Hi Andrea,
Just back from a trip. I will try your suggestion tonight. Most of the parts are from BOM (@Mouser), except 10pf (NP0, like C5), 100nf (NP0), 1N5711, BFR182, and all SMD resistors. I used SMD resistor for R11, and C3 joint is measured at 2.97V which I think is within tolerance. During the test, the circuit is powered from my lab power supply with 6V and 3.3V programmed. I will check all the parts again. Thanks for the input.
Poting
I was curious, I do a comparison with and without oven about the SC cut Driscoll.
Some people say (and I think it too), the oven just for long term stabiliy at SC cut, it dont affect the PN and SQ.
Last days I listened to little music details, music soundstage etc. It could be true what they say, I dont hear any different turn on/off oven, or it is unaudible.
I think it may running without oven too.
Some people say (and I think it too), the oven just for long term stabiliy at SC cut, it dont affect the PN and SQ.
Last days I listened to little music details, music soundstage etc. It could be true what they say, I dont hear any different turn on/off oven, or it is unaudible.
I think it may running without oven too.
My 11.2896MHz SC cut crystal is running fine in a modified KWAK-CLOCK (Colpitts osc.) without oven.
As I moved to 4.2336 MHz with TDA1540 DAC I no longer tried to improve that route.
Last edited:
You should Always use a Clapp oscillator (= important variation on the Colpitts) WITH THE SCHOTTKY DIODE for the avc.
What modification did you make on the lousy Kwak-clock?
Herbert.
No, compared it with a Citizen "tuning-fork" crystal as suggested by Jocko.
Last edited:
Hi all,
Hope the summer is pleasant where you may be
I've been experimenting a bit with different decoupling capacitors on my 6.144 MHz Driscoll oscillator and - somewhat surprising - found out that a 100 nF decoupling capacitor is quite some less efficient in decoupling the 6.144 MHz oscillating frequency than e.g 0.47 uF or 1 uF. The difference actually was about 35 dBs (!) as measured with my picoscope's FFT (screendump attached. Top is 100 nF C0G, second is 0.47 uF X7R, and third is 1 uF "acrylic" - FCA series from CDE; fourth is Vishay MKP1837).
Using Kemet's impedance vs. frequency tool (also to be found on the internet, link here: KEMET K-SIM: Capacitor Model Simulation Tool ) it can be seen that the lowest impedance for a 100 nF capacitor is found at ~12 MHz where the impedance is around 7 mohms. However, moving to 6.144 MHz the impedance is now ~190 mohms - quite a difference, I think.
On the other hand choosing e.g. a 0.47 uF X7R 1210 size the impedance at ~ 6.2 MHz is ~23 mohms. And for a 1 uF version it is some 32 mohms.
My point in this is that the ubiquitous 100 nF decoupling capacitor looks as if it is optimum for the 11 & 12 MHz frequencies whereas for other frequencies other sizes may give more damping of PSU noise (also on the inverter PSU - I can see that Andrea suggests 1 uF here for the Driscoll version).
I Would like to add that I haven't yet listened to the oscillator with the various capacitors so although I reckon the 35 dB difference somehow will be audible I don't yet know if it is so/what it may sound like.
As a curiosa I've also included a measurement from an MKP1837 100nF polypropylene capacitor which would normally not be used for digital decoupling e.g. due to a rather highish inductance. As can be seen from the FFT it also is not as efficient from a decoupling point of view - however, it looks as if the overall PSU noise level - outside of the 6.144 MHz frequency - is lower than that of the other capacitors.
Anybody else have experimented with this?
Cheers,
Jesper
Hope the summer is pleasant where you may be
I've been experimenting a bit with different decoupling capacitors on my 6.144 MHz Driscoll oscillator and - somewhat surprising - found out that a 100 nF decoupling capacitor is quite some less efficient in decoupling the 6.144 MHz oscillating frequency than e.g 0.47 uF or 1 uF. The difference actually was about 35 dBs (!) as measured with my picoscope's FFT (screendump attached. Top is 100 nF C0G, second is 0.47 uF X7R, and third is 1 uF "acrylic" - FCA series from CDE; fourth is Vishay MKP1837).
