replace 3.58MHz with 3.579545MHz

[ifrenide]

Hello, can anybody tell me, if i can replace 3.58MHz with 3.579545mHz?
best regards

[jeroen79]

It depends on what you want to do with it.

[ifrenide]

hi Jeroen,
I only have 3.579545Mhz crystal , I want to use it with Zilog microcontroller.
best regards

[mcovington]

Yes.  The person who specified 3.58 probably meant 3.579545 in the first place.  That is a standard crystal for American color TVs (old style, NTSC).  It is commonly called a 3.58-MHz crystal but is actually 3.579545 plus or minus some tolerance.

[ifrenide]

thank you Mcovington,but I saw some pcbs with 3.58MHz ceramic resonator or passive crystal with exacetly 3.58MHz,and other pcbs with 3.579545MHz,without problem.So when we need exact value and when we can use close value?

[groinksan]

I would hit the datasheet for the Zilog microprocessor. It should tell you the clock rates accepted. Personally, I think the 3.579545 MHz should be just fine. If the chip is used just to execute code (6502 or Z80 for example,) or virtually any analog-based application, then the precise value won't matter. If it was running something like a timing device, or some other application that requires precision, then theoretically the value would make a difference, and you'd then need to modify other components to match the speed. The thing to take away from this is basically how that Zilog chip will function, and then go from there.

[schmitt trigger]

Are you absolutely certain it was a 3.580000 Mhz crystal and not a 3.579545 one? Many people were lazy and used 3.58 as shorthand.

As others have described, the 3.579545 was the NTSC chroma burst frequency, and millions and millions of those crystals were sold each year. It was by far the cheapest and most ubiquitous crystal, and many commercial products unrelated to NTSC video used it.

[mcovington]

Perhaps more to the point, it is extremely unlikely that there will be any detectable difference in how a microcontroller works with one vs. the other.

[CatalinaWOW]

As said before this was a frequency used in US TV sets.  They were made to lowest possible cost, so I really doubt that all of those decimals really mean anything, particularly if you don't have the crystal mounted in the test manner, use the test circuit and have your room controlled to the test frequency.  It is the nominal frequency. 

If you are making a countdown clock for next Christmas or retirement you might care about all of those decimal places, but in almost all applications it won't matter.

[mcovington]

I am not even sure 3.58-MHz (exact, 3.5800000 MHz) crystals have ever been made.  3.579545-MHz crystals are extremely common and are commonly called 3.58-MHz.

I have even used them to substitute for 4-MHz crystals in microprocessor circuits.

[Brumby]

This is what I would expect:

3.579545-MHz crystals are extremely common and are commonly called 3.58-MHz.

I mean, if those crystals are as common as mud and there aren't any other frequencies in that vicinity which you would come across very often (if at all), then there is going to be a shorthand version of how it gets referred to.  Think of it ... if you're in a group off electronics people familiar with this area and you started referring to this common-as-anything crystal to 6 decimal places, then you're going to waste a lot of time reciting them and all those around you are going to look at you funny.  They know what you mean - and 2 decimals is all that needs to be mentioned to make it clear.

I have even used them to substitute for 4-MHz crystals in microprocessor circuits.

A lot of the time, the precise frequency isn't all that important when driving microprocessors unless you have critical time-keeping requirements.

[T3sl4co1l]

Yeah, worst possible case, NMOS Z80, it won't know the difference.  IIRC, they start getting loopy above 5 or 6MHz and below 2 or 1MHz.  CMOS CPUs work down to DC.

Anything to do with precise timing or output signals, that's an exercise for the student.

Tim

[schmitt trigger]

One of my favorite ICs back in the 80s, was a device from National Semi, which generated a 60 Hz signal from a 3.579545 Mhz crystal.

It was a NM-something, and very useful for a timing backup for clocks.

[mcovington]

MM5369AA, I believe it was.  I used it in, among other things, a power supply for astronomical telescopes (with 60-Hz motors, accurate frequency needed) in the first edition of my book Astrophotography for the Amateur.  The chip is no longer available.  I did a PIC-based circuit later to replace it.

[schmitt trigger]

Exactly, that was the part number.

To obtain 60 Hz from an oddball frequency such as 3.579545 MHz, it alternated the frequency’s division ratio.

As such, the output wasn’t a 50% duty.

Sadly it is now obsolete, but can be emulated with any modern microcontroller.

[boffin]

One of my favorite ICs back in the 80s, was a device from National Semi, which generated a 60 Hz signal from a 3.579545 Mhz crystal.

It was a NM-something, and very useful for a timing backup for clocks.

We used to use 8870 DTMF tone decoders which also used a colorburst crystal. Always thought it was odd.

Heck, even the original IBM PC ran at a multiple of Colorburst for it's systems clock.  The main crystal was 14.318 (4x colorburst - which was also fairly common) main crystal; which was divided by 4 for the video system (3.579...), and divided by 3 for the main clock @ 4.77MHz

[edavid]

(About the MM5369)

To obtain 60 Hz from an oddball frequency such as 3.579545 MHz, it alternated the frequency’s division ratio.

No, it did not, it just divided by 59659.  You had to tweak the crystal load to get it closer to 60Hz.

As such, the output wasn’t a 50% duty.

Yes, that is correct.

Sadly it is now obsolete, but can be emulated with any modern microcontroller.

