Some time ago I upgraded to Windows 7 and I then have to give up on my old EMU-0404 as the beta drivers for Windows 7 just didn't work.... Since then I have been trying to find a reasonable priced replacement with both Analog and Digital I/O's as the EMU-0404, the closest I got are the APX-515 or Dscope-III, and they still have limits.... The otherwise nice project being worked on here on diyaudio don't have the Digital I/O's needed when you're doing ADC and DAC designs....
So now it's time to finish the Audio Analyzer hardware I have been working on for some time, I'm announcing it now as I would like a little feedback on a couple of details.... So the main specs are:
Two channels Analog Input, fully balanced using opa1612 and opa1632 buffers, up to 100V RMS input, with a ltc2387-18 ADC per channel running at 12 Msps, feeding a Spartan-6 FPGA with decimation filters to downsample to 44K to 768K and archieve > 120 dB S/N ratio.
Two channels Analog Output, fully balanced using opa1612, opa1632 and discrete diamond buffer, up to 8V RMS output with 25R Zout single ended, up to 16V RMS with 50R Zout balanced, with a ak4490 per channel to arhieve > 120 dB S/N ratio.
Comprehensive Digital Inputs and Outputs, AES3/SPDIF/Toslink and I2S I/O, controlled by a SRC4393, using the ASRC to convert Digital input rates to USB rates.
USB interface by XMOS, for now up to 384K I/O, but target is to support 768K when shipping.... With licensed Thesycon drivers.
The Spartan-6 FPGA will handle signal routing and fixed sample rate conversion, so t.ex. you can run the ADC at 768K and the DAC at 48K for best performance....
A little STM32 uC to run the show, controlling input and output relays by serial port or switches, with plenty of status LED's. When a host driver is done, it can also be controlled over the USB port using the MIDI serial port....
This will be complete hardware with case and power supply, to be used with any ASIO software analyzer package, if/when software is available to run at 384K/768K speed and MIDI control it will function just like those high end analyzers....
Design target is THD better than -110 dB up to 20 Khz, -100 dB up to 100 Khz, and show harmonics up to 300 Khz.... The ltc2387-18 can do THD to -120 dB, but my focus is more on higher speed and flexibilty than lowest THD....
Time line: I have just completed schematics and the BOM, will start PCB layout soon, although there are many parts it don't seems to complex and I can reuse from other designs.... So prototype in probably 2 months from now, production in probably 5 months from now.
Pricing: Current estimate is USD 1200 / EUR 1100 per unit for complete finished unit incl. case, but without any Analyzer software. There are free software available, but are also planning to select a package and work with the vendor to add direct support for control and for up to 768K speed, maybe even bundle a software package.
This is a high quality commercial product with warranty, but will be DIY friendly and with each unit I plan to include schematics, component layout, BOM and other info needed for doing service yourself, a unit should last many years....
My questions are:
1) Is there anything better than the opa1612 and opa1632 ? And by better I don't mean 1 Khz THD but overall performance up to at least 100 Khz....
2) I plan to put a discrete diamond buffer inside the opa1632 feedback loop, anybody done that before ? Or know how to avoid oscillations ?
3) Right now I'm planning for 25R Zout (single ended) on output drivers, but are also considering 50R as that will lesson the load on the output driver and dividers.... Any opinions ?
4) Some Audio Analyzers have monitor outputs, anything that is actually needed or are useful ?
I will post more info as the design progresses. I might also change priority to other projects, but don't hope to....
So now it's time to finish the Audio Analyzer hardware I have been working on for some time, I'm announcing it now as I would like a little feedback on a couple of details.... So the main specs are:
Two channels Analog Input, fully balanced using opa1612 and opa1632 buffers, up to 100V RMS input, with a ltc2387-18 ADC per channel running at 12 Msps, feeding a Spartan-6 FPGA with decimation filters to downsample to 44K to 768K and archieve > 120 dB S/N ratio.
Two channels Analog Output, fully balanced using opa1612, opa1632 and discrete diamond buffer, up to 8V RMS output with 25R Zout single ended, up to 16V RMS with 50R Zout balanced, with a ak4490 per channel to arhieve > 120 dB S/N ratio.
Comprehensive Digital Inputs and Outputs, AES3/SPDIF/Toslink and I2S I/O, controlled by a SRC4393, using the ASRC to convert Digital input rates to USB rates.
