| |
Flavioweb
Registered: Nov 2011 Posts: 442 |
1541 Quartz speed on real device.
With the help of Ready. we measured the quartz frequency of a 1541 directly on the motherboard, with the drive running, to make a precise test by evaluating the values in a condition of real operation.
This is the quartz mounted on the drive we have tested:
The test lasted about 2h and 30 mins, during which the electronics were covered to simulate the presence of the top of the chassis.
The maximum temperature reached by the quartz was about 45c if covered and about 39 if uncovered.
For the whole duration of the test a frequency value oscillating between 16.000 and 16.0001 mhz was detected. No values higher or lower than these.
In the quartz tech specs we found a maximum drift of +/- 50ppm, which means:
50 * 16 = 800 = 16.000.800 / 15.999.200hz in the worst case.
During our test, instead, we found, in the worst case, a drift of 100hz, therefore 1/8 of the theoretical one found in specs, ie 50/8 = 6.25ppm.
Making a few calculations just to relate ppm to rpm, we have:
6.25 / 5 = 1.25 (ppm / (1mhz / hz per revolution)) = drift per revolution
1.25 / 200,000 = 0.00000625
300 * 0.00000625 = 0.001875 rpm calculation error due to the quartz drift in the worst case.
We do not know exactly, during the whole period of normal operation of the drive, what is the precise value of ppm of deviation, but we know with certainty that it is a number between 16,000,000 and 16,000,100 hertz, so 100/16 = 6.25ppm in the worst case. |
|
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
Not that you can draw any conclusions from this, especially not that generally all crystals in 1541s will have this kind of maximum deviation - you can only rely on what the manufacturer guarantees, thats why we have datasheets afterall :) |
| |
Flavioweb
Registered: Nov 2011 Posts: 442 |
Yes, of course.
Now we begin to get an idea of how things work in practice.
If we were to rely only on the datasheets we know, we would take for granted a deviation of 50ppm but, apparently, taking the first drive available, we found that the useful value to understand how things work is about 1/8 of the theoretical one.
A very reliable value would be obtained by measuring the quartz frequency on most of the existing drives but i don't think this is feasible.
From my point of view, whatever the routine that needs such precision, an error of just over one cycle in two hundred thousand is much more than acceptable.
Obviously, the more measurements are made, the more accurate the result will be.
Also from my point of view, measuring "something" in its context of real functioning is an always better condition than "extracting it" and carrying out the tests in other arbitrary conditions that are all to be defined. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
Measuring it on ONE drive tells you nothing, really :) Have you seen what Unseen measured on another drive? https://sourceforge.net/p/vice-emu/code/HEAD/tree/testprogs/dri..
(also please tell what exactly the equipment used was, how it was calibrated, and what its precision and accurracy is according to it's datasheet) |
| |
Flavioweb
Registered: Nov 2011 Posts: 442 |
Quoting GroepazMeasuring it on ONE drive tells you nothing, really :) Have you seen what Unseen measured on another drive? https://sourceforge.net/p/vice-emu/code/HEAD/tree/testprogs/dri..
Yes i saw it but (i admit through my ignorance) i don't understand 100% what it means. On the y axis what is it? the deviation expressed in ppm?
Quoting Groepaz(also please tell what exactly the equipment used was, how it was calibrated, and what its precision and accurracy is according to it's datasheet) The same thing should apply to this graph.
Under what conditions, with which instruments and with which calibrations were the measurements carried out?
Was the quartz running on a drive or was the measurement done on a "detached" quartz and under different conditions? The only info available about this is
Quoting GroepazWhat he told me is "The Frequency counter had a OCXO and was warmed up an hour before the measurement". I can ask for more details if you tell me what to ask :) (But i also assume no terrible mistakes were made, he knows his stuff) |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
check the readme in that directory in the repo, it has the info about the used equipment and conditions (as far as unseen could remember, he did this years ago)
Y is the deviation in Hz (so its also much less than 50ppm, but quite obviously temperature dependend) |
| |
Flavioweb
Registered: Nov 2011 Posts: 442 |
Quoting GroepazY is the deviation in Hz (so its also much less than 50ppm, but quite obviously temperature dependend)
Ah... ok!
So these two tests confirms the same results.
A deviation range from 16.000.000 to ~16.000.100 that means 100/16 = 6,25 ppm during actual use of the drive.
Hope Ready. take the time to come and specify some technical details on the instrumentation and the method used for the test. |
| |
ready.
