RightMark Soundcard Tests for SDR.
January 2017 Look here for some soundcard tests by Leif SM5BSZ with reference for use with SDR Among other parameters Leif tests noise and reciprocal mixing.
Some more tests of soundcards for SDR, noise dynamic range.
SOUND DEVICE TESTS USING RIGHTMARK WITH CONSIDERATION FOR USE WITH SDR.
I have some HTML files of tests I made.
The tests may not be representative of the performance of any card. They are just an illustration of results I achieved.
I used RMAA 6.2.3 found here http://audio.rightmark.org/index_new.shtml to help evaluate the performance of some soundcards for use with SDR. I believe I used optimum settings but I am not certain if different results could be obtained using other methods or different settings. Are my cards typical?
One setting I missed first time was "Analise Noise and distortion only in 20Hz to 20KHz range" This seems inappropriate for SDR. When I repeated the tests with the tick removed and at their maximum sample rate there was little difference on the Delta 44, Creative, Maya44 and Realtek. But the EMU showed poor results. The Asus Xonar D2X seems incapable of a level output above 30KHz. Unlike many other cards
I have often wondered about published specifications, now I wonder if they are really relevant for use with high sample rate SDR? High Fidelity is a numbers game biased towards the audio range, not the range above audibility..
SYSTEMS TESTED
EMU Tracker Pre this device is very similar to the EMU 0202 Sample rate 192KHz. USB2.
Xonar D2X Sample rate 192KHz PCIe I found this to be incompatible with other cards, even the on-board had to be disabled. All the others tested worked together with no conflicts.
M-Audio Delta 44 PCI Two stereo pairs input and output. 96KHz.
Creative SB24 Live! PCI Reasonable performance from this old SB0410 PCI card.
ESI Maya44 PCIe This is a "poor man's Delta 44" It has 4 ins and outs but not quite as good performance.
Realtek HD on-board audio
Creative SB24 Live! USB RX only. I have never been able to use these in duplex.
. Can anyone check what they find? Here is another test on the EMU (PDF file) This is based on only up to 20KHz and makes it look really good.
EMU 0404 PCI:- Several users have reported these are not suitable for SDR. Large sample offsets and unstable operation are quoted but others say it's good...
Both the Asus and EMU operate up to 192KHz sample rate. These tests show noise increasing at higher frequencies. A HDSDR screenshot on the ASUS test page confirms this. I not think these are properly usable for SDR higher than 96KHz sample rate. I have seen instructions about setting up EMU systems that recommend high settings of the input controls. This will mask the noise at each edge of the SDRs display but also reduce the dynamic range. With all cards to get the best performance the SDR noise floor should barely move when the audio plug is removed. Unless my tests are flawed the Asus is not suitable for transmitting at 192KHz with a fixed LO. The EMU output shows highish IMD but I'm not sure how that will affect the transmit signal. I suspect it is low enough not to be noticed.
Both these devices show low noise levels up to 50KHz audio. But if used without adjusting the gains of the SDR receiver no advantage may be obtained.
How good does the sound system need to be? I have listed the systems above with what I think is the best performing at the top. (But with reservations about the EMU and Xonar).
The On-board Realtek is better than some cards used satisfactorily for SDR.
Many seem happy with what they use, unless you must have the best try what you have first, possibly it will serve your purpose. Most SDRs seem setup for reasonable performance with average soundcards, I think they will have too much gain when used with low-noise cards. Utilising the whole range of the device will improve the overall dynamic range. I think that many users of expensive cards do not get the best from them.
June 2012. A recent discussion on the Mobokits group seems to confirm that the Softrock SDR is optimised for 16 bit soundcards. To gain any real advantage from low noise 24 bit cards the gain needs to be reduced, some have reduced it by 20dB. This possibly applies to other soundcard SDRs.
https://groups.yahoo.com/neo/groups/softrock40/conversations/topics/78727 A summary of how Softrock performance relates to the soundcard.
https://groups.yahoo.com/neo/groups/softrock40/conversations/topics/78727
https://groups.io/g/softrock40/message/77763
https://groups.io/g/softrock40/topic/sound_system_tests
/5642332?p=,,,20,0,0,0::Relevance,,Sound+System+Tests.+Especially+Concerning+EMU+and+Xonar+D2X+at+192KHz,20,2,0,5642332
https://groups.yahoo.com/neo/groups/softrock40/conversations/topics/57055
https://groups.io/g/softrock40/message/56418
These are all loop-through tests, simultaneously testing the combined input and output of the devices. The Xonar and, to a lesser extent, the Maya show a restricted range on their outputs. Consider whether this will restrict usefulness.
Transformers:-
Sometimes used to avoid the effects of ground loops. I have included various pairs of transformers my tests. In all cases IMD figures were poor. Maybe this is due to high levels used during tests but I do not think it they are needed with careful attention to grounding.
Here is an example of a test using transformers.
I started a thread on the Softrock40 group http://groups.yahoo.com/group/softrock40/message/57055 some comments, links, I'm beginning to think I was hoping for too much. Some seem satisfied, but are they really critical? Maybe a lot of SDR fans get dazzled by it all and forget performance? Of course there are other aspects affecting performance, how far do you go? http://groups.yahoo.com/group/softrock40/message/61479
Sample rate. I quote values in the often used term KHz. More correctly Ks/s (Samples per Second) or ksps.
November 2016 Behringer UMC404HD 4 channel sound specified 192ksps but noise rises over 96KHz "It seems it would be very satisfactory at 96 KHz sampling but that is probably not why we want it. Warren Allgyer - WA8TOD" Shirley, KE1L did more comprehensive tests and came to the same conclusion.
I have zip file that contains the HTM files of my tests that RightMark produced.