USB interface by XMOS, for now up to 384K I/O, but target is to support 768K when shipping.... With licensed Thesycon drivers.
The Spartan-6 FPGA will handle signal routing and fixed sample rate conversion, so t.ex. you can run the ADC at 768K and the DAC at 48K for best performance....
A little STM32 uC to run the show, controlling input and output relays by serial port or switches, with plenty of status LED's. When a host driver is done, it can also be controlled over the USB port using the MIDI serial port....
This will be complete hardware with case and power supply, to be used with any ASIO software analyzer package, if/when software is available to run at 384K/768K speed and MIDI control it will function just like those high end analyzers....
Design target is THD better than -110 dB up to 20 Khz, -100 dB up to 100 Khz, and show harmonics up to 300 Khz.... The ltc2387-18 can do THD to -120 dB, but my focus is more on higher speed and flexibilty than lowest THD....
Time line: I have just completed schematics and the BOM, will start PCB layout soon, although there are many parts it don't seems to complex and I can reuse from other designs.... So prototype in probably 2 months from now, production in probably 5 months from now.
Pricing: Current estimate is USD 1200 / EUR 1100 per unit for complete finished unit incl. case, but without any Analyzer software. There are free software available, but are also planning to select a package and work with the vendor to add direct support for control and for up to 768K speed, maybe even bundle a software package.
This is a high quality commercial product with warranty, but will be DIY friendly and with each unit I plan to include schematics, component layout, BOM and other info needed for doing service yourself, a unit should last many years....
My questions are:
1) Is there anything better than the opa1612 and opa1632 ? And by better I don't mean 1 Khz THD but overall performance up to at least 100 Khz....
2) I plan to put a discrete diamond buffer inside the opa1632 feedback loop, anybody done that before ? Or know how to avoid oscillations ?
3) Right now I'm planning for 25R Zout (single ended) on output drivers, but are also considering 50R as that will lesson the load on the output driver and dividers.... Any opinions ?
4) Some Audio Analyzers have monitor outputs, anything that is actually needed or are useful ?
I will post more info as the design progresses. I might also change priority to other projects, but don't hope to....
Soren, I also have a problem with EMU-0404 which no longer works under W7 so I decided to keep secondary XP system for analyzing with EM-U soundcard. Its output buffers are upgraded with OPA1612.
1) Yes, as a single op-amp buffer OPA1612 is one among the best and easiest to implement. I am using it in all of my project extensively.
If you are seekeng for ULTIMATE non-linear performance into 100kHz territory you should interest yourself in composite op-amps. In such a composite buffer classic VFA op-amp is driving CFA op-amp. Their open loop gains are summed and both are wrapped with common feedback loop. It will be tricky to design, though.
1) Yes, as a single op-amp buffer OPA1612 is one among the best and easiest to implement. I am using it in all of my project extensively.
If you are seekeng for ULTIMATE non-linear performance into 100kHz territory you should interest yourself in composite op-amps. In such a composite buffer classic VFA op-amp is driving CFA op-amp. Their open loop gains are summed and both are wrapped with common feedback loop. It will be tricky to design, though.
Last edited:
USB will be isolated, so will SPDIF I/O, with fully balanced analog inputs that should be fine. Did consider isolated analog inputs or outputs, but don't seems to be needed or worth it....
This looks interesting, especially the choice of LTC2387-18. The THD figures on page 7 of the datasheet looks promising, it will be interesting to see how easy it is to achieve those figures in practice.
I have used something inside the feedback loop of the opa1632 and could give some advice. But since this is posted in the vendor forums and you clearly intend to make a product for sale, I wonder what we get in return for helping you out. It would be nice if you could promise to either share the schematic and/or do a discounted group-buy for diyaudio members.
- OJG
I have used something inside the feedback loop of the opa1632 and could give some advice. But since this is posted in the vendor forums and you clearly intend to make a product for sale, I wonder what we get in return for helping you out. It would be nice if you could promise to either share the schematic and/or do a discounted group-buy for diyaudio members.
- OJG
You will get a better product faster by helping out.... It posted here because that's where it belong, and it's way to complex to be a diy project. As I already said, each unit will include full schematics.
Instead of doing hard to manage group buys or discount I just do a low price for everybody, like I usually do.... I might offer a discount to diy members for the first batch, but that's more to get an idea of how many to make, I'm running a small company and some of the parts are expensive, like the ADC's....
