Op-amp output impedance

[mike7877]

A couple weeks ago I opened up my Topping DX1 because I was curious which op-amp was chosen. Turns out, it's a Chinese creation from a company I've never heard of: RICORE TECHNOLOGY LIMITED. It's the RT6863D

I was both surprised and not surprised. Surprised that THD+n could be 0.0002-0.0003% from 20-20k, 32 multi-tone at an average of -128dB, IMD could be -100 to -110dB between -20 and 0dB, , 16 ohms could be driven to over 2V/ 340mW at 0.0003% THD+n, balance PERFECT from -40 to 0dB for 90% of the time and never more than 1/4dB out, output impedance completely flat from 35-20kHz at 0.785 ohms, and 20Hz to 35Hz climbs from 0.750 to 0.785 ohms. 1kHz/48kHz on the RCA outs at -115/-117dB THD+n and -113/-114 on the 1/4" / 1/8" HP outputs. And not surprised (can't forget about the not surprised...) Not surprised that something better (like the 1612) wasn't chosen because of cost - For the DX1 to be able to sell to consumers for just $100 USD using a $10 DAC, aluminum case, excellent potentiometer, etc., isn't doable if you nearly double the price with another $25 for op-amps (it uses 3)

This is a picture of one of the three op-amps I discovered inside the DX1. Because a specialty screwdriver is needed to remove the board from the inside, this picture of the op-amp nearest the opening is the best possible (at the least severe angle). The package is 1.7mm x 1.5mm x 0.3mm, so my photography is not as bad as it looks lol.

There was one op-amp by the RCA jacks, and two by the 1/4" / 1/8" jacks.

I theorized that because the packages are so small and no heatsinks are present, and that power output for the headphone section requiring up to 0.13A per channel, Topping chose to use two op-amps for ease of implementation (vs. adding/securing a heatsink on a 2.5mm2 chip surrounded by SMD components taller than it)

After reading the DX1 specs and noticing the RCA output impedance is different than the headphone output impedance (22 ohms vs 0.5 ohms), I'm starting to think that maybe I was wrong...

Below are the clippings of all the useful and applicable information in the datasheet. Some stuff is missing (like deg C heating per watt) which I find odd, but it's an obscure Chinese company, so...

Does anyone know why the output impedance is so vastly different between the headphone and RCA jacks? Did Topping just put a resistor on the RCA out op-amp, or?

As you can see, the AK4493S has a THD+n of -115dB, -2dB better than the 4493, but it's still not so much better than the non-S version that the surprisingly good cheap Chinese op-amp is a bad choice.

 

[fpitas]

the RCA output impedance is different than the headphone output impedance (22 ohms vs 0.5 ohms)

You should be aware that a series resistor is often inserted when driving an unterminated line like an RCA cable, to eliminate parasitic oscillations.

 

[sergeauckland]

Most likely that's the difference, a simple resistor in the line outputs. The reason is to protect the opamp from a high capacitance load such as long cables. The headphone output needs a very low output impedance, and anyway headphones are more inductive than capacitative, so less needing the protection.

Edit, fpitas beat me to it.

S

 

[mike7877]

You should be aware that a series resistor is often inserted when driving an unterminated line like an RCA cable, to eliminate parasitic oscillations.

Ah, one of my suspicions... in the sentence preceding the last image of the OP.
Good to know.

Do you think they used two op-amps on the headphone output for the reason I suspect as well? (since 2.5mm2 package shorter than SMD devices surrounding it and 0.13A output current, for heat dissipation)

 

[mike7877]

Most likely that's the difference, a simple resistor in the line outputs. The reason is to protect the opamp from a high capacitance load such as long cables. The headphone output needs a very low output impedance, and anyway headphones are more inductive than capacitative, so less needing the protection.

Edit, fpitas beat me to it.

Good to know. It's unlike Topping to over spec their stuff, and the 0.5 ohms they list as the output impedance is not the 0.785 that L7audiolab measured. It's still below 1 ohm, and not very different... still, I think it's the first spec of theirs that I've seen so off (+57%)

 

[solderdude]

Do you think they used two op-amps on the headphone output for the reason I suspect as well? (since 2.5mm2 package shorter than SMD devices surrounding it and 0.13A output current, for heat dissipation)

Doubling opamps usually is done to increase current output as opamps are always current limited.
When this is done output resistors need to be added so that DC offsets of 2 opamps don't cause high currents from one output to the other output. 10mV is already enough with a very low output resistance.
Or.... they used 2 in bridge mode (balanced drive only) in which case the output voltage is doubled but not the output current.

 

[MRC01]

Doubling opamps usually is done to increase current output as opamps are always current limited.
...

