Well one piece of advice I would give would be to match the source impedances at both op amp inputs because the input bias current on the OPA1622 is high (no Ib cancellation). So for a source impedance of 470 ohms on the non-inverting input, feedback resistors of 2800 ohms and 562 ohms, will get you pretty darn close to a gain of 6 (5.98) and a source impedance of about 468 ohms at the inverting input. This will minimize the offset error due to input bias current at the output.
As far as stability is concerned, that really depends on the load you are intending to drive with the OPA1622. There is a pole/zero pair in the open loop gain curve of the OPA1622 which maximizes loop gain at audio frequencies, but the net result is that at the loop closure point (loop gain is zero) the op amp is actually more stable in a gain of +1 than a gain of +6. For that reason, you could put a feedback capacitor across the 2800 ohm feedback resistor to reduce the gain of the circuit to unity at high frequency: 120pF or 150pF would be fine.
John,
Though this is bit off topic , I have a question.
TI's former top headphone driver TPA6120A2 which is gotten good evaluation on a lot of hi-fi communities, but on the datasheet, it shows higher distortion at low frequency than high when driving heavy load.
I experienced similar distortion on the other CFB opamp. I guess it is due to low DC open loop gain of CFB amp and high thermal distortion also related to CFB architecture.
What is the cause in reality?
Though this is bit off topic , I have a question.
TI's former top headphone driver TPA6120A2 which is gotten good evaluation on a lot of hi-fi communities, but on the datasheet, it shows higher distortion at low frequency than high when driving heavy load.
I experienced similar distortion on the other CFB opamp. I guess it is due to low DC open loop gain of CFB amp and high thermal distortion also related to CFB architecture.
What is the cause in reality?
Hi John,
In the datasheet plots it looks like the THD in inverting mode is covered up by the N. Is it safe to assume this part performs better in inverting mode like most op-amps?
Also on a related note - do you have any info on the common-mode input distortion of this op-amp? It looks to have been carefully addressed based on the performance at G=+1 but this is something that frequently appears with other op-amps in distortion measurements outside of the datasheet when used as a follower.
Thanks!
Looking at the TPA6120A2 datasheet, the increase in distortion at low frequencies seems to be a thermal effect. This would explain why it is only apparent on the high supply voltage case for the 2kOhm graph.
In a voltage feedback amplifier, symmetrical heating of the two input transistors prevents one cause of thermal distortion but the asymmetrical nature of the input stage in a current feedback amplifier prevents a similar cancellation from taking place. So those amplifiers are much more susceptible to thermal feedback effects. Note that these effects only occur at low frequency because the thermal time constants of the die are shorter than the period of the signal.
The TPA6120A2 comes from a different group within TI and I've never personally used it or looked into its topology.
There are a few sources of the classic common-mode input distortion:
1. Variation in the tail current with common-mode voltage. Really only an issue on older parts with low Early voltage processes.
2. Input impedance variation with common mode voltage. This is an issue on all op amps. In BJT-input amplifiers like the OPA1622 it comes from: beta variation in the input transistors with changing Vce, variation in the collector/base capacitance of the input transistors, variation in the junction capacitance of the ESD diodes, non-linear leakage currents in protection diodes, etc. Some of these can be addressed by cascoding the input devices to prevent variations in their Vce however the cascode will limit the input voltage range.
So far, the best amplifier's I've tested for input impedance variation are the OPA1602 (BJT-input) and OPA1642/OPA827 (tie, JFET-input).
The OPA1622 performance was comparable to the LM4562 with a 10k source impedance, G=+1 and 5Vrms input signal. Slightly worse than the OPA1602.
On the other hand, the inverting configurations have comparably low input impedance, reduces loop gain, increases noise gain, and increases the loading on the output stage. Therefore, it really depends on the application when deciding which topology will give better performance. For non-inverting circuits, with large signals and low gain, I recommend matching the impedances at both op amp inputs to cancel common-mode distortion due to input impedance variation.
Shinja,
TPA6120A2 was probably TI Dallas, not TI Tuscon (Burr Brown).
This member would know for sure, but doesn't post here often.
diyAudio - View Profile: ddapkus
There's a possibility the people I work for may have built some
of those TPA EVMs. If so, perhaps I can identify and contact the
relevant engineer for you. Can't do anything about it till Monday.
TPA6120A2 was probably TI Dallas, not TI Tuscon (Burr Brown).
This member would know for sure, but doesn't post here often.
diyAudio - View Profile: ddapkus
There's a possibility the people I work for may have built some
of those TPA EVMs. If so, perhaps I can identify and contact the
relevant engineer for you. Can't do anything about it till Monday.
Shinja,
We have built several EVMs where TPA6120A2 was included,
but not the EVM specifically intended for evaluating that part.
I may yet be able to track down a contact for you, but going
to TI's E2E forum and re-ask there would probably be faster.
--------
johnc,
After reading into OPA1622, particularly Page 11 of the spec,
I'm more impressed than ever. Good misbehavior when asked
to input or output the slightly impossible. Not a trivial feature!
Overkill for the personal projects I am building right this minute,
(guitar to PC), but tempted to request samples none the less...
We have built several EVMs where TPA6120A2 was included,
but not the EVM specifically intended for evaluating that part.
I may yet be able to track down a contact for you, but going
to TI's E2E forum and re-ask there would probably be faster.
--------
johnc,
After reading into OPA1622, particularly Page 11 of the spec,
I'm more impressed than ever. Good misbehavior when asked
to input or output the slightly impossible. Not a trivial feature!
