First the question, then the context:
In the case of active amplification (filter before amps) from digital source (file, CD), are there some advantages of analog filtering compared to DSP ?
Can the analog filtering be superior to DSP equivalent (objective and subjective perspectives) ?
The context of the question. I have my LXminis with DSP and Full Digital Amps (all digital). My amplifiers lack power and I don't find more powerfull FDA. I need to choose my new way.
1) One alternative is to try to work in same or similar direction: DSP=>multi DACs =>analog input amplifiers.
2) The other is to go to the DAC=>analog filter recently designed by Nelson Pass => analog input amplifiers
I wonder what are the pro/cons of each of those options. My beliefs is that DSP is mathematically perfect. But is it true ? An from sonic performance ?
Best regards,
JMF
In the case of active amplification (filter before amps) from digital source (file, CD), are there some advantages of analog filtering compared to DSP ?
Can the analog filtering be superior to DSP equivalent (objective and subjective perspectives) ?
The context of the question. I have my LXminis with DSP and Full Digital Amps (all digital). My amplifiers lack power and I don't find more powerfull FDA. I need to choose my new way.
1) One alternative is to try to work in same or similar direction: DSP=>multi DACs =>analog input amplifiers.
2) The other is to go to the DAC=>analog filter recently designed by Nelson Pass => analog input amplifiers
I wonder what are the pro/cons of each of those options. My beliefs is that DSP is mathematically perfect. But is it true ? An from sonic performance ?
Best regards,
JMF
You throw the term "DSP" around a lot. I'm not sure if you or others understand that there is more than one way to create a filter digitally: DSP can be done via hardware or in software. For audio work, DSP filtering is typically done either using an IIR filter (recursive) or FIR (non-recursive) algorithm. Both IIR and FIR filters can be implemeted with hardware or in software. Do you know which of these you are interested in?
To make the most fair comparison, you can compare IIR digital filtering against analog active filtering. The types of filters and the responses of these filters are very similar, whereas FIR filtering is another animal alltogether. So let's compare IIR DSP filtering and analog active filtering.
IIR DSP:
In theory, these are "mathematically perfect". But this is like saying that an analog circuit constructed out of idea amplifiers and idea passive components is perfect! The real world is not so perfect. For any DSP you can have errors of various types, and the circuitry itself (e.g. the DAC and possibly ADC) will impart some temporal and frequency error and inject some level of noise. I assume that this is no surprise to anyone. What I think you are getting at is the inherent ACCURACY of DSP filters, and that is one advantage of IIR DSP over analog active filters. By accuracy I mean the accuracy of the pole(s) and zero(s) of the filter(s), which determine the frequency domain response. This becomes more important as the order of the filter increases, since higher order filters are often constructed by placing multiple lower order filters in series. The other advantage is that you can implement many, many filter stages in DSP without much of a noise penalty and without much signal degradation, which mostly arises when coming from and going to the analog domain (e.g. ADC and DAC). Finally, and possibly the biggest motivation for DSP is the easy by which filter parameters and the number of filter stages can be changed. It's essentially instantaneous and on demand. Sure an analog filter can be build with some adjustable elements, but overall the adjustability really pales in comparison to DSP. On the downside, DSP processing is just what is happening on the "inside", and the ADC and DAC are what gets you in and out of the DSP filtering box. The quality of ADC and DAC are very important. In the DAC, the maximum output voltage is something that is often overlooked and compared to analog filters may be limited to 1Vrms or 2Vrms only, which limits dynamic headroom somewhat. Also, if your DAC can only output 1Vrms and your super duper power amp needs 2Vrms to reach full power, you have a problem.
Analog Active Filters:
The primary advantage of analog active filters, in my mind, is the lack of temporal effects to the signal. With no ADC or DAC to muck up the waveform you get no ringing, etc. with a properly designed circuit. If the circuit is not too complex and does not need adjustment (e.g. purposely designed like your intended use for the LXmini) then I can heartily recommend analog active circuitry. If not much gain is needed in the filters, then very clean active elements can be used (I believe Nelson's FET based buffers fall into this category). Where active analog filtering falls down is its flexibility and ability to be re-configured. Who wants to have to unsolder and resolder components to change the response? Also, the selection of the components used in the circuitry is very important to achieve the target response - if you are building the circuit, you must be aware of the sensitivity of the filter to the actual values of the component used to build the real-world circuit. This can require some careful testing of each component, selection of "best value" components out of a lot, or the purchasing of expensive tight-spec components. But if done well, you will get a very close match to the intended filter and it will perform well, as long as the filtering in total is not super complex and does not require high Q stages (where accuracy is very important). Also, analog active filters with gain stages can provide very high headroom and relatively high output voltage. This can be an advantage in certain situations and will provide good headroom at this stage in the playback chain.
I think the appropriate choice between analog-active and IIR DSP can be easily made between into two target audiences:
1) the listener: you intend to use the circuits with only one system/loudspeaker, and can build or buy an active analog filter of high quality
2) the builder/tweaker: you want to be able to re-use the filtering unit, possibly for widely different needs, and want lots of expandability and adjustability
If you fall under (1), then stick with an analog active crossover (not an adjustable type).
If you fall under (2), then use IIR DSP.
To make the most fair comparison, you can compare IIR digital filtering against analog active filtering. The types of filters and the responses of these filters are very similar, whereas FIR filtering is another animal alltogether. So let's compare IIR DSP filtering and analog active filtering.
IIR DSP:
In theory, these are "mathematically perfect". But this is like saying that an analog circuit constructed out of idea amplifiers and idea passive components is perfect! The real world is not so perfect. For any DSP you can have errors of various types, and the circuitry itself (e.g. the DAC and possibly ADC) will impart some temporal and frequency error and inject some level of noise. I assume that this is no surprise to anyone. What I think you are getting at is the inherent ACCURACY of DSP filters, and that is one advantage of IIR DSP over analog active filters. By accuracy I mean the accuracy of the pole(s) and zero(s) of the filter(s), which determine the frequency domain response. This becomes more important as the order of the filter increases, since higher order filters are often constructed by placing multiple lower order filters in series. The other advantage is that you can implement many, many filter stages in DSP without much of a noise penalty and without much signal degradation, which mostly arises when coming from and going to the analog domain (e.g. ADC and DAC). Finally, and possibly the biggest motivation for DSP is the easy by which filter parameters and the number of filter stages can be changed. It's essentially instantaneous and on demand. Sure an analog filter can be build with some adjustable elements, but overall the adjustability really pales in comparison to DSP. On the downside, DSP processing is just what is happening on the "inside", and the ADC and DAC are what gets you in and out of the DSP filtering box. The quality of ADC and DAC are very important. In the DAC, the maximum output voltage is something that is often overlooked and compared to analog filters may be limited to 1Vrms or 2Vrms only, which limits dynamic headroom somewhat. Also, if your DAC can only output 1Vrms and your super duper power amp needs 2Vrms to reach full power, you have a problem.
Analog Active Filters:
The primary advantage of analog active filters, in my mind, is the lack of temporal effects to the signal. With no ADC or DAC to muck up the waveform you get no ringing, etc. with a properly designed circuit. If the circuit is not too complex and does not need adjustment (e.g. purposely designed like your intended use for the LXmini) then I can heartily recommend analog active circuitry. If not much gain is needed in the filters, then very clean active elements can be used (I believe Nelson's FET based buffers fall into this category). Where active analog filtering falls down is its flexibility and ability to be re-configured. Who wants to have to unsolder and resolder components to change the response? Also, the selection of the components used in the circuitry is very important to achieve the target response - if you are building the circuit, you must be aware of the sensitivity of the filter to the actual values of the component used to build the real-world circuit. This can require some careful testing of each component, selection of "best value" components out of a lot, or the purchasing of expensive tight-spec components. But if done well, you will get a very close match to the intended filter and it will perform well, as long as the filtering in total is not super complex and does not require high Q stages (where accuracy is very important). Also, analog active filters with gain stages can provide very high headroom and relatively high output voltage. This can be an advantage in certain situations and will provide good headroom at this stage in the playback chain.
I think the appropriate choice between analog-active and IIR DSP can be easily made between into two target audiences:
1) the listener: you intend to use the circuits with only one system/loudspeaker, and can build or buy an active analog filter of high quality
2) the builder/tweaker: you want to be able to re-use the filtering unit, possibly for widely different needs, and want lots of expandability and adjustability
If you fall under (1), then stick with an analog active crossover (not an adjustable type).
If you fall under (2), then use IIR DSP.
It might also be said that since much source material is digital to begin with, one may as well do as much as possible in the digital domain. Certainly, that's how a lot of processing is done in most records made today.
Problem with doing digital crossovers is more than one high quality dac is needed, and more than one high quality power amp, too. If one goes top quality all the way with dacs and amps, then a digital cross over might make more sense.
However, most people seem to have a budget for such things and so lower quality dacs are often used to keep cost down. If one has to do that, my preference would be to have one really good dac and analog use cross-overs. Probably sound better despite the loss of some ability to play digital tricks such with time-alignment delays and room EQ.
Problem with doing digital crossovers is more than one high quality dac is needed, and more than one high quality power amp, too. If one goes top quality all the way with dacs and amps, then a digital cross over might make more sense.
