Cheap Fast Rise Time Oscillator

[BillyO]

The boards for the latest prototype are in and I built up a couple for testing.

I got some good results!  The last prototype (mentioned in this thread: https://www.eevblog.com/forum/testgear/best-brand-name-for-an-oscilloscope/msg4554020/#msg4554020) was reading about 2.15ns on my SDS2504XP.  The latest one reads 1.47ns.  Given that the scope has a rise time of about 680ps that would give the oscillator a rise time of about 1.3ns actual.

The rise time is dependent on the choice of 74AC14 too.  1.47ns is from the slowest of the two I built.  The fastest was 1.37ns or about 1.2ns actual.  Unfortunately I don't have a faster scope readily available to get closer to actual rise time.  I have a few ideas for the next (and possibly final) version such as using a 4-layer board, slightly shorter trace to the BNC and using micro-USB for power, but I don't expect these will speed it up much.  The 74AC technology is just not going to wiggle much faster.

[tggzzz]

If you want <300ps, have a look at https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

[BillyO]

Nice!  I might try that next using one inverter for a relaxation oscillator.  You could build it for about $3!

[BillyO]

Boards for the latest version  are in.  Below in purple.  Highlights of this version:
 

• Smaller board
• Shorter output trace
• Only 4 inverters on output to reduce PD skew
• Crystal replaced by relaxation oscillator to reduce size and cost to build
• 4-layer board with 2 full ground planes and a full Vcc plane

Will build one up tomorrow to test.

The 2nd device (black) is a TDR version in early stages.  Waiting on the proper parts to properly test the design so far.

[magic]

You will never get good results with DIP or even SO package. Too long pins, too long traces, too much inductance in series with the output, too much inductance in series with the bypass cap. Guaranteed severe overshoot and ringing, easily visible even on 100MHz scopes.

For the sort of results shown by tggzzz, SOT-23 (or smaller) package and low inductance layout are essential.
And LVC logic appears to be faster than AC. There may or may not be differences between manufacturers too.
But nothing is as important as small package and physical construction details.

My own attempt, never tested on "serious" gear but you can see its 200MHz performance by scrolling a few posts up:
https://www.eevblog.com/forum/projects/compact-74ac14-pulse-generator-pcb/msg3475680/#msg3475680

It's an open question whether this or tggzzz's single-gate chips on a pair of power/ground planes would be better


edit
And by the way, the chips you are using are not TI but Chinese mystery meat, just so you know.

[tggzzz]

You will never get good results with DIP or even SO package. Too long pins, too long traces, too much inductance in series with the output, too much inductance in series with the bypass cap. Guaranteed severe overshoot and ringing, easily visible even on 100MHz scopes.

Just for fun, my board also has 3*74LVC without series resistors. The overshoot is impressive (+7.5V peak on a 350MHz scope!) - and useless
(1V/div, 1ns/div, 50ohms load and 1Mohm//15pF load)

For the sort of results shown by tggzzz, SOT-23 (or smaller) package and low inductance layout are essential.
And LVC logic appears to be faster than AC. There may or may not be differences between manufacturers too.
But nothing is as important as small package and physical construction details.

Yes indeed.

If I ever respin the board, I'll probably opt for 0306 wide caps (not 0603) and will consider a 4-layer PCB with layers 1 and 2 being Vnd and Vcc planes. That will get the planes closer together than 1.6mm of FR4, which will improve the distributed capacitance and decoupling.

But I don't have access to time domain equipment that could tell the difference; I only have 21GHz spectrum analysers

My own attempt, never tested on "serious" gear but you can see its 200MHz performance by scrolling a few posts up:
https://www.eevblog.com/forum/projects/compact-74ac14-pulse-generator-pcb/msg3475680/#msg3475680

It's an open question whether this or tggzzz's single-gate chips on a pair of power/ground planes would be better

Yes

I suspect the MLCC's capacitance with Vcc across them will be at least as significant. For those that don't know, different flavours of MLCC dielectric can "lose" 80-90% of the capacitance when they have the rated voltage across them.

[BillyO]

And by the way, the chips you are using are not TI but Chinese mystery meat, just so you know.

?

Why the hostility?

I got the chips from Digi Key.  I didn't know they were selling junk these days.

So, it seems my results so far are a bit better than what you got with your SO effort.  1.2ns and minor ringing (voltage wise) that is done in less than 5ns.  We'll see what the latest one does when I get around to testing it.  Unfortunately I just discovered I don't have the right resistor value, so will have to wait for my next order.

In any case my intent here is not to get the fattest possible rise time.  Leo Bodnar already did a good job of that.  My intent is to design something that anyone can build up for next to nothing that will check out the rise time on their cheap scope or probes or whatever.  I'm really just having some fun passing time in little lab, not building the next missile guidance system.

What I do see often is people using a too narrow and too long output trace.  In this new version here is the bottom layout.  The bottom layer is the only layer with traces.  I'm also keeping the inductance and resistance low on the supply side low by using two full ground planes a full Vcc plane and a peripheral copper pour on the bottom layer connected to Vcc.  I've also taken some effort to make the traces feeding the drive inverters the same length.  Not sure if that will help much though.

[tggzzz]

In any case my intent here is not to get the fattest possible rise time.  Leo Bodnar already did a good job of that.  My intent is to design something that anyone can build up for next to nothing that will check out the rise time on their cheap scope or probes or whatever.  I'm really just having some fun passing time in little lab, not building the next missile guidance system.

It is good to have fun, and that's a worthwhile objective.

When measuring risetimes, it is also good to avoid overshoot in the signal. Sometimes the scope setup procedures require/allow a little overshoot in order to get the best useful risetime. Having ringing in the signal will affect those setup procedures, of course

[BillyO]

When measuring risetimes, it is also good to avoid overshoot in the signal. Sometimes the scope setup procedures require/allow a little overshoot in order to get the best useful risetime. Having ringing in the signal will affect those setup procedures, of course

Absolutely.  I take the internal risetime measurement with a grain of salt when looking at less than ideal waveforms (which is most of the time).  When I want to be accurate I use cursors.

[magic]

I got the chips from Digi Key.  I didn't know they were selling junk these days.

So switch to Mouser
Can you name the manufacturer of the 74AC14 shown in your fist post? Maybe you recognize that logo?
And the DIP8 chip on your TDR board also looks like typical fake junk from China.

Not necessarily saying such chips can't work, but they may work differently (either better or worse) than specified in original datasheets.

So, it seems my results so far are a bit better than what you got with your SO effort.  1.2ns and minor ringing (voltage wise) that is done in less than 5ns.

Sorry, I didn't look at your scope waveforms before saying "never". They look surprisingly good and I have no explanation for that
My SO14 build was crap with lousy resistors and not so great ground routing. But I have seen many other attempts with DIP/SO and they never looked much better.
In particular, rfmerrill's first board (the thread I linked) was almost identical to yours.

I just discovered I don't have the right resistor value, so will have to wait for my next order.

The right resistor value is 180~190Ω, depending on RDS(on) of the chip's internal FETs.

[tggzzz]

When measuring risetimes, it is also good to avoid overshoot in the signal. Sometimes the scope setup procedures require/allow a little overshoot in order to get the best useful risetime. Having ringing in the signal will affect those setup procedures, of course

Absolutely.  I take the internal risetime measurement with a grain of salt when looking at less than ideal waveforms (which is most of the time).  When I want to be accurate I use cursors.

How you measure the displayed waveform doesn't matter. What matters is what causes the displayed waveform to be the way it is.

To make the point with a ludicrous example... If you look at the waveforms I added to my earlier post, and use the unterminated waveform with overshoot to twiddle your scope's risetime so that it is nice and sharp+flat, then when you observe a decent edge it would be rather distorted and slower than expected.

[tggzzz]

The right resistor value is 180~190Ω, depending on RDS(on) of the chip's internal FETs.

7ohms was about right for a 74LVC1G output. Hence I used a 143ohms resistor to make it up to 150ohms.

[BillyO]

Well, being an impatient man, I went ahead and built one up anyway using the 220R resistors I have.  Yeah, the logo on the 74AC14 looks a little sloppy.  I bought a large batch of IC's from Digi Key back in 2010.  These, I think, were part of the batch.  I don't recall getting any others, but then 13 years is a long time.  Look at the last picture though.  It shows the logos on the ICs on the TDR version.  The '14 and the '02 look like proper TI logos.  The 555 did come from China.  It's only a 555 though, so meh.

However, on my scope with a measured risetime of 660 ps, this version did 1.27 ns.  Taking the scope RT in account that's a risetime of about 1.09ns!  Not too bad.  Yeah, but I'd not use a $5 oscillator to tweak my scope's front end. 

[tautech]

Good work Bill as you can't get much simpler than that !     
Send me one to look at with this 2 GHz SDS6204A. 

I had something very similar laid out in Altium and expected ~3ns RT but had a mission trying to do a toner transfer with a new printer (Yep, for some things I'm tight as !) so I gave up on it and ordered a Bodnar pulser that only took ~10 days to arrive in NZ from UK.

[BillyO]

Good work Bill as you can't get much simpler than that !     
Send me one to look at with this 2 GHz SDS6204A. 

Thanks and sure.  I'll send you a board from the final version.  I have a couple of tweaks in mind.

I had something very similar laid out in Altium and expected ~3ns RT but had a mission trying to do a toner transfer with a new printer (Yep, for some things I'm tight as !) so I gave up on it and ordered a Bodnar pulser that only took ~10 days to arrive in NZ from UK.

Mine took a month.

[jonpaul]

Bravo for the DIY Pulser. Please compare to Leo Bodnar 40ps pulser, BNC and 30 pS SMA versions

we use a 20 db wideband attenuator to reduce abberation due to VSWR.

Besides digital scopes, a 500 MHz or 1 ghz CRT analog scope photo is useful

We use Tektronix scope 7904, 7104 mainframe with 7A29 and 7B10 plug-ins.

Bon Soirée

Jon

[BillyO]

Bravo for the DIY Pulser. Please compare to Leo Bodnar 40ps pulser, BNC and 30 pS SMA versions

we use a 20 db wideband attenuator to reduce abberation due to VSWR.

Besides digital scopes, a 500 MHz or 1 ghz CRT analog scope photo is useful

We use Tektronix scope 7904, 7104 mainframe with 7A29 and 7B10 plug-ins.

Bon Soirée

Jon

Thanks!

I'd love a 7104.  There is actually one for sale for $800 CDN about 2 hours away from me right now with options 02 & 03 and 7A19, 7B92 and 7B15 plug ins.  A nice starter kit!

(I must stay strong!)

[Bud]

Yeah, the logo on the 74AC14 looks a little sloppy.  I bought a large batch of IC's from Digi Key back in 2010.  These, I think, were part of the batch.  I don't recall getting any others, but then 13 years is a long time. 

If you still have the Digikey bag they came in, the manufacturer name should be there.

[magic]

However, on my scope with a measured risetime of 660 ps, this version did 1.27 ns.  Taking the scope RT in account that's a risetime of about 1.09ns!  Not too bad.  Yeah, but I'd not use a $5 oscillator to tweak my scope's front end. 

This version looks more "ringy" than the first one, though.
Nevertheless, it's still very good performance for DIP14 package.

Maybe it's the Chinese chip being better than original AC logic.
One pattern I have seen in knockoff ICs is that they are often die shrinks on more advanced processes; apparently it saves them money somehow.

Yeah, the logo on the 74AC14 looks a little sloppy.  I bought a large batch of IC's from Digi Key back in 2010.  These, I think, were part of the batch.  I don't recall getting any others, but then 13 years is a long time. 

If you still have the Digikey bag they came in, the manufacturer name should be there.

My question about manufacturer was absolutely rhetorical. There is no Western manufacturer with this logo and it isn't TI logo either.
It's a vaguely TI-like blob intended to be different enough from TI that it's hard to prove trademark infringement.
Common practice in China. There is a new thread with some fake jellybean IC every other months here.

This chip probably (hopefully) didn't come from DigiKey, and 13 years is a lot of time indeed

[Bud]

Actually the logo looks familiar, but can't remember who it was. Given the long time when it was purchased, they may no longer be in business. And 13 years ago counterfeit chips were much less of a problem, i would not think Digikey had a reason to source such chips back then.

[blackdog]

Hi,

I tested and few years ago different cmos logic ICs for a nice fast pulse with few abberations.
I never managed to get it really nice due to such things as the length of the bonding wire in the IC and the various decoupling capacitors.
But the SO enclosure you still get and somewhat better result and think about your PCB pathways, make the output 50 Ohm with multiple resistors of different outputs, e.g. 4x a 200 Ohm resistor in parallel.

In the end I bought the Leo Bodnar pulser. 


PS
And just like jonpaul pointed out, use an attenuator 10 a 20dB to get the SWR as good as possible.

Kind regards,
Bram

[mawyatt]

....
Maybe it's the Chinese chip being better than original AC logic.
One pattern I have seen in knockoff ICs is that they are often die shrinks on more advanced processes; apparently it saves them money somehow.
....

Recurring chip cost to the Original Chip Manufacture (OCM) is directly related to the chip area, smaller chip lower cost. Think many generic chips are now fabbed in smaller feature size processes which yields smaller chips, wether the OCM offers this at a lower cost depends on the chip, market, technology and so on but their recurring cost is lower because of the smaller chip area. Of course these chips also benefit from a faster process which may benefit the end user!!

Best,

[BillyO]

Please compare to Leo Bodnar 40ps pulser, BNC and 30 pS SMA versions

  I don't have the SMA version, just the BNC version.

The Bodnar is the lower trace.  Twice as fast as mine and less ringing.

[tggzzz]

Please compare to Leo Bodnar 40ps pulser, BNC and 30 pS SMA versions

  I don't have the SMA version, just the BNC version.

The Bodnar is the lower trace.  Twice as fast as mine and less ringing.

The amplitude of the overshoot is pretty similar.

The frequency of your "oscillation" is much lower; as best as I can see, the peak-to-trough is 0.3div vs 1.1div.

Since it is the same scope, I think that points to your construction and components. "Excess" inductance or capacitance somewhere would, of course, lower the frequency; 10* capacitance or 10* inductance would be necessary to see that. I doubt the capacitance has changed that much, so my bet is on the inductance. That would be consistent with long leads between chip, PCB and decoupling capacitors.

[magic]

Actually the logo looks familiar, but can't remember who it was.

Of course it does. It's the logo of WanHungLo Semiconductor Shenzhen
Found an example: https://www.eevblog.com/forum/projects/watch-out-for-fake-74ls-logic-ics
They do analog too: https://www.eevblog.com/forum/beginners/lm741-testing/msg4466656/#msg4466656
All  DIP8 chips here are also fakes.

I tested and few years ago different cmos logic ICs for a nice fast pulse with few abberations.
I never managed to get it really nice due to such things as the length of the bonding wire in the IC and the various decoupling capacitors.

That's why I always advocate for using the smallest available package and most compact layout whenever this topic recurs.
The mikrocontroller.net thread clearly shows what's possible with a single LVC gate in SOT23.
Those things aren't any more expensive or harder to get, so why bother with DIP?