Two HC908QB8 in Master-Slave mode

[Richard777]

Hi, another question...

I'm planning use two (2) HC908QB8 in master-slave mode, via SPI interface. The master mcu will use a 4.9152 Mhz xtal, is possible use the slave mcu with the out OSC2 pin of the master MCU ?

Thanks in advance

[Encoder]

I enforce Peg's first answer.

There is absolutely no doubt in my mind that OSC2 output of the master Xtal-driven HC908QB8 may drive OSC1 input of the slave Xtal-less -QB8 preset in external clock driven mode. There is surely no violation of the drive fan-out of the first gate (the master one) and the load will be negligibly small compared to that of the oscillator circuitry itself. At such a low frequency the output swing will be very close to the power supply rails and very sharp: it will be consequently no possibility of any uncertain input drive or improper offset of the slave input oscillator gate, so there is not any reason to linearly bias the oscillator gate of the slave device as BigMac suggested (this leave you another pin available as I/O). As Peg pointed out, the two chips must be reasonably close one with the other (some centimeters, perhaps up to 10) to share the clock line.

But the main question remains WHY?

The HC908QB8 is a small, nice full featured 8bit wonder which I used in the past and impressed me at the time; it costs now 1.48$ for the cheapest version as reference price, with 12-13 available I/O pins (one or two are reserved for the oscillator, in your case). 3-4 pins for each MCU are anyway devoted to the serial SPI link and many many pins shares different functions which limits the possibility of using every function you may need to its full extent. The need of programming and possibly reprogramming the internal flash poses serious requirements on the design of the application circuitry vs. the on board programming port.

Even worse, the overheading of the firmware layer which provides the SPI link will cause operative slow down and programmer's headaches which I would avoid as much as possible.

If you need more I/O ports it would be wiser to use a bigger MCU, avoiding linking two -QB8. A natural choice in the past would be the HC908QC16, which is essentially a bigger -QB8 with up to 28 pins. But the Freescale's reference cost is a jealous secret (the 16 pin version is quoted at 1.94$) and Digi-Key quotes it at 2.228$. Availability is still a mistery. But a very more powerful MC9S08AW16CGFE is quoted 1.95$ in the 44pin version and only slightly more in the 64 pin version. This is definitely MUCH more powerful and easier to manage and program: I woudn't come back to the older -QB and -QC version after having tested it.

The only drawback I can see is the availability of the dip version for the QB8, usefull for tests and breadboards, while the AW16 is available only in small smd forms, difficult to manage without a proper pc board and some skill.

[Richard777]

Hi guys, sorry for my english ( it's not very well ).

My project must be need 2 UART, in my country ( Argentina ) Freescale distributor only sales a few models, also the relation between american dollar and my local currency is 1:3...1 dollar are 3 "Pesos Argentinos" then a few dollar represent a significant difference for me ... Another reason is dip version, smd mountaje is also very expensive for small production scales.

Best regards

[Encoder]

MC9S08AW16CFUE (64 pin 14x14mm QFP64) has 2 UARTs and is still much less costly than 2x HC908QB8, being rated at 2.20$. Obviously you do not need to use all the added pins!

Unfortunately it definitely is in smd package. But the added advantages may perhaps shift the more costly board manufacturing process.

[peg]

Hi,

The S08GTxxA is available in 42-pin shrink DIP and has two UARTS. Can run on internal oscillator as well!

I would check the availability/price of this one.

[Encoder]

Another hit for Peg!

I discarded from my horizon the S08GT being a 3.3V part, but no doubt that it is an impressive MCU with a very good price/performance ratio. It has 2xSCI and, whilst being much bigger and powerful than 908QB8 it costs less, at least for the 8k version.

From the Freescale reference price I see:

MC9S08GT8ACBE (42pin SDIP): 1.40$

MC9S08GT16ACBE (42pin SDIP): 1.70$

I'm a pretty old designer and tenaciously hanged to 5V logic world...

Encoder

[Curt]

I've run four 8051 cores from a single outboard clock in a testbed app. that needed 120 configurable I/Os.  By clocking everything in parallel, core-to-core skews were held to a minimum.  Worked great!

[peg]

Hi Richard,

You can do it although do review Motorola's EB396. (It's on the GP page for sure)

The MPU's would have to be close to each other, but then the question would arise as to WHY are you doing this?

Message Edited by peg on 2007-03-1509:18 AM

[Richard777]

Hi, thanks for your fast response. The idea is very simple, I want that my circuit has only one cristal for the two MCU.

Richard

[Alban]

A XTAL cannot supply/be supplied by two MCUs via OSC1 and OSC2.

I strongly discourage such application.

[Richard777]

Ok, thanks Alban I will be follow your suggestion.

Richard.

[peg]

Hi,

Well quite obviously you can otherwise EB396 would not have been made to warn against it!

Yes, it is probably a bad thing to do and that is why I referenced the official documentation specifically relating to this.

My query was not why use a common clock but, why have two MPU's almost side by side? Why not 1 bigger one?

Perhaps 1 canned oscillator is the answer here. (for two MPU's)

[Richard777]

Hi Peg, in my country ( Argentina ) for price ...

Regards

[peg]

Ok fair enough. And there I was trying to think of a good technical reason.

[bigmac]

Hello Richard,

A few years back I did successfully use a single 4MHz ceramic resonator with two HC705 devices that were located side by side.  Using a crystal should be similar.

The solution consisted of setting up the "master" as a conventional Pierce oscillator, as shown in the data sheet.  For the "slave" device, a 4M7 resistor was connected between OSC1 and OSC2, for biasing into the linear region, and then OSC2 of the master was capacitively coupled (say 100p) to OSC1 of the slave.  For the Pierce oscillator configuration, you might need to reduce the shunt capacitance at the OSC2 pin, to compensate for the additional input capacitance of the slave.

The two devices were low pin count types, and in one application the second device was an optional plug-in component, used for about 20 percent of cases.

In another application, one of the devices need to operate in a very tight loop, which then required another device to do other things.  With the availability of low end devices with input capture capability, this situation would now be different.

Regards,

Mac