Hi all,
I'm designing a preamp around the PGA2311 and I'm using the DRV134 and INA134 chips to facilitate balanced I/O. I want to maintain the largest signal swing I can within the PGA to keep the THD low but the added 6db of gain from the DRV134 is going to cause clipping with the low +/-5vDC rails I'm running the audio section from. I need to pad down the output of the PGA before entering the DRV to maintain unity gain. Since the front end of the DRV is an inverting opamp with 30k resistors can I just put an additional 30k in series to reduce the first stage gain to 1/2? The data sheet says to drive from a low source impedance but doesn't specify what parameters would suffer. If this is unwise I'll use a small resistive divider instead. Just wondering someone else has been here before since I don't have good measurement equipment.
Brian
I'm designing a preamp around the PGA2311 and I'm using the DRV134 and INA134 chips to facilitate balanced I/O. I want to maintain the largest signal swing I can within the PGA to keep the THD low but the added 6db of gain from the DRV134 is going to cause clipping with the low +/-5vDC rails I'm running the audio section from. I need to pad down the output of the PGA before entering the DRV to maintain unity gain. Since the front end of the DRV is an inverting opamp with 30k resistors can I just put an additional 30k in series to reduce the first stage gain to 1/2? The data sheet says to drive from a low source impedance but doesn't specify what parameters would suffer. If this is unwise I'll use a small resistive divider instead. Just wondering someone else has been here before since I don't have good measurement equipment.
Brian
Last edited:
Many of your answers can be found here. http://www.ti.com/lit/ds/symlink/drv134.pdf
Look at figure 4. This should tell you that you cannot use a voltage divider on the input! This is a precision balanced circuit that must be driven from a low impedance source. Any resistor or higher impedance will ruin the performance of the chip. This should be obvious.
And what is the input impedance? Once again, the datasheet tells us that it is 10K nominal. So even a 100 ohm source impedance introduces a 1% error - oops! Can't do that.
As far as running it off a low voltage like you mention, I suspect that 10 volts is too low. This is an old generation device. There are a lot of chips with rail to rail capability that will work great with 10 volts. I don't know if there is a modern chip to replace the DRV134 though.
I think the DRV134 is a great chip - lots of heavy lifting for cheap and simple. It does have limitations though.
I never tried to run a DRV134 on anything under 18 volts. I did try to run a 5532 on 9 volts though. It worked but its performance was unacceptable. On the other hand, some modern chips claim stellar performance on as little as 3 volts. I haven't used these devices at all, but they're in a lot of portable consumer products and some of these products are pretty good.
Look at figure 4. This should tell you that you cannot use a voltage divider on the input! This is a precision balanced circuit that must be driven from a low impedance source. Any resistor or higher impedance will ruin the performance of the chip. This should be obvious.
And what is the input impedance? Once again, the datasheet tells us that it is 10K nominal. So even a 100 ohm source impedance introduces a 1% error - oops! Can't do that.
As far as running it off a low voltage like you mention, I suspect that 10 volts is too low. This is an old generation device. There are a lot of chips with rail to rail capability that will work great with 10 volts. I don't know if there is a modern chip to replace the DRV134 though.
I think the DRV134 is a great chip - lots of heavy lifting for cheap and simple. It does have limitations though.
I never tried to run a DRV134 on anything under 18 volts. I did try to run a 5532 on 9 volts though. It worked but its performance was unacceptable. On the other hand, some modern chips claim stellar performance on as little as 3 volts. I haven't used these devices at all, but they're in a lot of portable consumer products and some of these products are pretty good.
Yikes. Not sure what happened with that thread title! If a mod can change it to "Reducing gain in a DRV134" that would great.
Slow down Fast Eddy, isnt figure 4 the THD corresponding to output load? It's labeled INA137 output. I didn't see any obvious implication of a higher source Z on the pin 4 of the DRV. Now that I think about it more I guess the input Z affects more than just the inverting stage at the front end. I guess I need a new plan. By the way, I had the datasheet in front of me when I posted, I was clear about that. I still wanted input. This should be obvious 🙂 thank you for replying.
Unfortunately it does seem that there are few "modern" alternatives. In fact, the DRV134 is the only one that is specified down to +/-5vDC that I can find.
Slow down Fast Eddy, isnt figure 4 the THD corresponding to output load? It's labeled INA137 output. I didn't see any obvious implication of a higher source Z on the pin 4 of the DRV. Now that I think about it more I guess the input Z affects more than just the inverting stage at the front end. I guess I need a new plan. By the way, I had the datasheet in front of me when I posted, I was clear about that. I still wanted input. This should be obvious 🙂 thank you for replying.
Unfortunately it does seem that there are few "modern" alternatives. In fact, the DRV134 is the only one that is specified down to +/-5vDC that I can find.
I'm talking about the "simplified schematic" (block diagram) on the very first page.
Input impedance of an inverting amplifier directly affects the gain, unlike a non-inverting amplifier. The input impedance is in series with the input resistor.
There is nothing wrong with an inverting amplifier. But in audio it must be driven by a low impedance source. This impedance must stay low through the audio band and often far beyond it to ensure stability.
Yes and especially in this circuit.Those resistors in the block diagram are laser trimmed to within 0.1%. This is what makes the chip's performance so awesome. Now if you put an impedance in series with one of those 0.1% resistors, you know what happens to the precision of the chip. The output won't be nearly so balanced.
