Can someone explain what am I scoping?

[miceuz]

As per advice in this thread https://www.eevblog.com/forum/projects/cheap-source-of-8mhz-square-wave/msg565436/#msg565436 I've built an inverter oscillator using SN74HC04 inverter and 12 Mhz ZTT ceramic resonator http://www.ecsxtal.com/store/pdf/zttr.pdf It's a standard minimal oscillator setup as in ZTT datasheet, with one inverter in oscillator and one for buffering, no dampening resistor.

At first I've built everything on a breadboard, the waveform was terrible, I thought, all the stray capacitance is affecting and 12MHz is a bit too much for it, so I've soldered everything on a piece of copper board.



The waveform was a lot better, but still with quite significant overshoot and undershoot. I've tried adding a dampening resistor, but that didn't work, I've tried adding a resistor between output of the first inverter and input of the second one, but that didn't affect the waveform anyhow.



Then I thought that maybe the problem is in my probing setup, I was already using probe in x10 mode, so I tried to get rid of ground lead, I've used a spring on a probe. The waveform is somewhat better, but still overshoots are there.



Then I've tried adding a 330 Ohm resistor at the output and scoping after this resistor. Whaoa.. overshoots are gone.



Another pecularity - since I have a resistor at the input of the second buffering inverter, I can see that the waveform entering the second inverter is without overshoots, but I can see overshoots at the output.

My question is - how do I interpret what I see here? Are the edges of the oscillator too steep for my scope to display? Or is the oscillator really producing the waveform with those overshoots and scope probe 15pF capacitance in conjunction with 330 Ohm resistor filtering those out?

[DmitryL]

As per advice in this thread https://www.eevblog.com/forum/projects/cheap-source-of-8mhz-square-wave/msg565436/#msg565436 I've built an inverter oscillator using SN74HC04 inverter and 12 Mhz ZTT ceramic resonator http://www.ecsxtal.com/store/pdf/zttr.pdf It's a standard minimal oscillator setup as in ZTT datasheet, with one inverter in oscillator and one for buffering, no dampening resistor.

At first I've built everything on a breadboard, the waveform was terrible, I thought, all the stray capacitance is affecting and 12MHz is a bit too much for it, so I've soldered everything on a piece of copper board.

Errr.. Why do you care at all ? You got your generator producing square waveform, it works.
Is you 10x probe properly compensated ? (tuning that tiny capacitor in the probe while wathching signal from scope's calibrator)

[miceuz]

Errr.. Why do you care at all ? You got your generator producing square waveform, it works.
Is you 10x probe properly compensated ? (tuning that tiny capacitor in the probe while wathching signal from scope's calibrator)

Err.. I want to be sure it's really square. Also the knowledge 

The scope has 1kHz calibration signal, no noticable ringing there.

[DmitryL]

There is no such thing as "really square", you can't achieve zero rise time
Also, there are a lot of very good video lessons explainig proper termination, problems with ringing, etc. etc. from forum members. e.g
https://www.youtube.com/user/w2aew/videos

[c4757p]

Driving capacitive loads with fast devices tends to cause overshoot, and your probe is a big ol' capacitor. Build yourself a Z₀ probe if you want to see an accurate picture of a digital signal.

[KJDS]

I'd check how much ringing you have on the supply and ground too, the decoupling looks to be via long leads and path lengths.

[Zero999]

Congratulations, your circuit is working. The rise/fall times seem to be consistent with the information on the datasheet.

[c4757p]

I'd check how much ringing you have on the supply and ground too, the decoupling looks to be via long leads and path lengths.

This too. FWIW, I got this with 74HC (Toshiba 74HC00), so you can definitely get a much cleaner rise with what you have:

[miceuz]

Yeah... decoupling is not the best - 150mV peak to peak... will have a look into it.



Unfortunately the coax cable I have is of some unknown origin, so Z₀ probe will have to wait till tomorrow. I have tried to make Z₀ with the cable I have and all I could see was just huge capacitive loading.

[miceuz]

After shortening capacitor leads, ripple on VCC bacame much better. Twice better really 



After adding a 1nF cap it became twice better again.



No joy at the output though.. As a side note - my scope gives me rise time of 4.3ns. It's a 50MHz Rigol DS1052. According to a formula I saw somewhere on the net while researching Z₀ probes, to display 4.3ns rise time I need 1/(pi*4.3ns)=74MHz of bandwidth. And it's 50MHz scope. Is this why I see those over/undershoots?

[AG6QR]

I can't tell for sure, but based on the first photo on the first photo on the first post of this thread, it looks like you might have a long ground lead on your scope probe.  This can cause issues when trying to see sharp transitions with lots of high frequency components.  W2AEW, who sometimes frequents this forum, has made a couple of good videos on the subject.

[miceuz]

I can't tell for sure, but based on the first photo on the first photo on the first post of this thread, it looks like you might have a long ground lead on your scope probe.  This can cause issues when trying to see sharp transitions with lots of high frequency components.  W2AEW, who sometimes frequents this forum, has made a couple of good videos on the subject.

Yeah I know, the third picture in the first post is a waveform captures with short spring ground lead  I know W2AEW and have learned a lot of stuff from his channel, but in this case I just want to understand - where does the overshoot come from? Another bit of knowledge about proper probing techniques 

[AG6QR]

I just want to understand - where does the overshoot come from?

