clem_o said:
clem_o, wouldn't the capacitive input of the ad825, if it did cause the trouble, happen regarless of which level db of the 1000hz digital was played, or only when the d/a chips analog output amp was being pushed to it's highest level.
I think it will be like a power amp when you put a 1uf across it's output to check stablity, it will ring the same whether at 2vpp or 20vpp no more no less.
I tend to think it's not the capacitive input of the AD825, it's a PITB to take out and exchange for the NE5532 only to see that it's going to be the same.
Cheers George
Hi georgehifi,
If it's a stability problem, then yes, but I think this is different. The DAC's output is a current which changes proportionately with the binary input amplitude. Since that is the variable that is supposed to be exactly held constant wrt the binary value input, capacitive loading of the output of the DAC should only result in a 'drooped' amplitude - exactly what is expected with a filter... duh... so yeah, it's probably not the capacitance of the AD825 inputs...
OTOH, what happens if virtual ground presented by the op-amp isn't 'true' - i.e. a non-zero voltage is created there... that would create the modulation effects... but that should only happen if the op-amp is too slow or it's outputs can't swing the required levels...
Arrrgh...
Clem
If it's a stability problem, then yes, but I think this is different. The DAC's output is a current which changes proportionately with the binary input amplitude. Since that is the variable that is supposed to be exactly held constant wrt the binary value input, capacitive loading of the output of the DAC should only result in a 'drooped' amplitude - exactly what is expected with a filter... duh... so yeah, it's probably not the capacitance of the AD825 inputs...
OTOH, what happens if virtual ground presented by the op-amp isn't 'true' - i.e. a non-zero voltage is created there... that would create the modulation effects... but that should only happen if the op-amp is too slow or it's outputs can't swing the required levels...
Arrrgh...
Clem
It might be an idea to establish whether or not you have a real issue or are simply running up against the limits of numerical representation.
Convert the offending track to a .wav file and look at its binary representation. If you see a series of maximum valued samples, the maximum positive value being 0111 1111 1111 1111 and maximum negative value, 1000 0000 0000 0000, then the top of the sine wave is clipping and there is nowt you can do about that.
Convert the offending track to a .wav file and look at its binary representation. If you see a series of maximum valued samples, the maximum positive value being 0111 1111 1111 1111 and maximum negative value, 1000 0000 0000 0000, then the top of the sine wave is clipping and there is nowt you can do about that.
rfbrw said:
rfbrw, I have no idea how to do that.
But trying to do it, I just played the tracks on my computer and they all sound the same after volume was adjusted. There was no overtone present at 0db digital, maybe my computers speakers are no able to give it to me is another thing, but they all souned pure. Bugger!!!!!!!!!!!!!!!!!
Cheers George
Download a program like Goldwave, install it. It'll allow you to edit wav files. I think it's capable of importing audio files from CD as well.
The editor lets you see the waveform, much like an oscilloscope, - closer to a DSO anyway, and you can move cursors around using your mouse...
Cheers
The editor lets you see the waveform, much like an oscilloscope, - closer to a DSO anyway, and you can move cursors around using your mouse...
Cheers
Since you have a pretty fancy Tek scope, and the distortion you want to see is probably in the high order harmonics, I suggest the following :
- use a sinewave at 0bDfs (where you can hear but not see distortion).
- have the output of your DAC pass through a second-order linkwith-riley high-pass filter (easily built with a dual opamp and a few passive components) whose corner frequency will be about 20x the test frequency (ie. 20 kHz if you use a 1 kHz sine to test).
This will attenuate the fundamental and let you see the harmonics. You can listen to the output, too.
Now, use the scope and zoom in on the filtered waveform to hunt the harmonics.
- use a sinewave at 0bDfs (where you can hear but not see distortion).
- have the output of your DAC pass through a second-order linkwith-riley high-pass filter (easily built with a dual opamp and a few passive components) whose corner frequency will be about 20x the test frequency (ie. 20 kHz if you use a 1 kHz sine to test).
