Speed ratings: 65C51

[BillO]

I'm not sure what they mean by the speed rating on a 65C51. I have lots of these from various manufacturers with speed ratings of 1mHz, 2mHz and 4mHz. I have a 1mHz unit working happily with a 6502 running at 4mHz and a 2mHz one running, as I write this, quite happily with a W65C02 running at 16mHz. Does everyone else see this with these chips?

[Dr Jefyll]

Tell us your secret, Bill! How do you do it?

[BillO]

I plugged it into the socket and hoped for the best!!??

The thought of having to write CPU clock speed/baud-rate dependent code to use W65C51s forced my hand, my shaking hand - but it turned out good.

Still, what do the specifications really mean? Do we trust them?

[Dr Jefyll]

Garth and Ed have some pertinent comments here. All the points are relevant, but this remark below was an eye-opener for me -- something I wouldn't have thought of.

BigEd wrote:

if 10% of the parts pass the test and go into the fastest bin, but the market only needs 5%, then some of the known fast parts can be marked as some slower grade and sold as such.

[floobydust]

I have a decent collection of both NMOS and CMOS 6551/65C51 chips. Vendors include Synertek, Rockwell, Intersil and WDC. Here's what I've found on speeds:

Rockwell NMOS parts: Good for rated speed and rarely more (1 and 2 MHz versions)
Rockwell CMOS parts: Good for rated speed and rarely more (4 MHz versions, some run @ 5MHz)

Syntertek NMOS part: I have one which will will run rock-solid at 6MHz

Intersil CMOS part: I have one rated at 4MHz. It runs at that speed but not any faster. It also has corrupt data from time to time.

WDC CMOS parts: I have three older LOT numbers which they sent me years ago.... all require a can oscillator to work, one or two run well up to 10MHz but occasionally have hiccups. The latest LOT number (with the XMIT bug) also runs at rated speeds, but it's a moot point (the XMIT bug).

So, all I can say is you have one special chip there... based on my findings, yours is a nice "fluke" to have. Do you have any additional chips you can test?

[BillO]

floobydust wrote:

Do you have any additional chips you can test?

That really fast one is an Intersil CDP65C51AE2 with a date/batch code of P0131ACWX and I have more than 5 full tubes of them (75+).

I'll be testing some of them...

[JohanFr]

Sorry to bring up a super old thread, but did you come to any conclusion about this?

Reason I ask is I have a bunch of WDC 65C51s that I bought new from Mouser in 2019. I am confused what is the actual speed rating of those chips. The part number ends with "-14", indicating it is a 14 MHz part (I think?), but the WDC documentation speaks only of speeds of 1, 2, 4, MHz. My current design runs at 3.68 MHz fine. But I am considering using a 65C51 again for a new design that runs at 7-8 MHz. The transmit bug doesn't really bother me much as the serial port serves mostly as an upload port. So I am trying understand the probability of success for such a combination.

[BigDumbDinosaur]

JohanFr wrote:

The part number ends with "-14", indicating it is a 14 MHz part (I think?), but the WDC documentation speaks only of speeds of 1, 2, 4, MHz.

The -14 suffix means the part is officially rated for a maximum Ø2 frequency of 14 MHz.  WDC’s documentation is legendary for having errors or just being out of date. 

Ever since WDC switched to the TSMC foundry for dice and 0.6µ geometry became the norm, the usable speeds have increased significantly over the official rating.  Several years ago, David Gray at WDC advised me in an E-mail that production testing of the 65C02 and 65C816 is at 20 MHz and that a part will be rejected if it cannot function at that speed (Bill Chen, aka plasmo, has overclocked the 65C02 into the 30 MHz range).  My discussion with David didn’t get into the peripheral chips, but it wouldn’t surprise me if the 65C22 and 65C51 are functional at 20 MHz.

That said, the W65C51’s stuck TxD ready flag is a nasty bit of hardware errata that makes using the device a pain in the derrière.  Also, the UART’s programming model leaves plenty to be desired—too many features are mapped into too few control bits, CTS/RTS behavior is annoying, etc..  There are much better UARTs around sans hardware bugs, and with better characteristics.

[JohanFr]

I am thinking about giving bitbanging a MAX3100 using the 65C22 a shot to simplify the bus-design, but I figured since I have 4 more 65C22s from my order I could at least give it another shot before giving up on them Thanks, then I know that at least in theory they should be able handle higher frequencies.

[floobydust]

Oddly, I was wondering if BillO was using the 6551 chips in polled mode only, i.e., no interrupt driven receive or transmit. I only use the 65(C)51 in interrupt driven mode and I've not been able to ramp the chips up to the high clock rates that BillO has gotten with his setup.

[BruceRMcF]

JohanFr wrote:

I am thinking about giving bitbanging a MAX3100 using the 65C22 a shot to simplify the bus-design, but I figured since I have 4 more 65C22s from my order I could at least give it another shot before giving up on them Thanks, then I know that at least in theory they should be able handle higher frequencies.

If you only have to capture the second MISO byte with a Max3100, I wonder how complicated the circuit would be with a direction pin on one of the CIA pins, to allow using the serial port to output the first byte, then turn it around to receive the MISO during the second byte with MOSI being bit banged out.

floobydust wrote:

Oddly, I was wondering if BillO was using the 6551 chips in polled mode only, i.e., no interrupt driven receive or transmit. I only use the 65(C)51 in interrupt driven mode and I've not been able to ramp the chips up to the high clock rates that BillO has gotten with his setup.

Is one of the upshots of the "feature" of the W65C51 that interrupt driven operation only works in one direction without the risk of stepping on its own toes? It may be that BillO has organized the timing of the output side so that it refrains from stepping on its own toes.