Hi all,
I know that is not the right place for that..
But I've asked this question in "digital forum" and have no question until now..
Anyway, I am expecting to have an answer in here as usual
The question is simple;
I am designing an USB DAC and using PCM2702.. The digital part is not the matter of here...
For the LPF stage, I thought that an AD8397 will be a good choice because of its output current capability. With this way I expect to have a headphone output (like Amb guys did) also.. So I have designed a simple 2nd order LPF circuit with some gain (2x) for AD8397. This gain is added to drive both 32/64R headphones..
Now the question comes;
I am not sure to use that op-amp as both LPF and headphone driver.. I dont know but it seems "too good to be true".. What do you say about that? Will I have good results with that configuration?
PS: LPF R/C values defined by trying on the simulation and the corner frequency is ~85Khz ..
I know that is not the right place for that..
But I've asked this question in "digital forum" and have no question until now..
Anyway, I am expecting to have an answer in here as usual
The question is simple;
I am designing an USB DAC and using PCM2702.. The digital part is not the matter of here...
For the LPF stage, I thought that an AD8397 will be a good choice because of its output current capability. With this way I expect to have a headphone output (like Amb guys did) also.. So I have designed a simple 2nd order LPF circuit with some gain (2x) for AD8397. This gain is added to drive both 32/64R headphones..
Now the question comes;
I am not sure to use that op-amp as both LPF and headphone driver.. I dont know but it seems "too good to be true".. What do you say about that? Will I have good results with that configuration?
An externally hosted image should be here but it was not working when we last tested it.
PS: LPF R/C values defined by trying on the simulation and the corner frequency is ~85Khz ..
Dxvideo:
- appears to a 2-pole Sallen-Key configuration
- at fc 85kHz, opamp GBP should be above 8.5mHz
Comments:
- I see ringing/oscillation on my sim, with your values (see phase plot exceeding 180deg rule)
- slight peaking, approx. +0.4db @ 24khtz
- consider increasing input coupling from 1uF to 4uF (improves low-freq phase response)
- consider Bessel filter configuration (better group delay and overshoot characteristics, look up "Blauert and Laws")
Hope this helps....
Attached:
- zipped sim Bode, transient plots and suggested Bessel schematics and comparison plots (823kB)
http://premium1.uploadit.org/bichi007//1-DxVideo-LPF-Schematics-Result-Plots.zip
REFS:
Active Low-Pass Filter Design (Rev. B)
http://focus.ti.com/general/docs/techdocsabstract.tsp?abstractName=sloa049b
Active Filter Design Application
http://focus.ti.com/docs/toolsw/folders/print/filterpro.html
- appears to a 2-pole Sallen-Key configuration
- at fc 85kHz, opamp GBP should be above 8.5mHz
Comments:
- I see ringing/oscillation on my sim, with your values (see phase plot exceeding 180deg rule)
- slight peaking, approx. +0.4db @ 24khtz
- consider increasing input coupling from 1uF to 4uF (improves low-freq phase response)
- consider Bessel filter configuration (better group delay and overshoot characteristics, look up "Blauert and Laws")
Hope this helps....
Attached:
- zipped sim Bode, transient plots and suggested Bessel schematics and comparison plots (823kB)
http://premium1.uploadit.org/bichi007//1-DxVideo-LPF-Schematics-Result-Plots.zip
REFS:
Active Low-Pass Filter Design (Rev. B)
http://focus.ti.com/general/docs/techdocsabstract.tsp?abstractName=sloa049b
Active Filter Design Application
http://focus.ti.com/docs/toolsw/folders/print/filterpro.html
I have no buffer before this circuit.. Directly connected to PCM2702s outputs..
Anyway,
This is what I see in my simulation;
The red line is frequency response and the blue dotted line is the phase..
I cannot see any oscillation, do you notice any?
And, concerning the source impedence; I have added a series 10K to the signal source and have no difference on both phase and frequency plot!
May you explain this issue a bit more please..
Thanks a lot for your comments..
Anyway,
This is what I see in my simulation;
An externally hosted image should be here but it was not working when we last tested it.
The red line is frequency response and the blue dotted line is the phase..
I cannot see any oscillation, do you notice any?
