On Thursday, November 15, 2018 at 10:18:51 AM UTC-5, Phil Martel wrote:> On 11/14/2018 18:57, gnuarm.deletethisbit@gmail.com wrote: > > On Wednesday, November 14, 2018 at 4:28:37 PM UTC-6, Clifford Heath wrote: > >> On 15/11/18 9:20 am, gnuarm.deletethisbit@gmail.com wrote: > >>> On Wednesday, November 14, 2018 at 3:52:51 PM UTC-6, Clifford Heath wrote: > >>>> On 15/11/18 4:48 am, gnuarm.deletethisbit@gmail.com wrote: > >>>>> On Wednesday, November 14, 2018 at 9:33:35 AM UTC-6, steve wrote: > >>>>>> Thanks having contacted Microchip/Atmel they agreed! > >>>>>> On Monday, November 12, 2018 at 1:41:58 PM UTC, gnuarm.del...@gmail.com wrote: > >>>>>>> On Monday, November 12, 2018 at 5:22:31 AM UTC-6, steve wrote: > >>>>>>>> I am using a Atmel4SD32C processor. > >>>>>>>> For the main clock crystal Atmel specifiey the following requirements: > >>>>>>>> CLEXT > >>>>>>>> Maximum External Capacitor on XIN and XOUT 17 pF > >>>>>>>> Allowed Crystal Capacitance Load From crystal specification 12.5 – 17.5 pF > >>>>>>>> CLOAD Internal Equivalent Load Capacitance Integrated load capacitance > >>>>>>>> (XIN and XOUT in series) 7.5min 9.5typical 10.5 pFmax > >>>>>>>> > >>>>>>>> Calculation for Cload external > >>>>>>>> CLEXT = 2 × (Ccrystal – CLOAD – CPCB). > >>>>>>>> Where CPCB is the capacitance of the printed circuit board (PCB) track layout from the crystal to the SAM4 pin > >>>>>>>> > >>>>>>>> If I use a 12pf load crystal it would appear that no external capacitors to ground are needed. > >>>>>>>> Atmel give a max load capacitance of 17pf but no minimum. > >>>>>>>> My pcb/pin tracks come in at about 2pf so I assume that is enough to start the oscillator and having tried it it seems to work fine > >>>>>>>> Is this ok? I assume this is what they intended when they designed in a internal load capacitor > >>>>>>> > >>>>>>> I researched the issue of crystal selection with an Atmel MCU some years ago and found not only issues with figuring out the required load capacitance, but just as important the crystal ESR. As a result of my conversations with the sales person the factory added an ESR spec to the data sheet, but they did it as a table for different frequencies with no info on how to interpolate the specified data. Still, that was a good start. > >>>>>>> > >>>>>>> Regarding your problem, I can't quite follow all the data you provided. Assuming you are interpreting the data correctly it does indeed seem that you need no additional capacitance. Just the same, I would recommend the addition of pads for adding capacitors, just in case. > >>>>>>> > >>>>>>> Rick C. > >>>>>>> > >>>>>>> https://ts.la/richard11209 - Tesla referral code > >>>>> > >>>>> What exactly did they agree with, that their data sheet is not clear and they would improve on it? > >>>> > >>>> This discussion reminds me of when my home-built MC68HC11 oscillator > >>>> wouldn't start. I'd built it according to the Pink Book, which showed > >>>> two series capacitors as load - but the recommended value for each was > >>>> the total series load expected. Or vice versa, I can't recall. I got an > >>>> answer from the guy at Motorola who worked with the main author, who > >>>> confirmed the error in the book. Anyhow I changed the capacitors and it > >>>> fired up right away. > >>> > >>> When you say "series capacitors" do you mean this? > >>> > >>> || |++| || > >>> o----||---||||---||---o > >>> || |++| || > >> > >> No > >> > >>> The caps should go to ground, in parallel with the crystal in essence but series with each other. > >> > >> The series combination still presents a C load to the crystal, > >> the fact that one node is earthed doesn't change that, because neither > >> of the other two is earthed. I recall needing two 33pF caps (series load > >> about 16pF), not two 18pF (series 9pF). > > > > I think we are saying the same thing. > > > > > >>> |++| > >>> o--+-----||||----+-o > >>> | |++| | > >>> | || || | > >>> +--||--+--||--+ > >>> || | || > >>> | > >>> --- > >>> V > >>> > >>> Rick C. > >>> > >>> Tesla referral code ----+ https://ts.la/richard11209 > >>> > >> > >> It was 23 years ago. The recommended crystal load was wrong anyhow, if > >> you followed the Pink Book, and the oscillator wouldn't start. I can dig > >> out the book and the facts if you care. > > > > I don't really need it as if I need an oscillator I usually use an oscillator rather than a crystal. But it is an interesting subject if you are interested in sharing. > > > > I spent a fairly large amount of time once researching crystals to try to get to the derivation of the design issues. I found an HP paper that actually discussed the various crystal cuts in great detail. Not sure where it is, but I likely could find it if I dig enough. > > > > > > Rick C. > > > > Tesla referral code ---+- https://ts.la/richard11209 > > > This? http://leapsecond.com/hpan/an200-2.pdf Google is your friend(tm)Yep, that's the one. Thanks. Rick C. Tesla referral code +-+ https://ts.la/richard11209
Crystal Load Capacitance
Started by ●November 12, 2018
Reply by ●November 15, 20182018-11-15
Reply by ●November 15, 20182018-11-15
On 11/15/18 2:04 PM, gnuarm.deletethisbit@gmail.com wrote:> On Thursday, November 15, 2018 at 10:18:51 AM UTC-5, Phil Martel wrote: >> On 11/14/2018 18:57, gnuarm.deletethisbit@gmail.com wrote: >>> On Wednesday, November 14, 2018 at 4:28:37 PM UTC-6, Clifford Heath wrote: >>>> On 15/11/18 9:20 am, gnuarm.deletethisbit@gmail.com wrote: >>>>> On Wednesday, November 14, 2018 at 3:52:51 PM UTC-6, Clifford Heath wrote: >>>>>> On 15/11/18 4:48 am, gnuarm.deletethisbit@gmail.com wrote: >>>>>>> On Wednesday, November 14, 2018 at 9:33:35 AM UTC-6, steve wrote: >>>>>>>> Thanks having contacted Microchip/Atmel they agreed! >>>>>>>> On Monday, November 12, 2018 at 1:41:58 PM UTC, gnuarm.del...@gmail.com wrote: >>>>>>>>> On Monday, November 12, 2018 at 5:22:31 AM UTC-6, steve wrote: >>>>>>>>>> I am using a Atmel4SD32C processor. >>>>>>>>>> For the main clock crystal Atmel specifiey the following requirements: >>>>>>>>>> CLEXT >>>>>>>>>> Maximum External Capacitor on XIN and XOUT 17 pF >>>>>>>>>> Allowed Crystal Capacitance Load From crystal specification 12.5 – 17.5 pF >>>>>>>>>> CLOAD Internal Equivalent Load Capacitance Integrated load capacitance >>>>>>>>>> (XIN and XOUT in series) 7.5min 9.5typical 10.5 pFmax >>>>>>>>>> >>>>>>>>>> Calculation for Cload external >>>>>>>>>> CLEXT = 2 × (Ccrystal – CLOAD – CPCB). >>>>>>>>>> Where CPCB is the capacitance of the printed circuit board (PCB) track layout from the crystal to the SAM4 pin >>>>>>>>>> >>>>>>>>>> If I use a 12pf load crystal it would appear that no external capacitors to ground are needed. >>>>>>>>>> Atmel give a max load capacitance of 17pf but no minimum. >>>>>>>>>> My pcb/pin tracks come in at about 2pf so I assume that is enough to start the oscillator and having tried it it seems to work fine >>>>>>>>>> Is this ok? I assume this is what they intended when they designed in a internal load capacitor >>>>>>>>> >>>>>>>>> I researched the issue of crystal selection with an Atmel MCU some years ago and found not only issues with figuring out the required load capacitance, but just as important the crystal ESR. As a result of my conversations with the sales person the factory added an ESR spec to the data sheet, but they did it as a table for different frequencies with no info on how to interpolate the specified data. Still, that was a good start. >>>>>>>>> >>>>>>>>> Regarding your problem, I can't quite follow all the data you provided. Assuming you are interpreting the data correctly it does indeed seem that you need no additional capacitance. Just the same, I would recommend the addition of pads for adding capacitors, just in case. >>>>>>>>> >>>>>>>>> Rick C. >>>>>>>>> >>>>>>>>> https://ts.la/richard11209 - Tesla referral code >>>>>>> >>>>>>> What exactly did they agree with, that their data sheet is not clear and they would improve on