I'm in the process of selecting components for a preamp that can double as a decent headphone amplifier.
The idea is that it should be able pass on an ultra-low-noise signal to a power amplifier, or directly drive even the fussiest of headphones if those are connected (I'll have a low/high gain switch to help here).
I'm keen to try an Opamp + Buffer type circuit, and have had great success in breadboard prototypes using the BUF634 chip. My choice is now about which opamp to use before the buffer.
In my stash I have the following bounty from Texas Instruments:
Given that I'll need...
...which of these four opamps would you choose, and why?
(I'll be driving it with +-15V, and have selectable gain of 1X, 5X and 10X).
The idea is that it should be able pass on an ultra-low-noise signal to a power amplifier, or directly drive even the fussiest of headphones if those are connected (I'll have a low/high gain switch to help here).
I'm keen to try an Opamp + Buffer type circuit, and have had great success in breadboard prototypes using the BUF634 chip. My choice is now about which opamp to use before the buffer.
In my stash I have the following bounty from Texas Instruments:
Given that I'll need...
- Low noise when acting as a preamp for a power amplifier
- High power when acting as a headphone amplifier
...which of these four opamps would you choose, and why?
(I'll be driving it with +-15V, and have selectable gain of 1X, 5X and 10X).
http://www.ti.com/lit/ds/symlink/lme49600.pdf
Ti suggest the lme49710
and figure 28 shows their suggestion for a headphone amplifier/driver.
Ti suggest the lme49710
and figure 28 shows their suggestion for a headphone amplifier/driver.
Last edited:
Thanks @Andrew! Yeah, LME49710 looks like a goodie, particularly as it seems purpose built for headphones.
That being said, I'm trying to do this build without spending of any more of my future children's inheritance money (sometimes I wish Mouser would mask their credit card statements with something less conspicuous, perhaps something like "STRIPPERS AND BEER").
Of my 4 choices (parts I currently own) which would be your pick, or what characteristics should I be looking at? They all seem low noise and highly efficient.
That being said, I'm trying to do this build without spending of any more of my future children's inheritance money (sometimes I wish Mouser would mask their credit card statements with something less conspicuous, perhaps something like "STRIPPERS AND BEER").
Of my 4 choices (parts I currently own) which would be your pick, or what characteristics should I be looking at? They all seem low noise and highly efficient.
Go for OPA1622, you'll get low noise, low distortion and you won't need a buffer ....and I still have some PCB left for free....🙂
Gentlebeings- Abza has those four chips in his stash, and is not inclined to spend precious money for other types.
IMO, the OPA1612 is fantastic for the stated purpose. Voltage hiss is really low. Current hiss is just low, but sources up to 10K are still low-low hiss (and in this application, probably lower than the hiss on the recording). Output current is ample to drive a buffer (the stated topology). Slew is far more than you need. Output swing is wide. There's a layout guide in the datasheet.
IMO, the OPA1612 is fantastic for the stated purpose. Voltage hiss is really low. Current hiss is just low, but sources up to 10K are still low-low hiss (and in this application, probably lower than the hiss on the recording). Output current is ample to drive a buffer (the stated topology). Slew is far more than you need. Output swing is wide. There's a layout guide in the datasheet.
OPA1612 is pretty well overkill here, and does run with a decent amount of input bias current. From a noise perspective, unless you're willing to get really down to small feedback resistances, the OPA2209 is going to be hard to beat (current noise on the super beta pair...)
Any of the last three put out similar current (2227 slightly less so) and stupidly low distortion. I'd probably save the OPA1612 for a more critical place, but why not use it if it's sitting there! OPA2209 would be my first choice though.
