Square Wave From Signal Generator on Rigol MSO1104 Not Square

[tggzzz]

Quite surprised to see that level of ringing on a Z0 probe with spear.

So am I. I think the 3*CMOS may be partially due to a longer track (<2cm) than desirable. Maybe I'll chop the track and see if it makes a difference. The other possibility is that a 100nF 0603 for each gate isn't sufficient decoupling.

OTOH, there's no such issue with the 1*CMOS, and the ringing is still there at the same frequency, albeit lower amplitude.

Hence I suspect it is the CMOS output being aggressive.

The obvious test is to use the low-impedance probe to observe the 3*(74LVC1G14 + 130ohms) signal output. The two traces are identical within the limits of the match of the two oscilloscope input channels. Conclusion: the ringing is a real artefact, not an artefact of the probe.

[Zero999]

Isn't there a risk of mismatch between the output impedance/propagation delays of the gates which can mess things up? If you're going to parallel them, it should be better if they're all on the same die.

[tggzzz]

Isn't there a risk of mismatch between the output impedance/propagation delays of the gates which can mess things up? If you're going to parallel them, it should be better if they're all on the same die.

Yes, but I think it is a lesser problem, especially when all the devices are from the same batch. Layout will be at least as intractable issue.

Having multiple gates in a package increases power supply lead lengths and consequent  ground bounce problems. Don't forget that di/dt is around  100000000A/s (100MA/s), I.e. 100mA in 1ns.

[Howardlong]

Is that 50 ohm termination inside the 485?

Yes it is, and all the better for  it   Presumably Tek learned from the HP1740a - and improved on it by disconnecting the 50ohms in the event of an overload.

Quite surprised to see that level of ringing on a Z0 probe with spear. TBH at that level because my crappy old Tek probes taper off at about 70MHz, and I'm too cheap to buy some nice ones, I built my own probe with a bit of quality RG58U and a couple of hand picked resistors and soldered it in circuit. No internal termination on my 465B though. Worked like a dream however

So am I. I think the 3*CMOS may be partially due to a longer track (<2cm) than desirable. Maybe I'll chop the track and see if it makes a difference. The other possibility is that a 100nF 0603 for each gate isn't sufficient decoupling.

OTOH, there's no such issue with the 1*CMOS, and the ringing is still there at the same frequency, albeit lower amplitude.

Hence I suspect it is the CMOS output being agggressive.

FWIW. the best I can do with an unterminated output is about 510ps with a single lvc1g14, probed with an Agilent 1130A 1.5GHz single ended FET probe with browser on an MSO7104B.

The input is AC coupled and terminated with 100 ohm to Vcc and 100 ohm to GND.

I did try a couple of other experiments such as making the output look like a 50 ohm source directly into a 20GHz sampling scope with a single series resistor and AC coupling cap but this made it worse at 670ps. It needs a T or pi match I think.

Is that 130 ohm a "sweet spot" for the LVC series? I can see three of them plus the inherent device's output impedances reasonably showing as a 50 ohm source.

[Howardlong]

This is an SN74AUC1G14, again with a 1130A with single ended browser probe + MSO7104B scope, probed directly on the unloaded output pin. (This scope & probe has a measured system rise time of 320ps).



I've found that as soon as I load the pin, the rise time deteriorates. A 270 ohm 0603 to ground soldered directly at the chip's lead, for example, increased the rise time to 570ps with the same probe/scope combination. As with the LVC1G14, using the lo-z probes I have for the 20GHz sampling scope again gave a slower rise time than using the 1130A FET probe

[tggzzz]

Is that 50 ohm termination inside the 485?

Yes it is, and all the better for  it   Presumably Tek learned from the HP1740a - and improved on it by disconnecting the 50ohms in the event of an overload.

Quite surprised to see that level of ringing on a Z0 probe with spear. TBH at that level because my crappy old Tek probes taper off at about 70MHz, and I'm too cheap to buy some nice ones, I built my own probe with a bit of quality RG58U and a couple of hand picked resistors and soldered it in circuit. No internal termination on my 465B though. Worked like a dream however

So am I. I think the 3*CMOS may be partially due to a longer track (<2cm) than desirable. Maybe I'll chop the track and see if it makes a difference. The other possibility is that a 100nF 0603 for each gate isn't sufficient decoupling.

OTOH, there's no such issue with the 1*CMOS, and the ringing is still there at the same frequency, albeit lower amplitude.

Hence I suspect it is the CMOS output being agggressive.

I've increased the decoupling capacitors from 100n to 4u7 0603s. This has reduced, but not eliminated, the switching perturbations on the 5V rail. On my scope 350MHz Tek485 it makes very little difference to the oscillograms, except that the overshoot on the 3*CMOS driving 50oms is reduced. That suggests the CMOS outputs are being aggressive.

FWIW. the best I can do with an unterminated output is about 510ps with a single lvc1g14, probed with an Agilent 1130A 1.5GHz single ended FET probe with browser on an MSO7104B.

Ah, the joys of having more than obsolete hobby equipment

I've made indirect measurements of the transition time by driving the 3*(LVC14+130ohms) from an AC coupled noise source and looking at the result on a spectrum analyser (an SDR dongle, since my Tek492 isn't immediately available). Eyeballing the transition between the 20dB/decade and 40dB/decade rolloff indicates a transition time of around 600ps, which isn't too far from your results.

Is that 130 ohm a "sweet spot" for the LVC series? I can see three of them plus the inherent device's output impedances reasonably showing as a 50 ohm source.

That's the intention, but quite frankly I don't have a means of validating it. Hopefully the 130ohms will not only reduce ground bounce and decoupling effects, but also reduce the effect the gate's output impedance changing with output voltage.

I've attached a picture of that output stage. Given the equipment available to me and the expected transition times:

• I think the BNC connector is sufficient
• there's no real attempt to get the PCB track to be 50ohms; I suspect it is around 75ohms

If I re-spin the 2 layer 1.2mm board, I might change that plus make the 5V rail into a more complete plane.

[Howardlong]

The decoupling cap I used was a 10nF 0402 directly on the Vcc pin to the groundplane, you can just see it if you zoom the photo.

[tggzzz]

I added 100nF 0603 directly across the top of each gate. It noticably reduced the PSU perturbations, but otherwise the signals were unaffected when measured with my much slower scope.