Look up the Gibbs Phenomenon

http://en.wikipedia.org/wiki/Gibbs_phenomenon

If you have a perfect square wave, and pass it through a filter with a finite bandwidth such that the higher frequency components are blocked, you'll get overshoot.

In the real world, you won't start with a perfect square wave, and your probing and oscilloscope bandwidth limits won't be a perfect "brick wall" low pass filter, so you won't get exactly the pretty displays shown in that wiki page.  But you will get overshoot and ringing.

[DmitryL]

I just want to understand - where does the overshoot come from?

Look up the Gibbs Phenomenon

http://en.wikipedia.org/wiki/Gibbs_phenomenon

If you have a perfect square wave, and pass it through a filter with a finite bandwidth such that the higher frequency components are blocked, you'll get overshoot.
t exactly the pretty displays shown in that wiki page.  But you will get overshoot and ringing.

.. how interesting... I always thought that any of the normal digital scope has an analog low-pass anti-aliasing filter before ADC (with infinite bandwidth ..
..and I also thought that Gibbs phenomenon relates to IFFT

[macboy]

I just want to understand - where does the overshoot come from?

Look up the Gibbs Phenomenon

http://en.wikipedia.org/wiki/Gibbs_phenomenon

If you have a perfect square wave, and pass it through a filter with a finite bandwidth such that the higher frequency components are blocked, you'll get overshoot.

In the real world, you won't start with a perfect square wave, and your probing and oscilloscope bandwidth limits won't be a perfect "brick wall" low pass filter, so you won't get exactly the pretty displays shown in that wiki page.  But you will get overshoot and ringing.

A properly compensated/calibrated oscilloscope input will not exhibit any significant overshoot or ringing, despite the limited bandwidth that effectively acts as a low pass filter. This is because the oscilloscope input has (by design) a Gaussian frequency response, which results in an "ideal" pulse response, with a slight tradeoff of a earlier and slower roll-off. In other words, you could get a flatter response out to higher frequencies, and a consequently sharper roll-off, but that would result in ringing and overshoot in the pulse response. Some oscilloscopes, especially high bandwidth digital ones, will have a response more like the latter, but most scopes are like the former, going to better pulse response rather than flatter passband.

[tggzzz]

I just want to understand - where does the overshoot come from?

Look up the Gibbs Phenomenon

http://en.wikipedia.org/wiki/Gibbs_phenomenon

If you have a perfect square wave, and pass it through a filter with a finite bandwidth such that the higher frequency components are blocked, you'll get overshoot.

In the real world, you won't start with a perfect square wave, and your probing and oscilloscope bandwidth limits won't be a perfect "brick wall" low pass filter, so you won't get exactly the pretty displays shown in that wiki page.  But you will get overshoot and ringing.

A properly compensated/calibrated oscilloscope input will not exhibit any significant overshoot or ringing, despite the limited bandwidth that effectively acts as a low pass filter. This is because the oscilloscope input has (by design) a Gaussian frequency response, which results in an "ideal" pulse response, with a slight tradeoff of a earlier and slower roll-off. In other words, you could get a flatter response out to higher frequencies, and a consequently sharper roll-off, but that would result in ringing and overshoot in the pulse response. Some oscilloscopes, especially high bandwidth digital ones, will have a response more like the latter, but most scopes are like the former, going to better pulse response rather than flatter passband.

Until, that is, you have probe with tip capacitace C and a ground lead inductance L - at which point you will see ringing due to the LC resonance. Since typical probes highZ *10 probes have 15-25pF capacitance, and a 6" ground lead has maybe 100nH inductance, the ringing will be  ~100MHz or so.

Go on, give it a try and see if you can duplicate the oscillograms in all the Tektronix HP/Agilent/Keysight app notes. It is quite easy to demonstrate!

[miceuz]

I've made myself a Z0 probel today. Yellow trace is Z0, blue trace is scope probe in 10X mode with short spring as a ground contact. Thanks to all, I've learned something new 

[macboy]

I've made myself a Z0 probel today. Yellow trace is Z0, blue trace is scope probe in 10X mode with short spring as a ground contact. Thanks to all, I've learned something new 

So the next thing to do is to adjust the HF compensation on your 10x passive probe so that the trace more closely matches the Z0 one. Not all probes and not all scope inputs are alike; they work as a system, and need to be adjusted to match each other.

[David Hess]

So the next thing to do is to adjust the HF compensation on your 10x passive probe so that the trace more closely matches the Z0 one. Not all probes and not all scope inputs are alike; they work as a system, and need to be adjusted to match each other.

I agree; that is too much discrepancy for 20 ns/div.  They should match much closer than that.

What oscilloscope and probes are you using?  Did the probes come with the oscilloscope?

[miceuz]

What oscilloscope and probes are you using?  Did the probes come with the oscilloscope?

It's Rigol DS1052E (non hacked), probes are native Rigol ones. The compensation looks ok on 1kHz calibration signal.

[David Hess]

What oscilloscope and probes are you using?  Did the probes come with the oscilloscope?

It's Rigol DS1052E (non hacked), probes are native Rigol ones. The compensation looks ok on 1kHz calibration signal.

The low frequency compensation adjustment found on all attenuating passive probes is not going to help in the 10s of nanoseconds domain.  Some probes have high frequency compensation adjustments for this.  High frequency stock probes, 200 MHz and above, often have fixed high frequency compensation made to work with one particular group of oscilloscope models.

One test I would do is to connect both probes simultaneously to the signal source.  That will remove any possibility that that difference being seen is because of a difference in loading.