I have an answer from Osram today on the PowerLED switching :
"Bei 20 mHz haben wir das Teil nicht getestet, dass sollte allerdings (meine Meinung nach) nicht funktionieren, da die Anstieg- & Abfallzeiten bei ~50ns liegen.
Da bedeutet, Sie können nur die Pulse von >100ns erzeigen.
Bei 20MHz wäre die Pulslänge (1:1 ratio) nur 25ns."
So I think you can expect a slower ramp on top of the rising / falling edge of the ECL logic of about 50ns rise time, and an amplitude of about 50mV. The transconductance of the LED is about 0.16S (assuming you use 100R termination resistors and 800mV ECL swing).
Patrick
"Bei 20 mHz haben wir das Teil nicht getestet, dass sollte allerdings (meine Meinung nach) nicht funktionieren, da die Anstieg- & Abfallzeiten bei ~50ns liegen.
Da bedeutet, Sie können nur die Pulse von >100ns erzeigen.
Bei 20MHz wäre die Pulslänge (1:1 ratio) nur 25ns."
So I think you can expect a slower ramp on top of the rising / falling edge of the ECL logic of about 50ns rise time, and an amplitude of about 50mV. The transconductance of the LED is about 0.16S (assuming you use 100R termination resistors and 800mV ECL swing).
Patrick
Last edited:
Why PowerLED ?
Do you think normal LEDs are too slow ?
Difficult for me to test, my scope has only 10 nS shortest sampling time.
There may be a difference between the optical and electrical rise time.
Do you think normal LEDs are too slow ?
Difficult for me to test, my scope has only 10 nS shortest sampling time.
There may be a difference between the optical and electrical rise time.
Last edited:
I googled a bit and found rise times between 1 ns and 100 ns for blue LEDs.
Not very helpful.
Looks like high speed fiber optical transmitters use laser diodes.
That would make the decision very easy:
Direct LVPECL, despite the fact that is sounds very good with the blue LED.
Not very helpful.
Looks like high speed fiber optical transmitters use laser diodes.
That would make the decision very easy:
Direct LVPECL, despite the fact that is sounds very good with the blue LED.
Power LED because I assume max current 16mA, and do not want to run too close to specified max current. But probably too conservative, so normal LED will work.
I do not think the 50mV change across the LED matters. But yes, LVPECL will not have that issue, whether it makes any difference or not.
Patrick
I do not think the 50mV change across the LED matters. But yes, LVPECL will not have that issue, whether it makes any difference or not.
Patrick
My simulation with a 3,2 V zener gives a current swing of 1 mA to 5 mA.
I did some measurements.
Swing across the LED: 140 mVpp.
Rise time with my slow scope manually measured with the cursor 0 % to 100 % is 30ns both on the ECL and on the DAC side.
Resolution is only 10 ns = one sampling point
So I guess the LED is below 10 ns.
I am looking forward to hear your opinion when you have built the thing.
Last edited:
I guess a smaller LED (than PowerLed) is likely to give you a faster rise time at the expense of a larger swing.
But then it does work in your circuit. So we can still use it as is, while searching for a "better" solution. Perhaps we should use laser diodes ??
Patrick
But then it does work in your circuit. So we can still use it as is, while searching for a "better" solution. Perhaps we should use laser diodes ??
Patrick
Anyway I planned to use 1,8 mm LEDs, by now those are 5 mm.
I read that LEDs have a capacitance, the rise time depends on the drive impedance and most likely on the chip area, so a SMD LED should have the fastest response, but 1,8 mm should come close.
Also PCM1704 has a settling time of 200 ns, PCM56 has 350 ns.
I read that LEDs have a capacitance, the rise time depends on the drive impedance and most likely on the chip area, so a SMD LED should have the fastest response, but 1,8 mm should come close.
Also PCM1704 has a settling time of 200 ns, PCM56 has 350 ns.
Fast LEDs can be found in opto couplers; but they won't take much current.
Laser diodes can take quite a bit more, and can work at MHzs easy (as in telecommunication). So I don't think we'll run out of choices.
Patrick
Laser diodes can take quite a bit more, and can work at MHzs easy (as in telecommunication). So I don't think we'll run out of choices.
Patrick
This gives an indication how LED works at high speed :
http://www.tytlabs.co.jp/japanese/review/rev402pdf/402_007kato.pdf
Patrick
http://www.tytlabs.co.jp/japanese/review/rev402pdf/402_007kato.pdf
Patrick
A couple of questions :
What termination resistor value are you using (50R) ?
And did you put the LED befpre or after the termination resistor ?
If after, there is little current flowing through the LED, and hence should have next to no voltage variation. Then who cares about the rise time of the LED.
Patrick
What termination resistor value are you using (50R) ?
And did you put the LED befpre or after the termination resistor ?
If after, there is little current flowing through the LED, and hence should have next to no voltage variation. Then who cares about the rise time of the LED.
Patrick
500 ohm from ECL to -5,2 V.
LED from ECL to DAC input.
430 or 500 ohm, not sure I'm not at home, from DAC to 5 V and 500 ohm from DAC to ground.
Swing across the LED 140 mV.
ECL has open Emitter outputs and needs a pulldown = termination to negative.
LED from ECL to DAC input.
430 or 500 ohm, not sure I'm not at home, from DAC to 5 V and 500 ohm from DAC to ground.
Swing across the LED 140 mV.
ECL has open Emitter outputs and needs a pulldown = termination to negative.
Last edited:
If there is no current through the LED (other than whatever the CMOS input might suck out, which is uA's I think), there should be little or no swing across the LED, right ?
Patrick
Patrick
Right, if the pull-up resistor on the DAC side is very large and I relocate the ECL terminations to the DAC boards.
I could do that on the next reclocker bord, this board with the translators will remain unmodified for some friend. That LED stuff is plugged into the translator sockets provisorical.
So for now I will order 1,8mm LEDs and hope I can have some nice glow 🙄 and stay below 1 ns.
If I use blue LEDs, they have to shine a bit, otherwise I will use direct LVPECL.
The only issue is if the LVPECL logic level is compatible with a PECL comparator.
Last edited:
You can always use a PECL to LVPECL translator. They are fast and introduce little jitter.
Only translation to CMOS or TTL is problematic.
I think I will terminate at DAC, then LED between termination and DAC input.
If necessarily bias the LED with a CCS diode (btw LED & DAC) to -5V.
The LED is then always on but draws constant current (say 2mA) and hence stable Vf.
Then I can use normal 0603 SMD LEDs.
😉
Patrick
Only translation to CMOS or TTL is problematic.
I think I will terminate at DAC, then LED between termination and DAC input.
If necessarily bias the LED with a CCS diode (btw LED & DAC) to -5V.
The LED is then always on but draws constant current (say 2mA) and hence stable Vf.
Then I can use normal 0603 SMD LEDs.
😉
Patrick
Bernhard,
I had a look again today at the oscillator schematics and have a stupid question.
In a "standard" single ended Colpitts, one uses the Xtal as an inductor in series with a cap to form a LC tank.
Does the Xtal functions differently here ? I am missing the cap in series, so to speak.
Thanks in advance,
Patrick
I had a look again today at the oscillator schematics and have a stupid question.
In a "standard" single ended Colpitts, one uses the Xtal as an inductor in series with a cap to form a LC tank.
Does the Xtal functions differently here ? I am missing the cap in series, so to speak.
Thanks in advance,
Patrick
I think you mean the 100 p / 120 p.
A small cap in series is normally used to pull the crystal frequency.
- Status
- Not open for further replies.