And, concerning the source impedence; I have added a series 10K to the signal source and have no difference on both phase and frequency plot!
May you explain this issue a bit more please..
Thanks a lot for your comments..
By the way,
Dear Bichi, I ve just noticed that you made a simulation for me.. Thanks a lot for this.
As I see there is no problem with my simulator since I have exactly the same results..
The slight peaking on 24hz you mentioned is calculated by me to reduce the slope affect on 20Khz band.. (May be because I like treble details too much ) Otherwise there a -0,5dB affect on 20Khz point.. Or the worst is I have to extend the frequency response till 200Khz to reduce this... So to put a peak on 24Khz seemed a good idea to me..
Anyway, another thanks for the software.. Now I am using it and already calculated some other combinations..
Any special recommendation on filter type for especially audio circuits?
Dear Bichi, I ve just noticed that you made a simulation for me.. Thanks a lot for this.
As I see there is no problem with my simulator since I have exactly the same results..
The slight peaking on 24hz you mentioned is calculated by me to reduce the slope affect on 20Khz band.. (May be because I like treble details too much
) Otherwise there a -0,5dB affect on 20Khz point.. Or the worst is I have to extend the frequency response till 200Khz to reduce this... So to put a peak on 24Khz seemed a good idea to me..Anyway, another thanks for the software.. Now I am using it and already calculated some other combinations..
Any special recommendation on filter type for especially audio circuits?
DxVideo:
- you will not "see" real-time oscillation, on Bode or Phase plots.
- indicator is phase rotation, over frequency. (phase margin)
Tips:
- input a frequency, say 7.5kHz at 1Vrms and watch it on "virtual" sim oscilloscope on your simulator.
- or take a look at the "transient" plots I made for you, and compare your values with the Bessel values.
- or input the Bessel values on your simulator and see the differences.
Easier:
- input your values on "online simulator tool," Sallen-Key and look at line: "Oscillation frequency."
- if the design is not prone to oscillation, it will state: "This system does not oscillate."
- I see "Oscillation frequency: f = 63527.1933944[Hz]" with your values, indicating a possible oscillation problem.
Page Navigation to Sallen-Key Tool:
Tech Tools -> Engineering Design Utilities -> Filter Tool -> Sallen-Key Active Filter -> Sallen-Key Low-pass Filter Tools
Online Sallen-Key Low-Pass Filter Design Tool:
http://translate.google.com/transla...ult&prev=/search?q=okawa-denshi.jp&hl=en&sa=G
Good Luck!
- you will not "see" real-time oscillation, on Bode or Phase plots.
- indicator is phase rotation, over frequency. (phase margin)
Tips:
- input a frequency, say 7.5kHz at 1Vrms and watch it on "virtual" sim oscilloscope on your simulator.
- or take a look at the "transient" plots I made for you, and compare your values with the Bessel values.
- or input the Bessel values on your simulator and see the differences.
Easier:
- input your values on "online simulator tool," Sallen-Key and look at line: "Oscillation frequency."
- if the design is not prone to oscillation, it will state: "This system does not oscillate."
- I see "Oscillation frequency: f = 63527.1933944[Hz]" with your values, indicating a possible oscillation problem.
Page Navigation to Sallen-Key Tool:
Tech Tools -> Engineering Design Utilities -> Filter Tool -> Sallen-Key Active Filter -> Sallen-Key Low-pass Filter Tools
Online Sallen-Key Low-Pass Filter Design Tool:
http://translate.google.com/transla...ult&prev=/search?q=okawa-denshi.jp&hl=en&sa=G
Good Luck!
Dxvideo said:
Ah, just saw your reply and you are welcome!
- ummm, I prefer Bessel with MFB topology, instead of Sallen-Key or Butterworth.
- if you search "Audio Bessel Filter," you will find detailed papers/theories and opinions.
- lastly, I like the sound Bessel filters produce, especially filtering/reconstructing after high-performance (up to 192kHz sample rate) Delta-Sigma type DACs.
- last comment, SMT (surface mount), careful layout and decoupling required for best performance.
An externally hosted image should be here but it was not working when we last tested it.
This is the new one.. Based on TI Filter Pro calculations for
Bessel+MFB (as you recommended) and x2 gain...
By the way,
I am already designed a PCB for it.. However I am not good at working with SMD resistors and caps
And the PCB is mostly fullfill with "normal" components.. Do you think it causes any problem?PS: I will post my PCB design in the day.
DxVideo,
- hard to predict "leaded component" performance vs. SMT, without seeing and testing physical design.
- 1% resistors, NPO/COG 5% capacitors are preferred. (SMT PPS film caps sounded "harsh" to me)
- capacitors should not be below 20pf and larger value resistors (50k+), can add Johnson (thermal) noise.
- attention to groundplane is key.
- opamp power rail decoupling is an important factor (10 to 100nF, X7R ceramic, close to opamp power pins to groundplane)
- larger, power rail decoupling capacitors (10 to 100uF electrolytics) should be less than 1cm from opamp power pins.
- depending on your power-supply decoupling, quality and location, some prefer low-ESR, electrolytic without smaller ceramic.
- hard to predict "leaded component" performance vs. SMT, without seeing and testing physical design.
- 1% resistors, NPO/COG 5% capacitors are preferred. (SMT PPS film caps sounded "harsh" to me)
- capacitors should not be below 20pf and larger value resistors (50k+), can add Johnson (thermal) noise.
- attention to groundplane is key.
- opamp power rail decoupling is an important factor (10 to 100nF, X7R ceramic, close to opamp power pins to groundplane)
- larger, power rail decoupling capacitors (10 to 100uF electrolytics) should be less than 1cm from opamp power pins.
- depending on your power-supply decoupling, quality and location, some prefer low-ESR, electrolytic without smaller ceramic.
I think I have a situation with this opamp!
With this design (the new one) the simulator shows ~19mV DC offset on outputs which is quite high for a headphone amp.. If I reduce the resistor values by 10 times (680R/330R) then the offset becomes acceptable values (~1mV).. But in this case the LPF will reach the PCM2702s R(load) limits (which is >5K)..
I know these values are just on the paper.. But I am sure the real life will be worst..
So in this case, I have two alternatives; another opamp or a buffer... But I dont want to make more complex it.
What do you think?
With this design (the new one) the simulator shows ~19mV DC offset on outputs which is quite high for a headphone amp.. If I reduce the resistor values by 10 times (680R/330R) then the offset becomes acceptable values (~1mV).. But in this case the LPF will reach the PCM2702s R(load) limits (which is >5K)..
I know these values are just on the paper.. But I am sure the real life will be worst..
So in this case, I have two alternatives; another opamp or a buffer... But I dont want to make more complex it.
What do you think?
An externally hosted image should be here but it was not working when we last tested it.
So,
This is the last status of the design..
- Bessel with MFB advices considered.
- Some gain (I am not sure but the simulator shows 6dB with 10K and ~3dB with 32R loads ???) still exists..
- 2Hz to 85Khz with -3dB
- Considered the possible (simulated) DC offset problem with AD8397 opamp and exchanged with OPAx134 opamp.. To keep outputs high current capability I use two paralelled (according to Intersil AN-1111.1).
- Power supply voltage for the opamps increased to +/-8v, because theyre not rail to rail..
Any more suggestions or critics?
PS: Both input caps should be 47u not 22u..
DxVideo,
- welcome to the "Design TradeOff/Compromise Monkey House!" (and you thought it was a simple question...)
- don't know what sim package you are using and don't know if the AD8397 model is accurate.
- I don't have the AD8397 model.
- from spec sheet, AD8397 input offset = 2.5mV (so, gain of x2 = possible 5mV DC offset)
- my sim with current opamp, THS3061, (similar TI current opamp) DC offset with your values shows 11.99mV.
- adjusting TIFilter Pro, seed to 5k, 85kHz fc, to match your required input impedance, shows 8.98mV.
- simulating with OPA2134, shows 5.1uV DC offset and NE5534 shows 5.1mV DC offset.
- OPA2353 sim shows 4.5nV DC offset, with: C1 = 100p; C2 = 680p; R2=4.5k; R3=7k; R4=2.3k (Rimp adjusted for 5k)
You might be better off duplicating LPF, shown on TI's PCM2702 Evaluation Board.
