A question or two to those who have used this FDA in non-ADC applications. I'm building a preamp/volume control for a listening system.
I will be running it on +/-5vDC supplies so the bias will be 0vDC. Since I am not trying to force the COM pin to particular voltage other than ground I am trying to figure out what to do with the pin. The data sheet says, "For a VOCM voltage at mid-supply, make no connection to the VOCM pin." But later says to "Depending on the intended application, a decoupling capacitor is recommended." I have seen implementations that are mid supply and ground biased with the bypass cap and others with the pin tied to ground. Tying it to ground makes the most sense to me but I wanted to ask what others had done.
Second question: How high of a resistor value can I use and keep the chip stable? I'm using the chip as a unity gain buffer that create a differential signal from a single ended signal and I'm trying to keep the input impedance high. Is 10k for the input and feedback resistors too high? I know noise will go up but 1k or 2k2 resistors seem too low for an input impedance. I just wanted to make sure the device is stabile with higher resistor values.
Thanks in advance,
Brian
I will be running it on +/-5vDC supplies so the bias will be 0vDC. Since I am not trying to force the COM pin to particular voltage other than ground I am trying to figure out what to do with the pin. The data sheet says, "For a VOCM voltage at mid-supply, make no connection to the VOCM pin." But later says to "Depending on the intended application, a decoupling capacitor is recommended." I have seen implementations that are mid supply and ground biased with the bypass cap and others with the pin tied to ground. Tying it to ground makes the most sense to me but I wanted to ask what others had done.
Second question: How high of a resistor value can I use and keep the chip stable? I'm using the chip as a unity gain buffer that create a differential signal from a single ended signal and I'm trying to keep the input impedance high. Is 10k for the input and feedback resistors too high? I know noise will go up but 1k or 2k2 resistors seem too low for an input impedance. I just wanted to make sure the device is stabile with higher resistor values.
Thanks in advance,
Brian
I have used opa1632 to convert single ended signal into differencial signal with 33k input resistor and 33k feedback resistor(unity gain ).As yo say,1k input resistor is too small.
My usage is for channel divider for multi speaker system.
33k feedback resistor is paralleled with 10nF or 1500pF cap.It works well without problem.I think probably 100k is also available.
The functionality of Vocm is not certain for me. As long as connecting to GND (+V=-V ) ,it make sense.When connecting to anather voltage,I still can't understand the true meaning of this PIN.
My usage is for channel divider for multi speaker system.
33k feedback resistor is paralleled with 10nF or 1500pF cap.It works well without problem.I think probably 100k is also available.
The functionality of Vocm is not certain for me. As long as connecting to GND (+V=-V ) ,it make sense.When connecting to anather voltage,I still can't understand the true meaning of this PIN.
I will be running it on +/-5vDC supplies so the bias will be 0vDC. Since I am not trying to force the COM pin to particular voltage other than ground I am trying to figure out what to do with the pin. The data sheet says, "For a VOCM voltage at mid-supply, make no connection to the VOCM pin." But later says to "Depending on the intended application, a decoupling capacitor is recommended." I have seen implementations that are mid supply and ground biased with the bypass cap and others with the pin tied to ground. Tying it to ground makes the most sense to me but I wanted to ask what others had done.
I never used the OPA1632, but I often use similar amplifiers at work. Do you prefer the DC voltage at the outputs to be as close as possible to ground, or do you prefer it to be as close as possible to the average of your +5 V and -5 V lines? Ideally there would be no difference, but in practice your supplies will have inaccuracies; for example, maybe your +5 V will be +5.1 V and your -5 V will be -4.9 V, so the average is +0.1 V.
If you want the DC voltage at the outputs to be close to ground, ground the VOCM pin, if you want it to be close to the average of your +5 V and -5 V lines, only decouple VOCM.
I don't know if it could be a concern but the common mode range is well inside the supply rails.
For supplies set to +-5Vdc the +ve common mode limit is +4Vpk and the -ve common mode limit is -3.5Vpk.
Protection against exceeding this tight common mode range could be had from input connected transdiodes tied into +-2.8Vdc
If this seems too low, then you would need to increase the supply rail voltage to some higher value.
For supplies set to +-5Vdc the +ve common mode limit is +4Vpk and the -ve common mode limit is -3.5Vpk.
Protection against exceeding this tight common mode range could be had from input connected transdiodes tied into +-2.8Vdc
If this seems too low, then you would need to increase the supply rail voltage to some higher value.
1k0 has a noise output equivalent to 4.1nV/root Hertz
10k has ~ sqrt(10k/1k) * 4.1nV = 6.5nV/rtHz
The opamp claims 1.3nV/rtHz which is roughly the same as a 105r resistor.
If you added a 105r resistor into the feedback, then you would roughly double (+3dB) the output noise of the opamp.
I'm not sure of that arithmetic, can someone confirm?