it? >>>>>> >>>>>> This discussion reminds me of when my home-built MC68HC11 oscillator >>>>>> wouldn't start. I'd built it according to the Pink Book, which showed >>>>>> two series capacitors as load - but the recommended value for each was >>>>>> the total series load expected. Or vice versa, I can't recall. I got an >>>>>> answer from the guy at Motorola who worked with the main author, who >>>>>> confirmed the error in the book. Anyhow I changed the capacitors and it >>>>>> fired up right away. >>>>> >>>>> When you say "series capacitors" do you mean this? >>>>> >>>>> || |++| || >>>>> o----||---||||---||---o >>>>> || |++| || >>>> >>>> No >>>> >>>>> The caps should go to ground, in parallel with the crystal in essence but series with each other. >>>> >>>> The series combination still presents a C load to the crystal, >>>> the fact that one node is earthed doesn't change that, because neither >>>> of the other two is earthed. I recall needing two 33pF caps (series load >>>> about 16pF), not two 18pF (series 9pF). >>> >>> I think we are saying the same thing. >>> >>> >>>>> |++| >>>>> o--+-----||||----+-o >>>>> | |++| | >>>>> | || || | >>>>> +--||--+--||--+ >>>>> || | || >>>>> | >>>>> --- >>>>> V >>>>> >>>>> Rick C. >>>>> >>>>> Tesla referral code ----+ https://ts.la/richard11209 >>>>> >>>> >>>> It was 23 years ago. The recommended crystal load was wrong anyhow, if >>>> you followed the Pink Book, and the oscillator wouldn't start. I can dig >>>> out the book and the facts if you care. >>> >>> I don't really need it as if I need an oscillator I usually use an oscillator rather than a crystal. But it is an interesting subject if you are interested in sharing. >>> >>> I spent a fairly large amount of time once researching crystals to try to get to the derivation of the design issues. I found an HP paper that actually discussed the various crystal cuts in great detail. Not sure where it is, but I likely could find it if I dig enough. >>> >>> >>> Rick C. >>> >>> Tesla referral code ---+- https://ts.la/richard11209 >>> >> This? http://leapsecond.com/hpan/an200-2.pdf Google is your friend(tm) > > Yep, that's the one. Thanks. > > Rick C. > > Tesla referral code +-+ https://ts.la/richard11209 >I think that was written by Rick Karlquist, designer of the HP 10811 series OCXOs. He's a former SED regular and is still quite active on the time-nuts list. He also did a super interesting direct frequency synthesizer based on cheap 455 kHz IF filters. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●November 18, 20182018-11-18
On 11/15/18 19:38, Phil Hobbs wrote:> > I think that was written by Rick Karlquist, designer of the HP 10811 > series OCXOs. He's a former SED regular and is still quite active on the > time-nuts list. He also did a super interesting direct frequency > synthesizer based on cheap 455 kHz IF filters. > > Cheers > > Phil Hobbs >A good book on crystal oscillators: Crystal Oscillator Circuits, Robert J. Matthys John Wiley, 1983, ISBN 0-471-87401-9 Chapters on crystals, various oscillator circuits and plenty of practical examples. Not needed that often but good to have in the bookshelf. My copy came from abe books... Chris
Reply by ●November 18, 20182018-11-18
On Sunday, November 18, 2018 at 7:01:00 PM UTC-6, Chris wrote:> On 11/15/18 19:38, Phil Hobbs wrote: > > > > I think that was written by Rick Karlquist, designer of the HP 10811 > > series OCXOs. He's a former SED regular and is still quite active on the > > time-nuts list. He also did a super interesting direct frequency > > synthesizer based on cheap 455 kHz IF filters. > > > > Cheers > > > > Phil Hobbs > > > > A good book on crystal oscillators: > > Crystal Oscillator Circuits, Robert J. Matthys > John Wiley, 1983, ISBN 0-471-87401-9 > > Chapters on crystals, various oscillator circuits > and plenty of practical examples. Not needed that > often but good to have in the bookshelf. My copy came > from abe books...I've yet to see an analysis of digital oscillator circuits. I've seen one described but no real analysis of it. Performance is ok at lower frequencies (for undetermined values of "OK"), but no real measurements other than it was operable. By "digital" I'm not talking about an inverter pressed into service as an amplifier. I mean a digital circuit mediated by software in the above case or it could be digital logic in other cases. Rick C. Tesla referral code - https://ts.la/richard11209