Nice !
I will try to help by suggesting cheap but useful features, in order of importance according to me.
1) Platform
If it is Windows-only, then product value diminishes greatly, because then it will stop working when M$ decides to make drivers incompatible like they did with the E-MU. Whereas an older version of linux will usually be much easier to get to work.
Since you use UAC2 I guess you agree
2) Scripting. Cost : 0
Let's consider a product like Quantasylum, for example. The first thing I'd want to do with it is write a python script using the arb generator to output a waveform of my design, run it through a DUT, then digitize the results. Simple, right ?
However, it isn't even clear I can do it from reading the docs. It would appear I'd have to use a stinking DLL, on windows, which essentially scripts their app and provides an extremely restricted API whose functionality is only very briefly documented.
I didn't buy this product :S:
If you use UAC2 then the only think you need to do to enable full scripting is to publish the specs of your custom UAC2 controls. Cost : 0
3) Use full sample rate.
Your ADC runs ar 12 Msps. Include a separate FPGA configuration option which does not create a UAC2 device, but a simply a Bulk USB2 device that the user can hit using libusb. FYI, in the past I used a python script with libusb to stream full bandwidth (ie close to 50 MBytes/s) to the PC. No problem, it is quite simple.
Cost : 0 (I believe XMOS can handle the throughput, but it needs to be connected to the FPGA the right way).
Increase in product value : it becomes a 12Msps PC oscilloscope
4) GPIO (and especially outputs)
Some FPGA pins used as GPIO would be great. It needs to be synchronized to the main signal being played. The way to do this is to add a dunny audio channel, and assign bits of it to each digital output pin.
This would be useful to control switches or modes in the DUT for automated testing, or pulsed measurements, for example. This is why it needs to be synced to the sampling.
Cost : a few FPGA pins and a 74AHC04. Possibly a few optos, unless you isolate the USB.
5) Expansion connector for unused FPGA pins
(because you never know)
Have fun !
I will try to help by suggesting cheap but useful features, in order of importance according to me.
1) Platform
If it is Windows-only, then product value diminishes greatly, because then it will stop working when M$ decides to make drivers incompatible like they did with the E-MU. Whereas an older version of linux will usually be much easier to get to work.
Since you use UAC2 I guess you agree
2) Scripting. Cost : 0
Let's consider a product like Quantasylum, for example. The first thing I'd want to do with it is write a python script using the arb generator to output a waveform of my design, run it through a DUT, then digitize the results. Simple, right ?
However, it isn't even clear I can do it from reading the docs. It would appear I'd have to use a stinking DLL, on windows, which essentially scripts their app and provides an extremely restricted API whose functionality is only very briefly documented.
I didn't buy this product :S:
If you use UAC2 then the only think you need to do to enable full scripting is to publish the specs of your custom UAC2 controls. Cost : 0
3) Use full sample rate.
Your ADC runs ar 12 Msps. Include a separate FPGA configuration option which does not create a UAC2 device, but a simply a Bulk USB2 device that the user can hit using libusb. FYI, in the past I used a python script with libusb to stream full bandwidth (ie close to 50 MBytes/s) to the PC. No problem, it is quite simple.
Cost : 0 (I believe XMOS can handle the throughput, but it needs to be connected to the FPGA the right way).
Increase in product value : it becomes a 12Msps PC oscilloscope
4) GPIO (and especially outputs)
Some FPGA pins used as GPIO would be great. It needs to be synchronized to the main signal being played. The way to do this is to add a dunny audio channel, and assign bits of it to each digital output pin.
This would be useful to control switches or modes in the DUT for automated testing, or pulsed measurements, for example. This is why it needs to be synced to the sampling.
Cost : a few FPGA pins and a 74AHC04. Possibly a few optos, unless you isolate the USB.
5) Expansion connector for unused FPGA pins
(because you never know)
Have fun !