Overkill for the personal projects I am building right this minute,
(guitar to PC), but tempted to request samples none the less...
Last edited:
Johnc124 when TI marketing emailed me the their blurb "Burr-Brown Audio™ line expands: Highest-performance audio op amp" announcing OPS1622 it naturally caught my attention. Than I saw this:
Oh, be still my beating heart. When laying out PCBs I let my obsessive tendencies out and get manic about parts and signal layout. Seeing those tiny current loops allowed by the VSON-10 μPackage and signal pin out takes me to my Happy Place.
The only glitch is the conflict I have between the tight current loops SMD passive components allow vs. the better secondary characteristics one can sometimes have with thru-hole parts. Specifically ceramic SMD caps vs. plastic film DAs and DFs. Care to comment on those trade-offs when a design has space to allow the choice and preferred SMD types when only they will fit? Or perhaps your experience is the tight current loops SMD parts allow outweighs advantages thru-holes parts?
An externally hosted image should be here but it was not working when we last tested it.
Oh, be still my beating heart. When laying out PCBs I let my obsessive tendencies out and get manic about parts and signal layout. Seeing those tiny current loops allowed by the VSON-10 μPackage and signal pin out takes me to my Happy Place.
The only glitch is the conflict I have between the tight current loops SMD passive components allow vs. the better secondary characteristics one can sometimes have with thru-hole parts. Specifically ceramic SMD caps vs. plastic film DAs and DFs. Care to comment on those trade-offs when a design has space to allow the choice and preferred SMD types when only they will fit? Or perhaps your experience is the tight current loops SMD parts allow outweighs advantages thru-holes parts?
Thank you for your comment!
In terms of resistors, small SMD variants pretty much have to be thin-film types. Thick film SMD resistors suffer from poor voltage coefficient characteristics as the electric field intensity across their resistive element increases (smaller package --> higher field). At low frequencies and low resistance values, thick film resistors can also show thermal distortion effects. But I usually see this in larger power amplifiers, not small signal circuits. With thin-film types, I haven't had any issues with non-linearity and thankfully the cost of thin-film surface mount resistors is decreases. For reference, the measurements in the application section of the OPA1622 datasheet were performed on a circuit built with 0603 thin-films. I haven't personally tested 0402 or smaller, but several customers have been able to achieve very high performance with passives of that size.
For SMD capacitors, its C0G/NP0 types all the way of course, however there are some interesting new SMD film types being offered on the market. The voltage coefficient of the high-k ceramics (X7R, Y5V, Z5U, etc.) make them unusable in the signal path for Hi-Fi audio. Not to mention the microphonic effects in high-k ceramics can make for some interesting debugging... The circuit in the application section of the OPA1622 used 0603 C0G ceramic capacitors. I also tested film types, but couldn't see any measurable difference in performance on my APx555.
Thanks! We were really excited with the behavior we achieved going into and coming out of saturation. Very smooth, no latching, minimal rail sticking, and best of all no burst oscillations.
Request samples! You'll find a use for them someday
I was just looking through the datasheet and found an inconsistency with rail voltage labeling. Specifically on the suggested layout on p.22, the thermal pad copper is marked Vee, and I started looking for Vcc until I noticed the power input pins are marked V+ and V-. Most of the datasheet uses V+ and V-, but some places use Vcc and Vee.
We can usually figure these things out, but it would be easier and faster for everyone if the whole datasheet (and even the whole TI catalog going forward) would stick to one or the other nomenclature.
We can usually figure these things out, but it would be easier and faster for everyone if the whole datasheet (and even the whole TI catalog going forward) would stick to one or the other nomenclature.
Thanks for pointing this out, I'll correct this when I do some datasheet revisions later this year.
You might be thinking of Jim Solomon's 1974 paper The Monolithic Operational Amplifier: A Tutorial Study which goes into great detail about thermal feedback from output to input. National Semi made it into an application article and TI now has it online as Literature Number SNOA737. That's the Google search string to use.
Don't be misled by the for-profit corporation slapping a couple of part numbers on the title page. This was originally a scholarly paper published in a refereed journal, namely the IEEE Journal of Solid State Circuits, Volume 9, Issue 6 (Dec 1974), pp. 314-332. A journal in which I and quite a few other DIYA members have published; their editors are extremely serious about referee reviews.
This is a really great resource, thank you for posting it. It's also important to consider that thermal feedback from the output transistors to the input transistors is not isolated to an individual amplifier channel. Or to put it another way: the output transistors of one amplifier may asymmetrically heat the input transistors of an adjacent amplifier in a monolithic dual amplifier. Looking at the pinout of the OPA1622 it should be fairly obvious that the output stages and input stages of both channels reside along the same line of thermal symmetry.
We really need an enterprising manufacturer to make up some boards about 20mm square with this IC already mounted and SMD pads for reasonably sized passives say 0805 or larger, generally following the layout in the datasheet. Or, TI could make a few thousand in SOIC-10 or SOIC-16. How about it Johnc?
BrianL, I can be a lot more unrealistic than that! I may be wrong but I doubt that the tiny DRC package, with 0.5mm contact pitch and minute thermal mass, could be reliably hand-soldered without damaging it, especially as the thermal pad must be connected to the most negative supply voltage. Are we going to hand solder that too? How hot is the silicon going to get while we are soldering it? TI would not approve I'm sure.
But my current headphone amp design is coming along nicely, based on the veteran TPA6111A2. It will serve my purpose quite well.
But my current headphone amp design is coming along nicely, based on the veteran TPA6111A2. It will serve my purpose quite well.