| Jocko Homo |
I think there was a thread about this some time ago.......just can't find it. Anyway, it is probably too old for any of us to remember what we were arguing, er, ah, discussing.
Anyway......here is a slightly different version. The idea was inspired (or ripped off.....) from yet another ham radio book. The change is that it uses a common-base stage (Elso.....are you reading this......?) at Q2; instead of a 22 ohm resistor to pick off the crystal current. The input Z of this stage is around 5 ohms, so there is somewhat less Q degradation.
Jocko |
|
|
| HarryHaller |
Doesn't C8 go to the emmiter and not the collector of Q3?
H.H. |
|
|
| Jocko Homo |
I know that sooner or later, some wiseguy will fuss about a capacitor in series with the crystal. So for those of you who are new at this.....and for those that ought to know better, but don't.......
This circuit operates in the series-resonant mode. When ordering a crystal, you can either specify series-resonant frequency, or a parallel-resonant one.
Both cases refer to the frequency at which the crystal exhibits a series resonant frequency, usually in the range of tens of ohms. This is not the anti-resonant frequency, which is the equivalent of a parallel tuned circuit, and therefore in the range of many kilohms.
For a crystal spec'ed for series resonance, they measure the resonant frequency with only a resistive load, usually 50 ohms. The only capacitance is the stray holder capacitance, etc., usually around 6 pF.
When you spec for "parallel resonant", they measure the series resonant point using the specified load capacitance in parallel with the crystal. Typical values are 20 and 32 pF. This frequency is a few kilohertz higher when measured this way.
Most crystals you see in parts catalogs are spec'ed for parallel resonance, as the most common clock circuit around uses that type. It might be easier to order a crystal that way, because it will work in any circuit. In the case of a Butler oscillator, you can simply put a capacitor of the proper value in series with the crystal. The circuit will work just as well.
Jocko |
|
|
| HarryHaller |
Do I win a prize or anything?
H.H. |
|
|
| Jocko Homo |
You might want to help me complain how many times I tried uploading the correct version.
Oh well, it is not my server that will be clogged up. |
|
|
| Jocko Homo |
The common-base stage runs at around 50 mA. (Are you reading this, Elso?)
So........is anyone going to try to build this? I know what it sounds like, but what do I know.
Jocko |
|
|
| mrfeedback |
Jocko, I believe you that this circuit sounds good. ;)
Q - the output is sinewave ?
If so, is it better to feed your oscillator output into a squaring circuit first before feeding into the oscillator pin of DSP/DAC to avoid internal sqaring/timing error ?.
Eric. |
|
|
| Jocko Homo |
And low harmonics too. That way you can put it outside of your player, feed it in at a low level without radiating **** all over the place. Then square it up once inside.
Also will let you feed it into a D/A box using.....say....a DIR1703 to sync up things.
Jocko |
|
|
| mrfeedback |
Reciever info recieved and thankyou.
If using your oscillator circuit inside a one box cd machine (CD-94), would you square the sinewave before feeding the DSP/DAC osc input pin, and if so how ?. (some kind of data slicer/schmitt/74HCT etc or something or discrete ? - keeping in mind your previous snippets/advice of some gate devices causing zero/little backfeeding - can you remember ?).
A two box CD player is on the future wish list - I have various and plenty standard mechanisms and servo/dsp/system junked machines to play with and experiment with.
On this subject, I reckon different Laser pickups cause different sounds.
Eric. |
|
|
| Jocko Homo |
I think it only puts out about 1 V rms or so. (I forget........yet another senior citizen moment.....)
The first go-round uses a '04 inverter set for a gain of 4 to square it up. Maybe next time I'll come up with something more exotic.
Jocko |
|
|
| pjkunz |
At what point (if any) would one need to adjust values for higher freq. crystals (~24Mhz) and is this freq. still in a good space for this design? I'm trying to put together a circuit with the CS8420 on the front and the output freq. is a fixed 256Fs. I'd like to be able to 'play' with the higher Fs ranges.
Also, Jocko mentions gain on a 7404, this is something I don't understand? It as simple as adding a R2/R1 input/feedback resistor ratio for an inverting opamp? I thought output level was independent of input level in TTL?
Thank you. |
|
|
| Jocko Homo |
Scale C1 and C2. For 24 MHz, cut them in half. C3 & C4 are adjusted so that the crystal will resonate into the load it was cut for. You will need to know that.
As for the inverter........yep, you can configure it just like an op-amp. The output level of that circuit is too low to make a square wave without some gain somewhere.
Make sure that whatever you do has a stable PSU, or all that work to get low phase noise will be gobbled up by jitter in that circuit.
