High frequency voltage translation using the LSF0102?

[AndrewP]

I cannot make it work

I have a 25MHz clock driving an 'LVC1G34 buffer / line driver jobbie whose output is fed into the LSF0102. That seems a reasonable approximation of what ICs I might really use.

It does terrible things to my 0V line and *still* has a slow rise time.

I mean I can kinda see it trying but I guess the 'LVC1G34 just cannot supply enough current for the 330Ohm pull-up resistor on the output side of the '0102. It does make it all the way up to 4.5V before getting driven back down the logical 'low' of 1.4V.

With the 'LSF disconnect I get a pretty decent 25MHz signal from the '34. The slightly slow rise time you see is because the probe is after a 100Ohm resistor that limits current slightly and stops the '0102 (or at least its pull resistor) from flattening the signal completely.

Am I using the wrong device for the job? Can 3.3V to 5V translation with decent rise times be done using LSF ICs? They work fine for slower (sub-1MHz) signals for things like the SNES controllers but it doesn't seem they should be anywhere near LVC circuits...

(below the schematic and circuit).

[GARTHWILSON]

I wonder if those long leads are causing the problem.

[ttlworks]

Well, it might be worth a try if the output of a 3.3V powered Renesas XLH535025 oscillator
is able to drive the input of a 5V powered 74ACT logic gate... 74ACT125 for instance.

To be on the safe side, because of the two different power supplies better place a 100 Ohm resistor between both chips.
If that resistor affects the shape of the clock signal, try wiring something like a 100pF capacitor in parallel to the resistor.

[CountChocula]

Howdy! Assuming the schematic is correct, why do you have Vrefb going through a 20k resistor to +5V? Won't that limit the amount of current that the LSF0102 can drive on its output?

[Dr Jefyll]

For what it's worth, I notice TI has a series of videos dealing specifically with these products.

voltage-level-translation-lsf-family

-- Jeff

[AndrewP]

Thanks for all the replies!

GARTHWILSON wrote:

I wonder if those long leads are causing the problem.

I don't know but I don't thinks so. I originally built this on a breadboard and a got a *very* similar signal that at the time I thought was because of the breadboard and jumper wires. Just to prove I was right and it was the breadboard I then built it on the strip-board you see. But I was wrong. It wasn't the breadboard. This is why I have a gut feeling it's not wire length or the circuit build but something I am either a) misunderstanding or b) that I am using the wrong component for the job.

ttlworks wrote:

Well, it might be worth a try if the output of a 3.3V powered Renesas XLH535025 oscillator ... 100pF capacitor in parallel to the resistor.

Oo! I tried this! Kind of. And it worked perfectly. I used an XLHxxx080 to drive VHC circuitry and it was fine. Just for the sake of it I then tried upping the clocks voltage to 4.1V and left it running for four days. It didn't seem too have suffered but the current draw on the clock was more than doubled.

CountChocula wrote:

Howdy! Assuming the schematic is correct, why do you have Vrefb going through a 20k resistor to +5V? Won't that limit the amount of current that the LSF0102 can drive on its output?

Dr Jefyll wrote:

For what it's worth, I notice TI has a series of videos dealing specifically with these products.

Now this was interesting. I was using a 20K because I found some random circuit on the interwebs that suggested a 20K resistor and a 0.1uF capacitor for the enable Bias circuit.

Watching the the videos Dr Jefyll suggested they really stress that getting the bias circuit right is important (probably why it is now built into the '0204). I swapped the 20K out for the suggest 200K and got this:

Still not great in terms of rise time but notice that the yellow probe low level voltage has dropped from 1.4V to 0.7ishV. I want to use that yellow signal in other LVC devices and if you sort of squint 0.7V is okay.

I think the actual solution is to use a 74LVC4245 3.3V to 5V up translator and put all my signals through it to keep the timings similar. It's basically the '816 and a bunch of 5V tolerant LVC circuitry after the '4245 so in terms of timing and not mucking around with LSF devices this is my best choice. And the '4245 is the same price as the LSF0102 anyway

[BillO]

Is there a significant voltage drop along the ground wire going back to the power supply? Or from where you attach your scope ground to the ground on any of the modules?

If you scope the clock signal going into the buffer does the '0' level come up there too?

[AndrewP]

BillO wrote:

Is there a significant voltage drop ... If you scope the clock signal going into the buffer does the '0' level come up there too?

Nope, this is the same circuit as before but with the yellow probe connected to the clock output (and the purple probe disconnected because it was distracting).

I then fiddled a bit more and had an epiphany. I replaced the 330Ω pull-up resistor with a 75Ω one and got the signals below (yellow is still the same probe on the clock output).

I'm basically creating a voltage divider using the 100Ω and 75Ω resistors (with the potential difference between 5V and whatever the '34 is outputting). The LSF0102 is barely even involved anymore.

I think the lesson here is that anything with pull resistors should not be used with high frequency signals*.

*[EDIT]To clarify a bit. It seems to me that signals through large resistors will change too slowly to be useful and small resistors will stomp all over the small current (LVC) ICs can drive.

[BillO]

I think you're right. Do you have s trace without the 100 Ohm resistor?

[AndrewP]

BillO wrote:

I think you're right. Do you have s trace without the 100 Ohm resistor?

Yup, after a few snags * here are the traces with:

No 100Ω resistor (output of the '34 straight into the LSF0102) and a 330Ω pull-up on the output of the LSF0102.

No 100Ω resistor and a 75Ω pull-up on the output of the LSF0102.

* South Africa is still under load shedding - that's rolling national brownouts where the electricity is out for about 9 hours every day. That's a problem but not THE problem. When the power is out is a great time to steal THE WHOLE $%#&ing minisub. They're not small and are incredibly heavy. Like, just how? And really!? What!

The one pictured isn't the one I'm connected to but still a group stole cables out of one on my street. Which is a nuisance. No, the problem is they tried to steal another minisub on my street and messed it up. When the power came on there were massive voltage spikes on the lines and that damaged pretty much every appliance plugged in.

Not good for soldering or running oscilloscopes.

[Dr Jefyll]

[BillO]

Nobody sees this happening? HSA, where in SA do you live?

Lo0ks like that happened right in Johannesburg.

600,000 rand is about $35K USD.

[AndrewP]

I'm in Midrand, Johannesburg. The picture looks like it was taken somewhere Wynberg to Alexandra (also Jo'burg) - nobody is going to 'see' it happening there. The one where the cables were stolen on my street? I have no idea how no one noticed, it's on a street corner with five houses facing it.