Hi Jens,
You're right about the clocks. They are far better these days than they once were.
To use an external clock, you would normally feed it a 10 MHz lab reference clock signal. That's the only "normal" high quality clock sources out there. Only people using lab equipment would be affected by that level of stability that is more stable than the clock you are probably using.
-Chris
You're right about the clocks. They are far better these days than they once were.
To use an external clock, you would normally feed it a 10 MHz lab reference clock signal. That's the only "normal" high quality clock sources out there. Only people using lab equipment would be affected by that level of stability that is more stable than the clock you are probably using.
-Chris
Unfortunately a 10 MHz clock would throw all the sample rates off and make a mess of the software in most cases. In some cases the software won't work. What you are really looking for is phase noise that generates artifacts below the minimum usable signal levels so its inconsequential. 5 PPM is really good enough but .5 PPM for a clock is not hard and the NDK oscillators meet both requirements handily. An external master clock would allow syncing several modules together but you would need special software to exploit such a situation. If I get one I will see if a better oscillator will make a difference but I doubt it. The best results from my AK5394A demo board is using an NDK or a Vectron OCXO and no reliable differences between them.
Hi Demian,
I agree with you to a point. But if you're going to use a better external clock, why not use the extremely good and common lab source to start with? Now we are talking about laboratory grade clocks which will be well beyond the needs of this system. Stratum 1 performance anyone? Distribution amplifiers are also common at this frequency (the filtered ones, not a video dist amp).
I'm sure you could convert the 10 MHz input to the desired frequency with a DDS based converter. Those chips are cheap and cheerful with some performing at levels beyond what is needed. There is no need for long term frequency accuracy as short term jitter and drift are the only important issues, as long as the clock is in the ballpark.
One vendor was selling a GPS disciplined CD master clock system for those who must have what they think is the best. But an ovenized clock (OCXO) is very probably the lowest reasonable low jitter source you can find, as you've pointed out. Not expensive either.
I'll back you 100% in your guess that there shouldn't be any meaningful improvement beyond what you are using right now. At this point it's just a numbers game with experimental error obscuring the results. I don't have an HP 5370A or equivalent, so who knows which might be better. The ultimate ABX test for sure.
-Chris
I agree with you to a point. But if you're going to use a better external clock, why not use the extremely good and common lab source to start with? Now we are talking about laboratory grade clocks which will be well beyond the needs of this system. Stratum 1 performance anyone? Distribution amplifiers are also common at this frequency (the filtered ones, not a video dist amp).
I'm sure you could convert the 10 MHz input to the desired frequency with a DDS based converter. Those chips are cheap and cheerful with some performing at levels beyond what is needed. There is no need for long term frequency accuracy as short term jitter and drift are the only important issues, as long as the clock is in the ballpark.
One vendor was selling a GPS disciplined CD master clock system for those who must have what they think is the best. But an ovenized clock (OCXO) is very probably the lowest reasonable low jitter source you can find, as you've pointed out. Not expensive either.
I'll back you 100% in your guess that there shouldn't be any meaningful improvement beyond what you are using right now. At this point it's just a numbers game with experimental error obscuring the results. I don't have an HP 5370A or equivalent, so who knows which might be better. The ultimate ABX test for sure.
-Chris
I do have an HP5370. These clocks are all below its residual. The DDS chips have spurs that can really get in the way. I have two high performance synthesizers but neither are as clean as a good oscillator. The absolute stability is a nonissue so no need for GPS. My 10 MHz GPS reference (Wenzel based) is nice but a locking chain would probably not improve the Vectron (which does have a voltage trim) over periodic manual adjusting in this application.
Hi Demian,
I agree with you, you lucky fella! (having an HP 5370)
Anyway, my point was that a 10 MHz reference is far more likely to be in existence than anything at a sampling clock frequency. To purchase a specific OCXO means you would very likely install that inside the instrument. It would be necessary to run the OCXO all the time to keep it stable. The rest of the instrument could be powered up and down as required. Of course using the "lab clock" at 10 MHz, you just plug it in and go as these systems run all the time - as you know. Not unless you would want to create a special OCXO system just for this instrument. That doesn't seem to make a whole lot of sense to me. That's why I suggested the use of an existing lab frequency. A hobbyist is less likely to need or even know how to set up an external clock that would improve performance (a Time Nut would). Someone in the biz might wish to use what already exists.
I wonder if an LC circuit would reduce the spurs to levels that didn't matter. Thoughts? What else would you do, last stage divide by two and use that pulse train for your clock? Might work.
-Chris
I agree with you, you lucky fella! (having an HP 5370)
Anyway, my point was that a 10 MHz reference is far more likely to be in existence than anything at a sampling clock frequency. To purchase a specific OCXO means you would very likely install that inside the instrument. It would be necessary to run the OCXO all the time to keep it stable. The rest of the instrument could be powered up and down as required. Of course using the "lab clock" at 10 MHz, you just plug it in and go as these systems run all the time - as you know. Not unless you would want to create a special OCXO system just for this instrument. That doesn't seem to make a whole lot of sense to me. That's why I suggested the use of an existing lab frequency. A hobbyist is less likely to need or even know how to set up an external clock that would improve performance (a Time Nut would). Someone in the biz might wish to use what already exists.
I wonder if an LC circuit would reduce the spurs to levels that didn't matter. Thoughts? What else would you do, last stage divide by two and use that pulse train for your clock? Might work.
-Chris
Interesting, what switch is that? Is it operated with a rod from the front panel or how does it work?
The choice of SW depends on what you need to measure and personal preference.
Basically the SW should support the ASIO interface for optimum performance.
Some examples include HpW Works, ARTA (ARTA, STEPS, LIMP), AudioTester, REW, RightMark Audio.
There are others out there. Perhaps other forum members can extend the list and give recommendations.
I have mainly used AudioTester, but you should really look at what features you need/want. The prices on the SW mentioned above go from freeware (versions) to a few hundred USD.
Basically the SW should support the ASIO interface for optimum performance.
Some examples include HpW Works, ARTA (ARTA, STEPS, LIMP), AudioTester, REW, RightMark Audio.
There are others out there. Perhaps other forum members can extend the list and give recommendations.
I have mainly used AudioTester, but you should really look at what features you need/want. The prices on the SW mentioned above go from freeware (versions) to a few hundred USD.
Is it really bound to ASIO only? I thought Jens' design is basically just a USB Audio device, which will work with any software layer that supports USB Audio. Mac OS X has CoreAudio, Windows and Linux have different options for USB Audio layers (I believe). So, you should be fine as long as your test/analysis software is designed to work with the audio hardware accessed via the USB Audio software layer of the operating system.
It is a USB Audio device. I have not tried it with Mac OS or Linux, but that should work also. With Windows a driver is needed (and will be included).
I use ASIO to bypass the audio processing in Windows (resampling etc.), which would otherwise limit the performance considerably.
There may be other possibilities as well. I am not an expert on the different audio interface standards.
I use ASIO to bypass the audio processing in Windows (resampling etc.), which would otherwise limit the performance considerably.
There may be other possibilities as well. I am not an expert on the different audio interface standards.
Windows later than XP is a problem for audio measurement software. You can get it to work well BUT you need to make sure the device, the Win audio system and the application are all running at the same sample rate and depth if possible. Otherwise Windows inserts a lousy SRC in the chain. ASIO makes it all cleaner however none of the software i have tried allows two separate ASIO links or ASIO + non ASIO. Usually its is a non issue but I need to often do Bluetooth in and analog from the DUT. For this I need to use the Windows audio interfaces. The XMOS/Thesycon drivers are quite good and not as expensive as they once were. Jens does have a chore to integrate the features of his interface with the driver and expose as appropriate an API that applications like HPWorks and ARTA can use. This is not real easy for the XMOS stuff.
Yes, some cases can be a problem to handle, like the ones you describe with either separate ASIO links or ASIO + non ASIO. It does e.g. make it difficult in some cases to make frequency response measurements. But for purely analog systems this is not an issue.
Regarding the API that has already been implemented on the XMOS. At the moment a simple user interface program can be used to read out or control the attenuator settings etc. The purpose of this small program is mainly to test the API.
I hope that some of the providers of applications like HpW Works and ARTA etc. will integrate drivers for the API. This will allow proper scaling of measurements results as well as autoranging etc.
