robert bristow-johnson <rbj@audioimagination.com> wrote:

(snip, I wrote)

>> The rules were pretty complicated to make sure everything has the
>> right ratios, but such crystals are in every NTSC color TV.
 
> i knew all of that, Glen.  including how (or why) f_H got bumped down to 
> 15734 Hz (which i think was one of the most obscure factoids) which 
> brought some of us audio guys the wonderful sampling frequency of 44.056 
> kHz.

I thought so, but decided to write it anyway.

Even more interesting is how the 4.77MHz original IBM PC came to be.

They had 5MHz processors, and could have run them at that.
The original CGA card generates NTSC signals similar to the way the
Apple II did, by switching the output at various phases of 4x
the color subcarrier frequency. (That is in addition to the RGB 
output for the IBM CGA monitor.)

Pretty much they run a shift register at 4*3579545Hz, and the 
different bit combinations will look like different phases of
the subcarrier, and so show up at specific colors.

The 8088 requires a clock with a 33% duty cycle (to meet the 
required high and low times at the rated speed) and the 8284
clock generator will divide its input by three. So, with one
crystal they could take care of both, with the 14.1818MHz signal
on one pin of the (now) ISA bus.

Later processors used different clock sources, but the ISA bus
still requires the 14.1818MHz signal in case you use a CGA card.

When I got my first 286 based AT clone around 1987, I bought a
used CGA card and used the composite video output to test
the convergence on our TV set. I had a long enough (composite)
video cable to reach the TV, and a long keyboard cable (probably
violating the standard, but it did work) to reach the next room.

-- glen

On 7/19/13 8:31 PM, glen herrmannsfeldt wrote:
> robert bristow-johnson<rbj@audioimagination.com>  wrote:
...
>
>> i remember in the olden daze that 3.579545 MHz were the cheap xtals t
>> to run your microprocessors (like the venerable MC6809) on.
>> can't imagine how they came up with that frequency.
>
> The rules were pretty complicated to make sure everything has the
> right ratios, but such crystals are in every NTSC color TV.

i knew all of that, Glen.  including how (or why) f_H got bumped down to 
15734 Hz (which i think was one of the most obscure factoids) which 
brought some of us audio guys the wonderful sampling frequency of 44.056 
kHz.

anyway, i need to be careful about Poe's Law (i thought smiley emoticons 
might protect me from it).  this isn't exactly that, but it's similar.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

claysturner@gmail.com wrote:
> On Friday, July 19, 2013 9:11:30 PM UTC-4, robert bristow-johnson wrote:

(snip)
>> i remember in the olden daze that 3.579545 MHz were the cheap xtals to 
>> run your microprocessors (like the venerable MC6809) on.  can't imagine 
>> how they came up with that frequency.

(snip)

(snip)
> Robert, the 3.57945MHz xtal is a color burst crystal (NTSC). 
> They were very cheap due to every color tv (US) having one. 
> The freq was 455 * (horizontal scan rate)/2.
> 
> The original scan rates for NTSC were 60Hz and 15750HZ. 
> When color was added the rates were changed to 59.94Hz and 
> 15734Hz. The luminance and chroma info were interleaved among 
> each other via the use of comb filters (made with a physical 
> delay line using coax cable in the early tvs). 

The reason for the odd half is that the phase alternates between
lines and between successive scans of the same line. That makes
it much less visible than it might be.

> Locking the horizontal scan rate to the color burst, 
> kept the color phase from dancing all over the place.

Yes, that, too. 

There are three requirments. 

One is an odd half multiple of the line rate, as you noted.

The line rate has to be an odd half multiple of the field rate
to make interlaced scanning work.  

Since there are an odd number of lines per frame (525), the
subcarier will naturally also be an odd half multiple of the
frame rate.

The last requirement is that the line rate be an integer fraction
of the audio carrier offset of 4.5MHz. 4.5MHz/15734, rounded to 
the nearest integer, is 286, so the new line rate is 4.5MHz/286,
or about 15734Hz. 

Also interesting is where the original line and frame rates
come from. Early power supply filtering wasn't so good, so
running the field (half the frame rate for interlace) at 60Hz
(power line synchronous) helps. Instead of dividing down from
a higher frequency (with a PLL) like we might do now, they
mutliplied up from 60Hz using vacuum tube based multipliers.

   525 = 3 * 5 * 5 * 7

-- glen

robert bristow-johnson <rbj@audioimagination.com> wrote:
> On 7/19/13 5:51 PM, glen herrmannsfeldt wrote:
>> robert bristow-johnson<rbj@audioimagination.com>  wrote:

(snip)
>>> does it count if you have an on-chip oscillator (with an on-chip cap)
>>> running at some high frequency (perhaps some multiple of 7600 Hz) and
>>> you divide the output frequency down with the old standard logic?

>> No, completely different.

>> In this case, that is a reason why it might have a higher frequency
>> divided down. Also, often the power used is related to the frequency,
>> and so low power requires low frequency.
 
> power is also related to the number of transistors or gates hooked up 
> to it.  so you could have a small portion of the chip driven by a 
> high-frequency oscillator, divide the sucker down, and have the rest 
> of the chip driven by the low-frequency clock derived from the original 
> high-frequency clock.

You could. Maybe that works. Also, it sounds like they can turn
off the sampling clock when it isn't needed. It can then turn on
when appropriate motion is detected.
 
>> Watches usually use a 32768Hz crystal, as the power needed is much
>> lower than for higher frequencies.
 
> how did they arrive at that value?  :-)

A nice, but not too big, power of two.
 
> also 32 kHz sounds like a pretty big xtal.  i woulda thunked that the 
> electronics in a watch might be running at a higher clock, but i surely 
> don't know first hand.

It is interesting. The usual crystals, like used for radios and older
computer clocks are in a thickness mode, such that the period is 
proportional to the thickness of the (usually round) crystal.

The 32768Hz are called tuning fork mode. I don't know it more than
that, but the mode is very different. 
 
> i remember in the olden daze that 3.579545 MHz were the cheap xtals t
> to run your microprocessors (like the venerable MC6809) on.  
> can't imagine how they came up with that frequency.

The rules were pretty complicated to make sure everthing has the 
right ratios, but such crystals are in every NTSC color TV.

>> Maybe some trick with an op-amp and small capacitor to make it
>> look larger.
 
> still can't see why they can't just divide a high frequency down with a 
> very few gates and let all the other gates run on the slow clock.
 
They might do that. I was considering the ability not to do it.

-- geln

On Friday, July 19, 2013 9:11:30 PM UTC-4, robert bristow-johnson wrote:
> On 7/19/13 5:51 PM, glen herrmannsfeldt wrote:
> 
> > robert bristow-johnson<rbj@audioimagination.com>  wrote:
> 
> >
> 
> > (snip, I wrote)
> 
> >
> 
> >>> I don't think you can build an on-chip (no external capacitor)
> 
> >>> oscillator down to 400Hz, but for low power reasons you want it
> 
> >>> as low as possible.   LCM(380,400) is 7600Hz if that helps.
> 
> >
> 
> >> does it count if you have an on-chip oscillator (with an on-chip cap)
> 
> >> running at some high frequency (perhaps some multiple of 7600 Hz) and
> 
> >> you divide the output frequency down with the old standard logic?
> 
> >
> 
> > No, completely different.
> 
> >
> 
> > In this case, that is a reason why it might have a higher frequency
> 
> > divided down. Also, often the power used is related to the frequency,
> 
> > and so low power requires low frequency.
> 
> 
> 
> power is also related to the number of transistors or gates hooked up to 
> 
> it.  so you could have a small portion of the chip driven by a 
> 
> high-frequency oscillator, divide the sucker down, and have the rest of 
> 
> the chip driven by the low-frequency clock derived from the original 
> 
> high-frequency clock.
> 
> 
> 
> >
> 
> > Watches usually use a 32768Hz crystal, as the power needed is much
> 
> > lower than for higher frequencies.
> 
> 
> 
> how did they arrive at that value?  :-)
> 
> 
> 
> also 32 kHz sounds like a pretty big xtal.  i woulda thunked that the 
> 
> electronics in a watch might be running at a higher clock, but i surely 
> 
> don't know first hand.
> 
> 
> 
> i remember in the olden daze that 3.579545 MHz were the cheap xtals to 
> 
> run your microprocessors (like the venerable MC6809) on.  can't imagine 
> 
> how they came up with that frequency.
> 
> 
> 
> > Maybe some trick with an op-amp and small capacitor to make it
> 
> > look larger.
> 
> 
> 
> still can't see why they can't just divide a high frequency down with a 
> 
> very few gates and let all the other gates run on the slow clock.
> 
> 
> 
> but then i hadn't done anything with hardware since 1991.
> 
> 
> 
> -- 
> 
> 
> 
> r b-j                  rbj@audioimagination.com
> 
> 
> 
> "Imagination is more important than knowledge."

Robert, the 3.57945MHz xtal is a color burst crystal (NTSC). They were very cheap due to every color tv (US) having one. The freq was 455 * (horizontal scan rate)/2.

The original scan rates for NTSC were 60Hz and 15750HZ. When color was added the rates were changed to 59.94Hz and 15734Hz. The luminance and chroma info were interleaved among each other via the use of comb filters (made with a physical delay line using coax cable in the early tvs). Locking the horizontal scan rate to the color burst, kept the color phase from dancing all over the place.