Jocko |
|
|
| pjkunz |
Thanks for the reply,
I'll breadboard it up and play with it a bit. Thanks for the circuit and thanks for the help. I'll do a little homework on this before I ask any follow up questions...
-Pete Kunz |
|
|
| pjkunz |
| Ah, so this is designed for a parallel res. Xtal?, I was confused when it was described as operating in the series res. mode. I looked closer at your posts a saw where i missed it. Mia Culpa. |
|
|
| Pjotr |
Hi,
Here is some good reading about noise in oscillators.
:cool: |
|
|
| Jocko Homo |
Almost all modern circuits place the crystal in the series resonant mode. The confusion arises when you start to specify the frequency.
A crystal that is said to be series resonant has its series resonant mode measured into a resistive load, say 50 ohms.
When you see a crystal with its resonant frequency specified at a parallel load of X pF (usually in the 20 - 32 pF range), the series resonant mode is measured with a load of that specifed capacitance, instead of the resisitive load.
Jocko |
|
|
| jwb |
| Hi Jocko. Guess I'm glad you didn't take that 90-year sabbatical, since I'm building your oscillator now. What is the lowest-jitter way to turn this into a square wave? An inverter is pretty simple, but I was thinking of buying some AD8611 for testing. |
|
|
| Jocko Homo |
I dunno........soemone send me an expensive 'scope that measures jittter, and I'll tell everyone. (That's life when the lab you used to work in is gone..........)
Your approach is the same as mine. I would imagine a clean supply for the inverter, and not using exotic high-speed logic, would be good starting points.
Jocko |
|
|
| jwb |
My[1] scope can't measure jitter directly either (despite having the "advanced math module"[2]), but there's no shame in the cursor method.[3]
-jwb
1: "My" being the scope I have at the office, not the amazing Tek 465 at home.
2: Advanced math apparently means addition, subtraction, and multiplication.
3: Set the scope to accumulate or high intensity and use the cursors to find the excursion on the time axis. |
|
|
| Pjotr |
Huh, err... And that way you want to measure down to 1 ps ??? Hopefully your scope’s time base has that low jitter too. A better approach would be to build two identical oscillators. Feed them to a mixer and measure the sideband noise with your soundcard and a FFT program. I am just thinking of this. But maybe it is easier to contact some old colleagues from the university to dig up a nice new Tek TDS scope.
;) |
|
|
| Jocko Homo |
That will give you a phase noise measurement, which will not be the same as a jitter measurement. Jitter measurements take into account voltage variations from ground bounce, Vcc shift, and stuff like that. Those fancy 'scopes have fancy A/Ds that are supposed to be good for that. But then they can cost $40K. For that price, they ought to be good.
Jocko |
|
|
| Pjotr |
True Jocko,
But about what are we talking here? Jitter of the oscillator or conversion-clock jitter at the DAC?
RMS jitter and sideband noise are closely related if it concerns an oscillator. If you measure the sideband noise you can calculate the RMS value of the jitter. It’s true you can’t measure the pp value of the jitter that way.
Note that if an oscillator is specified for say 5 ps jitter it is most time given in RMS. The pp value can easily be as high as 20 ps to 50 ns.
But why does anyone want to measure the numerical value of the jitter if you can listen to the musical result? Measuring is only handy to see if you make progress on a testbed. Relative measurements will suffice then. If you look at the down converted sideband spectrum, the sideband lobe need to be as narrow as possible and the white noise floor as low as possible. If you want to investigate psu-related noise you need to power each oscillator from different psu’s to avoid correlation of the psu noise. |
|
|
| Jocko Homo |
The thermal phase noise of the oscillator, and the jitter introduced by the circuitry that follows, that can degraded its performance.
I prefer the noise spectrum method, but it would be nice to have a fancy 'scope that could give a quick check of the amount of deviation in pS. The noise spectrum will give you more insight, from a design standpoint. Which I would find more interesting.
Jocko |
|
|
| ALW |
Unless anyone has a 'scope that is capable of measuring jitter, don't bother.
Triggering the 'scope on the jittered signal gives misleading results, related to the 'scope trigger interval and the transition interval being measured.
The frequency spectrum also has an effect, at some frequencies jitter will not be shown at all, and at others it could appear doubled in amplitude, what one is really displaying is interval variations.
Andy. |
|
|
| jwb |
Seems like the main trick here is going to be clean power supplies. The oscillator obviously needs one (9V) and the inverter or comparator needs a different one (5V). I guess I'll derive the latter from the former.
Also I suspect that the comparator will have better PSRR than the inverter for turning sine into square. With the inverter, any supply noise is going to translate directly into timing errors on the output. But with the comparator the reference voltage can be set with something quiet (LED, buried zener, etc.). |
|
|
| pjkunz |
Hi Jocko,
I've etched a small board for the oscillator and destroyed yet another shirt with ferric chloride stains...(the schematic with the power supply) and am in the processs of stuffing it. But I have a question on BJT's...
I'm using Zetex 694b's and 795a's in place of the 2sc2240's and 2sa970's, these seem to match up pretty well, but what is the reason for using these very high hfe types, just curious.
Also is there anything specific to keep in mind if substituting for the 2sd438's? I have a ton of MPS8099's from a Leach amp project....(yes I've read the new law) I'd be more than happy to use something better, but I'm not sure what I'm looking for in this application?
Thanks again,
Pete Kunz |
|
|
| sonnya |
Those Zetex have pretty high cbe... about 10 - 20 times higher than 2sc2240...
I would guess that they would mess up the oscillator circuit Jocko have designed pretty much!!!!
You would have to recalculate c1 and c2
Sonny |
|
|
| HpW |
Well,
some years ago I build my own XVCO-oscillator (for my DAC).
I suggest :
1 : no power noise should influence each other gain stage
- R/C each power/gain element separatly
- R/C may also the base separatly
2 : Grounding & shielding: build each gain part in an separated 0.5...1mm cu box (military alike) and use the box as the ground
result : no magnetic or electric RF should go in/out and interference with other parts.
3: use smd parts (trans. I used 2n918 alike smd)
4: no trimm-elements, first measure and then replace with related
smd cap's
5: use for power connection a pass-trough caps to keep
the nois within the cu box
well life it's not so easy in the rf field...
may I helped in the right direction....
Cheers,
HpW |
|
|
| Jocko Homo |
The 'D438 is there because a little package will get too warm. At least it will if you put 50 mA through it. You could use whatever NPN that works well in the oscillator if you can get the heat out of it.
The Zetex ones should work. You will not have to recalculate anything. Except maybe the base divider, to get lots of current through th common-base stage.
The reason is that they have narrow base regions, and should have good gain, bandwidth, and low noise.
SMT would be nice.
Someday I will post a Butler version. Problem is getting enough voltage out of one with low distortion. (Yeah, I know.....it will be squared up somewhere, and lots of harmonics then. I want it to be low harmonic, so that it can be piped all over the shop without generating EMI.)
Jocko |
|
|
| pjkunz |
Does anyone now of a North American source for these Toshiba BJT's in small quantities (under 50 pieces). So far I've tried:
Digikey, Mouser, Allied, Newark, Jameco, my usual suspects...no luck. That was my reason for searching out substitutes.
Thanks,
Pete Kunz |
|
|
| Jocko Homo |
Should be less than $0.50 each.
Jocko |
|
|
| pjkunz |
Thank you. I forgot about MCM. Is it worth (a realtive term, I know) ordering the toshiba parts instead of th Zetex parts I have now? The Zetex datasheets don't list noise specs. How much, if any, difference/degredation will there be with the Zetex parts. I'd rather do it once and do it relatively right (drilling boards is a pain..) , eventually I may do this in SMD but the other item on my plate (trying to graduate...finally, 12 years of college is a bit much for anybody) is dominating my landscape right now. Any insight is appreciated as is the amount I'm learning from this little project.
Thanks,
Pete Kunz |
|
|
| Pjotr |
Don't forget the good old BF240 and BF199. There is really no need to put more than 10 mA to 20 mA trough them concerning noise. R_be of the output transistor is low enough then. R4 does not to be a low 10 ohms, it is just a waste of signal. 100 ohms wil do better.
;) |
|
|
| Elso Kwak |
Hi All, I find this thread quite amusing as if there never excisted a KWAK-CLOCK.
Powersupply does make make a difference but it is only a minor effect. The "secret" of the KWAK-CLOCK is in the FET and the comparator in other words in the schematic.
Just hit the e-mail button underlining this post and you will get the schematic. Professionals get version 6. Others version 5.:) Decide for yourself which catagory you belong.;) ;) ;) |
|
|
| jwb |
Hi Elso. You may take some education from this. The reason there is interest in this oscillator is because it is out in the open. The schematic is posted by Jocko, and we can all evaluate it. We can discuss it on common terms.
With your Kwak-Clock, the schematic is only available by email. It is very mysterious. It is not "out there". |
|
|
| jwb |
| quote: | Originally posted by Elso Kwak
jwb,
My clock was discussed in quite some detail here:
http://www.diyaudio.com/forums/show...+AND+oscillator
Here you will also see that Jocko's one is a development of a excisting circuit from
The reason I do do not post it is that I do make some effort to protect my intellectual property. Do you know how many hours went into the design and testing and loosing my temper etc.? |
Don't be a bitterman, Elso. You can't deny that posting a schematic generates discussion. If you want people to discuss your oscillator instead of Jocko's, you need to post the schematic.