Using Kemet's impedance vs. frequency tool (also to be found on the internet, link here: KEMET K-SIM: Capacitor Model Simulation Tool ) it can be seen that the lowest impedance for a 100 nF capacitor is found at ~12 MHz where the impedance is around 7 mohms. However, moving to 6.144 MHz the impedance is now ~190 mohms - quite a difference, I think.
On the other hand choosing e.g. a 0.47 uF X7R 1210 size the impedance at ~ 6.2 MHz is ~23 mohms. And for a 1 uF version it is some 32 mohms.
My point in this is that the ubiquitous 100 nF decoupling capacitor looks as if it is optimum for the 11 & 12 MHz frequencies whereas for other frequencies other sizes may give more damping of PSU noise (also on the inverter PSU - I can see that Andrea suggests 1 uF here for the Driscoll version).
I Would like to add that I haven't yet listened to the oscillator with the various capacitors so although I reckon the 35 dB difference somehow will be audible I don't yet know if it is so/what it may sound like.
As a curiosa I've also included a measurement from an MKP1837 100nF polypropylene capacitor which would normally not be used for digital decoupling e.g. due to a rather highish inductance. As can be seen from the FFT it also is not as efficient from a decoupling point of view - however, it looks as if the overall PSU noise level - outside of the 6.144 MHz frequency - is lower than that of the other capacitors.
Anybody else have experimented with this?
Cheers,
Jesper
Yes, 10 years ago...
< http://www.hoffmann-hochfrequenz.de/downloads/experiments_with_decoupling_capacitors.pdf >
I should repeat that now with the 18 GHz TDR.
regards, Gerhard
Oh, and could you please keep these B*nchm@rk advertisements
away from this thread, one of the few here that has at least a weak
connection to reality?
@gerhard: Thanks again for commenting ... I think I've actually seen the paper you link to before. Besides looking into Kemet's simulator it's one of the sources that made me aware that finding the right decoupling capacitor is quite key to efficient decoupling at specific frequencies.
These decoupling phenomena sometimes make me wonder if one of the (main) reasons why D/A converters may sound differently at different sampling frequencies is that the DAC's decoupling probably is only marginally effective at some of the sampling frequencies ...
Cheers,
Jesper
These decoupling phenomena sometimes make me wonder if one of the (main) reasons why D/A converters may sound differently at different sampling frequencies is that the DAC's decoupling probably is only marginally effective at some of the sampling frequencies ...
Hmmm... I've been looking at the CSA-803 from tektronix to get a better view into what happens in the high frequencies with digital transmission, decoupling, ground bounce etc. But I don't really have any experience with such equipment ... Maybe I can ask you for a slightly off-topic advice: Would you say that the CSA-803 is a "good/the best" device for RF/TDR measurements (and if so "A" or "C" or no suffix version) or is there a better one at a reasonable price (reasonable price should apply to this device's accessories)?
Cheers,
Jesper
I have access to a DSA8200 with the 20 GHz TDR modules. I would advise against owning this stuff unless its really cheap (under $300) since its essentially unrepairable. We had to get a TDR module repaired and the quote was $5K. It got over 5V on its input. The whole setup is now obsolete (5 years an $150K down the tubes) because HDMI moved to using a VNA for validating cables. The VNA for this is another $150K. Not good for the bottom line.
However I have a 7854 + 7S12 + S52 pulse generator and S6 TDR sampler modules good for 12 GHz. You could put together similar in the $600 range pretty easily. What you will learn is that even a short in microwaveland is not a short. And that its probably not important for audio.
Access to a VNA can do the same. (I have an 8753E in my garage right now but its too big and clumsy to set up for a few measurements.) It takes way more time to calibrate and de-imbed the fixtures etc. than to make a measurement.
I would focus on the best layout practices and use components that are up to the task: Leads bad. Ground plane good. Vias are to be avoided if possible, high inductance. Look to the chip vendors reference designs and layouts. use the smallest footprint parts that are good. Many ceramics change their values a lot with voltage, not good. The better cap's datasheets have self-resonant plots on them. I think Gerhard's paper validates those datasheets.
However I have a 7854 + 7S12 + S52 pulse generator and S6 TDR sampler modules good for 12 GHz. You could put together similar in the $600 range pretty easily. What you will learn is that even a short in microwaveland is not a short. And that its probably not important for audio.
Access to a VNA can do the same. (I have an 8753E in my garage right now but its too big and clumsy to set up for a few measurements.) It takes way more time to calibrate and de-imbed the fixtures etc. than to make a measurement.
I would focus on the best layout practices and use components that are up to the task: Leads bad. Ground plane good. Vias are to be avoided if possible, high inductance. Look to the chip vendors reference designs and layouts. use the smallest footprint parts that are good. Many ceramics change their values a lot with voltage, not good. The better cap's datasheets have self-resonant plots on them. I think Gerhard's paper validates those datasheets.
Hi 1audio ... thanks once more for your feedback - which actually made me smile while I was looking for the equipment that you mention ... I happened to come by a tektronix wiki on the internet and when searching for the 7854 I found the attached image. And looking at it I just happened to smile because my work space is not that large and this setup (and I reckon the setup is more than the 7854) is HUGE (IMHO)!!!
One of the reasons I somehow fancy the CSA-803 is that it seems to be just not too big ... It's also got a double probe with a 100k input impedance which makes it usable for real-world circuitry measurements up to 3 GHz.
Anyway besides HF/RF measurements what I'd really like to be able to measure is the close-in phase noise/jitter that has been discussed in this thread. So somehow I hope that Andrea's phase noise thread will take off. Could be really interesting to see what could be done here ...
And then a clarifying question: What do you mean by "Leads bad. Ground plane good"? Are you hinting at the prioritization of a good ground relative to the other leads?
Hmmm ... it seems to me that low-Q components really are the interesting ones as they would be less "narrow resonance" and thus more equally efficient with a wider range of frequencies ... But I reckon that they don't come as low inductance, low impedance, low DK, low diffraction factor, low distortion, low price, no size, and high capacitance ... (just wishing here) ...
Cheers,
Jesper
One of the reasons I somehow fancy the CSA-803 is that it seems to be just not too big ... It's also got a double probe with a 100k input impedance which makes it usable for real-world circuitry measurements up to 3 GHz.
Anyway besides HF/RF measurements what I'd really like to be able to measure is the close-in phase noise/jitter that has been discussed in this thread. So somehow I hope that Andrea's phase noise thread will take off. Could be really interesting to see what could be done here ...
And then a clarifying question: What do you mean by "Leads bad. Ground plane good"? Are you hinting at the prioritization of a good ground relative to the other leads?
Hmmm ... it seems to me that low-Q components really are the interesting ones as they would be less "narrow resonance" and thus more equally efficient with a wider range of frequencies ... But I reckon that they don't come as low inductance, low impedance, low DK, low diffraction factor, low distortion, low price, no size, and high capacitance ... (just wishing here) ...
Cheers,
Jesper
Attachments
I just want to report that I've managed to solve the problem by reducing C3 to 2pf, C1 to 12pf+3.9pf and Q1 = BFR182. I used a trimmer to find those values (measured with bench top LCR meter), and it seems this is the most stable setup for my crystal. My MMBT5179 never works though. It just won't start. By the way, it's a 24.576M SC-cut Driscoll.
Poting
I don't have much experience with TEK samplers, but the gurus on usenet group
sci.electronics.design use them as a standard tool (Larkin, Hobbs..) and I value their
opinion. Tek has the advantage that you can get new old sampling heads at a
relatively low price, the market for used stuff seems to be bigger for Tek than HP.
