Today, I had a very detailed look at :
MAX9601
ADCMP581
And I shall probably go for MAX9601 if I can get MAX9401 as well. 😉
Patrick
MAX9601
ADCMP581
And I shall probably go for MAX9601 if I can get MAX9401 as well. 😉
Patrick
Thanks,
MAX9601: 300 fs jitter looks very good. Are there adapters ? ssop = tssop ?
ADCMP581: 200 fs, even better.
But when I read about microstrip and stripline...
Also 50V/µs on the inputs is required.
What do you think ?
My preferred option at the moment would be differential LVPECL all the way.
No translators, no LEDs, no tricks.
That means I need a comparator that has LVPECL outputs.
I would be satisfied with 1 - 3 GHz devices.
Who knows what the DAC chip does with it.
I would be more comfortable with 8 pin single flipflops as I need buffers.
No, I like blue color more on LEDs.
I have no idea what they can do or if they add jitter. No coupling caps. Not possible for data line.
The PCM1704 has no bit clock ? It is the bitclock that latches the DAC on 4th pulse after data entry IIRC.
Last edited:
Firstly the comparators.
The ADCMP581 looks great on paper, but it is LVPECL out (suits you), and it can only take 2V differential input. So some attenuation required from the Colpitts.
The MAX9601 is easier and more flexible, so I shall go with that first.
I have a different need to you. I need to clock dsPIC33 (3.3V CMOS), PMD100 (5V CMOS), and the reclocker. The first two I am going to drive with AD8561 feed by the same Colpitts (with JFET Buffer). If I can get MAX9401, then life is wonderful. One clock line in PECL, and no output translator. Just level shift 2Vs, and I am going to test with Osram Power TOPLED first before going decoupling cap & resistor bias. The MAX9401 can take any diff input, including LVPECL.
As usual, I am going to sit on this for a month or so and see whether I change things after a while. And I need to wait for Maxim to confirm availability of MAX9401.
Class A balanced logic. This is really getting crazy, but real fun ...... 😉
Patrick
The ADCMP581 looks great on paper, but it is LVPECL out (suits you), and it can only take 2V differential input. So some attenuation required from the Colpitts.
The MAX9601 is easier and more flexible, so I shall go with that first.
I have a different need to you. I need to clock dsPIC33 (3.3V CMOS), PMD100 (5V CMOS), and the reclocker. The first two I am going to drive with AD8561 feed by the same Colpitts (with JFET Buffer). If I can get MAX9401, then life is wonderful. One clock line in PECL, and no output translator. Just level shift 2Vs, and I am going to test with Osram Power TOPLED first before going decoupling cap & resistor bias. The MAX9401 can take any diff input, including LVPECL.
As usual, I am going to sit on this for a month or so and see whether I change things after a while. And I need to wait for Maxim to confirm availability of MAX9401.
Class A balanced logic. This is really getting crazy, but real fun ...... 😉
Patrick
> I have no idea what they can do or if they add jitter.
People use LEDs for opto-couplers.
And any other active device (level shifters or diodes) would have the same problem, if any.
Anyway I have emailed Osram to ask.
Blue LEDs drop too much (>3V)..... 🙁
> No coupling caps. Not possible for data line.
Possible. Read the data sheet of 10EL89.
Patrick
PS PCM1704 has bitclock (BCLK), but not a separate MCLK (master clock), as some other does have (e.g. AK4396).
People use LEDs for opto-couplers.
And any other active device (level shifters or diodes) would have the same problem, if any.
Anyway I have emailed Osram to ask.
Blue LEDs drop too much (>3V)..... 🙁
> No coupling caps. Not possible for data line.
Possible. Read the data sheet of 10EL89.
Patrick
PS PCM1704 has bitclock (BCLK), but not a separate MCLK (master clock), as some other does have (e.g. AK4396).
Last edited:
> The output level can also be shifted by shifting the supply voltages.
Output swing is proportional to Vcc, as I remember from some of the OnSemi datasheets.
I don't want to loose swing.
Or you mean I should shift both rails, then it gets complicated with the CMOS to PECL translators at the source side .....
Patrick
Output swing is proportional to Vcc, as I remember from some of the OnSemi datasheets.
I don't want to loose swing.
Or you mean I should shift both rails, then it gets complicated with the CMOS to PECL translators at the source side .....
Patrick
Last edited:
Yes, like this.
I have EL11 buffers and could make adapters for use in my 10125 sockets / reclocker board for evaluation of that PECL - DAC chip level issue.
Should be no problem with coupling caps & bias on the EL11 inputs.
Last edited:
Good morning,
as another option a 3,1 V - 3,2 V blue LED
with bypass cap - we don't want no noise - should work perfectly to translate from NECL to DAC.
as another option a 3,1 V - 3,2 V blue LED
with bypass cap - we don't want no noise - should work perfectly to translate from NECL to DAC.
Gooood morning ...... (very early morning) 😉
Yes, sounds like good idea. No use for me though as I want to stick to PECL.
I'll let you know when I hear from Osram about driving them at 11MHz.
Patrick
Yes, sounds like good idea. No use for me though as I want to stick to PECL.
I'll let you know when I hear from Osram about driving them at 11MHz.
Patrick
My concern with the LED is that the voltage drop depends on the current and the current changes with the swing of the N/PECL output.
So I would not expect the voltage drop to be constant, it could introduce jitter.
Also the LEDs have tolerances which would result in some skew.
To keep the drop constant I would propose a bypass cap on the LED.
So I would not expect the voltage drop to be constant, it could introduce jitter.
Also the LEDs have tolerances which would result in some skew.
To keep the drop constant I would propose a bypass cap on the LED.
The cap was great idea, but it needs to have resonance well above 25MHz. So probably 0603 MLCC.
I meant the blue LED was no use for me, as I am not going for NECL (too complicated with an extra -ve supply, especially for the translators).
I had a feedback from Maxim today. The MAX9401 has been stopped since August 2009 due to low demand .... 🙁
There are still inventory around with dealers, and I am trying to source some.
But I am afraid that they are only interested in selling 1k or so in quantity, which is ly wish a bit out of reach at 20 USD per piece.
I made an excel sheet today comparing the MAX9401 with a "discrete" solution using Onsemi ICs.
The latter is both much more complicated, expensive and poorer performance.
Oh, what a pity !!!!!!!!! Only wish you had started this 6 months earlier.....
Cheers for now,
Patrick
I meant the blue LED was no use for me, as I am not going for NECL (too complicated with an extra -ve supply, especially for the translators).
I had a feedback from Maxim today. The MAX9401 has been stopped since August 2009 due to low demand .... 🙁
There are still inventory around with dealers, and I am trying to source some.
But I am afraid that they are only interested in selling 1k or so in quantity, which is ly wish a bit out of reach at 20 USD per piece.
I made an excel sheet today comparing the MAX9401 with a "discrete" solution using Onsemi ICs.
The latter is both much more complicated, expensive and poorer performance.
Oh, what a pity !!!!!!!!! Only wish you had started this 6 months earlier.....
Cheers for now,
Patrick
Last edited:
A blue 5 mm LED works.
It comes very close to the desired 1,7 V switching point.
Swing is 700 mVpp, a bit reduced by heavy load.
A 100n cap distorts the 44100 Hz square waveform.
I will make another one for the other channel with a bigger cap and do a listening test.
It comes very close to the desired 1,7 V switching point.
Swing is 700 mVpp, a bit reduced by heavy load.
A 100n cap distorts the 44100 Hz square waveform.
I will make another one for the other channel with a bigger cap and do a listening test.
Welcome to the jungle...
I had to add an EL11 as inverter for the other channel because the translator was in inverting mode, fed from Q_ of the flipflop.
No listening test yet.
I had to add an EL11 as inverter for the other channel because the translator was in inverting mode, fed from Q_ of the flipflop.
No listening test yet.
What is the problem with -5,2V ?
My DAC has four transformers and 14 regulators plus a pair of regulators for the psu voltage reference.
Impressive. No problem with -5.2V; just personal choices.
I keep thinking that no cap is fast enough for such sharp rising and falling edges ?
But then why do they recommend 100n in Fig. 2 of EL89 datasheet ?
Maybe we should just use the LED without a cap.
Patrick
BTW, do you have a good source of X'tal 22.5792 other than Segor ?
I keep thinking that no cap is fast enough for such sharp rising and falling edges ?
But then why do they recommend 100n in Fig. 2 of EL89 datasheet ?
Maybe we should just use the LED without a cap.
Patrick
BTW, do you have a good source of X'tal 22.5792 other than Segor ?
If caps are fast enough as bypass on the supply pins of the chips, why not as coupling cap ?
The appl. in the EL89 is a special case, "polarity free", I don't know....
No source for 22.5792 MHz.
I did not have much time for listening tests yet but what I heard, the LED translator sounds as good or better than the 10125. Without caps.
Standard termination for ECL is 50 ohm to Vcc -2 V.
I use alternative termination 500 ohm to -5,2 V to spare the -2 V supply.
Here also the blue LED could be used, -5,2V + 3,2V gives -2V.
So perhaps in future I will have a lot of blue LEDs in my reclocker.
Another alternative termination found in appl. notes is diode termination.
So I believe this is save.
So I have the choice, NECL with blue LED, PECL with red LED or direct LVPECL.
Anyway I believe there will not be a sonic difference.
Last edited:
I just ordered some ICs from Onsemi.
Still need to do PCB, and still trying to get MAX9401.
Have ordered 22.5792 from Segor. (My BCLK is 11.2896).
Will keep you posted,
Patrick
Still need to do PCB, and still trying to get MAX9401.
Have ordered 22.5792 from Segor. (My BCLK is 11.2896).
Will keep you posted,
Patrick
- Status
- Not open for further replies.