You might also consider this for an amplifier on the TT. The 200fA of current noise is the greatest interest. I'm going to try one, actually two as a differential output amp, on the right channel of an AK5394 and something else on the left channel. Making up a discovery board.
Really? A lot of clock recovery system are not that low.
I can't find any isolator IC's below 100ps pulse jitter.
My research shows there are a few different measures of jitter.
Which are you referring to?
a paper i refered others to but would take awhile to retrace and find it. If you cant do 20ps then do the best you can to get as low as possible. It doesnt look easy... Benchmark does some magic to get their jitter into the unknownn with their UltraLock2 clock system. I think 100ps would be passable, though.
-RM
-RM
Last edited:
AKM has some new products that can do 300fs and 400 fs but you need cloud computing to run them.
Jitter is way more complex. The isolators are high. They are fine for the data lines but use a transformer for the clock. I can provide details.
20 pS should get really good SNR at 20 KHz. FWIW that's what I measured on the master clocks on the Juli@ card. I would measure it differently now.
The datasheet for the http://www.analog.com/static/imported-files/data_sheets/AD7760.pdf has what you need on page 25. In their example 108 dB at 1 MHz needs just a little less that 2 pS. A second example shows 19 KHz, 120 dB needs 133 pS.
Benchmark uses an async sample rate converter to "process" (hide) the jitter. They are not as wonderful it seems as they are reputed to be. Benchmark converts to an off frequency that is supposed to have lower artifacts. I have not seen a lot of uptake on the concept and once converted any jitter or jitter artifacts in the data are baked in.
20 pS should get really good SNR at 20 KHz. FWIW that's what I measured on the master clocks on the Juli@ card. I would measure it differently now.
The datasheet for the http://www.analog.com/static/imported-files/data_sheets/AD7760.pdf has what you need on page 25. In their example 108 dB at 1 MHz needs just a little less that 2 pS. A second example shows 19 KHz, 120 dB needs 133 pS.
Benchmark uses an async sample rate converter to "process" (hide) the jitter. They are not as wonderful it seems as they are reputed to be. Benchmark converts to an off frequency that is supposed to have lower artifacts. I have not seen a lot of uptake on the concept and once converted any jitter or jitter artifacts in the data are baked in.
This can't be answered in general. Different converter architectures show different sensitivities to jitter, and the signal degradation very much depends on the jitter spectrum. If the jitter is broadband/white, all that happens is decreased SNR. It takes quite a bad clock (or an exceptionally good converter) until this becomes detectable at audio frequencies. More significant is serious 1/f noise content or periodic content (spectral lines).
Quoting a single integrated figure is more or less useless. You want a spectral density plot in dBc/rtHz or s/rtHz.
Transformer? I'm not sure why you want to isolate the clock signal (I might have missed that bit as I've not read all posts, my apologies if so); best practice is to place the clock near the converter, and isolate its signal for data transmission if needed.
Samuel
True; about the only place you want an async SRC is for DACs (yes, the ESS stuff has those built-in!).
Samuel
Hi Samuel,
You didn't miss anything. The isolation is for an eventual usb connection. I have an I/F from miniDSP which wants to be mater clock. If you've been following along there is no practical USB isolator which will work with high speed. It's easier to isolate the data and clock lines or so I thought. But it's looking like a greater challenge than I anticipated.
What I'm doing is experimental. Please do provide details. Also explain jitter measurement proceedure. What's needed for equipment. If it's off topic and clouding the thread then just email it to me. You should have my address.
The transformer isolation trick works pretty well. Samuel is right about having the clocks close to the ADC however I have seen evidence of loading, power and ground noise modulating the clock enough to degrade things. I use an RF transformer (usually one designed for cable TV isolation) cap coupled to the source. it only needs to pass the fundamental so the cap can be small (100 pF) and at the other end DC bias the transformer to the logic threshold and couple it to a fast inverter (74AC04 or equivalent). This should get a clean clock with very low additional phase noise.
One trick I worked on to look at phase noise without the full on setup was using a good FM tuner. Some details here: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html Absolute calibration was beyond what I was up for but its a good relative measurement tool. And orders of magnitude cheaper.
SG mentioned close in phase noise being important. What is interesting is that the AES recommendation is to ignore phase noise below 100 Hz. Some people claim that 10 Hz and below matters. I'm really not sure. However I would say that no turntable or tape machine ever made comes close to the wow and flutter or lf phase noise of even a simple crystal oscillator.
One trick I worked on to look at phase noise without the full on setup was using a good FM tuner. Some details here: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html Absolute calibration was beyond what I was up for but its a good relative measurement tool. And orders of magnitude cheaper.
SG mentioned close in phase noise being important. What is interesting is that the AES recommendation is to ignore phase noise below 100 Hz. Some people claim that 10 Hz and below matters. I'm really not sure. However I would say that no turntable or tape machine ever made comes close to the wow and flutter or lf phase noise of even a simple crystal oscillator.
Since I work for a cable company and we no longer carry FM channels I can help myself to the box full of transformer. But these are common mode ballast transformers and don't really provide isolation. They're intended to convert impedance. Can I used the transformers for digital audio isolation or ethernet pulse transformers. I have those as well.
The on board clock and counter can be place near the convertor. The external clock isolated.
These are what I'm talking about for transformers: Mini-Circuits I would really isolate the clock so external influences are minimized (including common ground stuff). Minicircuits has several multiple secondary transformers that could be used to drive multiple clock domains.
Apparently according to Agilent I can use my HP5335 universal counter to measure jitter.
I'll read the article later.
I'll read the article later.
@ davada
Which I/F from miniDSP do you have? If it is the USBStreamer, then take a look at this:
http://www.diyaudio.com/forums/digital-source/244042-master-clock-isolator-minidsp-usbstreamer.html
It makes it possible to put the master clock at the ADC/DAC and feed it back to the PC.
Which I/F from miniDSP do you have? If it is the USBStreamer, then take a look at this:
http://www.diyaudio.com/forums/digital-source/244042-master-clock-isolator-minidsp-usbstreamer.html
It makes it possible to put the master clock at the ADC/DAC and feed it back to the PC.
yes this is the one. I will look later. The test board will accommodate much more than just the miniDSP. I'm not looking for a product.
Thanks for the info.
Yes, but not to the resolution necessary. 2 nS resolution is the limit. Even the HP5370 with 20 pS is not really enough (I know). I would not go back down that road. I have a shelf of this stuff that doesn't see use any more.
Using a reference oscillator and a double balanced mixer etc. would be the right way but gets pretty involved.
Is there anything you would do differently or better since then?
What did the jitter measure?
THx-RNMarsh