It was an old PMOS part that wouldn't run at 5V, let alone 3.3V, so it wouldn't be very useful today.

[edavid]

I am not even sure 3.58-MHz (exact, 3.5800000 MHz) crystals have ever been made.  3.579545-MHz crystals are extremely common and are commonly called 3.58-MHz.

Here is a datasheet for a 3.580MHz crystal sold by Mouser:

https://www.mouser.com/datasheet/2/741/LFXTAL032257Bulk-997159.pdf

I have no idea why it exists.

[james_s]

I've always called the colorburst crystals "3.58 MHz", it's pointless to use gratuitous precision when it's not necessary to do so.

Some applications did make use of this frequency to work with standard definition TV and monitors as displays. In many cases though it was simply the cheapest crystal available since they were made in vast quantities for the TV industry. If you needed something in that general range of frequency it made sense to select 3.58MHz because it was dirt cheap and readily available.

[nigelwright7557]

If its just a Z80 clock I cant see why any xtal around that speed wont do.
Unless the same xtal drives a UART or some other time critical device.

[schmitt trigger]

Here is a datasheet for a 3.580MHz crystal sold by Mouser:


I have no idea why it exists.

My favorite theory? Because of non-technical purchasing people and lazy engineers who did not accurately  specified it.

That is why our engineering director required that we should always call  it as a NTSC Chroma frequency crystal.

[mcovington]

I've never seen it referred to as "3.580" either, but that's simply 3.579545 rounded to 3 decimal places.

[amyk]

A lot of early microcomputers in the late 70s/80s used this as a timing reference, allowing direct connection to a colour TV.

[Zero999]

(About the MM5369)

To obtain 60 Hz from an oddball frequency such as 3.579545 MHz, it alternated the frequency’s division ratio.

No, it did not, it just divided by 59659.  You had to tweak the crystal load to get it closer to 60Hz.

As such, the output wasn’t a 50% duty.

Yes, that is correct.

Sadly it is now obsolete, but can be emulated with any modern microcontroller.

It was an old PMOS part that wouldn't run at 5V, let alone 3.3V, so it wouldn't be very useful today.

You're right about the duty cycle, but according to the data sheet, it was CMOS, not PMOS. It will work at 5V, but not at 3.579545MHz, which requires a 10V supply.
http://pdf.datasheetcatalog.com/datasheet/nationalsemiconductor/DS010820.PDF

[SeanB]

Only reason those crystals are around is because they had 2 major uses, TV IF oscillators and filters since 1954, and then, because they were a cheap and available item, they were then used in DTMF tone generators, developed by 1960, as a low cost precision reference oscillator for tone generation. Thus they were both a very cheap, and very precise, quartz crystal oscillator, and thus were used in the first microprocessors as a clock generator, because they were cheap, worked well with a single transistor oscillator, could be divided down easily with a 7474 to give you a roughly 1MHz clock with 50% duty cycle, and ran the processor at a decent speed.

That the clock was so close to TV line rates also helped with video generation, but the leading contributor to use was that you could get them very cheaply, and they were widely available, unlike any other crystal, and were a no set up device unlike a LC oscillator or any other form of oscillator. you have the design, apply power and, within the limits of your test equipment, it would work, and you could test it just using a regular TV set.

[ifrenide]

the exact 3.58 MHz exist in ceramic or passive crystal.
I want to know what is the  MAX tolerance or close replacement of crystals in microcontroller application?
Can we for example replace 4MHz with 12 MHz if we can't why?
I want limits of internal circuitary of microcontrollers "timmers" when we use external crystals, how can a value like  3.579545MHz will be usefull while there is a pre defined prescale timming.we can't get for example 2546.5 clock in digital electronics so ,I think that there is a hardware limit for crystal application.
best regards

[amyk]

You would have to look up the specifications of the components and take into account the firmware of the system too. There is no general answer.

[ifrenide]

thank you amyk.

[Kilrah]

Without knowing what the program on the device is doing it's impossible to say. If the program uses that as reference for precise timing then it can be a problem, if not it won't care a single bit.

[CatalinaWOW]

Amyk is correct.  To do this correctly you have to evaluate both the parts involved and the purpose you will put them to. 

If you want some general rules of thumb I would give you the following:

1.  If your goal is just to get the processor running you will be generally safe with crystals with 10-25% of the nominal design (more on the low end, less on the high.  You may be successful with much larger departures but the odds go down the further you go.  Again, data sheets are your friend.

2.  If you are doing PWM and other similar processes with your processor things will generally work with small variations (again 10-25%) from the nominal value.  But be prepared for odd cases that don't work, and behavior that is not optimal.

3.  If you are doing clocks and timers you really do have to be guided by the needs of your application.  While some of these applications may be OK with a few percent error, others will demand errors that are a tiny fraction of a percent.

[mikeselectricstuff]

If its just a Z80 clock I cant see why any xtal around that speed wont do.
Unless the same xtal drives a UART or some other time critical device.

UART comms can accommodate a total ( tx+rx) error of at least 4%, so the ~0.01% error would be insignificant.
For anything other than long-term timekeeping or frequency measurement, 0.01% error is well below the weeds

[westfw]

Note that the difference is about 0.01%, which is less than the tolerance of the ceramic resonators used in many modern designs, and probably on the same order as the tolerance of random cheap crystals that you buy from China.It's really hard for me to imagine a "normal" application that would care about that level of error.