Registered: Feb 2003 Posts: 441 |
I used Rigol DS1054Z oscilloscope to measure the frequency. I relayed on the hardware frequency counter feature to measure the 1541 crystal frequency. Accuracy for this measurement should be less than 10ppm according to the tests I found here: http://nerdralph.blogspot.com/2015/07/rigol-ds1054z-frequency-c.. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
I updated the readme with the info posted here, please check :) https://sourceforge.net/p/vice-emu/code/HEAD/tree/testprogs/dri.. (at the bottom of the file) |
| |
Flavioweb
Registered: Nov 2011 Posts: 442 |
Nice ! =) |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Nice digression.
As I said before (without even doing all calculations and tests) the quartz drift is totally irrelevant to the results of 1541 Speed Test :D
And that's why I chose to accurately measure up to a cent of RPM.
I also checked the rpm page on sourceforge and noticed your rpm3.asm "inspired by 1541 speed test".
You even copied the (totally arbitrary) value I use for the byte I write. Why don't you do the right thing and take all the drive code from 1541 speed test and just call it from the test app?
If you like I can provide the drive code as stand-alone on github in the format you prefer. |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Quoting ready.I used Rigol DS1054Z oscilloscope to measure the frequency. I relayed on the hardware frequency counter feature to measure the 1541 crystal frequency. Accuracy for this measurement should be less than 10ppm according to the tests I found here: http://nerdralph.blogspot.com/2015/07/rigol-ds1054z-frequency-c..
Very well done. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
Quote:Why don't you do the right thing and take all the drive code from 1541 speed test and just call it from the test app?
It *is* equivalent code. |
| |
ChristopherJam
Registered: Aug 2004 Posts: 1359 |
This is fascinating. I wonder if the c64 crystal is as accurate? |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Quoting ChristopherJamThis is fascinating. I wonder if the c64 crystal is as accurate?
Hmmm... we could measure it using the crystal in the drive, but I think it's just the same crystal.
Maybe I am wrong. |
| |
ChristopherJam
Registered: Aug 2004 Posts: 1359 |
Nah, crystal in a PAL c64 is nominally 17.734475MHz, which is divided by 18 to give a CPU clock of around 985kHz.
But yes, you could measure it against a drive crystal, which is exactly what RPM Test 1.0 does :) |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Quoting ChristopherJamNah, crystal in a PAL c64 is nominally 17.734475MHz, which is divided by 18 to give a CPU clock of around 985kHz.
I know that.
By "same crystal" I meant same inaccuracy of about 6ppm. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
Just don't make the mistake of misinterpreting those two measurements into the general case - generally we still have to assume what the datasheet is true, more or less. Because if the manufacturer would have been able to guarantee ~6ppm deviation, you can be very sure he would have not written 50ppm into the datasheet :) Right now those measurements can be taken as anecdotal - no more no less. |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Quoting GroepazJust don't make the mistake of misinterpreting those two measurements into the general case - generally we still have to assume what the datasheet is true, more or less. Because if the manufacturer would have been able to guarantee ~6ppm deviation, you can be very sure he would have not written 50ppm into the datasheet :) Right now those measurements can be taken as anecdotal - no more no less.
Wrong:
the manufacturer can't guarantee weather conditions or ambient temperatures.
The ppm is hence calculated, so probably at the north pole in the cold it will go slower than in the middle of the deser at 1 pm.
But in most cases the quarts will be in a 18-38 degrees environment.
Also it may happen that some "unlucky" quartz are bad but still within the 50ppm "specs" so they will be safe from legal actions about mislabeling. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
All you can actually rely on are the specs in the datasheet, under the conditions listed in the datasheet. Not doing so is amateurish tinkering. There are no "unlucky" components, there are only components that operate within the specs, and components that do not.
And no, not wrong. If they could guarantee certain things within certain more narrow conditions - they'd write it in the datasheet. |
| |
ChristopherJam
Registered: Aug 2004 Posts: 1359 |
Quoting Zibrithe manufacturer can't guarantee weather conditions or ambient temperatures.
Of course not. But what they do do, is specify a range of ambient conditions under which they will guarantee the clock speed is within a given tolerance. |
| |
Oswald
Registered: Apr 2002 Posts: 5007 |
the error of the measuring device should be also accounted for. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
which is why it's listed already, indeed. |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Yes. I realize this and both Oswald and Christopher are right.
What I am saying is that is totally irrelevant to the RPM speed test which express the speed in cents of rpm...
The "differences" in quartz do not influence the results if nor very marginally. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
This thread is about the crystal speed though, not about RPM testing. |
| |
Zibri Account closed
Registered: May 2020 Posts: 304 |
Quoting GroepazThis thread is about the crystal speed though, not about RPM testing.
Right, but the crystal speed was tested because "someone" implied that my RPM test was anyway inaccurate because of the crystal. Which was so proven not to be relevant in the calculation. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11088 |
everyone understood that by now, and its irrelevant for this discussion. |