In the thread you linked, the only schematics were posted by Jocko, ftorres, and som others. There are exactly zero schematics of the Kwak-Clock. |
|
|
| Elso Kwak |
| jwb, I don't complain about lack of interest. I get several e-mails a day since two years. Most notably the best hints to improve the KWAK-CLOCK I got off-line by private e-mail, not from comments/discussion on this forum. ;) Please don't lecture me.:o |
|
|
| peranders |
| quote: | Originally posted by Jocko Homo
Almost all modern circuits place the crystal in the series resonant mode. The confusion arises when you start to specify the frequency.
A crystal that is said to be series resonant has its series resonant mode measured into a resistive load, say 50 ohms.
When you see a crystal with its resonant frequency specified at a parallel load of X pF (usually in the 20 - 32 pF range), the series resonant mode is measured with a load of that specifed capacitance, instead of the resisitive load. |
Aren't you a little bit unclear here?
Serie resonance frequency = when the crytal is purely resistive, 10-20 ohms for AT-crystal
Parallel resonance frequency = when the crystal is highly inductive, around Henrys. How much inductive depends on the load capacitnce.
Every crystal has both modes but they can't be tuned to the "wrong" mode. You get very bad Q if a parallel crystal is tuned to series mode and vice versa. If the crystal is made for 4 MHz parallel mode, don't tune it to 4 MHz series mode.
I have practical experince of that. I made long ago a 4.194304 Hz ocsillator with a parallel crystal and put it in a series one. The oscillator became rather sensitive due to much tuning. Without tuning the result was quite OK, but with the wrong frequency. I used it as an 1 Hz oscillator to an ordinary clock. |
|
|
| pjkunz |
So I fired up my board,
the output looks good, freq looks good, but the output level is only 200-300 mV p2p, it was mentioned earlier that it should be ~1Vrms. I made the following changes:
Freq. ~24Mhz 18pF crystal
C4 =10pF
C3 = 5-18pF
As per Jocko's recommendations:
C1 from 470pF to 220 pF
C2 from 86pF to 33pF
My own doing:
C6 C7 from 18 to 15 pF (all I could get locally)
Q2,Q4 MPS8099
Q1,Q3,Q5 to zetex parts listed previously
Resistor values have all been checked, Vrail ~12V, ~50mA total draw.
Any Ideas?
-Pete Kunz |
|
|
| pjkunz |
There appears to be alot of attenuation over the 15 (originally 18)pf capacitors C6, C7. Paralleling another 15 to each brings the signal level back up... Earlier schematics from other threads:
http://www.diyaudio.com/forums/show...ight=oscillator
show 47pf values for the blocking caps? Is there a reason they were reduced for this design, and is increasing the value a proper solution to the low output I see with the spec'd values?
Thanks,
Pete Kunz |
|
|
| jwb |
| I'm afraid I don't understand the situation with the crystal here either. My source for crystals (Digi-Key, naturally) has 24.576MHz crystal specified for series load. I'll need to read up on crystals before I understand what that means for this circuit. |
|
|
| jwb |
| If I reduce this to the equivalent circuit, it starts to look like the 24.576MHz crystal with 36pF caps to ground on either side. So the crystal would be operating in parallel mode, and therefore going approximately 11kHz faster than spec. |
|
|
| pjkunz |
Mine my well be that far off, I've only had a chance to look at it so far on a 20Mhz ASO (the crystal is a 24.576MHz 18pf parallel from digikey), the 100Mhz DSO is in another lab right now.
-Pete Kunz |
|
|
| mlloyd1 |
While looking for something else, I found this link for a high performance sine to square converter.
Hopefully this satisfies "exotic" :D
mlloyd1
| quote: | Originally posted by Jocko Homo
....
The first go-round uses a '04 inverter set for a gain of 4 to square it up. Maybe next time I'll come up with something more exotic.
Jocko |
|
|
|
| jwb |
| Using a transformer to square a sine wave is silly and expesive, especially if you need to use two-supply comparator. I am using a single-supply comparator with a DC-coupled input: |
|
|
| mlloyd1 |
Hmmm... sounds like the audio definition of "exotic" to me :)
| quote: | Originally posted by jwb
Using a transformer to square a sine wave is silly and expensive .... |
|
|
|
| Jocko Homo |
How is a transformer supposed to square up anything?
Jocko |
|
|
| jwb |
| Well it isn't really, but check the link posted by mlloyd. It's got a trafo-coupled 3-supply comparator. Which is about the single most expensive method I can think of. |
|
|
|