Any of the last three put out similar current (2227 slightly less so) and stupidly low distortion. I'd probably save the OPA1612 for a more critical place, but why not use it if it's sitting there! OPA2209 would be my first choice though.
besides opa2134, they will all do the job.
opa1612 is the best. 1.1nV/rHz, 1.7pA/rHz.
the noise is so low when Rs is 80 ohm,
the noise generated by Rs will be equal to 1.1nV/rHz.
it starts to affect the noise performance of opa1612,
same goes to Rg resistor if there's any.
opa1612 is the best. 1.1nV/rHz, 1.7pA/rHz.
the noise is so low when Rs is 80 ohm,
the noise generated by Rs will be equal to 1.1nV/rHz.
it starts to affect the noise performance of opa1612,
same goes to Rg resistor if there's any.
I think reasonable design will have to consider source device noise before deciding preamp stage. Because to push noise performance to extreme you'll end up with ridiculously low resistors in feedback which is heavy load, and brings other issues, but won't work according to your plan, bcs the source device's noise might be much higher.
Reasonable means the lowest noise preamp generate will be 1/3 -> same of source device noise.
Reasonable means the lowest noise preamp generate will be 1/3 -> same of source device noise.
Have a read of this article and the following 4 parts of development of their chip to suit headphone driving.
Amp up your cans: How much power do headphones need? (Part 1) - Precision Hub - Archives - TI E2E Community
Amp up your cans: How much power do headphones need? (Part 1) - Precision Hub - Archives - TI E2E Community
I have now read the 5 parts of the Ti advertising blub on developing the opa1622 to suit headphone amplifier duty.
Some raised opamp noise specification as a concern in choosing a headphone amplifier opamp.
Using some of the data from those papers.
Max Vout = 2.12Vac for 110dB SPL at the ear.
Unity gain would require a maximum input of 2.12Vac
opa1622 noise is given as 2.8nV/sqrtHz
For a unity gain arrangement, the opamp noise contribution at the output will be the same 2.8nV/sqrtHz
Summing this for the 20kHz audio band, we get the noise output due to the opamp alone to be 0.396 uVac
The maximum signal to opamp noise ratio will be 134.6dB
If we have a maximum output of 110dB, then the opamp noise contribution will be 134.6dB below that, i.e ~24.6dB below audibility.
Let's increase the maximum output to 120dB for those users who require exceptional peak transient signal handling capability and change the gain from unity to 3.16times to get the 2.12Vac signal upto 6.7Vac.
The opamp noise is now 2.8nV/sqrtHz * 3.16 * sqrt(20000) = 1.25uVac
The max signal to noise ratio is still 134.6dB
The output noise due to the opamp is now 14.6dB below audibility.
It seems that opamp noise for that particular combination of headphone impedance, sensitivity and required SPL is not a concern in choosing a suitable opap.
If you need convincing, then choose some other headphones and repeat the opamp noise predictions.
I am sure you will find that noise is not a problem.
I guess that is why noise never got considered by the design development team, even though the 2.8nV is much worse than the 1612's figure of 1.1nV
Some raised opamp noise specification as a concern in choosing a headphone amplifier opamp.
Using some of the data from those papers.
Max Vout = 2.12Vac for 110dB SPL at the ear.
Unity gain would require a maximum input of 2.12Vac
opa1622 noise is given as 2.8nV/sqrtHz
For a unity gain arrangement, the opamp noise contribution at the output will be the same 2.8nV/sqrtHz
Summing this for the 20kHz audio band, we get the noise output due to the opamp alone to be 0.396 uVac
The maximum signal to opamp noise ratio will be 134.6dB
If we have a maximum output of 110dB, then the opamp noise contribution will be 134.6dB below that, i.e ~24.6dB below audibility.
Let's increase the maximum output to 120dB for those users who require exceptional peak transient signal handling capability and change the gain from unity to 3.16times to get the 2.12Vac signal upto 6.7Vac.
The opamp noise is now 2.8nV/sqrtHz * 3.16 * sqrt(20000) = 1.25uVac
The max signal to noise ratio is still 134.6dB
The output noise due to the opamp is now 14.6dB below audibility.
It seems that opamp noise for that particular combination of headphone impedance, sensitivity and required SPL is not a concern in choosing a suitable opap.