But at unity gain, does 1.3nV, or 4.1nV, or 6.6nV really matter? sqrt(6.5^2+1.3^2)= 6.6nV
10k has ~ sqrt(10k/1k) * 4.1nV = 6.5nV/rtHz
The opamp claims 1.3nV/rtHz which is roughly the same as a 105r resistor.
If you added a 105r resistor into the feedback, then you would roughly double (+3dB) the output noise of the opamp.
I'm not sure of that arithmetic, can someone confirm?
But at unity gain, does 1.3nV, or 4.1nV, or 6.6nV really matter? sqrt(6.5^2+1.3^2)= 6.6nV
Last edited:
Thanks for sharing your experience. I'll be at 10k or lower and unity gain so it sounds like I'll be OK in terms of stability.
The Vcom pin is there so external devices can set the bias of the OPA1632. This is a common need with ADC chips so the bias is set perfectly at the input - any mismatch in bias will limit signal swing in one direction and create a DC offset in the digital conversion. Have a look at page 32 of the CS4272 ADC/DAC Codec (https://d3uzseaevmutz1.cloudfront.net/pubs/proDatasheet/CS4272_F1.pdf), figure 12 and 13. Here the ADC is providing a voltage that is delivered to two driving amps via a resistive divider. Compare to figure 14 of the OPA1632 datasheet - it's meant to simplify the connections.
Well that is a great way to summarize the thought process. I appreciate it. I want my signal at ground regardless of supply variation since I'm DC coupling. So grounded it is!
I have measured signal noise ratio of several audio systems so far.The best S/N in usual situation(single ended) is around 100dB which is almost equal to CD. I think measurement data from IC makers are not in actual usage situation but in special situation for measurement.So,100dB is my reference for audio system.
opa1632 has very low noise figure of 1.3nV/rtHz.I guess in order to emphasize on this figure, TI wants to use low register like 1k.
But from my view point(S/N=100dB),conditions are more rough.In usual 2Vrms output,tolerable noise is 20 microVrms.
If you are in 20kHz frequency range,you can accept roughly 20micro/sqrt(20000) =130nV .
From my standard(S/N=100dB),thermal noise and inherent OPamp noise are less than from DAC chip or analog record.High register with opa1632 is acceptable for me.
Of cource another standard wants another requirement.
opa1632 has very low noise figure of 1.3nV/rtHz.I guess in order to emphasize on this figure, TI wants to use low register like 1k.
But from my view point(S/N=100dB),conditions are more rough.In usual 2Vrms output,tolerable noise is 20 microVrms.
If you are in 20kHz frequency range,you can accept roughly 20micro/sqrt(20000) =130nV .
From my standard(S/N=100dB),thermal noise and inherent OPamp noise are less than from DAC chip or analog record.High register with opa1632 is acceptable for me.
Of cource another standard wants another requirement.
+1
When noise is a concern, use low ohmic resistors and place buffers in front, else use high ohmic resistors preferably with bypass caps, because these are very fast amps.
Hans
That signal swing is sufficient for this application. The PGA2311 is limited a Vs of +/-5vDC and at that it can handle about +/-2.8v peak swing... a litle more than +8dbu IIRC. As long as the opamp can meet that it is sufficient for this circuit. I'm planning the board and components so that I can stuff a second one later with +/-15vDC supply and a PGA2320 for larger signals - assuming I don't botch the board design.
In terms of input protection, I don't really know a good scheme for protecting op amps in the inverting configuration. I always thought they were in better shape than non-inverting due to the series resistance on the input to limit current in over-voltage scenarios. Can you link to a schematic as an example?
It's not signal swing. It's the common mode limit even if there were no signal to be processed.
Could there ever be a start up or shutdown condition that allowed a common mode to get up to half rail voltage?
The series resistor combined with some shunting semiconductor is what protects the inputs from excessive voltage.
The series resistor is usually kept very low to minimise noise.
A Zener, or a stack of signal diodes, or a transdiode can be used as the shunt, but that still needs a resistor.
An alternative is a signal diode, or transdiode to both supply rails (this still needs a series resistor). But that allows the input to go one diode drop beyond the supply rail.
Setting up a supply rail referenced current sink at say 2V inside the supply rails would allow a diode, or similar to start shunting before the input voltage got as high as the supply.
Could there ever be a start up or shutdown condition that allowed a common mode to get up to half rail voltage?
The series resistor combined with some shunting semiconductor is what protects the inputs from excessive voltage.
The series resistor is usually kept very low to minimise noise.
A Zener, or a stack of signal diodes, or a transdiode can be used as the shunt, but that still needs a resistor.
An alternative is a signal diode, or transdiode to both supply rails (this still needs a series resistor). But that allows the input to go one diode drop beyond the supply rail.
Setting up a supply rail referenced current sink at say 2V inside the supply rails would allow a diode, or similar to start shunting before the input voltage got as high as the supply.
