[Garbz]

I have been wondering recently what some of the advantages of multiple opamps in a stage would be. What I mean is like an OPA134 followed by another OPA134, not by a dedicated buffer.

I can only think of a few which is allowing higher order analogue filters, multiple feedback loops, biasing without inducing DC offset I think, loading up a previous stage with a high impedance allowing different choices of capacitors.

What else can people here come up with? Specifically I am wondering about applications in DAC design. I see a lot of DACs where a second opamp is used to convert a signal to single ended, as well as applying additional filtering.

 

[rickcr42]

impedance converters and no more.

provies any previous stage with and easy to drive without losses high impedanc then providing a non-varying low output impedance to any passive networks which again need to see a buffer/impedance converter for the network while providing good output drive to the next "thing" in line.

think RIAA networks,tone controls,filters,crossfeed networks.etc.

anything other than the need to match stage impedances is throwing active stages in just to do it since it makes little sense to add multiple stages to do something a single stage will do just as well and add in the fact that any active stage adds noise and you actually take a step back not forwrd...at least that is my opinion and I'm sticking to it

out

 

[kramer5150]

Doesn't it also double the current capacity of the OP amp?

 

[Voodoochile]

Only in parallel, not in series. (re: current capacity)
And it's not recommended to parallel op-amps, generally.

 

[FallenAngel]

I know you guys are talking mainly about a DAC, but doesn't the A47 design in uses an OPA2134 followed by another OPA2134?

 

[rickcr42]

Quote:

but doesn't the A47 design in uses an OPA2134 followed by another OPA2134?

nope.

Parallel connected to double the output current delivery to the load,straight from the TI/BB OPA604 app notes (see current doubled opamp).The reason why the A47 has the name it does is because Mike AKA "Apheared" the "A" in the A47 used parts that all had the numbers 4 and or 7 in their value (47ohms,4,7 ohms,4.7 uf ,etc) so Apheared's 47 or A47 Amp

 

[Garbz]

Hey rick stumbled in again by accident I see

Can someone help me illustrate this by example. Page 9 Figure 5 of the PCM1704 datasheet http://focus.ti.com/lit/ds/symlink/pcm1704.pdf shows an analogue filter comprising of 2 opamps in series.

The task of the first opamp is perfectly clear. I/V conversion with a lowpass filter tuned to a stupidly high value probably only for stability purposes. This is followed by a 2nd order low pass filter.

The question is this. Assuming that the digital filter takes the stop band attenuation near enough to zero that it becomes irrelevant, does it still make sense to use a filter like this. So suppose only a single order low pass filter tuned to 50khz is sufficient to remove all digital crap would it make more sense to simply use a single stage for both I/V and low pass filtering by paralleling the I/V resistor with a much larger cap?

I am just trying to understand why designers still opt for these multi stage filters when the stop band attenuation is so low nowadays that it is almost possible to eliminate low pass filtering altogether.

 

[jcx]

you need to read the datasheets real careful and apply some external knowledge, delta sigma dacs do have a region of very high image attenuation but there is inevitable breakthrough when you reach the modulator clock frequency - often a few MHz

SACD dsd single bit modulators are really bad - modulator noise is shaped to give 120 dB S/N up to 20KHz but then the modulator noise starts rising at ~ 5 th order rate with increasing frequency - to +6 dB around 1-2 MHz

image rejection filters are usually biquad sections - 2 poles per op amp, >5 th order atten is desirable for dsd, usually the I/V can have 1 pole to implement odd order filters with the following biquad stages

of course typical 8 MHz GBW audio op amps make terrible MHz filters - there isn't enough feedback gain at high frequencies to keep their intrinsic 50-100 Ohm output impedance low enough to soak up I/V "glitch" or dsd modulator energy - this is where discrete or cfb op amp buffers in the loop can help and multiple feedback low pass filter topologies which have a "passive" RC up front

 

[rickcr42]

Quote:

Hey rick stumbled in again by accident I see

Nah,this one was by intent dude.Figured you guys could use a little shaking up and I am the rickmonster afterall sooooooo

Quote:

I am just trying to understand why designers still opt for these multi stage filters when the stop band attenuation is so low nowadays that it is almost possible to eliminate low pass filtering altogether

either you can have a slow rolloff beginning at a lower frequency (single pole filter) or a higher rate of rolloff,multiple-stage filters with the F3 at a higher frequency where there is less chance of upper octave requncy response rolloff in the audible bands.More likleyhood of phase response abberations so as with most such things choose your poison.

the bessel filter is an attempt to minimize this but nothing is perfect and to fix a thing many times you screw something else up.Why the great DACs are audibly superior even when they use the same "engine" at the heart of the design.They get the post DAC chip parts right (and all modern DAC chips have built in filters on the voltage outputs,to releive the responsibility of having to design an adequate output section

)

 

[Garbz]

Ahh I see. Now I am paranoid that my first order LP filter is insufficient. Oh well at least my tweeters have not blown yet.

Looks like my next DAC will need a couple of opamps and a well designed VCVO filter

 

[rickcr42]

Quote:

Ahh I see. Now I am paranoid that my first order LP filter is insufficient. Oh well at least my tweeters have not blown yet.

the poles do not have to be at exactly the same F3 to get big time attenuation where you need it most so for instance you can do an LCR bessel (see BB/TI App.notes) then buffer it THEN add your line level trafos box after THIS (should roll off around 30-40khz) which will add another pole of -6dB per octave rolloff that when combined with the intial filter will be a full 2 pole filter at the upper end

 

[JahJahBinks]

hey rickcr42, nice to see you back.

 

[jcx]

Your tweeter is likely safe – few accept any energy above 40 KHz

the MHz grunge in your signal is more likely to blow up you amplifier - I have seen claims that early SACD dsd did just that to a few "audiophile" power amps - now the spec calls for starting the roll off at 50 KHz vs the original 100 KHz

even a reasonable amp that can't be provoked into instability by the HF noise will rectify high level HF at its diff pair input and increase the noise floor and possibly add correlated distortion in the audio band

 

[rickcr42]

Quote:

hey rickcr42, nice to see you back.

hey yourself

not exactly back though,still in exile across the big pond until further notice.Could take years..........................

back OT

Single pole filter with an F3 at 30k kHz falls at a rate of -6dB per octave so by 60 kHz is only 12 dB down.That same filter with a second pole will fall at a rate of -12dB per octave so that same 60kHz would now be -24dB.

That is a HUGE difference.

The problems come in with the phase response of higher order filter combined with the need for active filters when you get to these higher orders which as in anything using an active stage will have audible consequences so it is always a tradeoff and why the digital filter was invented.

By oversampling the digital signal where being a one or zero there is less damge that can be done to the original signal (in theory) the aliasing is moved up in frequency by a factor of the oversampling so for a X8 digital filter the shift is up in the analog domain as well making the analog filter F3 higher as well,again theoretically making it inaudible and having no effect on the below 20khz freqs AND relaxing need for high order roloffs which intrude at the upper octaves...on paper.......

If the solution were as perfect as it is claimed there would be NO zero oversampling DACs yet there are many so what does that tell you ?

no free rides and nothing is perfect in the "perfect forever" medium so choose your poison then live with it