Hello,
I'd like to know, when using an operational amplifier in a transimpedance conf. for a current output dac chip, what the best strategy to choose the value of the feedback resistor decoupling cap.
Assuming this cap is used for stability of a high bandwith OPA (> 50 Mhz or >100 M hz for illustration), how would you choose the low passmade by this capacitor ?
Is the phase angle an important item in that choice ?
I readed high frequency bandwidth was something important for amps, is it too for I/V stage of a current output DAC chip as well or better at the opposite not to play too much high, mostly if using NOS DAC with no reconstruction filter at the output to avoid nasties to come back in the audio band ?
Should be filtered both with the I/V operational amplifier, then in the next operational amplifier used as buffer for instance ?
Thanks
I'd like to know, when using an operational amplifier in a transimpedance conf. for a current output dac chip, what the best strategy to choose the value of the feedback resistor decoupling cap.
Assuming this cap is used for stability of a high bandwith OPA (> 50 Mhz or >100 M hz for illustration), how would you choose the low passmade by this capacitor ?
- maximum flat : what is shown in datasheet at 0.1% flat low passs
- something in the kilo hz instead as a 6db filter for the NOS DAC : illustration 30 K hz
- 200 K hz (I often this number in conversations)
- no feedback capacitor but a passive RC low pass before the next stage (buffer) ?
Is the phase angle an important item in that choice ?
I readed high frequency bandwidth was something important for amps, is it too for I/V stage of a current output DAC chip as well or better at the opposite not to play too much high, mostly if using NOS DAC with no reconstruction filter at the output to avoid nasties to come back in the audio band ?
Should be filtered both with the I/V operational amplifier, then in the next operational amplifier used as buffer for instance ?
Thanks
My understanding of the feedback cap in an I/V stage is that its to prevent slew-rate limiting in the opamp. I seem to recall Hawksford had a paper which covered this (amongst other things), I'll see if I can look it out.
Subjectively I've found the smaller the cap the better the sound, but not so small that the opamp goes into slewing.
Subjectively I've found the smaller the cap the better the sound, but not so small that the opamp goes into slewing.
Hi, you mean to prevent the OPA slews ? Or that it is stable on high level signals whatever the slew rate number given in the OPA datasheet ? Both same things ?
For the moment I use the value for the maximum flat when it is shown in the datasheet but do not know if the best strategy...
For instance with an opa828 and 1k5 R feedback, I used 20 pF, which was giving a high bandwidth (this opa is given up to 50 M hz if I am correct, but some jump much higher 200 M Hz and more... which becomes RF territory = fear for the enthusiast not EE like me !)
For the moment I use the value for the maximum flat when it is shown in the datasheet but do not know if the best strategy...
For instance with an opa828 and 1k5 R feedback, I used 20 pF, which was giving a high bandwidth (this opa is given up to 50 M hz if I am correct, but some jump much higher 200 M Hz and more... which becomes RF territory = fear for the enthusiast not EE like me !)
Last edited:
Yes I mean the cap prevents the OPA reaching its maximum slew-rate. As regards stability I think introducing a cap tends to decrease stability once any parasitic DAC output capacitance has been compensated for.
Having now skimmed @rayma's app note, its clear the choice of cap in that document is mainly dictated by the need to compensate for the DAC and layout parasitics. This leads to a cap value orders of magnitude below what's typically seen in audio I/V applications.
Having now skimmed @rayma's app note, its clear the choice of cap in that document is mainly dictated by the need to compensate for the DAC and layout parasitics. This leads to a cap value orders of magnitude below what's typically seen in audio I/V applications.
Last edited:
Excerpt from TI note SBAA150A:
"The remaining element in the design is the required feedback capacitor CF across the feedback resistor RF . Contrary to most DAC descriptions of this capacitor, it is not really there to limit the fast edge rates of the DAC output current (while it may appear to have this effect); rather, this capacitor is included to control the closed loop frequency response of the amplifier circuit. Without this capacitor, most high-speed amplifiers will show a very peaked closed loop frequency response when driven from a capacitive source (CS = DAC output capacitance + op amp and layout parasitic). If the closed loop frequency response is peaking, then each DAC step will cause an overshoot and ringing in VO. "
"The remaining element in the design is the required feedback capacitor CF across the feedback resistor RF . Contrary to most DAC descriptions of this capacitor, it is not really there to limit the fast edge rates of the DAC output current (while it may appear to have this effect); rather, this capacitor is included to control the closed loop frequency response of the amplifier circuit. Without this capacitor, most high-speed amplifiers will show a very peaked closed loop frequency response when driven from a capacitive source (CS = DAC output capacitance + op amp and layout parasitic). If the closed loop frequency response is peaking, then each DAC step will cause an overshoot and ringing in VO. "
@ abrax. I use very thin 0.2 to 0.26 width mm short trace between the dac and the I/V ( 5 mm length in my last board) with no ground below to figth the stray capacitance here But I do not know if enough. I try also to not put any ground pours in the 5 mm arounds.
Some people are not using this cap here, but not sure it is the best strategy, should be not stable with that modern fast OPAs?!
Have to read this TI paper
Some people are not using this cap here, but not sure it is the best strategy, should be not stable with that modern fast OPAs?!