Yes and you can have your cake and eat it too if you put a non-inverting buffer on the input of the chip. Put your attenuation circuit on the input of the buffer. Keep the values as low as practical or else your buffer will be a low pass filter.😉
No mass produced products would use a "modern" alternative. Only pro gear uses chips like this. They want to develop devices that manufacturers will purchase in lots of 1000+, not for people like us that might buy 10.
I didn't see the voltage spec but there you go. The chip should provide adequate performance for you, if you're careful about gain structure (which you obviously are).
Except it hasn't changed in the index, maybe my bumping it to the top will do the trick.
Nope,ok. You can get the attention of the moderators by reporting your own post
Nope,ok. You can get the attention of the moderators by reporting your own post
OK, we're on the same page. I was looking at the simplified schematic as well. Since the input is an inverting stage I figured that any additional resistance on the input would simply lower that gain of the first stage. If we were talking about the output of the DRV I certainly wouldn't have asked since clearly anything I add will be less matched than the laser trimmed resistors inside and mess with CMRR. For what it's worth I simulated the simplified schematic and a 10k added to the input reduced the output by 1/2 and didn't mess with the amplitude matching of the two outputs.... whether or not the real innards of the DRV react the same way who knows... or that I'm not messing with the impedance balance in some way that I can't see in sim I don't know.
Maybe your simulation treats the DRV134 as a "black box." This would not highlight the error introduced by the input impedance. I still think that it's the right idea to drive the DRV134 with a low impedance source and let it do what it does best.
I still suspect that any significant source impedance will ruin the precision and ultra low distortion characteristics of the DRV134.
If you are driving it with a known low impedance source, then you can alter your gain structure elsewhere in the circuit. If you intend to drive the DRV134 directly from an external source (CD player etc) then I strongly recommend adding a buffer at the input. This will ensure consistent performance with various sources.
I still suspect that any significant source impedance will ruin the precision and ultra low distortion characteristics of the DRV134.
If you are driving it with a known low impedance source, then you can alter your gain structure elsewhere in the circuit. If you intend to drive the DRV134 directly from an external source (CD player etc) then I strongly recommend adding a buffer at the input. This will ensure consistent performance with various sources.
Yes you can put 30k ohm before input to reduce output gain, but the 30k ohm will be source resistance for the op amp inside DRV134, which means johnston noise.
I simulated it,
0 ohm source resistance , 11uV 20khz, not too great.
30k ohm source resistance , 169uV 20khz, well, say goodbye to HIFI.
maybe try OPA1632?
I simulated it,
0 ohm source resistance , 11uV 20khz, not too great.
30k ohm source resistance , 169uV 20khz, well, say goodbye to HIFI.
maybe try OPA1632?
The buffer makes all good things happen.
This chip was intended to be driven by a low impedance source. It most certainly was not intended to be driven by an unbuffered volume control or level matching network.
Use of a non-inverting input buffer will allow the adjustment of gain structure without introducing errors or excessive noise. Use of a non-inverting input buffer will allow the DRV134 to perform to specification.
Don't fight it.
This chip was intended to be driven by a low impedance source. It most certainly was not intended to be driven by an unbuffered volume control or level matching network.
Use of a non-inverting input buffer will allow the adjustment of gain structure without introducing errors or excessive noise. Use of a non-inverting input buffer will allow the DRV134 to perform to specification.
Don't fight it.
So this little problem has sent me down the rabbit hole of "Fully Differential Amplifiers" - a device I had not yet heard of. It seems there is a robust market of devices from many manufacturers with extraordinary specs for a low cost.... they just happen to be tiny SMT devices. Some are in 8 pin SOIC which I can manage but the bulk seem to be MSOP and only 3mm wide.
This whole project was about putting a balanced front/back end on a PGA2311 volume control chip. I've since learned that somepeople use 2 chips and attenuate the four signal in a dual differential mode rather than unbalance > attenuate > balance. The benfit is cancellation of even-order harmonic distortion. I'm now looking into doing just that with FDA chips before and after the PGA... looking at the LMP8350.
Brian
Is this simulation correct?
If a 1k0 perfect resistor has 4.1nV/rtHz of Johnson noise, then a 30k has sqrt(30) or 5.48times more, i.e. 21.9nV/rtHz
over the 20kHz audio bandwith that accounts for 3.1uV of JN compared to 0.6uV for the 1k0
That's only an extra Ein of ~2.5uV contributed by increasing the 1k0 to 30k
If instead we compare Ein of 30k to 10r we get 3.1uV : 0.058uV an extra 3uV Ein of JN due to increasing the source resistance from 10r to 30k
Last edited:
After seeing your calculation, I would say the simulation is wrong.
could be 10uV (-98dBu noise floor , 70nV/rHz) + 3uV (30k ohm) *2
so 16uV in ideal situation.
I feel so silly
but actually I'm not sure about this, after testing other models. it looks like the simulation should be right. I try to do hand calculation on this, but datasheet doesn't have enough information on it.
The number is so off, could be caused by current noise which isn't in datasheet, so we might never know.
The number is so off, could be caused by current noise which isn't in datasheet, so we might never know.
- Status
- Not open for further replies.