If you need convincing, then choose some other headphones and repeat the opamp noise predictions.
I am sure you will find that noise is not a problem.
I guess that is why noise never got considered by the design development team, even though the 2.8nV is much worse than the 1612's figure of 1.1nV
Last edited:
Found it. I love this series Articles, super practical and packed with information.
Last edited:
The main drawback of the OPA1622 is that it's not hand-solderable. As a minimum, you'll need a solder paste stencil, solder paste, and a reflow system (which can be a toaster oven or frying pan) to solder the QFN/LLP package.
Tom
Tom
I managed to successfully solder more than ten OPA1622 using cheap chinese hot air station and solder paste. I only fried one IC, the first one🙂 I did not use solder paste stencil, just gently spread a bit of solder paste by a toothpick, position the IC as good as I can and the rest surface tension of the molten solder paste did by itself. The key is to use the right (small) amount of solder paste. If there are solder bridges between the pins, just apply a lot of soldering flux and drug hot soldering iron tip across the pins to clean up the excess solder. The way it looks after such soldering can be seen here.
Everything is possible with a bit of effort!
Regards,
Oleg
Everything is possible with a bit of effort!
Regards,
Oleg
You need to account for the feedback resistance/current noise, which, while isn't much with the modest values recommended in the DS, still adds on!
I too have hand soldered soic devices and smaller.
I have a cheap temp controlled iron, 63/37 solder in two sizes, 22swg and 24swg and some solder wick to remove any shorts between solder pads.
I have not fried any.
Thank you gents for the suggestions! That article from T.I. was also really interesting (cheers @AndrewT).
I've got it in mind to use a DIP-8 socket in the design so I can test them all out and see what sticks. Although to be honest I worry then about obsessively IC-swapping (desperately trying to ascertain almost imperceptible differences late into the night, while forgetting to eat and slowly wasting away).
However, given that this preamp will be my main port of call for future designs (a DAC and power amplifier, certainly) I'll probably end up using the OPA1612 - as @DPH states, it's just sitting there, may as well use it. Just finished the rest of the schematic, will build a chop-shop breadboard version before a protoboard one and see what happens.
I've got it in mind to use a DIP-8 socket in the design so I can test them all out and see what sticks. Although to be honest I worry then about obsessively IC-swapping (desperately trying to ascertain almost imperceptible differences late into the night, while forgetting to eat and slowly wasting away).
However, given that this preamp will be my main port of call for future designs (a DAC and power amplifier, certainly) I'll probably end up using the OPA1612 - as @DPH states, it's just sitting there, may as well use it. Just finished the rest of the schematic, will build a chop-shop breadboard version before a protoboard one and see what happens.
Right, but the current noise of all these devices changes by an order of magnitude, which adds geometrically to e_n. Anyhow, looks like the OP is sorted out.
+1 on your refutation of noise as a design problem.
Additionally, hiss is most a problem at the lowest-level parts of the chain. Microphone or pickup amplifier. For similar technology, hiss at later points is discounted by nearly the amount of gain before it. A 2mV mike signal is brought to 200mV out of preamp, to 2V after line-amp and power amp. Gain over 100 from the critical point to the power amp input. The power amp input would have to be awful crappy to approach 100X the hiss of the mike amp. By the time we get to the power amp, hiss should be a non-issue.
There can be exceptions. If the power amp has too much gain, so you must turn-down a Volume control way-far, power amp hiss can dominate. The cure is simple: don't put too much gain in the power amp.
Another exception is that there are a few VERY sensitive earphone drivers (derived from hearing aid systems). These really can make microVolt hiss audible. They also need VERY low levels to get LOUD. The simple answer (if not using button-batteries) is to pad-down signal from power amp to earphone. 100r series works. 47r+10r divider does, and damps the resonances better. "Wasteful" but these earphones are so very sensitive that the waste is negligible.
- Status
- Not open for further replies.