This will attenuate the fundamental and let you see the harmonics. You can listen to the output, too.
Now, use the scope and zoom in on the filtered waveform to hunt the harmonics.
As far as that Piere Varnay disc goes. Any sine wave greater than 0 dB FS on the disc will/must be distorted by any DAC. If a DAC makes it clean, the DAC is broken!! Sometimes digital domain amplitude values are expressed as "FFS" Fraction of Full-Scale. A 0 dB FS, or 1.00 FFS sine wave uses all of the quantization steps. The positive peak and the negative peak of the sine are A Full-Scale square wave measures 1.41 FFS, but still uses all of the quantization steps. It has to do with the crest factor of the signal. The crest factor for a pure sine is 3.01 dB, the crest factor for a pure square is 0 dB. If a sine wave is 0 dB FS (1.0 FFS), that is the maximum level that a sine wave can be produced cleanly. If you scale your binary bits through multiplication (say X2), that will increase the sine wave by 6 dB. There are no more bits to quantize the new level. It will flat top or "clip". Think of the bit depth as power supply rails. Once you hit the limit, it's out of gas.
Now there may be a problem if you hear distortion playing the 0 dB FS tone. I think you're on the right track for skulling that one out. .....but you will never get the the +3, or +6 dB FS sine waves to ever play cleanly.
Now there may be a problem if you hear distortion playing the 0 dB FS tone. I think you're on the right track for skulling that one out. .....but you will never get the the +3, or +6 dB FS sine waves to ever play cleanly.
Hi mrshow4u,
I use the Pierre Verany disc all the time to check muting transistors on CD players. As you say, it works fine. I have not tested any new CD players, but I do have some Cyrus units around I can test. They are current.
George,
Any digital signal over the maximum level will give you a clean clipped signal. It is possible you are chasing a bug in a DSP chip. Nice if that's the case. Too bad you can't see the samples before the reconstruction filter gets the signal. That would be at the input of the op amp. Too small to see I think. I was thinking you may see a sample or two wrapping around due to an over range condition.
-Chris
I use the Pierre Verany disc all the time to check muting transistors on CD players. As you say, it works fine. I have not tested any new CD players, but I do have some Cyrus units around I can test. They are current.
George,
Any digital signal over the maximum level will give you a clean clipped signal. It is possible you are chasing a bug in a DSP chip. Nice if that's the case. Too bad you can't see the samples before the reconstruction filter gets the signal. That would be at the input of the op amp. Too small to see I think. I was thinking you may see a sample or two wrapping around due to an over range condition.
-Chris
Hi George,
Sorry. I am lazy and therefore abbreviate my posts where possible.
Digital Signal Processing (DSP) can occur in the transport or the external DAC. An filter does this as does the oversampling bits. Error correction may as it interpolates.
Fixing a bug in a DSP chip is not easy. Either the manufacturer of the chip releases a pin for pin replacement chip (or someone else), or an entirely new board is required. A smaller "daughter board" may do the trick. A "daughter board" would contain a new chip and support circuit to adapt it for use in another chip's footprint.
But first, you need to find out where the problem exists. from there you look at options.
-Chris
Sorry. I am lazy and therefore abbreviate my posts where possible.
Digital Signal Processing (DSP) can occur in the transport or the external DAC. An filter does this as does the oversampling bits. Error correction may as it interpolates.
Fixing a bug in a DSP chip is not easy. Either the manufacturer of the chip releases a pin for pin replacement chip (or someone else), or an entirely new board is required. A smaller "daughter board" may do the trick. A "daughter board" would contain a new chip and support circuit to adapt it for use in another chip's footprint.
But first, you need to find out where the problem exists. from there you look at options.
-Chris
Hi Chris and George. George, can you offset your scope channel and increase the vertical sensitivity to just zoom in on either the positive or negative peak? ...while listening, just to make sure the setup loading etc. doesn't change.