Clay
 

On 7/19/13 5:51 PM, glen herrmannsfeldt wrote:
> robert bristow-johnson<rbj@audioimagination.com>  wrote:
>
> (snip, I wrote)
>
>>> I don't think you can build an on-chip (no external capacitor)
>>> oscillator down to 400Hz, but for low power reasons you want it
>>> as low as possible.   LCM(380,400) is 7600Hz if that helps.
>
>> does it count if you have an on-chip oscillator (with an on-chip cap)
>> running at some high frequency (perhaps some multiple of 7600 Hz) and
>> you divide the output frequency down with the old standard logic?
>
> No, completely different.
>
> In this case, that is a reason why it might have a higher frequency
> divided down. Also, often the power used is related to the frequency,
> and so low power requires low frequency.

power is also related to the number of transistors or gates hooked up to 
it.  so you could have a small portion of the chip driven by a 
high-frequency oscillator, divide the sucker down, and have the rest of 
the chip driven by the low-frequency clock derived from the original 
high-frequency clock.

>
> Watches usually use a 32768Hz crystal, as the power needed is much
> lower than for higher frequencies.

how did they arrive at that value?  :-)

also 32 kHz sounds like a pretty big xtal.  i woulda thunked that the 
electronics in a watch might be running at a higher clock, but i surely 
don't know first hand.

i remember in the olden daze that 3.579545 MHz were the cheap xtals to 
run your microprocessors (like the venerable MC6809) on.  can't imagine 
how they came up with that frequency.

> Maybe some trick with an op-amp and small capacitor to make it
> look larger.

still can't see why they can't just divide a high frequency down with a 
very few gates and let all the other gates run on the slow clock.

but then i hadn't done anything with hardware since 1991.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On Fri, 19 Jul 2013 16:38:28 -0700, robert bristow-johnson
<rbj@audioimagination.com> wrote:

>On 7/19/13 3:10 PM, glen herrmannsfeldt wrote:
>> Randy Yates<yates@digitalsignallabs.com>  wrote:
>>
>> (snip)
>>
>>> No. They are from an LSM330 chip - one signal is the
>>> acceleration at ~400 Hz, the other is the gyro data at
>>> ~370 Hz.
>>
>> Interesting chip.
>>
>> The CAP input (C1) looks suspiciously like a clock generating
>> capacitor, but the data sheet doesn't say anything about it.
>>
>> Seems to not say much at all about the clocks.
>>
>> I don't think you can build an on-chip (no external capacitor)
>> oscillator down to 400Hz, but for low power reasons you want it
>> as low as possible.   LCM(380,400) is 7600Hz if that helps.
>
>does it count if you have an on-chip oscillator (with an on-chip cap) 
>running at some high frequency (perhaps some multiple of 7600 Hz) and 
>you divide the output frequency down with the old standard logic?

They're implemented that way very often.   Web search Integer-N and
Fractional-N Synthesizer.


>
>-- 
>
>r b-j                  rbj@audioimagination.com
>
>"Imagination is more important than knowledge."
>
>

Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

robert bristow-johnson <rbj@audioimagination.com> wrote:

(snip, I wrote)

>> I don't think you can build an on-chip (no external capacitor)
>> oscillator down to 400Hz, but for low power reasons you want it
>> as low as possible.   LCM(380,400) is 7600Hz if that helps.
 
> does it count if you have an on-chip oscillator (with an on-chip cap) 
> running at some high frequency (perhaps some multiple of 7600 Hz) and 
> you divide the output frequency down with the old standard logic?

No, completely different.

In this case, that is a reason why it might have a higher frequency
divided down. Also, often the power used is related to the frequency,
and so low power requires low frequency. (Popular use is for handheld
devices, such as cellphones.)

Watches usually use a 32768Hz crystal, as the power needed is much
lower than for higher frequencies. 

Maybe some trick with an op-amp and small capacitor to make it
look larger. 

-- glen

Tim Wescott <tim@seemywebsite.really> writes:

> On Fri, 19 Jul 2013 17:49:18 -0400, Randy Yates wrote:
>
>> I just wanted to say to everyone that's responded, thank you! I have
>> read all your posts, even though I have only responded to a few, and I
>> appreciate all of them.
>> 
>> --Randy
>
> You don't think that's going to stop us, do you?

I hope not! :)
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

On 7/19/13 3:10 PM, glen herrmannsfeldt wrote:
> Randy Yates<yates@digitalsignallabs.com>  wrote:
>
> (snip)
>
>> No. They are from an LSM330 chip - one signal is the
>> acceleration at ~400 Hz, the other is the gyro data at
>> ~370 Hz.
>
> Interesting chip.
>
> The CAP input (C1) looks suspiciously like a clock generating
> capacitor, but the data sheet doesn't say anything about it.
>
> Seems to not say much at all about the clocks.
>
> I don't think you can build an on-chip (no external capacitor)
> oscillator down to 400Hz, but for low power reasons you want it
> as low as possible.   LCM(380,400) is 7600Hz if that helps.

does it count if you have an on-chip oscillator (with an on-chip cap) 
running at some high frequency (perhaps some multiple of 7600 Hz) and 
you divide the output frequency down with the old standard logic?


-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."