Last edited:
I'm not sure. Certainly the input buffer could see an incoming signal before startup, but it would either be single ended or differential. Maybe I"m confused about "common mode"... does it mean the maximum signal that can be present on both inputs?
Would something like this work?
https://goo.gl/photos/82t5S7zDiC3n3X5o8
I could use 2.8v zeners.
The whole point is through hole packaging. Even if not PDIP, it could be something else, like those metal cases or whatever, but not SMT.
I have yet to find anything worthwhile in through hole.
The alternative might just be to mock up a fully diff opamp using regular opamps, like the LM4562, but the performance won't be quite the same, even with the 4562 having really good ones.
Why can't they offer through hole parts? They're cutting out a lot of the DIYers this way...
I have yet to find anything worthwhile in through hole.
The alternative might just be to mock up a fully diff opamp using regular opamps, like the LM4562, but the performance won't be quite the same, even with the 4562 having really good ones.
Why can't they offer through hole parts? They're cutting out a lot of the DIYers this way...
Through hole is expensive, hampers performance for anything more demanding than audio and everyone wants small footprints.
An option is to solder an SMT to one of those DIP08 adapters, and then plug that into a socket. I am using that routinely to avoid being cut off from the best newer parts.
Jan
An option is to solder an SMT to one of those DIP08 adapters, and then plug that into a socket. I am using that routinely to avoid being cut off from the best newer parts.
Jan
New chips are seldom if ever in PDIP. The OPA1612 is at least in SOIC8, rather than VSOP10, which isn't too hard to work with.
The economics of making a separate package style for < 1% of the volume sales isn't there I'm afraid - breakout boards are one way to go for prototyping/DIY, some companies sell breakouts and populate them with part of your choice...
The economics of making a separate package style for < 1% of the volume sales isn't there I'm afraid - breakout boards are one way to go for prototyping/DIY, some companies sell breakouts and populate them with part of your choice...
But SMT is out of reach for many of us DIYers. I certainly could never handle SMDs any more. I could've done that some 30-40 years ago, but no more.
That's one idea, and I did think of this, but then who would solder that SMD on that adapter? That's the question.
Do you have any potential sources in mind for this?
Can you point somewhere?
Would they put an OPA1632 on one for a DIYer? Without breaking the bank of course.
Very fine-pitch and leadless SMDs are increasingly hampering the DIY efforts of many of us. Especially, those of us whose vision is no longer as sharp and whose hands are no longer as steady. This situation will become increasingly problematic as market forces push IC vendors to reduce chip package cost, and also support high PCB stuffing density to enable smaller form-factor end products. My worst fear is that those market forces will eventually render hobbyist DIY impractical, if not outright impossible.
I always picture the ideal solution as being that the PCB manufacturer places and solders the SMDs of my design to the board which they produced - and, here's the key - at pricing affordable by hobbyists in the single unit quantities typical of DIY. Another solution I can imagine would be a benchtop pick-and-place machine based on low cost ink jet printer technology. Perhaps, the print-head mechanism could spray some sort of room temperature UV activated conductive adhesive on to the PCB pads, like it were printing with ink. Then the print 'print-head' mechanism is reconfigured to place the SMDs on the board. Finally, maybe the user shines an UV lamp on the PCB to activate the conductive adhesive, gluing the SMDs in place. While probably not suitable for commercial production, it might be fine for single piece DIY projects. Oh, well, I can dream. If someone knows of such affordedable solutions that are already in available, please advise us.
In the meanwhile, some of our community are managing to do a very fine job of manually stuffing even leadless SMDs. XRK971 has posted some interesting details on how he does this, and has suggested a thread dedicated to discussing his methods. https://www.diyaudio.com/forums/analog-line-level/338607-balanced-output-single-4.html#post5820115
I always picture the ideal solution as being that the PCB manufacturer places and solders the SMDs of my design to the board which they produced - and, here's the key - at pricing affordable by hobbyists in the single unit quantities typical of DIY. Another solution I can imagine would be a benchtop pick-and-place machine based on low cost ink jet printer technology. Perhaps, the print-head mechanism could spray some sort of room temperature UV activated conductive adhesive on to the PCB pads, like it were printing with ink. Then the print 'print-head' mechanism is reconfigured to place the SMDs on the board. Finally, maybe the user shines an UV lamp on the PCB to activate the conductive adhesive, gluing the SMDs in place. While probably not suitable for commercial production, it might be fine for single piece DIY projects. Oh, well, I can dream. If someone knows of such affordedable solutions that are already in available, please advise us.
In the meanwhile, some of our community are managing to do a very fine job of manually stuffing even leadless SMDs. XRK971 has posted some interesting details on how he does this, and has suggested a thread dedicated to discussing his methods. https://www.diyaudio.com/forums/analog-line-level/338607-balanced-output-single-4.html#post5820115
Last edited: