IIRC, there has been an appnote by TI for paralleling ADCs 14 bit / 100 MHz or so.
I have only a weak cell phone internet connection where I'm now,
no fun searching it.
Gerhard
(just paralleling 10 pairs of ADA4898-2 for low voltage noise
)
Last edited:
I am afraid that stacking two converters won't quite work as intended here.
But adding the outputs of two (or more) ADC's does work. I have done it with a CS5361. I achieved an improvement of 2.8dB. In this case the op-amp buffer was shared, so the noise from the buffer was not reduced. Using separate buffers might have improved it by the last 0.2dB.
I am afraid that stacking two converters won't quite work as intended here.
But adding the outputs of two (or more) ADC's does work. I have done it with a CS5361. I achieved an improvement of 2.8dB. In this case the op-amp buffer was shared, so the noise from the buffer was not reduced. Using separate buffers might have improved it by the last 0.2dB.
How did you do it, exactly? We have two channels to work with... could one be used to make mono ADC?
AKM indicated their ADC IC's with 2-4 per package can be used for mono with excellent results. But I havent got to the detail stage of implimentation yet and am wondering how you did it with the CS5361.
[gathering info stage]
Thx-RNMarsh
Last edited:
Even if they do start synchronously there will still be a bus conflict if you just connect the outputs together. They have to be added in a DSP or similar (e.g. FPGA). And then shifted one position to the right to avoid clipping, since the addition is equal to a gain of 2. This is the "dsp magic" needed.
For DAC's the stacking may work, depending on the output type. But not for ADC's.
For DAC's the stacking may work, depending on the output type. But not for ADC's.
For the intrepid- I have been working with a team to pull this together- https://groups.google.com/forum/?fromgroups=#!topic/audio-widget/uZx-SLY-0R4 which uses a TI miniDSP chip to make an async USB interface. The chip also has several I2S interfaces that can be used for input or output. It would be possible to use the on board DSP to add multiple I2S inputs together. The DSP is 16 bit but could be used in a double precision way to handle a simple task like this. It would take someone with the necessary software skills to implement. Alex has done some very impressive work on the DAC side.
It looked to me like paralleling-- with the proper (DSP magic) glue to make it work-- is the next step for a test instrument (at reasonable cost) for THD measurements to get lower levels than a single ADC can do.
I hope this gets going as we are done with single ADC circuits and have reached/realized its potential for now.
Technologists, take a look at what caught my eye: www.AKM.com/AppsNotes/AN120516.pdf
Thx-RNMarsh
I hope this gets going as we are done with single ADC circuits and have reached/realized its potential for now.
Technologists, take a look at what caught my eye: www.AKM.com/AppsNotes/AN120516.pdf
Thx-RNMarsh
Last edited:
The way I did it with the CS5361 was this:
I connected the two inputs of the ADC in parallel, using an existing differential op-amp circuit to drive both inputs. I connected the ADC to an SPDIF transmitter feeding an SPDIF input of a sound card. Then I used AudioTester V2.2 to measure the noise. V2.2 has a function to add the two channels. Using the adder function the noise increased by around 3dB. But the signal increases 6dB.
For some reason the AudioTester V3 does not have this adder function.
Using ADC's in parallel can reduce the noise level. But it will probably not reduce the THD if we assume that the THD is the same for all channels. If they are different they may cancel to some degree though.
I connected the two inputs of the ADC in parallel, using an existing differential op-amp circuit to drive both inputs. I connected the ADC to an SPDIF transmitter feeding an SPDIF input of a sound card. Then I used AudioTester V2.2 to measure the noise. V2.2 has a function to add the two channels. Using the adder function the noise increased by around 3dB. But the signal increases 6dB.
For some reason the AudioTester V3 does not have this adder function.
Using ADC's in parallel can reduce the noise level. But it will probably not reduce the THD if we assume that the THD is the same for all channels. If they are different they may cancel to some degree though.
If there is going to be an a DSP involved why not just use one convertor. Store the period of a single cycle and sum it with the next cycle. This can be done recursively for any number of cycles. The output can be any convenient audio format.
Or just do what the 725 does and output that to a convenient audio format rather than reconstructing the signal and analog processing.
LT has a app note for this.
Look at figure 4. This can be modified to do the job.
http://cds.linear.com/docs/en/lt-journal/Sample-and-Hold_1100_Mag.pdf
Or just do what the 725 does and output that to a convenient audio format rather than reconstructing the signal and analog processing.
LT has a app note for this.
Look at figure 4. This can be modified to do the job.
http://cds.linear.com/docs/en/lt-journal/Sample-and-Hold_1100_Mag.pdf
That could work as well. This sort of thing is the natural next step for a hyper-low distortion test system.
The Panasonic VP-7722A has just arrived to see what it's potential is and how well it works.and what is the design used. has great specs, uses DSP etc.
[>-140dB 2H rejection spec. THD <.0002% spec. Designed for to be fully automated for production testing of 10 parameters or more]
Thanks,
Richard
The Panasonic VP-7722A has just arrived to see what it's potential is and how well it works.and what is the design used. has great specs, uses DSP etc.
[>-140dB 2H rejection spec. THD <.0002% spec. Designed for to be fully automated for production testing of 10 parameters or more]
Thanks,
Richard
Last edited:
The VP-7722A is similar to the ShibaSoku AD725D in THD resolution (using its own generator output to it's 7722 analyzer input displays a THD level of .00005%. So thats a good beginning. Have to look at the monitor port residuals with FFT. Measures individual harmonics like the 725 does. It also does IM and a host of measurements.... might be the one for you guys to find before prices go thru the roof on it. I paid $1k for it on eBay. I had found the 725D on eBay a microsec after it was put there asking only $450 and snatched it up. Later, I saw a refurbished 725d on another site going for $12K ! Good deals do exist from time to time.]
-RNM
-RNM
Last edited:
This is what the melt noise is doing on the QA400. Unfortunately the trigger/clock must be precisely synchronized with the signal or both the signal and the noise "melt". If the QA400 had a continuous mode it would be possible to lock an external oscillator.
If you use it in loopback the signal should not change. Any synchronous component will be reinforced. Any noise/randomness will get reduced. It does seem to work for me.
Depends- The windows sound engine supports a standard list of sample rates and resamples everything to the default rate of your system. Its not a particularly good resampler and should not be used in any measurement process. If you use something like ASIO you can go around the sound engine somewhat, if your soundcard driver and application supports ASIO.