About your design and my 2 cents:
1) It will be hard to have 2 different SR at same USB device and made thing too complicated also the SW have to deal with ASIO Multi-Client
2) As you recognized, a different SR is required for ADC & DAC and requires usual 2 HW devices to deal with OS driver's limitations
3) 768kHz on the input is may too low, to measure out of band noise, as on DSDxxx DAC's speeds, or use x-Speed interface (what I support) to have higher SR as 2-Speed as 1.5Mhz
4) For tube user a higher AC protection is required as up 1000V
5) May a modular input would be nice as tube or as for low noise measurements
6) Digital I/O usually stops at 192kHz and 2-Speed will allow 384kHz as seen on the RME ADI-2 Pro
Just my 2 cents
Hp
1) It will be hard to have 2 different SR at same USB device and made thing too complicated
also the SW have to deal with ASIO Multi-Client 2) As you recognized, a different SR is required for ADC & DAC and requires usual 2 HW devices to deal with OS driver's limitations
3) 768kHz on the input is may too low, to measure out of band noise, as on DSDxxx DAC's speeds, or use x-Speed interface (what I support
) to have higher SR as 2-Speed as 1.5Mhz4) For tube user a higher AC protection is required as up 1000V
5) May a modular input would be nice as tube or as for low noise measurements
6) Digital I/O usually stops at 192kHz and 2-Speed will allow 384kHz as seen on the RME ADI-2 Pro
Just my 2 cents
Hp
As I wrote:
The Spartan-6 FPGA will handle signal routing and fixed sample rate conversion, so t.ex. you can run the ADC at 768K and the DAC at 48K for best performance....
Meaning the USB interface will run at the same rate for both directions, typically the ADC rate, the FPGA will convert down as needed for the DAC and Digital out, the ASRC for Digital in.
Even as the ADC chip can do higher I don't plan to go further than 768K, which already is pretty good, to see t.ex. DSD out of band noise, just use you digital scope, most have a FFT function nowadays....
The Spartan-6 FPGA will handle signal routing and fixed sample rate conversion, so t.ex. you can run the ADC at 768K and the DAC at 48K for best performance....
Meaning the USB interface will run at the same rate for both directions, typically the ADC rate, the FPGA will convert down as needed for the DAC and Digital out, the ASRC for Digital in.
Even as the ADC chip can do higher I don't plan to go further than 768K, which already is pretty good, to see t.ex. DSD out of band noise, just use you digital scope, most have a FFT function nowadays....
Hi,
I use an AMBER5500 audio analyser and also an external USB interface with SpectraPlus software and other software. There are plenty of SW out there! BUT, there are not really any really good HW!!
With really good HW, I mean something with extremely low THD and high SNR over a wide range of input voltages and frequencies.
For audio, I see not much reason to have low THD up to 100kHz. But at 1k and 10k I want to have -130dB THD up to 10 harmonics, at around 100mVAC to 100VACrms input voltage.
I would like to have two channels (balanced and unbalanced).
DC coupled if possible (or say 5Hz lower limit).
Monitor out: yes.
Amplitude flatness: +-0,005dB from 20 - 20kHz. (For phono amp RIAA eq accuracy measurements).
Floating outputs possible.
Different output resistances: 50, 600,
Different input resistances: 600, 50, 100k
Calibration possibility.
A ground lug connector.
Best regards,
S.
I use an AMBER5500 audio analyser and also an external USB interface with SpectraPlus software and other software. There are plenty of SW out there! BUT, there are not really any really good HW!!
With really good HW, I mean something with extremely low THD and high SNR over a wide range of input voltages and frequencies.
For audio, I see not much reason to have low THD up to 100kHz. But at 1k and 10k I want to have -130dB THD up to 10 harmonics, at around 100mVAC to 100VACrms input voltage.
I would like to have two channels (balanced and unbalanced).
DC coupled if possible (or say 5Hz lower limit).
Monitor out: yes.
Amplitude flatness: +-0,005dB from 20 - 20kHz. (For phono amp RIAA eq accuracy measurements).
Floating outputs possible.
Different output resistances: 50, 600,
Different input resistances: 600, 50, 100k
Calibration possibility.
A ground lug connector.
Best regards,
S.
I agree with everything staccatis has mentioned above.
Also this thread documents the development of an audio analyser - lots and lots of useful info...
http://www.diyaudio.com/forums/equipment-tools/277808-diy-audio-analyzer-ak5397-ak5394a-ak4490.html
Also this thread documents the development of an audio analyser - lots and lots of useful info...
http://www.diyaudio.com/forums/equipment-tools/277808-diy-audio-analyzer-ak5397-ak5394a-ak4490.html
Don't seems to be much interest, so project has been put on low priority....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.