Reply by ●November 19, 20182018-11-19
On 19.11.18 05:30, gnuarm.deletethisbit@gmail.com wrote:> I've yet to see an analysis of digital oscillator circuits. I've seen one described but no real analysis of it. Performance is ok at lower frequencies (for undetermined values of "OK"), but no real measurements other than it was operable. > > By "digital" I'm not talking about an inverter pressed into service as an amplifier. I mean a digital circuit mediated by software in the above case or it could be digital logic in other cases. > > Rick C.You're hunting unicorns here ... The digital frequency generation circuits, aided by software or not, are running behind some kind of clock. The clock is ultimately from an analog time/frequency reference part, e.g. a quartz crystal. -- -TV
Reply by ●November 19, 20182018-11-19
Microchip response. Incidentally Atmels development board has crystal with 22pf load capacitance, which does not meet the chip requirements of max 17pf **** AUTOMATED MESSAGE - SEE RESPONSE METHODS BELOW **** Below is a proposed resolution from Microchip Engineering Support Team for your Case 00352596. Subject: Crystal Reason: Hardware/Firmware Support Product: ATSAM4SD32C Problem Description: I am using a ATSAM4SD32C processor. For the main clock crystal Atmel specifiey the following requirements: CLEXT Maximum External Capacitor on XIN and XOUT 17 pF Allowed Crystal Capacitance Load From crystal specification 12.5 – 17.5 pF CLOAD Internal Equivalent Load Capacitance Integrated load capacitance (XIN and XOUT in series) 7.5min 9.5typical 10.5 pFmax Calculation for Cload external CLEXT = 2 × (Ccrystal – CLOAD – CPCB). Where CPCB is the capacitance of the printed circuit board (PCB) track layout from the crystal to the SAM4 pin If I use a 12pf load crystal it would appear that no external capacitors to ground are needed. Atmel give a max load capacitance of 17pf but no minimum. My pcb/pin tracks come in at about 2pf so I assume that is enough to start the oscillator and having tried it it seems to work fine Is this ok? I assume this is what they intended when they designed in a internal load capacitor Thanks **** Proposed Resolution Begin **** For the crystal capacitance used in this case(12 pF), with the typical load capacitance and a 2 pF CPCB, the CLEXT comes about 1 pF. Since the value of CLEXT is more or less negligible, this should work fine. However, we do recommend to use a crystal load capacitance between 12.5 to 17.5 pF just to be on the safe side. **** Proposed Resolution End ******
Reply by ●November 19, 20182018-11-19
On 11/18/18 8:00 PM, Chris wrote:> On 11/15/18 19:38, Phil Hobbs wrote: >> >> I think that was written by Rick Karlquist, designer of the HP 10811 >> series OCXOs. He's a former SED regular and is still quite active on the >> time-nuts list. He also did a super interesting direct frequency >> synthesizer based on cheap 455 kHz IF filters. >> >> Cheers >> >> Phil Hobbs >> > > A good book on crystal oscillators: > > Crystal Oscillator Circuits, Robert J. Matthys > John Wiley, 1983, ISBN 0-471-87401-9 > > Chapters on crystals, various oscillator circuits > and plenty of practical examples. Not needed that > often but good to have in the bookshelf. My copy came > from abe books... > > Chris >I have Parzen's book of about the same vintage. A pretty good read. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Reply by ●November 19, 20182018-11-19
On Monday, November 19, 2018 at 4:01:56 AM UTC-5, Tauno Voipio wrote:> On 19.11.18 05:30, gnuarm.deletethisbit@gmail.com wrote: > > > I've yet to see an analysis of digital oscillator circuits. I've seen one described but no real analysis of it. Performance is ok at lower frequencies (for undetermined values of "OK"), but no real measurements other than it was operable. > > > > By "digital" I'm not talking about an inverter pressed into service as an amplifier. I mean a digital circuit mediated by software in the above case or it could be digital logic in other cases. > > > > Rick C. > > You're hunting unicorns here ... > > The digital frequency generation circuits, aided by software > or not, are running behind some kind of clock. The clock is > ultimately from an analog time/frequency reference part, e.g. > a quartz crystal.You are assuming facts not in evidence... in other words, you are talking through your hat. Drop all your misconceptions and think about it again. software and logic do not need to be clocked. If you were going to drive a crystal with a narrow pulse, what point in the cycle would it most beneficially be applied? Some of my thinking about this was inspired by reading about the Shortt–Synchronome free pendulum clock. It uses a gravity arm to supply an impulse every thirty swings of the main pendulum. I believe it is applied at the bottom of the swing. Whether that corresponds to the top of the crystal waveform or the midpoint depends on whether you consider position or velocity of the main pendulum. I'm thinking the electrical analog would be the midpoint of the voltage sinewave. Rick C. Tesla referral code ---++ https://ts.la/richard11209
Reply by ●November 20, 20182018-11-20
On 20.11.18 02:24, gnuarm.deletethisbit@gmail.com wrote:> On Monday, November 19, 2018 at 4:01:56 AM UTC-5, Tauno Voipio wrote: >> On 19.11.18 05:30, gnuarm.deletethisbit@gmail.com wrote: >> >>> I've yet to see an analysis of digital oscillator circuits. I've seen one described but no real analysis of it. Performance is ok at lower frequencies (for undetermined values of "OK"), but no real measurements other than it was operable. >>> >>> By "digital" I'm not talking about an inverter pressed into service as an amplifier. I mean a digital circuit mediated by software in the above case or it could be digital logic in other cases. >>> >>> Rick C. >> >> You're hunting unicorns here ... >> >> The digital frequency generation circuits, aided by software >> or not, are running behind some kind of clock. The clock is >> ultimately from an analog time/frequency reference part, e.g. >> a quartz crystal. > > You are assuming facts not in evidence... in other words, you are talking through your hat. Drop all your misconceptions and think about it again. software and logic do not need to be clocked. > > If you were going to drive a crystal with a narrow pulse, what point in the cycle would it most beneficially be applied? Some of my thinking about this was inspired by reading about the Shortt–Synchronome free pendulum clock. It uses a gravity arm to supply an impulse every thirty swings of the main pendulum. I believe it is applied at the bottom of the swing. Whether that corresponds to the top of the crystal waveform or the midpoint depends on whether you consider position or velocity of the main pendulum. > > I'm thinking the electrical analog would be the midpoint of the voltage sinewave. > > Rick C. > > Tesla referral code ---++ https://ts.la/richard11209So - you have the pendulum, which is equivalent to a resonator. I like to quote Don Vonada, once the main designer at Digital Equipment Corp. 'All digital parts are made of analog components'. The oscillators made of digital gates are using the analog delays in the gates a timing reference. It is a completely different question, how the necessary energy feed to the timing reference is applied, like kicking the crystal. -- -TV
Reply by ●November 20, 20182018-11-20