Have to read this TI paper
Last edited:
@diyig That's good practice to minimize circuit parasitics at the opamp summing node. However the contribution that the DAC chip itself makes is generally an unknown as audio DAC chip manufacturers don't specify output capacitance, though they do often tell us the output resistance.
<Hawksford paper now attached> He says in this paper that the feedback cap does limit the slew-rate at the output of the OPA however there's going to be internal (within OPA) distortion due to large instantaneous voltage swings at the OPA input terminals which the feedback cap doesn't significantly suppress.
<Hawksford paper now attached> He says in this paper that the feedback cap does limit the slew-rate at the output of the OPA however there's going to be internal (within OPA) distortion due to large instantaneous voltage swings at the OPA input terminals which the feedback cap doesn't significantly suppress.
Last edited:
The DAC output analog low pass filter has the task to reject all images (and noise) which are not in the passband. You only want to amplify your signal and not ultrasonic noise. In addition ultrasonic signals could couple back into your passband by intermodulation distortion of your op amp.
Images start to appear at f_sampling/2, thus you want your low pass to set in at this frequency. There is a tradeoff when selecting the 3db cut off frequency, because low passes aren't ideal. Either cutting off a bit of your pass band or loosing rejection in the stop band.
For NOS dacs with sampling frequencies of 44.1kHz or 48kHz anything above 22.1kHz or thereabout should be rejected.
Most dac chips use 8x oversampling. Then your image frequencies shift up 8 times in frequency.
8 x 22.1 kHz = 176.8 kHz
8 x 48 kHz = 192 kHz
This is how ~200 kHz as cut off frequency arises.
If you use a current feedback amplifier for IV you will have to rely on a passive RC low pass. To me there is nothing inherently wrong in active or passive filtering.
I do not know when phase angle deviation becomes critical.
How much filtering/attenuation you need/want (1st order, 2nd order,... 8th order) certainly depends on your DAC and is not universal.
Images start to appear at f_sampling/2, thus you want your low pass to set in at this frequency. There is a tradeoff when selecting the 3db cut off frequency, because low passes aren't ideal. Either cutting off a bit of your pass band or loosing rejection in the stop band.
For NOS dacs with sampling frequencies of 44.1kHz or 48kHz anything above 22.1kHz or thereabout should be rejected.
Most dac chips use 8x oversampling. Then your image frequencies shift up 8 times in frequency.
8 x 22.1 kHz = 176.8 kHz
8 x 48 kHz = 192 kHz
This is how ~200 kHz as cut off frequency arises.
If you use a current feedback amplifier for IV you will have to rely on a passive RC low pass. To me there is nothing inherently wrong in active or passive filtering.
I do not know when phase angle deviation becomes critical.
How much filtering/attenuation you need/want (1st order, 2nd order,... 8th order) certainly depends on your DAC and is not universal.
Depending on how you dimension it, I think it's either a filter capacitor or a frequency compensation capacitor or both.
If I were to design a circuit like this, I would choose some odd order for the reconstruction filter and then use this capacitor to make the first-order section (realize the filter's negative real pole).
Thanks Marcel for chimming in that thread. (I still have to try your last no dac chip DAC but monney gets a little low here,,so I deleted the try for the moment but big thanks you for the sharing I willl go later, seems crazy improvment over DAC chip based DAC devices.)
If one side of the capacitor see the virtual ground side and the other one the other polarity like the resistor does at the output side of the loopback, it could be called also decoupling (cause the virtual ground.... okay decoupling is more for real ground on the non inverting input as here not virtual reference like the inverting input that is "floating") ? Well for me that is a rethoric level as I do not understand all what happening electronically here inside the OPA circuitry.
I know transimpedance OPA are not liked for that transimpedance task, but some for my ears give crazy good results, few but stilll some are very good here.
Btw I have certainly a stupid question yet :
Some datasheet shows in gain configuration (not transimpedance conf) that the other input pin needs a small resistor compensation, often around 100R.
If so in the datasheet for an OPA used the transimpedance conf. should one lift the ground level onn the non inverting input by putting this pin at 100R ground level too instead zero R (what a ground is) and as the rest of the pcb ground is for the others devices : dac chips, others OPA like the folllowing buffer for illustration ?
If one side of the capacitor see the virtual ground side and the other one the other polarity like the resistor does at the output side of the loopback, it could be called also decoupling (cause the virtual ground.... okay decoupling is more for real ground on the non inverting input as here not virtual reference like the inverting input that is "floating") ? Well for me that is a rethoric level as I do not understand all what happening electronically here inside the OPA circuitry.
I know transimpedance OPA are not liked for that transimpedance task, but some for my ears give crazy good results, few but stilll some are very good here.
Btw I have certainly a stupid question yet :
Some datasheet shows in gain configuration (not transimpedance conf) that the other input pin needs a small resistor compensation, often around 100R.
If so in the datasheet for an OPA used the transimpedance conf. should one lift the ground level onn the non inverting input by putting this pin at 100R ground level too instead zero R (what a ground is) and as the rest of the pcb ground is for the others devices : dac chips, others OPA like the folllowing buffer for illustration ?
Last edited: