Hi All,
I hope this is in the right section and forgive me if this is a 'dumb' question but I have yet to get a satisfactory answer.
What is the difference between Over-Sampling and Up-Sampling?
I run a TDA1541A DAC and I can feed it via USB from my PC running JRiver. If I get JRiver to output everything at 352.8kHz, this is 'Up-Sampling' (?).
In a Philips CD Player using the Digi Filter SAA7220P set to 8x Over-Sampling, the TDA1541A will be seeing DATA at a Sample Rate of 352.8kHz. This, though is called 'Over-Sampling' by Philips.
So what's the difference?
What is the difference between the DATA the TDA1541A sees from JRiver compared to that which it sees from a SAA7220P?
Thanks in advance,
P.
I hope this is in the right section and forgive me if this is a 'dumb' question but I have yet to get a satisfactory answer.
What is the difference between Over-Sampling and Up-Sampling?
I run a TDA1541A DAC and I can feed it via USB from my PC running JRiver. If I get JRiver to output everything at 352.8kHz, this is 'Up-Sampling' (?).
In a Philips CD Player using the Digi Filter SAA7220P set to 8x Over-Sampling, the TDA1541A will be seeing DATA at a Sample Rate of 352.8kHz. This, though is called 'Over-Sampling' by Philips.
So what's the difference?
What is the difference between the DATA the TDA1541A sees from JRiver compared to that which it sees from a SAA7220P?
Thanks in advance,
P.
I also think there is no apparent difference between Over-sampling and Up-sampling. Hardware chips like digital filter and DAC tend to use Over-sampling. IMHO, the word Over-sampling implies there is no creation of data even if the sampling frequency is changed. Up-sampling is usually used in software conversion like JRiver. I don't know it has some data modification or not. But there is a more chance to be modified if you use software Up-sampling or software process. It's impossible to know inside Windows OS which must Up-sample or down-sample many audio sources to compatible with one speaker. I suppose it needs some modification for proper operation not for excellent SQ. But hardware chips are probably so simple that they don't need to modify data.
Wrt terminology; if a sample rate in the digital domain it is called resampling.
If the new sample rate is higher than the original one, it is called upsampling; if the new sample rate is lower than the original oen, it is called downsampling.
In the case of oversampling it is upsampling because the new sample rate is higher than the original one.
As it was Philips who used the trick of upsampling in their first CD players (driven by the fact that Philips at that time didn´t have a 16-Bit DA converter and had therefore to use their 14-bit DA converter) to achieve similar good measurement numbers compared to the japanese CD players and used the term oversampling.
Since then the term "oversampling" will be most often used (at least in the audio field) if the ratio between original and new sample rate is an even integer number while the term "upsampling" is used in the case of other ratios.
If the new sample rate is higher than the original one, it is called upsampling; if the new sample rate is lower than the original oen, it is called downsampling.
In the case of oversampling it is upsampling because the new sample rate is higher than the original one.
As it was Philips who used the trick of upsampling in their first CD players (driven by the fact that Philips at that time didn´t have a 16-Bit DA converter and had therefore to use their 14-bit DA converter) to achieve similar good measurement numbers compared to the japanese CD players and used the term oversampling.
Since then the term "oversampling" will be most often used (at least in the audio field) if the ratio between original and new sample rate is an even integer number while the term "upsampling" is used in the case of other ratios.
You have effectively answered your own question. The tendancy is to oversample by and upsample to. Everything sent to the SAA7220 has the sample rate increased by 4. Jriver, OTOH, converts everything to 352.8K
This seems mostly a matter of semantics, and the definition appears to vary from industry to industry. Within the audio industry, the term 'oversampling' typically refers to integer synchronous sample rate conversion associated with digital filtering. In synchronous rate conversion, the output rate is an an constant and (usually, but not necessarily) an whole number multiple of the input sample rate. The ratio of the input sample rate to the output sample rate are almost always some base-2 multiplication factor. So, we typically see digital audio oversampling factors of x2, x4, x8, x16, and so forth
'Upsampling', on the other hand, typically refers to asynchronous and/or (usually, but not necessarily) non-integer rate conversion ratios. These factors can largely be arbitrary, such as x1.088, x2.17, x4.35 and so forth. Such asynchronous non-integer conversion is needed for applications such as converting an 44.1kHz music file to 48kHz, which is a factor of x1.088, or to 96kHz, which is a factor of x2.17. The circuit function associated with upsampling is known as 'Asynchronous Sample Rate Conversion' (ASRC), which is itself based aound a digital filter. Hence much of the confusion, I suspect.
'Upsampling', on the other hand, typically refers to asynchronous and/or (usually, but not necessarily) non-integer rate conversion ratios. These factors can largely be arbitrary, such as x1.088, x2.17, x4.35 and so forth. Such asynchronous non-integer conversion is needed for applications such as converting an 44.1kHz music file to 48kHz, which is a factor of x1.088, or to 96kHz, which is a factor of x2.17. The circuit function associated with upsampling is known as 'Asynchronous Sample Rate Conversion' (ASRC), which is itself based aound a digital filter. Hence much of the confusion, I suspect.
Last edited:
I think my earlier post (#2) was wrong.
In the contexts that I've heard or seen it used, upsampling usually means increasing the sample rate of a discrete-time signal. Normally you use some sort of filter to get rid of images of the signal, in which case it becomes interpolation.
The term oversampling is typically used for data converters, and it does not necessarily involve adding more samples to a discrete-time signal. An oversampling DAC is a DAC that gets its input signal from an interpolation filter or interpolation filter chain, but an oversampling ADC is an ADC followed by a decimation filter or decimation chain. The ADC runs at a higher rate than the sample rate of the system, and the digital filter reduces rather than increases the number of samples.
Anyway, you have five contradictory answers now...
In the contexts that I've heard or seen it used, upsampling usually means increasing the sample rate of a discrete-time signal. Normally you use some sort of filter to get rid of images of the signal, in which case it becomes interpolation.
The term oversampling is typically used for data converters, and it does not necessarily involve adding more samples to a discrete-time signal. An oversampling DAC is a DAC that gets its input signal from an interpolation filter or interpolation filter chain, but an oversampling ADC is an ADC followed by a decimation filter or decimation chain. The ADC runs at a higher rate than the sample rate of the system, and the digital filter reduces rather than increases the number of samples.
Anyway, you have five contradictory answers now...
Yup, I sort of expected it'd be so !!
Thanks to all anyway, but am I any nearer to an answer ?!?!? ;-)
P.
Yeah, sorry; i somehow missed the real question at the end.....
Afterall, the propper semantics is much more important......
Short answer, you´ll never know until you (or someone else) compare the data.
Different realisations of digital filters are possible, different dither schemes (if in use) are usable as the wordlength has to be reduced before transferred to the DAC chip, so most likely the data would be slightly different even if the sample rate is identical.
But afair the SAA7220P was still just a 4 times oversampling filter, so unable to reach 8 times oversampling.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.