开云体育

Hi. I'm hoping someone in the UK can spare the time to give me a little practical help with a pulse generator. I don't have leveled oscillators, standardizers etc, and generally only very occasionally need high speed stuff. 400MHz is 399MHz more than I need most of the time. I have a 7a13 with relay problems, more on that in another post. A decent pulse generator should give me some indication of any problems with the replacement relays. I've made one. The problem is I don't know how decent it is, I just don't have the equipment. I've tweaked the vertical amp in the 7834 (haven't touched L1667 as I can't mark its original position easily) - pre tweak post tweak which looks like it's just about in spec but for all I know I could be over compensating for overshoot in the pulse generator .

I'm guessing I'm probably not the only group member with a need for a decent pulse generator. It uses two NC7SZ14 chips. One for the oscillator which gives a trigger out before the pulse if you need it (albeit inverted) and one for the output. If someone could test it and if necessary tweak it I'd package it up and make the files available to the group, either on here or maybe on OSH. No need to return the generator. Just measure, tweak if necessary and provide results - please:) 50 Ohms out by the way. Variable from ~0.5V to ~1V.

Cheers

Gordon

Gordon,

I'm in Swindon and will happily look at it using my 7104 scope, if
that is going to help.

Regards,

Norman

On 01/04/2016 10:24, Gordon tekscopes@... [TekScopes] wrote:

I'm hoping someone in the UK can spare the time to give me a little
practical help with a pulse generator.

Thanks to all who offered to help both on and off list. What a great place this is. Craig is giving it a coat of looking at so we'll see how it performs in due course no doubt.

Thanks again

Gordon

I'm in Kenilworth, and also have a 7104 and a 7S11 with S6 and 7T11 sampler - also available to assist.

Dave

On 01/04/16 10:24, Gordon tekscopes@... [TekScopes] wrote:

I'm guessing I'm probably not the only group member with a need for a
decent pulse generator. It uses NC7SZ14 chips. 50 Ohms out by the way. Variable from ~0.5V to ~1V.

FWIW, I have something similar based on three parallel (74LVC1G14 + 130ohm). The three parallel outputs lessens the effects of reduced drive strengths as the output approaches the supply rails.

At 5V and using a 350MHz scope (Tek 485) the risetime/falltime is fractionally under 1ns. Looking at the spectrum leads me to believe the risetime is 600-650ns. Without calibrating the 485, it looks like <>

It works down to a supply voltage <1.5V, with correspondingly lower output voltages and much slower transition times, of course.

In contrast, 3 NC74SZ04 have a marginally slower risetime of about 1ns.

Don't you mean "looking at the spectrum leads me to believe that the risetime is 600-650pS" (i.e. not nS).

Dave

On 01/04/16 16:00, 'David C. Partridge' david.partridge@...
[TekScopes] wrote:

Don't you mean "looking at the spectrum leads me to believe that the risetime
is 600-650pS" (i.e. not nS).

Oh..... rats :) And it is too late to claim an All Fools day joke :(

But I definitely don't mean picoSiemens (nor even picoSamples)!

-----Original Message-----
From: TekScopes@... [mailto:TekScopes@...]
Sent: 01 April 2016 15:38
To: TekScopes@...
Subject: Re: [TekScopes] UK practical help with pulse generator

At 5V and using a 350MHz scope (Tek 485) the risetime/falltime is

fractionally under 1ns. Looking at the spectrum leads me to believe the
risetime is 600-650ns. Without calibrating the 485, it looks like

On 01/04/2016 15:37, Tom Gardner tggzzz@... [TekScopes] wrote:

At 5V and using a 350MHz scope (Tek 485) the risetime/falltime is fractionally
under 1ns. Looking at the spectrum leads me to believe the risetime is
600-650ns.

That's the scope and generator rise summed of course. The 485 is spec'd at 1ns so one, the other or both are faster but that's part of what I'd like to know. If you know one for definite you can calculate the other, within reason. Rightly or wrongly I thought it would be easier to 'tighten up' the layout with a single chip and I wanted something that could be put together by anyone with reasonably predictable results. David Hess pointed me to what I think was the origin of the paralleled 7414 circuit and I did try it but the overshoot/ringing bothered me a bit. OK for the tdr which it was originally designed for I think but not really good enough for calibrating overshoot!

