Interesting feature of an inexpensive function generator

Chris Jones <lugnut808@spam.yahoo.com> wrote:

> On 02/07/2013 09:28, Charles wrote:
 
> If you are satisfied with rise and fall times of about 250 - 300ps,
> then you can use a cheap 74LVC04AD or similar.
> 
> http://www.youtube.com/watch?v=oUiEeOp8ynY#t=698s
 
> Unfortunately that particular board didn't have sufficient
> low-frequency decoupling (it was intended for use at much higher
> frequencies where the decoupling was adequate), so the power supply
> cable inductance rings badly with the small decoupling capacitors on
> the board. If he had looked at the fall-time instead of rise time, and
> left my ac-coupling and back termination in, it might have been a bit
> cleaner. The best option would be to improve the local low frequency
> decoupling. 
 
> Chris
 
Chris, how do you get 250 ps risetime? I tried to find the datasheet on 
Octopart and they have started using captchas for rate limiting. I could 
never get past that screen.

I did get into Mouser and found the part is made by TI and NXP. Neither 
datasheet said clearly what the output rise and fall times were. The best 
I could find was a table that had numbers like <2.5ns, but it wasn't 
clear if that was for the input or the output.

I'd like to find out for sure if that family really does 250ps risetime, 
but I'd be surprised if it does. Do you have any more information?

Thanks,

JK

On 03/07/2013 20:16, John K wrote:
> Chris Jones <lugnut808@spam.yahoo.com> wrote:
>
>> On 02/07/2013 09:28, Charles wrote:
>
>> If you are satisfied with rise and fall times of about 250 - 300ps,
>> then you can use a cheap 74LVC04AD or similar.
>>
>> http://www.youtube.com/watch?v=oUiEeOp8ynY#t=698s
>
>> Unfortunately that particular board didn't have sufficient
>> low-frequency decoupling (it was intended for use at much higher
>> frequencies where the decoupling was adequate), so the power supply
>> cable inductance rings badly with the small decoupling capacitors on
>> the board. If he had looked at the fall-time instead of rise time, and
>> left my ac-coupling and back termination in, it might have been a bit
>> cleaner. The best option would be to improve the local low frequency
>> decoupling.
>
>> Chris
>
> Chris, how do you get 250 ps risetime? I tried to find the datasheet on
> Octopart and they have started using captchas for rate limiting. I could
> never get past that screen.
>
> I did get into Mouser and found the part is made by TI and NXP. Neither
> datasheet said clearly what the output rise and fall times were. The best
> I could find was a table that had numbers like <2.5ns, but it wasn't
> clear if that was for the input or the output.
>
> I'd like to find out for sure if that family really does 250ps risetime,
> but I'd be surprised if it does. Do you have any more information?
>
> Thanks,
>
> JK
>

The ~280ps risetime was found by measurement. If you want it guaranteed 
then you'd need a more expensive chip. It is only a typical value on 
even ADI's fast comparators ADCMP572 etc. In practice rather than paying 
a manufacturer extra to guarantee it, you might be better off buying a 
lot of 74LVC chips from the same batch, measuring a few and then knowing 
that this is what you have.

You can see the risetime figure briefly on the screen of the scope in 
the video I linked. I measured it previously with a similar scope and I 
got about 280ps, depending on supply voltage etc. That was for an old 
Philips chip, so now that would be called NXP. I have a few of those old 
chips left over, and I know they are fast. No idea whether they use the 
same chip layout and fab now, maybe JT can advise.

Chris

Chris Jones <lugnut808@spam.yahoo.com> wrote:

> The ~280ps risetime was found by measurement. If you want it
> guaranteed then you'd need a more expensive chip. It is only a typical
> value on even ADI's fast comparators ADCMP572 etc. In practice rather
> than paying a manufacturer extra to guarantee it, you might be better
> off buying a lot of 74LVC chips from the same batch, measuring a few
> and then knowing that this is what you have.
 
> You can see the risetime figure briefly on the screen of the scope in 
> the video I linked. I measured it previously with a similar scope and
> I got about 280ps, depending on supply voltage etc. That was for an
> old Philips chip, so now that would be called NXP. I have a few of
> those old chips left over, and I know they are fast. No idea whether
> they use the same chip layout and fab now, maybe JT can advise.
 
> Chris
 
It was very difficult to follow the video. He kept rotating the time axis 
knob back and forth. Too fast to actually look at the waveform and try to 
understand what it is saying.

I am surprised the risetime can be that low. It takes very fast 
transistors to do that. If that is true, why is the propagation delay so 
long? You'd expect with transistors that fast, the prop delay would be 
down in the picosecond range instead of 4.5ns at 3.3V as shown here

"http://datasheet.octopart.com/SN74LVC04APWR-Texas-Instruments-datasheet-
114799.pdf"

I don't have any of these devices on hand and do most of my stuff in 
10GHZ ecl so I really don't have a need for them. If you have access to a 
fast scope, could you take a picture of the risetime?

Thanks,

JK

On 04/07/2013 19:40, John K wrote:
> Chris Jones <lugnut808@spam.yahoo.com> wrote:
>
>> The ~280ps risetime was found by measurement. If you want it
>> guaranteed then you'd need a more expensive chip. It is only a typical
>> value on even ADI's fast comparators ADCMP572 etc. In practice rather
>> than paying a manufacturer extra to guarantee it, you might be better
>> off buying a lot of 74LVC chips from the same batch, measuring a few
>> and then knowing that this is what you have.
>
>> You can see the risetime figure briefly on the screen of the scope in
>> the video I linked. I measured it previously with a similar scope and
>> I got about 280ps, depending on supply voltage etc. That was for an
>> old Philips chip, so now that would be called NXP. I have a few of
>> those old chips left over, and I know they are fast. No idea whether
>> they use the same chip layout and fab now, maybe JT can advise.
>
>> Chris
>
> It was very difficult to follow the video. He kept rotating the time axis
> knob back and forth. Too fast to actually look at the waveform and try to
> understand what it is saying.
>
> I am surprised the risetime can be that low. It takes very fast
> transistors to do that. If that is true, why is the propagation delay so
> long? You'd expect with transistors that fast, the prop delay would be
> down in the picosecond range instead of 4.5ns at 3.3V as shown here
>
> "http://datasheet.octopart.com/SN74LVC04APWR-Texas-Instruments-datasheet-
> 114799.pdf"
>
> I don't have any of these devices on hand and do most of my stuff in
> 10GHZ ecl so I really don't have a need for them. If you have access to a
> fast scope, could you take a picture of the risetime?
>
> Thanks,
>
> JK
>

Note that the rise and fall time of the 74LVC04 is quoted in the 
datasheet with a 30pF or 50pF capacitive load. Instead, I measured the 
risetime with the output connected through a resistive back-terminating 
network to a matched coaxial cable, which would be expected to give much 
faster edges.

Each inverter in those 74LVC04 chips would contain several (an odd 
number of...) inverter stages in a chain. They would probably also have 
ESD resistors at the input which would somewhat slow down the 
propagation delay to make it likely to be several times greater than the 
best possible rise and fall times. I think they probably use a 0.5um 
CMOS process guessing from the absolute max. supply voltage rating. On a 
0.18um process I think a single inverter stage has about 100ps 
propagation delay.

Jim Thompson might know a lot more about it than that, and might even be 
allowed to tell us.

I no longer have access to a fast scope so I can't give you any plots, 
but I can confirm that you could get a pretty clean looking square wave 
with <300ps rise and fall times out of that very board that was on the 
video.

Chris

Chris Jones <lugnut808@spam.yahoo.com> wrote:

> Note that the rise and fall time of the 74LVC04 is quoted in the 
> datasheet with a 30pF or 50pF capacitive load. Instead, I measured the
> risetime with the output connected through a resistive
> back-terminating network to a matched coaxial cable, which would be
> expected to give much faster edges.

Where do you see a spec for the rise and fall times? I tried several 
different manufacturer's datasheets and all I can find is the prop delay.

I can see the mfgr's wanting to eliminate every possible spec parameter 
that they can get away with. This reduces cost and removes the loss for 
devices that don't meet the spec. 

But I'm old school. I want as many spec parameters I can get:)

JK

On 04/07/2013 21:12, John K wrote:
> Chris Jones <lugnut808@spam.yahoo.com> wrote:
>
>> Note that the rise and fall time of the 74LVC04 is quoted in the
>> datasheet with a 30pF or 50pF capacitive load. Instead, I measured the
>> risetime with the output connected through a resistive
>> back-terminating network to a matched coaxial cable, which would be
>> expected to give much faster edges.
>
> Where do you see a spec for the rise and fall times? I tried several
> different manufacturer's datasheets and all I can find is the prop delay.
>
> I can see the mfgr's wanting to eliminate every possible spec parameter
> that they can get away with. This reduces cost and removes the loss for
> devices that don't meet the spec.
>
> But I'm old school. I want as many spec parameters I can get:)
>
> JK
>

http://www.nxp.com/documents/data_sheet/74LVC04A.pdf
Table 8

In any case, removing the capacitive loading will speed up the output 
waveform, both risetime and propagation delay.

Chris