However, most people seem to have a budget for such things and so lower quality dacs are often used to keep cost down. If one has to do that, my preference would be to have one really good dac and analog use cross-overs. Probably sound better despite the loss of some ability to play digital tricks such with time-alignment delays and room EQ.
+1
OP if you are into computer audio, it becomes easier if using a software player like JRiver, it contains a convolution engine, which allows you to host your IIR and/or FIR filtering, all in the digital domain.
Using shareware programs like rePhase and REW or commercial DSP software programs like Acourate or Audiolense, you can create digital crossovers of any type/slope, you can time align the drivers, linearize the drivers, apply over all eq to shape the response to your preferred target. One can even apply excess phase correction...
The above is very difficult to do (and some of it impossible) from an analog active XO perspective...
Good luck with your project!
Kind regards,
Mitch
Thanks all for your feedbacks, and especially Charlie for his detailed and experienced based long answer.
To ne more precise, I was speaking of IIR filters, implemented in software. I programmed my filters in a Stm32 Microcontroller, using ST libraries. As proposed by some of you, my set up is 100% digital: files on RPi => USB => Stm32 performing DSP => 2xSpdif => 2xFX-Audio D802 (that use STA326).
Works fine except lack of power :-(
Charlie, I'm in your familly (1) class, except that I like to integrate electronics and assemble lean solutions.
I understand that IIR filters can have a lot of advantages, especially if we stay in the digital domain all the ways.
Well designed analog filter can have some budget advantage using only one DAC and a much larger offer as of analog input amplifiers.
JMF
To ne more precise, I was speaking of IIR filters, implemented in software. I programmed my filters in a Stm32 Microcontroller, using ST libraries. As proposed by some of you, my set up is 100% digital: files on RPi => USB => Stm32 performing DSP => 2xSpdif => 2xFX-Audio D802 (that use STA326).
Works fine except lack of power :-(
Charlie, I'm in your familly (1) class, except that I like to integrate electronics and assemble lean solutions.
I understand that IIR filters can have a lot of advantages, especially if we stay in the digital domain all the ways.
Well designed analog filter can have some budget advantage using only one DAC and a much larger offer as of analog input amplifiers.
JMF
Ah, I see if your files/source are already in the digital domain then that eliminates the ADC.
If you can live with non-esoteric DACs, then using a pro-audio recording interface is a good solution. Look with one having 8 analog outputs. These will often be at 2Vrms or higher max output voltage, which is high enough to match most any amp's required input voltage for full power.
I have used:
Behringer FCA610 - 8 analog, 2 digital outputs. Inexpensive. Sounds OK. Make s a "pop" sometimes when starting playback, especially if sample rate has changed
Presonus 1818VSL - 8 analog, 2 digital, 8 ADAT. Moderately expensive. Better all around compared to the FCA610. Sometimes a faint "pip" (as opposed to a pop) when changing sample rates and beginning playback, but nothing that bothers me.
Behringer ADA8200 - 8 analog inputs, 8 analog outputs, I/O to computer via ADAT only. Because the I/O is ADAT only, you need a computer or computer peripheral that can send/receive ADAT. THere is only 1 that I know of (that's affordable), and that is the $95 USBstreamer by miniDSP operated in ADAT mode. Sample rate is only 48kHz max, but the ADA8200 ain't bad and has 4Vrms max output IIRC.
The above are mostly USB interfaces (the USBstreamer is USB), so the USB bus on the computer should be sufficiently capable. I have used the FAC610 on a Raspberry Pi 3 and it works, barely. When I move the system over to a BayTrail or Atom CPU based computer, no problem. Same for the Presonus.
If this is not exactly what you are after, then hopefully my ramblings are helpful for others thinking about doing IIR DSP in software. Perhaps your particular problem is related to your "digital input amps" not having enough power and not to DSP or analog per se?
Thanks Charlie,
Youn pointed it: in my case, what is frustrating is that I reach the point where my amps have too much distorsion. So It is the amps I have to change first... and then this triggers concerns about the whole chain.
If I want to stay on the same route, for more power, I need to have dedicated PWM controller and more powerfull PWM input Class D amps. The chips exist, But I don't seee affordable boards implementing them. I will continue to investigate that path before switching back to partly analog.
JMF
Youn pointed it: in my case, what is frustrating is that I reach the point where my amps have too much distorsion. So It is the amps I have to change first... and then this triggers concerns about the whole chain.
If I want to stay on the same route, for more power, I need to have dedicated PWM controller and more powerfull PWM input Class D amps. The chips exist, But I don't seee affordable boards implementing them. I will continue to investigate that path before switching back to partly analog.
JMF
- Status
- Not open for further replies.