Make a very simple test for me

[_q12x_]

Im asking this question to everyone but in particular to my friend here mister @marconi .
I am a professional artist but I am not an electronist like you guys here. That doesnt mean I dont know anything. I know something but I can't raise to some of your standards and knowledge. I'm happy (sometimes fun) to learn something new here and there.
- Recently, it was suggested to me to buy a frequency counter, because I got into some crystal oscillators I have in my stock and they have no markings anymore and the reason is a bit too long story. I already buy a cheap one from ebay, exactly this yellow version (not the red one)

but it is on the road. I have about 2-3 months (usually) to wait, until it arrives.
I also have a dinky DSO138 osciloscope that is trembling of Parkinson all the time. So you can imagine, I can't put my 100% trust in it all the time.

So, my first circuit for testing a crystal oscillator I find is this:
"Oscillator Circuit of The First Quartz Wrist Watch"

I had high hopes for this circuit. I used 10k for both Rc(c=collector) and 1k for both Re(e=emitor). And I used BC548 for both Tr.I used a known value of a Quartz of 20MHz. And I used 2V (VB=Voltage Battery). But the oscilloscope just showed me some very weird and random readings that I can not even put head to tail. I build this circuit on my breadboard, and that may had influence the results.

---So this circuit didn't work for me---. But I bet my as it must be a good one and I blame my DSO138 for being crappy.
And also not having (yet) a frequency counter.
- In short, this is more a curiosity for me. I hope it is for you as well.
- My request for you is to help me with the following:
- Because you are a better electronist, you must have better tools than I have. So, using your normal oscilloscope and your normal frequency counter, (I say normal, comparative to my ebay measuring tools), please make this very quick and simple circuit and measure it for me. And confirm to me with some images or a short video, that everything is working as I imagine and hope. It must be. The idea is to measure 20MHz on the "out" pin in respect to the ground (if you used the same values as I used). That's it. Also, feel free to change the resistors or the transistors. It must be GPT (general purpose transistors), but the resistors I used I just guessed their values. I didnt had the values from the page with the circuit. So I had to invent something. And those values are my best guess.
Thank you and hope to hear good news from you.

 

[pc1966]

PC1966: I had another try with Rc = 1kR and Re = 4.7kR, 10kR and 22kR over a range of voltages from 3-15V. Not a twitter at 20MHz. I also tried two other crystals.

Next I will try keeping Re at 4.7kR and reducing Rc but at the moment it is too hot in my shed to do so.

OK, I will see if I have the parts to try it as well.

At one point we had loads of stuff for this sort of experiment but sadly it was thrown out when my last work ended. We saved some stuff "just in case" but space did not permit all of the items I wish we had kept.

 

[pc1966]

@marconi Was the DC bias condition much as expected? I guess around (Vcc - 0.7) / (Rc + Re) per transistor current, at 5V supply around 3.5V over emitter resistors?

 

[marconi]

@marconi Was the DC bias condition much as expected? I guess around (Vcc - 0.7) / (Rc + Re) per transistor current, at 5V supply around 3.5V over emitter resistors?

I will check tomorrow morning. We are heading for thunder and lightning this evening and it feels very muggy right now.

 

[pc1966]

This is a handy web site for current activity:

Lightning & Thunderstorms - United Kingdom, England, Scotland, Wales, Ireland

Blitzortung.org provides lightning and thunderstorm information in real-time on maps for USA, United Kingdom, Australia, new Zealand, Europa, Africa, Asia and other Countries.

www.blitzortung.org

However we have seen the odd flash here that did not register on their detectors.

 

[_q12x_]

Im loughing how everyone is preoccupied: me with the fv reading, marconi with lightning and pc1966 with stuff that doesnt find or have anymore. Hahaha.

 

[marconi]

Regarding the circuit Symmetrischer Oszillator I have included below again. I don't think this is the working circuit - it is an outline schematic of a circuit missing some key detail. I reckon the circuit idea is a classical astable transistor circuit - see second circuit - which has resistance and capacitance to make it oscillate at a frequency close to that of the quartz crystal as in the second image. The purpose of the crystal connected between emitters is to 'force' operation at its frequency. This is what I am going to pursue.

 

[Lucien Nunes]

I was just going to say, it looks like a circuit from a patent which shows enough to define the principle that is claimed, but not enough to construct a replica directly. I too would be inclined to provide AC feedback and bias in the usual way, and probably make the emitter coupling somewhat tunable.

It seems a very long time ago that I had the opportunity to make with circuits that might not function, to see whether something can be done some particular way. In my work I am so often against time pressure that the solution has to be guaranteed to work first time even it is technically ridiculous or extravagant. I have used a 14-bit 40-pin PIC running at 20MHz as a monostable to generate a 10 second pulse, because it was already built and reprogramming took 2 minutes vs. 5 minutes to make something specific.

If you asked me to test the frequency of a vanilla-flavour crystal now I would reach for the PIC prototyping board again and have an answer in 20 seconds.

 

[_q12x_]

We know that is a symbol of a BJT NPN but try with 2 N-ch Mosfets.
Also... Im thinking to use different types of npn's. 1 BC548 and another BD139. To induce a instability in the circuit...
Just putting some ideas on the table.

 

[pc1966]

A crude simulation attempt, did not work at 20MHz with 2N2222 but OK with BC548, also used low-Q crystal model so faster to build up (along with 1mV kick at t=0)

Time-domain for test point in Tr1 collector:

 

[Lucien Nunes]

I gave myself 5 minutes to do this, just to prove we can all build a discrete transistor crystal oscillator that doesn't work...
First transistors I came to - ZTX753s - should have enough zing to work. The circuit has a name but I can't remember it, a common-base stage driving an emitter follower. I made the values up as I went along, they won't be far out but it still doesn't work. I've no interest in simulating it or debugging it, time's up!

 

[Lucien Nunes]

So then I decided to give myself another 5 minutes to build a crystal oscillator the way I would have done if somebody had said: "Build a simple crystal oscillator." I.e. with an 'HC00. Of course this one works just fine. I didn't bother tuning it, I used the first capacitors that I picked up from the drawer, which turned out to be 22pF. Build time was just under 3 minutes including stripping some wire to make links.

 

[Lucien Nunes]

And this is how it looks with a 4MHz xtal.

*BONUS POINTS for anyone who noticed I had the scope's 20MHz bandwidth limiting switched on. The actual output waveform is very ringy because I did a lousy job of decoupling the power supply (stood a ceramic 0.1 across the IC without chopping its leads down). You can change the W/F just by bending this component about. It could be fixed but... time's up!

E2A, I semi-fixed it (relocated the bypass cap and poked the probe direct into the breadboard)

 

[_q12x_]

very interesting update.
I managed to split 1/10 of my test PWM original fv. From its 850Hz down to 85Hz
Now Im playing with some capacitors values to bring it down to 1/100.
My target end result will be 1/1000.
And, I did managed to --average-- the 1/100 values. I say average because the fv counter is jumping a couple of Hz up and down. It should be 8.5Hz but is heavily fluctuating with diverse values between 10Hz and 6Hz. So I am theoretically at 8 but because is jumping all around these values, it was hard in the beginning to understand where I stand. But staying and watching these values for a while, a couple of minutes, I concluded it is jumping around these margins. That might be from the High resolution of the splitting, all the circuit(s) combined might present some tolerances, especially the capacitors and resistors and why not the chip(s).
At 1/10, it does not jump at all, it stays put at that fv. So.... I think im good at this point.
I might have to make a second circuit that will split another 1/10 from this one that is 1/100 and in the final output will get 1/1000. And I expect it will be extremely jumpy at that Very High resolution. If it is a way of "average" all these jumping, especially at 1/100 stage I am right now. That will give a good clear chance for the 1/1000 stage. Hmmm. But it is in plan. For now, I have to reach that 1/1000 target and Im GOOD, literally. Haha.

 

[marconi]

A crude simulation attempt, did not work at 20MHz with 2N2222 but OK with BC548, also used low-Q crystal model so faster to build up (along with 1mV kick at t=0)
View attachment 97652

Time-domain for test point in Tr1 collector:
View attachment 97653

I had another go this morning after thinking about the small signal model of this circuit and why the ratio of Rc to Re matters so much- see first attachment. I tried Rc = 1kR and Re = 4.7kR and even exchanged the BC549s for two new but in my construction it did not oscillate. Not enough feedback methinks so I swapped Rc for 100R and the symmetrischer oszillator found its 20MHz voice. We did it Sir and self starting to boot!

 

[pc1966]

Not enough feedback methinks so I swapped Rc for 100R and the symmetrischer oszillator found its 20MHz voice. We did it Sir and self starting to boot!

Excellent!

While the E-E impedance should be roughly -2*Rc quite possibly the higher Rc values along with stray capacitance might cause too much phase shift. I seem to recall that 10-20MHz crystals are below 100R series resistance so indeed 100R values for Rc would be adequate, as you have so well demonstrated!

 

[marconi]

q12x. : My next step is to take the small amplitude 20MHz sine wave and turn it into 20MHz pulse stream of TTL logic 0 and 1s to feed into the divider circuit.

 

[_q12x_]

Congratulation mister @marconi I am really happy you made it !
But I am not sure I understand completely your values.
So I grabbed all the information you provided so far:

Rc for 100R and the symmetrischer oszillator found its 20MHz voice.

did not work at 20MHz with 2N2222 but OK with BC548

I tried Rc = 1kR and Re = 4.7kR

and I redraw the circuit with the new values:
- Are these the values you used?

 

[marconi]

Congratulation mister @marconi I am really happy you made it !
But I am not sure I understand completely your values.
So I grabbed all the information you provided so far:



and I redraw the circuit with the new values:
- Are these the values you used?
View attachment 97682

Yes. VB is 5 Volts not 1.35V. The npn transistors are BC548B types. the letter A, B or C after BC548 indicates the so-called direct current gain of the transistor - the hFE - and relates the emitter-collector current flow to the base-emitter current flow eg: Ice/Ibe. Which BC548s do you have? Cs have higher gain than Bs than As so I used the mid-range gain type.

See top of page 2 of:

BC548B datasheet - https://datasheetspdf.com/pdf-file/1292183/Kingtronics/BC548B/1

 

[marconi]

Which operational amplifiers do you have please and I will use the same ones to amplify the small amplitude 20MHz sine wave.

 

[_q12x_]

Mine are BC548 B - coincidence! like yours.
I have uA741(not that used anymore) and LM358(most used)
Updated circuit after your last specifications:
Ive added + and 5V and gnd symbol. Also B after tr name.

 

[_q12x_]

My last night progress:

The fv Counter is reading the output of the frequency divider and is reading 1/100 = (around) 8.5Hz
(The osciloscope reading is intermittent for this 8.5Hz output, not the same every time)
The osciloscope shown here is reading the output of the PWM at 850Hz

And now, Im building the board of the frequency divider:

 

[_q12x_]

Mister @marconi and @pc1966 , in my post #73 I had a problem:
"..I might have to make a second circuit that will split another 1/10 from this one that is 1/100 and in the final output will get 1/1000. And I expect it will be extremely jumpy at that Very High resolution. If it is a way of "average" all these jumpings, especially at 1/100 stage I am right now. That will give a good clear chance for the 1/1000 stage. Hmmm. But it is in plan. For now, I have to reach that 1/1000.. "
- Can you give me an answer to this problem?
I have an idea but I think is a bit too complicated.... and is incomplete as well.
I want to see what are your thoughts.

 

[marconi]

Impressive progress and great dedication.

 

[_q12x_]

deleted post ______

 

[_q12x_]

Impressive progress and great dedication.

True, true,
The same to you
haha

 

[marconi]

To use a 555 as a frequency divider of a known and constant frequency pulse train you have to make sure that the 555 is ready to be triggered again after n input pulses where n is the division factor.

This means that one has to ensure the 555 output (shown in green) has changed to zero after a further n input pulses (shown in yellow) since it was last triggered so the 555 is ready to be triggered again by first pulse of the next group of n pulses.

The way the 555 retriggerable monostable circuit operates to divide by n means that the time constant/period of the monostable 555 - x - must satisfy this equation:

(n x T) - t < x < (n x T)

What this says is the 555 output must transition to zero after the trailing edge of the nth pulse and before the leading edge of the (n+1)th pulse - that is in the gap between them when the input is zero.

Attached is an example for a divide by 3 scheme which shows the meaning of n, T, t and x.

It is probable better then for your PWM input to keep t short with respect to T so that there is a relatively big time gap between input pulses so that variations in the 555 timer function which creates the interval x will generally fall clearly after n pulses and nicely before the (n+1)th.

Or something close to this explanation - I am bit rushed.

 

[_q12x_]

So... no answer. Ok.
And I finished the board. it took me exactly 3h to make it (continuous work, no brakes).

Is the small right (card)board

Not bad !
Now I can make another division, 1/10 to get a total of 1/1000.
It would be nice if you could look over the "average" problem, but... you are busy as well I guess.
I will probably not implement it, but it will had give me a very clean result in the final stage....
As I said, we'll see after I finish the whole 1/1000 process then.

 

[marconi]

So... no answer. Maybe not the complete answer you were expecting but something for you to consider. The explanation suggests pointers on what you might do to 'tweak' your design so that division by n actually occurs in the real circumstances of the circuit when the input period T and pulse duration t are varying randomly and the monostable duration x is subject to random lengthening and shortening in time. In plain English/Romanian make sure the on to off ratio of the input pulse stream is less than 5%(say) and n is high.

 

[_q12x_]

I still don get it.... make me a clear circuit... or a continuation of the one that I have already, Like another module to add after the last 1/100 output (that Im actually measure with the fv counter).
Draw on paper and make a picture and post it.

 

[marconi]

I and you do not know whether the variation in division result is due to variation of the input from your PWM circuit or in your 555 divider circuit or both. All I can say is the variation becomes less significant if the mark-space ratio of the input is low and n is higher. That is how the mathematics works.

To do the averaging one would need to store values over an interval which can be done but at the moment I do not think you have the ics to do that. When my parcel arrives maybe but then you can do the division in a much more practical way to not bother.