On Tuesday, November 20, 2018 at 4:21:23 AM UTC-5, Tauno Voipio wrote:> On 20.11.18 02:24, gnuarm.deletethisbit@gmail.com wrote: > > On Monday, November 19, 2018 at 4:01:56 AM UTC-5, Tauno Voipio wrote: > >> On 19.11.18 05:30, gnuarm.deletethisbit@gmail.com wrote: > >> > >>> I've yet to see an analysis of digital oscillator circuits. I've seen one described but no real analysis of it. Performance is ok at lower frequencies (for undetermined values of "OK"), but no real measurements other than it was operable. > >>> > >>> By "digital" I'm not talking about an inverter pressed into service as an amplifier. I mean a digital circuit mediated by software in the above case or it could be digital logic in other cases. > >>> > >>> Rick C. > >> > >> You're hunting unicorns here ... > >> > >> The digital frequency generation circuits, aided by software > >> or not, are running behind some kind of clock. The clock is > >> ultimately from an analog time/frequency reference part, e.g. > >> a quartz crystal. > > > > You are assuming facts not in evidence... in other words, you are talking through your hat. Drop all your misconceptions and think about it again. software and logic do not need to be clocked. > > > > If you were going to drive a crystal with a narrow pulse, what point in the cycle would it most beneficially be applied? Some of my thinking about this was inspired by reading about the Shortt–Synchronome free pendulum clock. It uses a gravity arm to supply an impulse every thirty swings of the main pendulum. I believe it is applied at the bottom of the swing. Whether that corresponds to the top of the crystal waveform or the midpoint depends on whether you consider position or velocity of the main pendulum. > > > > I'm thinking the electrical analog would be the midpoint of the voltage sinewave. > > > > Rick C. > > > > Tesla referral code ---++ https://ts.la/richard11209 > > > So - you have the pendulum, which is equivalent to a resonator. > > I like to quote Don Vonada, once the main designer at Digital > Equipment Corp. 'All digital parts are made of analog components'. > > The oscillators made of digital gates are using the analog delays > in the gates a timing reference.There is no use of any timing delays of logic.> It is a completely different question, how the necessary energy > feed to the timing reference is applied, like kicking the crystal.Very much like the Shortt clock. I suggest you read a bit about it. It is very interesting and educational. In the short clock there are two pendulums to decouple the ancillary components of the secondary pendulum from the free pendulum which actually controls the timing of the overall circuit. On detecting that the secondary pendulum is at its mid-swing position, an electrical signal trips a relay to release a gravity arm which gives an even push to the free pendulum. Because the timing of the gravity arm is controlled by the free pendulum and not the release relay the current can also be used to apply feedback to the secondary pendulum. Depending on whether the secondary pendulum is running too fast or too slow a spring is engaged to change the time constant for that one swing. This complexity is not required in an electronic oscillator. The point is that the oscillator signal can be used to trigger a digital circuit to deliver an impulse to the crystal to keep it running. I don't know much about the losses in a crystal circuit. I assume it can be self limiting in amplitude, but this may work differently in a digital and analog circuit where the amplifier loses gain when the signal reaches full swing. In a digital oscillator variation in amplitude corresponds to the energy stored in the crystal. I don't know this has any gain factors, but I suppose the Q of the crystal defines the losses and so the energy impulse required for a given level of output. This is why I would like to find a book discussing such a circuit. I know these have been built, but I don't know if anyone has studied them. Rick C. Tesla referral code --+-- https://ts.la/richard11209