At Infineon Fiber Optics (RIP) we used HP for daily work and a TEK CSA8000 (?)
for the data sheet pictures, that order gave the nicest eye diagrams.
I use a 54750 mainframe; started with 20 GHz plugins, later 50 GHz and a 54754A
differential TDR. Adding another scope family won't add capabilities to that
what I already have. Maybe a newer mainframe for comfort if I can get it cheaply,
but the RF specs are determined by the plug ins, apart of triggering.
There used to be an app note from Picosecond Pulse Labs in Boulder, Co
that compared the different models. The "best" thing they found was Agilent
TDR + 50 GHz plug in + pulse former from, you guess, PSPL.
PSPL has been bought by TEK; if you cannot find the app note I can search
it in the data pile on my disks.
I don't think that this all is of much use for audio; for me it is more important
for FPGA gigabit links.
Cheers, Gerhard
Last edited:
@1audio:
Thanks for clarifying - I somehow thought this was what you meant ...
Hmmm ... I might send you a PM about this ... will just consider.
@Gerhard:
Thank you also for your feedback ...
Searching the internet for the application note I found this website - apparently set up by the founder of PSPL - with some of their application notes and matlab programs etc. Could it be that the AP notes that you refer to are the AN-2* series?
https://kh6htv.com/pspl-app-notes/
Cheers,
Jesper
P.S.: Thanks for the tip on
Thanks for clarifying - I somehow thought this was what you meant ...
Hmmm ... I might send you a PM about this ... will just consider.
@Gerhard:
Thank you also for your feedback ...
Searching the internet for the application note I found this website - apparently set up by the founder of PSPL - with some of their application notes and matlab programs etc. Could it be that the AP notes that you refer to are the AN-2* series?
https://kh6htv.com/pspl-app-notes/
Well, sometimes I wonder about this myself and then leans towards something like the Tektronics TDS784 - 1 GHZ BW, realtime sampling, and it also has 50 ohm inputs - so as to reasonably mingle the "normal" audio world with a glimpse into the HF/RF region. Could be that it's a better solution ...
Cheers,
Jesper
P.S.: Thanks for the tip on
... looks like an interesting venue ... ;-)
Not an expert here, but I just want to share some of my findings about this clock.
1. I found PO74G04 to be a better choice than NXP 74LVC04.
2. Regulator is most critical. I currently use 2 LT3042s for the clock and they are clearly better than TI TPS7A47XX. (Salas 1.2R as pre-reg)
3. Use independent power supply for the oven. Do not share the VCC with clock board. I used different windings for the oven and clock, the sound improved immediately.
Poting
1. I found PO74G04 to be a better choice than NXP 74LVC04.
2. Regulator is most critical. I currently use 2 LT3042s for the clock and they are clearly better than TI TPS7A47XX. (Salas 1.2R as pre-reg)
3. Use independent power supply for the oven. Do not share the VCC with clock board. I used different windings for the oven and clock, the sound improved immediately.
Poting
Hi Jesper,
Nope, I wasn't able to find a stable 6V battery source for the XO part. The 6V battery could be 7V or more. The 3.3V rail is ok though. In fact, I prefer regulators. No need to take care of recharge issues.
Well, it seems that not many members talk about it. I've waited 2 years to complete the XO. The improvement of Andrea's work over NDK is hugh, IMHO. I feel the clarity got improved, too.
Nope, I wasn't able to find a stable 6V battery source for the XO part. The 6V battery could be 7V or more. The 3.3V rail is ok though. In fact, I prefer regulators. No need to take care of recharge issues.
Well, it seems that not many members talk about it. I've waited 2 years to complete the XO. The improvement of Andrea's work over NDK is hugh, IMHO. I feel the clarity got improved, too.
Last edited:
- Status
- Not open for further replies.