Without calibrating the 485, it looks like
<>

It works down to a supply voltage <1.5V, with correspondingly lower output
voltages and much slower transition times, of course.

In contrast, 3 NC74SZ04 have a marginally slower risetime of about 1ns.

I based my choice on the minimum tpd which is lower for the NC7... assuming the rise time has to support the minimum tpd but I'm not that experienced at these things so we'll see.

Cheers

Gordon

Oh..... rats :) And it is too late to claim an All Fools day joke :(

My April Fool's day event was having root canal work done. Round 1. Round 2 on 18th.

Craig

On 01/04/16 17:38, 'Craig Sawyers' c.sawyers@... [TekScopes] wrote:

My April Fool's day event was having root canal work done. Round 1. Round 2 on 18th.

Curiously I never found it to be as bad as people claimed. I tend to fall asleep in the dentists chair and only wake up if they want me to do something. I think being an engineer helps, since if anything I concentrate on the unusual sensations and try to imagine what's causing them :)

I've no reason to believe that I can't feel pain; some things certainly hurt!

On 01/04/16 16:51, Gordon tekscopes@... [TekScopes] wrote:

On 01/04/2016 15:37, Tom Gardner tggzzz@... [TekScopes] wrote:

At 5V and using a 350MHz scope (Tek 485) the risetime/falltime is fractionally
under 1ns. Looking at the spectrum leads me to believe the risetime is
600-650ns.

That's the scope and generator rise summed of course. The 485 is spec'd
at 1ns so one, the other or both are faster but that's part of what I'd
like to know. If you know one for definite you can calculate the other,
within reason. Rightly or wrongly I thought it would be easier to
'tighten up' the layout with a single chip and I wanted something that
could be put together by anyone with reasonably predictable results.
David Hess pointed me to what I think was the origin of the paralleled
7414 circuit and I did try it but the overshoot/ringing bothered me a
bit. OK for the tdr which it was originally designed for I think but not
really good enough for calibrating overshoot!

I understand and accept that you aren't making definitive claims; neither am I :)

Overshoot is usually associated with inductance in ground leads, especially when using scope probes. I use it as an example of why all circuits are analogue and why probes matter.

There shouldn't be any overshoot if the outputs are directly connected to the scope with a 50 ohm transmission line. Using SMD gates allows the trace on the PCB to be very short, and I couple the board directly to the scope with a 2" BNC male to BNC male adaptor, which should be OK to 1GHz, maybe more.

None of which constitutes proof, so I would like to test it on a decent scope - that's why I didn't offer to help you.

Without calibrating the 485, it looks like

<>

It works down to a supply voltage <1.5V, with correspondingly lower output
voltages and much slower transition times, of course.

In contrast, 3 NC74SZ04 have a marginally slower risetime of about 1ns.

I based my choice on the minimum tpd which is lower for the NC7...
assuming the rise time has to support the minimum tpd but I'm not that
experienced at these things so we'll see.

I didn't care about the propagation delay, which is usually how fast the output gets halfway (i.e. to the threshold voltage). I was more concerned about the ability to drive a 50 ohm load all the way to the supply rails. The LVC family seemed to have a marginally better current driving capability near the supply rails, and the load is further reduced by having 3 in parallel plus resistors.

"The Art of Electronics" indicates the LVC and SZ are both "nice" at 5V, and that you can get a lower tpd with ALVC at 3.3V and AUC at 2.5V. Unfortunately AUC's current driving capability is also lower . I'm tempted to spin another variant of my board, sometime.

The 485 also has a very nice calibrator with a 1ns or less risetime. It is used to calibrate the 485's front end. It is effectively an ECL output with a 450ohm impedance thus forming a 10:1 divider when attached to a 50ohm load. The only "surprise" is that voltages incident on the 20pF (plus 50ohm) scope input cause a small reflection, which is again reflected when that hits the 450ohm output and finally appears as a short glitch at twice the cable's propagation delay.

Its output levels are calibrated by removing the transistors and using a DMM to measure and tweak the output voltages. Sneaky.

Curiously I never found it to be as bad as people claimed. I tend to fall asleep in the dentists

chair and

only wake up if they want me to do something. I think being an engineer helps, since if anything I
concentrate on the unusual sensations and try to imagine what's causing them :)

Yes - actually me too. Because there was zero pain, just the weird discomfort of someone offering
abuse to the internals of my tooth with motorised files, I too was feeling oddly chilled.

Helped by my dentist of 20 years being a wizard with anaesthetic - absolutely no part of my face is
dead - he manages to just numb the tooth itself.

Before I found him, previous dentists used to numb most of your face. I've even ended up with a
numb nose with a particular knock-em-dead dentist in the past.

I too have a variant of the NC7SZ14 fast edge generator. Mine follows a 28MHz DIP crystal oscillator so the rep rate is fast enough to use with a7T11. It is air wired with SMD components directly on a SMA socket, there is a 220R / 50R divider to minimise the current requirement (200mV into 50 Ohm terminated both ends) I see 500psec rise and 200 psec fall with an S4 sampler. There is a few percent overshoot and I plan to build another with a better ground plane.

I am sure a decent PCB layout could give better results but haven't branched into PCB making yet.

Roger

On 01/04/16 19:23, very_fuzzy_logic@... [TekScopes] wrote:

I too have a variant of the NC7SZ14 fast edge generator. Mine follows a 28MHz DIP crystal oscillator so the rep rate is fast enough to use with a7T11. It is air wired with SMD components directly on a SMA socket, there is a 220R / 50R divider to minimise the current requirement (200mV into 50 Ohm terminated both ends) I see 500psec rise and 200 psec fall with an S4 sampler. There is a few percent overshoot and I plan to build another with a better ground plane.

Interesting.

My new variant will have a external input and a internal RC schmitt trigger relaxation oscillator. The external inputs will be unterminated, 50ohm to ground, 120/82 ohms to Vcc/Gnd for 50 ohms to the threshold voltage. The input will also be AC or DC coupled. There will be some reflections etc, but that won't be a major problem since the external input will drive a schmitt trigger.

I am sure a decent PCB layout could give better results but haven't branched into PCB making yet.

Don't be put off by SMD. I recently returned to PCB designing and assembling after 25 years of avoiding it. Nowadays design is easy, and SMD is remarkably easy and doesn't require expensive equipment. I now prefer it to through hole components.

You can get excellent free PCB/schematic programs, and 10 5cm*5cm double-sided through hole plated boards with silk screens and solder masks for ?10, within a couple of weeks.

There are many many guides on the web. My experiences are noted at <>

Using a clean pulse works great for tracking down the bad relays and
any other problems in a 7A13. Since it is only 100 MHz, the bandwidth
requirements on the pulse are relatively low.

Since any significant problems in the 7A13 will cause gross
distortion, pulse fidelity is also not critical. You should be able
to get away with using the TTL sync output on a function generator if
you connect directly to it using a x10 probe or use a terminated 50
ohm coaxial cable.

If you have a personal computer with a VGA connector available, the
horizontal sync output will provide a clean and fast (less than 1
nanosecond) edge if you keep the x10 probe connections very short.

Your home made pulse generator may not be good enough for the 400 MHz
7834 but it should work fine for a 100 MHz 7A13.

On Fri, 1 Apr 2016 17:51:31 +0100, you wrote:

...

Overshoot is usually associated with inductance in ground leads, especially when
using scope probes. I use it as an example of why all circuits are analogue and
why probes matter.

There shouldn't be any overshoot if the outputs are directly connected to the
scope with a 50 ohm transmission line. Using SMD gates allows the trace on the
PCB to be very short, and I couple the board directly to the scope with a 2" BNC
male to BNC male adaptor, which should be OK to 1GHz, maybe more.

None of which constitutes proof, so I would like to test it on a decent scope -
that's why I didn't offer to help you.

There is a way to improve the flatness of the pulse commonly used in
reference level pulse generators; drive a 50 ohm termination through a
series diode which disconnects the logic output from the termination
on one edge. This way the unterminated stub between the output
transistor and transmission line is disconnected at the transition.
Many of the Tektronix reference level pulse generators like the PG506
and the NBS design do this.

The diode needs to be fast though because its reverse recovery time
will delay disconnecting the logic output. For that reason small
signal schottky diodes which have no recovery time are idea but the
base-emitter junction of a transistor like a 2N3904 is also suitably
fast (better than 600 picoseconds?) and their low reverse breakdown
voltage is not a problem in this application. The 1N4148 and similar
switching diodes are too slow.

Low diode capacitance is also important to prevent feedthrough. Faster
designs use more than one diode and a second inverted pulse to
compensate for charge delivered through the diode's junction
capacitance. Tektronix did this in their reference level pulse
generators intended for 1 GHz oscilloscope calibration.

On Fri, 1 Apr 2016 at 18:30 David davidwhess@... [TekScopes] <
TekScopes@...> wrote:

Low diode capacitance is also important to prevent feedthrough. Faster

designs use more than one diode and a second inverted pulse to
compensate for charge delivered through the diode's junction
capacitance. Tektronix did this in their reference level pulse
generators intended for 1 GHz oscilloscope calibration.

Hey David,

would you happen to be able to point to a schematic of such a thing?

Siggi

On Fri, 01 Apr 2016 22:40:23 +0000, you wrote:

On Fri, 1 Apr 2016 at 18:30 David davidwhess@... [TekScopes] <
TekScopes@...> wrote:

Low diode capacitance is also important to prevent feedthrough. Faster

designs use more than one diode and a second inverted pulse to
compensate for charge delivered through the diode's junction
capacitance. Tektronix did this in their reference level pulse
generators intended for 1 GHz oscilloscope calibration.

Hey David,

would you happen to be able to point to a schematic of such a thing?

Siggi

Take a look at the schematics for the 7000 series standardizers, the
PG506, and the Type 106 square wave generator.

I can host the documentation to the NBS design if you would like to
see it as well.

On 01/04/2016 23:13, David davidwhess@... [TekScopes] wrote:

Since any significant problems in the 7A13 will cause gross
distortion, pulse fidelity is also not critical. You should be able
to get away with using the TTL sync output on a function generator if
you connect directly to it using a x10 probe or use a terminated 50
ohm coaxial cable.

Some interesting stuff coming out here, as usual. Here's some preliminary results with the relay adapters in place -



I haven't got stuck into it properly just yet. I'd like to know what the pulse generator's doing first but that's what it looks like so far as it is. I've replaced k60 and 61, one of which was bad but it seemed sensible to keep them matched and K47 and 48 (both bad). I've since got some more attenuators which will allow me to check the lower ranges but haven't tried them yet. I managed to rejuvenate one of the relays but don't really trust it. The fact that it's recovered suggests it's knocked the flaking crap off the polepieces which is now floating about inside the case.

Cheers

Gordon

On Sat, 2 Apr 2016 01:04:50 +0100, you wrote:

On 01/04/2016 23:13, David davidwhess@... [TekScopes] wrote:

Since any significant problems in the 7A13 will cause gross
distortion, pulse fidelity is also not critical. You should be able
to get away with using the TTL sync output on a function generator if
you connect directly to it using a x10 probe or use a terminated 50
ohm coaxial cable.

Some interesting stuff coming out here, as usual. Here's some
preliminary results with the relay adapters in place -



I haven't got stuck into it properly just yet. I'd like to know what the
pulse generator's doing first but that's what it looks like so far as it
is. I've replaced k60 and 61, one of which was bad but it seemed
sensible to keep them matched and K47 and 48 (both bad). I've since got
some more attenuators which will allow me to check the lower ranges but
haven't tried them yet. I managed to rejuvenate one of the relays but
don't really trust it. The fact that it's recovered suggests it's
knocked the flaking crap off the polepieces which is now floating about
inside the case.

Cheers

Gordon

When my 7A13 had relays which were not closing, the pulse was
massively distorted. I did like you and replaced both x10 relays;
since they are SPDT, it was possible to directly use a modern DPDT to
replace them by just bending the pins around. Also the relay which
switches the bandwidth limit on the trigger output was bad but of
course that was not apparent except for the trigger point on the
display not matching up. That relay I repaired by operating it on my
function generator for a couple of hours.

A fast edge pulser built using surface mount parts with moderate care
is going to settle faster than the 100 MHz 7A13 can respond so I am
inclined to believe that what you are seeing is a problem with the
7A13. On the other hand, a reference flat pulse generator is the only
way to be sure.