- then, substituting opamps to simulate/predict performance. (see page 3)
- and right, might have to add output coupling cap or offset correction circuit (divider/servo) or driver/buffer for your 32ohm headphone load.
off to bed now, see how you go tomorrow....
edit: ha! another new design, eh? - (you're making me work too hard... hahahaaaa)
DEM-PCM2702 EVALUATION FIXTURE:
http://www.ti.com/litv/pdf/sbau025
- welcome to the "Design TradeOff/Compromise Monkey House!" (and you thought it was a simple question...)
- don't know what sim package you are using and don't know if the AD8397 model is accurate.
- I don't have the AD8397 model.
- from spec sheet, AD8397 input offset = 2.5mV (so, gain of x2 = possible 5mV DC offset)
- my sim with current opamp, THS3061, (similar TI current opamp) DC offset with your values shows 11.99mV.
- adjusting TIFilter Pro, seed to 5k, 85kHz fc, to match your required input impedance, shows 8.98mV.
- simulating with OPA2134, shows 5.1uV DC offset and NE5534 shows 5.1mV DC offset.
- OPA2353 sim shows 4.5nV DC offset, with: C1 = 100p; C2 = 680p; R2=4.5k; R3=7k; R4=2.3k (Rimp adjusted for 5k)
You might be better off duplicating LPF, shown on TI's PCM2702 Evaluation Board.
- then, substituting opamps to simulate/predict performance. (see page 3)
- and right, might have to add output coupling cap or offset correction circuit (divider/servo) or driver/buffer for your 32ohm headphone load.
off to bed now, see how you go tomorrow....
edit: ha! another new design, eh? - (you're making me work too hard... hahahaaaa)
DEM-PCM2702 EVALUATION FIXTURE:
http://www.ti.com/litv/pdf/sbau025
My AD8397 model is from Analog website.. So it must be accurate enough (I hope)..
However I am not sure about my simulator.. I am using "5spice Analysis" for over 5 years.. And I cannot use another one! (believe me I have tried severeal simulators but failed) Its a free one and has very limited functions.. But in real life, if it says there is 10mV offset on outputs I always see a bit more or less 10mV..
So from the beginning it was a mistake to select a BJT input opamp, so it was my fault!
However, I still need some gains... And this is the modifyed circuit, accorded to TI's reference design + my comment
And this is the frequency + phase response of it.. Of course plus my simulators comment....
However I am not sure about my simulator.. I am using "5spice Analysis" for over 5 years.. And I cannot use another one! (believe me I have tried severeal simulators but failed) Its a free one and has very limited functions.. But in real life, if it says there is 10mV offset on outputs I always see a bit more or less 10mV..
So from the beginning it was a mistake to select a BJT input opamp, so it was my fault!
However, I still need some gains... And this is the modifyed circuit, accorded to TI's reference design + my comment
An externally hosted image should be here but it was not working when we last tested it.
And this is the frequency + phase response of it.. Of course plus my simulators comment....
An externally hosted image should be here but it was not working when we last tested it.
As mentioned that would be "a simple question"... So the simple part is coming 
When I simulate a basic inverted configuration of OPA134 with 24K/12K resistors it gives 6dB gain.. Thats weird.. It seems no matter what kind of load you use 32R or 10K! Thats another weird situation..
However, when I simulate this (the above) circuit it gives ~4dB with 10K load and reduces to ~3dB when load is 32R..
Now!
- Should I trust that? Whats the diffrerence between a basic inverted amplifier and this?
- Am I wrong or in inverted amplifier AV = Rf/Ri so 24K/12K must give 2x = 3dB of gain.. Which is true?
I am confused!
When I simulate a basic inverted configuration of OPA134 with 24K/12K resistors it gives 6dB gain.. Thats weird.. It seems no matter what kind of load you use 32R or 10K! Thats another weird situation..
However, when I simulate this (the above) circuit it gives ~4dB with 10K load and reduces to ~3dB when load is 32R..
Now!
- Should I trust that? Whats the diffrerence between a basic inverted amplifier and this?
- Am I wrong or in inverted amplifier AV = Rf/Ri so 24K/12K must give 2x = 3dB of gain.. Which is true?
I am confused!
at 200Khz!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.