I have to clrify a couple things I said. I said something or other about +3 and +6 "always will distort" and "never play cleanly". Just to be clear, I mean the waveform will be faithfully reproduced (distortion free representation) of a distorted signal. Does that make sense? garbage in (+3 and +6 dB sines), garbage out.
Have you tried putting the NE5532 back in yet, just to see? Maybe a good place for a machined pin socket
Good Luck, I hope for the good outcome.
...Is your prime reason for using the AD amp to remove coupling caps due to the better DC drift performance?
I have to clrify a couple things I said. I said something or other about +3 and +6 "always will distort" and "never play cleanly". Just to be clear, I mean the waveform will be faithfully reproduced (distortion free representation) of a distorted signal. Does that make sense? garbage in (+3 and +6 dB sines), garbage out.
Have you tried putting the NE5532 back in yet, just to see? Maybe a good place for a machined pin socket
Good Luck, I hope for the good outcome.
...Is your prime reason for using the AD amp to remove coupling caps due to the better DC drift performance?
Hi mrshow4u,
I think to actually see what is going one, you would need to subtract the fundimental frequency. A THD meter or FFT would show the issues nicely. The newer 'scope input circuits do not take kindly to being over ranged.
One of the mixed signal DSO's from Agilent would be perfect for this. You could see the digital sample and the analog output word for word. I think you can even set the trigger for a specific digital word.
-Chris
That's it exactly, however I wouldn't call the reproduced signals at +3 and +6 garbage. They are exact, clean representations of the expected waveform. They just don't sound nice.
I think to actually see what is going one, you would need to subtract the fundimental frequency. A THD meter or FFT would show the issues nicely. The newer 'scope input circuits do not take kindly to being over ranged.
One of the mixed signal DSO's from Agilent would be perfect for this. You could see the digital sample and the analog output word for word. I think you can even set the trigger for a specific digital word.
-Chris
mrshow4u, yes I can zoom in and have a closer look at the +&- peaks.
I also understand what you said about the +3 and +6 that they should be clipped.
And yes the reason for the AD825 is one I found they sounded better than the NE5532's and also they only gave out 2-4mv offset which was easy to trim out to zero where the NE5532's had 20-50mv dc offset.
And when I tried DC coupling with the AD825 the sound was fast tight and grain free then I thought I'll put the coupling cap back in a Blackgate NX Bi-Polar 6.3v 220uf and it sound clearly softened in comparsion, so dc coupling for me is a must have.
I have no NE5532's to try at the moment, but I will look at the scope again tomorrow, it's 1am now here in Aus and I'm starting to see two keyboards.
Cheers and good night George
I also understand what you said about the +3 and +6 that they should be clipped.
And yes the reason for the AD825 is one I found they sounded better than the NE5532's and also they only gave out 2-4mv offset which was easy to trim out to zero where the NE5532's had 20-50mv dc offset.
And when I tried DC coupling with the AD825 the sound was fast tight and grain free then I thought I'll put the coupling cap back in a Blackgate NX Bi-Polar 6.3v 220uf and it sound clearly softened in comparsion, so dc coupling for me is a must have.
I have no NE5532's to try at the moment, but I will look at the scope again tomorrow, it's 1am now here in Aus and I'm starting to see two keyboards.
Cheers and good night George
mrshow4u, here are the two shots + and - tops of the 1001hz 0dfs sine waves, looks good to me, there must be something else.
Have any of you a test disc you can listen to with -10dfs
-5dfs and 0dfs. You will not reconize the slight overtone unless you listen to the -10dfs first and it may come across as ok unless you pre condition your self with the undistorted one first, it's that slight, and of course a pair of esl's would help as they have the least distortion at that fequency.
Cheers George
Have any of you a test disc you can listen to with -10dfs
-5dfs and 0dfs. You will not reconize the slight overtone unless you listen to the -10dfs first and it may come across as ok unless you pre condition your self with the undistorted one first, it's that slight, and of course a pair of esl's would help as they have the least distortion at that fequency.
Cheers George
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.