I've seen devices that use BUF634 for higher current output. Is that what you could call "doubling opamps", or is it a different circuit?

 

[solderdude]

I meant more like the Schiit Heresy, O2 etc, that uses the same principle like the circuit below (A2 and A3)

The BUF634 is just a couple of power transistors (gain is only 1x) that can be integrated in a feedback loop with an opamp.
It is a buffer and not an opamp.

 

[AnalogSteph]

In the world of inexpensive ESS-powered DACs, these RT6863D opamps are actually very common (even sold together as a bundle if memory serves). For example, you can find them on the Khadas Tone Board or lower-end Fiio dongle DACs. There's obviously no reason why you couldn't use them with a classic AKM, on the contrary you wouldn't have any issues trying to deal with the "ESS hump".

It's a good thing you can find a datasheet nowadays. Most conspicuously absent is a spec for input bias current or input current noise performance. Given the fairly low input voltage noise density, I rather assume you're best off sticking to rather low-impedance circuitry.

BTW, an opamp's output impedance is open-loop output impedance (which is generally low, <30 ohms) divided by loop gain.

 

[mike7877]

Doubling opamps usually is done to increase current output as opamps are always current limited.
When this is done output resistors need to be added so that DC offsets of 2 opamps don't cause high currents from one output to the other output. 10mV is already enough with a very low output resistance.
Or.... they used 2 in bridge mode (balanced drive only) in which case the output voltage is doubled but not the output current.

It seems impossible that there would be a 0.78 ohm output impedance with opamps doubled that way, at least to any usable voltage

I think a picture might be required to say for sure what's going on - I thought deduction was possible... Unfortunately a Topping only screwdriver (or enclosure destruction) are the only two ways for me to get a picture ar an angle of more than 20 degrees, and the 2 HP op-amps are furthest in from the opening, making them the most obscured

The op-amp operation voltage is given as up to +-17V, and I think the DX1 is 11V p-p, and the DX1 isn't balanced.

The chip is rated for "more than" 100mA, and the DX1 puts about 130mA out before distortion into impedances 30 ohms and less (11V p-p), which suggests it's operating mormally.

I wonder if they chose to run them bridged... the capability isn't on the datasheet, but it's missing a few things so it could be one of them... Performance difference between RCA and HP is 2-3dB

Tere are two headphone outputs, 1/4 and 1/8 - it'd be bizarre if they were individually amp'd!

If some op-amps have worsening stereo separation as their output current increases? If so, I could see why they might opt to use one chip for RCA, and two for HP

 

[DVDdoug]

It seems impossible that there would be a 0.78 ohm output impedance with opamps doubled that way, at least to any usable voltage

Low impedance doesn't mean it can drive an equal-impedance load. We don't "match impedance" It can still have current limits and it can overheat. Power amps (for speakers) are similar. They are designed to drive 4 or 8-Ohm loads but they have output impedance of a fraction of an Ohm.

I wonder if they chose to run them bridged...

You can't have a bridge output with regular unbalanced headphones because you can't have a common left & right ground. (there is no ground with balanced connections.)

 

[solderdude]

The chip is rated for "more than" 100mA, and the DX1 puts about 130mA

You would have to take in mind that if a chip is rated for 100mA the maximum current output (peak) is around that value.
More often than not there is an asymmetry for the negative and positive side of the output as well.
It also depends a bit on the temperature of the chip and on the output voltage vs current.

When you measure a sine-wave you measure the RMS value and using a load resistor you determine the RMS current.
But the opamp limits at the peak currents.
So 130mA x (SQRT 2) = 184mA = very close to double that of the max. current of the opamp.
So most likely there are 2 opamps in parallel that must have extra output resistors.

You can also do this trick in the feedback loop but this can present some stability problems.

and another trick is the master-slave option:

 

[egellings]

A couple weeks ago I opened up my Topping DX1 because I was curious which op-amp was chosen. Turns out, it's a Chinese creation from a company I've never heard of: RICORE TECHNOLOGY LIMITED. It's the RT6863D

I was both surprised and not surprised. Surprised that THD+n could be 0.0002-0.0003% from 20-20k, 32 multi-tone at an average of -128dB, IMD could be -100 to -110dB between -20 and 0dB, , 16 ohms could be driven to over 2V/ 340mW at 0.0003% THD+n, balance PERFECT from -40 to 0dB for 90% of the time and never more than 1/4dB out, output impedance completely flat from 35-20kHz at 0.785 ohms, and 20Hz to 35Hz climbs from 0.750 to 0.785 ohms. 1kHz/48kHz on the RCA outs at -115/-117dB THD+n and -113/-114 on the 1/4" / 1/8" HP outputs. And not surprised (can't forget about the not surprised...) Not surprised that something better (like the 1612) wasn't chosen because of cost - For the DX1 to be able to sell to consumers for just $100 USD using a $10 DAC, aluminum case, excellent potentiometer, etc., isn't doable if you nearly double the price with another $25 for op-amps (it uses 3)

