a board for

it with their audio codecs on there. That would be like hauling the

garbage pail with a 5-ton truck but maybe it has to be that way.
BTW, I wouldn't put 100% faith in the idea that the I2S *has* to work.
Certainly this approach has the best chance of working the way you want.
Just make sure the eval board will work in daisy chained mode with the
two devices. TI isn't the only company to ship product with errata. I
had a design scrapped once because the chip maker had a fatal flaw in
the device and had no intention of fixing it.

They claim in the manual that they can address 16 channels in TDM mode
if you throttle the clock to 48kHz which is all we need. At 96kHz it's
still eight channels.
That does make the bit clock 24.576MHz (assuming 32b samples). Point
to point, that's not too difficult but it can be problematic in large
systems. The bit clocks tend to have high edge rates so can cause
problems with EMI.

It's not a problem if you ac-terminate properly and/or bury the traces.
The trick is to always know and design towards the trace impedance. I've
run north of 200MHz clear across teh large backplane and family-pizza
size boards in ultrasound machines which all had to pass class B.

Sure, you can easily do Gb across boards, too, but that's a
predictable system with dynamic compensation at both ends. The edge
rates on many of these parts are much faster than you might expect,
they're single ended, and there is no dynamic impedance control.
tolerances can eat you alive. Just watch it, high speed I2S can get
you into trouble.

Usually only one site of a trace needs to be terminated but it has to be
precisely the trace impedance. A few gBits are of, when it goes towards
10Gits FR4 might reach the end of its usefulness for longer runs.

AC termination is often sufficient, saving power. RF effects are the
least of my worries because that field is my home turf, all I need is a
processor that can stomach the workload. So far the ADSP-CM40x looks the
most enticing and if that doesn't do the trick then the Blackfin.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sun, 19 Oct 2014 08:04:38 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 09:26:58 -0700, Joerg <news@analogconsultants.com
wrote:

Lasse Langwadt Christensen wrote:
Den lřrdag den 18. oktober 2014 02.12.40 UTC+2 skrev Joerg:
rickman wrote:

On 10/17/2014 4:03 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 2:40 PM, Joerg wrote:
rickman wrote:
On 10/17/2014 10:20 AM, Joerg wrote:
rickman wrote:
On 10/16/2014 8:56 PM, Joerg wrote:
krw@attt.bizz wrote:
[...]


but I don't know that you can get the blackfin in anything with pins,
and it is expensive, you could probably get an M4 and an FPGA for the
same

They come in QFP which is what I'd prefer.
I prefer QFNs (space). The ones with the wettable flanks make for
easy hand soldering, too.

I am a striong believer in compliance. A completely rigid structure on
top of FR4 which by nature is flexible is IMHO just not a good thing.

I've never seen problems with BGAs. We did have problems with QFNs
for a while but it was self-inflicted. Wetable flanks would have
solved those problems.

I (and laptop as well as video game makers) have. The gear we are
designing here will be used in very rough environments. For example,
smacks from 4-5ft onto a tile floor are considered normal.

I've got the perfect test rig here. All I have to do at the end is tie a
few units to the rear-wheel section of my mountain bike and do a few of
my tougher local rounds. Anything that survives can be considered robust.

I'd be tempted at the FPGA solution because I know that it would be
quick and easy to do and is guaranteed to work.

Right now I am looking for the solution that require the leat in
engineering effort. Price is not very important as long as it isn't
outrageous. So even $100 more is fine.
Holy crap! I thought I'd never hear *you* say that! $100 is a
fortune for this stuff. You're in high-end SHARC territory, with the
ADCs thrown in, too.

This is a feasibility project but there must be a path to later turn it
into a marketable product where cost begins to matter. So even a fancy
BGA is (grudgingly ...) acceptable now, just not long term.

If it's just a feasibility project, why all the hand-wringing about
BGAs? I don't understand your problem anyway, but...

As I said for feasibility BGA is ok but we must show a path towards
series production. There I would not advocate BGA for various reasons.

I think you mentioned something about the DAC being driven with a DDS
that would also be very easy to off-load to the FPGA

One of the DACs could be replaced by a DDS which is addressed via SPI.
Why?

Only in case we'd run out of I2S resources. Which now with the Blackfin
or the ADSP-CM40x series doesn't look like a problem anymore.

You can get CODECs with single I2S ports. The ADC would then be free.

AFAIU many can be daisy-chained. My preference would be the AD1939 or
AD1938.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Joerg]

krw@attt.bizz wrote:

On Sun, 19 Oct 2014 08:12:27 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Sat, 18 Oct 2014 12:32:49 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 17:50:25 -0700, Joerg <news@analogconsultants.com
wrote:

krw@attt.bizz wrote:
On Fri, 17 Oct 2014 07:58:10 -0700, Joerg <news@analogconsultants.com
wrote:

rickman wrote:
On 10/16/2014 4:22 PM, Lasse Langwadt Christensen wrote:
The STM32f429 will do, it has two dual Serial audio interfaces so
it'll run
two ad1939 with no trickery.

It'll do the 256 bits, 8x32bit slots (in/out) shown in the AD1939
datasheet
and each slot can optionally be set as unused so it'll output zero for
that
slot
I don't think this will do it. The ADC and DAC interfaces run with
different bit rates so it will need three I2S interfaces. Maybe the ADC
interface can be daisy chained to give 8 channels and the DAC interface
not chained which could ideally work off of one I2S.

Not really. My plan is to at least initially run them all at the same
standard 48ksps.
That's a perfect environment for I2S. Forget SPI. You won't have any
hair left.
I already almost don't have any left
OK, forget about your head, then. ;-)
It's dizzy right now. Something stung or bit me in the leg which I
didn't find out until I almost fell asleep on a mountain bike ride
yesterday. Back home I checked the temp and found that I ran a fever.
Crashed into bed a couple hours later and dozed off for 12h. Still not
quite there yet. If I just knew what bit me.
You haven't been to Dallas, have you? ;-) Might want to get that
checked but it's probably too late to worry now.

Haven't been there in years but I like listening to West African music.

The fever is gone but it's still all red down there and when I get out
of bed the first 10-20 steps on that leg hurt like heck, then the pain
always vanishes completely until I get up from a horizontal position the
next time.

If it's still bothering you that much, it might be a good idea to seek
help. The time to do it was when the fever started, though. A day
later, it's not now likely to kill you. ;-) I know people who have
had nasty encounters with spiders and lost good chunks out of their
limbs. Muscle doesn't grow back.

The fever is gone, my wife says the redness has receded some. There is
still pain but only right after I get out of a horizontal position (bed,
couch) onto my left foot and then only for 10-20 steps. Doesn't seem too
bad.

Muscle actually does seem to grow. After I started intense mountain
biking over a year ago the leg muscles seem to have almost doubled in
size and took on a different shape. Also grew in horsepower when going
up steep hills that I used to have to push the bike.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/