Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Increasing oscillator output to almost rail to rail?

Status
Not open for further replies.

neazoi

Advanced Member level 6
Joined
Jan 5, 2008
Messages
4,118
Helped
13
Reputation
26
Reaction score
15
Trophy points
1,318
Location
Greece
Activity points
36,907
Hi, I am trying to build a clock for a microcomputer using discrete components, the less components the better.
An oscillator that worked is the attached one, but the output signal swing was about 3.5Vpp when loaded by an external 74LS74 flip flop and also a 74LS165, which seems is not enough to clock the 165.
The ceramic resonator and the 100nF capacitor cannot be changed and the 100nF is connected to the ground, so I do not want to change that (so as to not modify the micro PCB.

What are my options to bring the swing close to rail to rail?
 

Attachments

  • Capture.PNG
    Capture.PNG
    5.3 KB · Views: 128

Hi,

Do you mean "microcontroller"?
--> They usually have built in oscillator circuit.
* some have built in RC oscillators --> they don´t need external parts
* some have built in RC oscillators --> with external parts you may determine clock frequency
* some have built in XTAL oscillators --> they usually just need 2 external Cs and one external XTAL

"Less components" and "discrete circuit" are contradictorily somehow.

when loaded by an external 74LS74
Why do you use outdated "LS" type logic? they draw a lot of input current. --> use "HC" type instead.

Isn´t useful for logic inputs. You may use it for DC decoupled HF circuits.
But logic circuits need absolute values: like <0.7V for LOW and >2.0V for HIGH (example).

****
Without the information about your voltage levels.. hard to say.
* a comparator should work.
* or a FET as buffer

Confusing to me: you say you "want to build" but on the other hand there seems to be an existing circuit.

Klaus
 

Hi,

Do you mean "microcontroller"?
--> They usually have built in oscillator circuit.
* some have built in RC oscillators --> they don´t need external parts
* some have built in RC oscillators --> with external parts you may determine clock frequency
* some have built in XTAL oscillators --> they usually just need 2 external Cs and one external XTAL

"Less components" and "discrete circuit" are contradictorily somehow.


Why do you use outdated "LS" type logic? they draw a lot of input current. --> use "HC" type instead.


Isn´t useful for logic inputs. You may use it for DC decoupled HF circuits.
But logic circuits need absolute values: like <0.7V for LOW and >2.0V for HIGH (example).

****
Without the information about your voltage levels.. hard to say.
* a comparator should work.
* or a FET as buffer

Confusing to me: you say you "want to build" but on the other hand there seems to be an existing circuit.

Klaus


No I mean a microcomputer and this is the schematic **broken link removed**
I am replacing the oscillator IC (20, three gates) with a discrete oscillator, but I do not want to modify the existing PCB hence the limitation of the ceramic resonator and the series capacitor connected to the ground.

3.5Vpp swing should be enough I think for the computer to work. The 74LS74 seems to be ok, but the 74LS165 seems to have problems driven with 3.5vpp, hence some video elements are not displayed.
 

It won't work - the circuit uses C9/R2 to create a short 'spike' to clock the '165. It isn't the voltage levels that matter, it is the speed the edges rise and fall.
I haven't checked but if there is a spare logic gate in there somewhere, you could use it to 'square up' your existing clock generator.

[edit - there should also be a diode, cathode to +5V across R21 and I'm not sure the 'ear' and 'mic' aren't crossed over]
[even more edit - they refer the socket markings on the cassette recorder the connections go to.]

Brian.
 
  • Like
Reactions: neazoi

    neazoi

    Points: 2
    Helpful Answer Positive Rating
It won't work - the circuit uses C9/R2 to create a short 'spike' to clock the '165. It isn't the voltage levels that matter, it is the speed the edges rise and fall.
I haven't checked but if there is a spare logic gate in there somewhere, you could use it to 'square up' your existing clock generator.

[edit - there should also be a diode, cathode to +5V across R21 and I'm not sure the 'ear' and 'mic' aren't crossed over]
[even more edit - they refer the socket markings on the cassette recorder the connections go to.]

Brian.

All right, I see. So it depends on a square wave clock, so that the C9/R2 can then generate the fast rising spike, am I getting it right?

Any ideas of what kind of crystal oscillator topology to search for the purpose?

Maybe a simple germanium anti-parallel diodes shunted to the GND at the output of the oscillator could help squaring the signal up, I will test this.
 

Pick up the signal from the resonator and you may get the level you want (instead from the emitter as it is now, pick the signal from the transistor base).
 
  • Like
Reactions: neazoi

    neazoi

    Points: 2
    Helpful Answer Positive Rating
Maybe a simple germanium anti-parallel diodes shunted to the GND at the output of the oscillator could help squaring the signal up, I will test this.
I would doubt that solution would work - the diodes may well stop the oscillator running and in any case would clip the amplitude to ~0.5V when it still has to be more than about 3V to guarantee clocking the LS165. The original design uses logic gates to produce and buffer the clock specifically to ensure fast logic level edges.

Picking the signal from the base might work but it would still be very borderline. At risk of the clock being unstable, you could try adding a choke, or better still a tuned circuit in the collector and taking the signal from there instead. It would probably give more amplitude but whether the edges would be fast enough is debatable. The simplest solution is to add a logic gate (buffer or inverter) in series with your oscillator output.

Are R4 to R11 really 1K? I can see why a resistor is needed, it is due to the horrible way of forcing the data bus low to make the Z80 execute NOP instructions but the value seems inordinately high, even for the slow data speed. I would have thought 47 Ohms more than adequate.

Brian.
 
  • Like
Reactions: neazoi

    neazoi

    Points: 2
    Helpful Answer Positive Rating
I might recommend a hex "single stage inverter"
(as commonly used for XTAL oscillators for logic)
capacitively coupled to the oscillator output, the
first inverter held linear by resistor feedback against
the coupling cap, drives the second, second drives
third, third drives the other 3 in parallel for a nice
strong bang-bang output. At least, up to where the
gate delay approaches the reciprocal of frequency.
I'd look for a logic family that has a delay about 1/20
of reciprocal frequency just for some slack.
 

Are R4 to R11 really 1K? I can see why a resistor is needed, it is due to the horrible way of forcing the data bus low to make the Z80 execute NOP instructions but the value seems inordinately high, even for the slow data speed. I would have thought 47 Ohms more than adequate.

Brian.

Yes they are. And in the original hardware too.
I have managed to bring the swing to about 4vpp and the zero logic starting from near zero. Here is the schematic of it.

It is not what I want thought because the rising edge is smooth, but at list I got a large swing at that point. Any buffer I had initially tried was killing the oscillator. but when I included an intermediate differentiator formed by the 10nF and the shunt resistor this bjt buffer could be driven ok. It is weird that when I remove the 150R the oscillation is killed! I will tweak the buffer resistors to see if I can improve the swing more and the waveform. Maybe an anti-parallel diode set at the output of the buffer now can do the trick, or any other suggestions/ideas?
 

Attachments

  • Capture.PNG
    Capture.PNG
    6 KB · Views: 124
Last edited:

The base resistor of the first BJT is 820R not 8.2k, its a typo.
 

Diode clipping just removes the top and/or bottom of the waveform, it doesn't increase the rise or fall time of the waveform. Your problem is simply that the waveform should be TTL level with fast rise and fall edges, you will never get that directly from a single transistor oscillator even if it is followed by an analog buffer stage. I think as you will need to produce logic levels, it is probably easiest to abandon your existing plan and go back to the original design. You will end up with something far more complicated following your present policy.

Brian.
 

Diode clipping just removes the top and/or bottom of the waveform, it doesn't increase the rise or fall time of the waveform. Your problem is simply that the waveform should be TTL level with fast rise and fall edges, you will never get that directly from a single transistor oscillator even if it is followed by an analog buffer stage. I think as you will need to produce logic levels, it is probably easiest to abandon your existing plan and go back to the original design. You will end up with something far more complicated following your present policy.

Brian.

Yes, I have thought it so, it just clips the edges it does not change the waveform.
Brian I was thinking if these crystal controlled multivibrators, there are 1-2 circuits on the net to try. Do you think they could work better?
http://www.seekic.com/circuit_diagram/Basic_Circuit/CRYSTAL_CONTROL_IMPROVES_STABILITY.html
http://www.next.gr/circuits/Low-Frequency-Crystal-Controlled-Oscillator-l44057.html
http://www.uploadarchief.net/files/download/butler.png

- - - Updated - - -

or this **broken link removed** which is an elektor one (they always work)
 

They won't work better than the original. The 'seekic' and 'circuitswiring' designs will probably work but you would have to remove C4. I don't think the other circuits stand much chance. Basically, you want a square wave with fast edges and that's exactly what a logic gate solution will do. Your other option is a ready built oscillator module but the 6.5MHz frequency isn't standard so you might have trouble finding one. The frequency has to be exact or the video and audio will not work.

Brian.
 
  • Like
Reactions: neazoi

    neazoi

    Points: 2
    Helpful Answer Positive Rating
They won't work better than the original. The 'seekic' and 'circuitswiring' designs will probably work but you would have to remove C4. I don't think the other circuits stand much chance. Basically, you want a square wave with fast edges and that's exactly what a logic gate solution will do. Your other option is a ready built oscillator module but the 6.5MHz frequency isn't standard so you might have trouble finding one. The frequency has to be exact or the video and audio will not work.

Brian.

After a few hours of hair pulling problems, I replaced the buffer bjt on post #9 with a 2n2222 and the waveform became more square-ish. This seemed to make the microcomputer happy and now everything works ok. The base to gnd resistor is 680R and the collector 120R. The base to vcc resistor is not needed at all and it was removed (class-C ???).
The min voltage is 0 and the max 5.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top