View attachment 322454

This is a picture of one of the three op-amps I discovered inside the DX1. Because a specialty screwdriver is needed to remove the board from the inside, this picture of the op-amp nearest the opening is the best possible (at the least severe angle). The package is 1.7mm x 1.5mm x 0.3mm, so my photography is not as bad as it looks lol.

There was one op-amp by the RCA jacks, and two by the 1/4" / 1/8" jacks.

I theorized that because the packages are so small and no heatsinks are present, and that power output for the headphone section requiring up to 0.13A per channel, Topping chose to use two op-amps for ease of implementation (vs. adding/securing a heatsink on a 2.5mm2 chip surrounded by SMD components taller than it)

After reading the DX1 specs and noticing the RCA output impedance is different than the headphone output impedance (22 ohms vs 0.5 ohms), I'm starting to think that maybe I was wrong...

Below are the clippings of all the useful and applicable information in the datasheet. Some stuff is missing (like deg C heating per watt) which I find odd, but it's an obscure Chinese company, so...


View attachment 322463
View attachment 322464
View attachment 322465
View attachment 322470

Does anyone know why the output impedance is so vastly different between the headphone and RCA jacks? Did Topping just put a resistor on the RCA out op-amp, or?

View attachment 322493

As you can see, the AK4493S has a THD+n of -115dB, -2dB better than the 4493, but it's still not so much better than the non-S version that the surprisingly good cheap Chinese op-amp is a bad choice.

Thing is that both are so preposterously good that you'd never hear a difference between them. So except for laboratory use, both are vastly better than needed for home audio, which has been a solved problem for some time now.

 

[mike7877]

You would have to take in mind that if a chip is rated for 100mA the maximum current output (peak) is around that value.
More often than not there is an asymmetry for the negative and positive side of the output as well.
It also depends a bit on the temperature of the chip and on the output voltage vs current.

When you measure a sine-wave you measure the RMS value and using a load resistor you determine the RMS current.
But the opamp limits at the peak currents.
So 130mA x (SQRT 2) = 184mA = very close to double that of the max. current of the opamp.
So most likely there are 2 opamps in parallel that must have extra output resistors.

You can also do this trick in the feedback loop but this can present some stability problems.

and another trick is the master-slave option:

Since there are only 2 it seems it'd have to be the master slave... but the output impedance is just 0.785 ohms - is it possible with such a low output impedance?
edit: and if hooked up that way, wouldn't noise increase by 3dB? The DX1's measurements matches the op-amp datasheet. Topping does do good things though - maybe they're able to get 3dB better than datasheet performance

The "more than 100mA" rating is for AC (afaik..) the ~1.4x peak current is taken into account for it

 

[mike7877]

Thing is that both are so preposterously good that you'd never hear a difference between them. So except for laboratory use, both are vastly better than needed for home audio, which has been a solved problem for some time now.

Why do so many people have to state "can't hear the difference", "solved problem" whenever well spec'd stuff is being discussed? It's redundant and annoying. If your attitude is that home audio is a "solved problem", why read or write about it? It didn't add to the conversation, right?

I do agree with you that this op-amp + DAC combo is very transparent. It's not perfect though. A lot of people say -100 is good enough. "solved problem" etc. etc. But that's at full power and being considered on its own! Put it in series with the other components of an entire audio system, say 7 other steps, all "good enough" (-100) as well.. Then use a couple of them (pre and power amps) at less than full power, add some efficient speakers to the mix, treat the room properly and use some good source material with a high dynamic range, and "good enough" isn't good enough anymore (that perfect -100 ain't near-100 anymore...)

 

[mike7877]

Low impedance doesn't mean it can drive an equal-impedance load. We don't "match impedance" It can still have current limits and it can overheat. Power amps (for speakers) are similar. They are designed to drive 4 or 8-Ohm loads but they have output impedance of a fraction of an Ohm.


You can't have a bridge output with regular unbalanced headphones because you can't have a common left & right ground. (there is no ground with balanced connections.)

I didn't think it could... what I meant was the outputs of the op amps themselves would have to be like 0.005 ohms for their combined total with resistors on their outputs to be only 0.785 ohms. And if that was the case, since 100mA is their limit, voltage on the output couldn't be as high as it is, and loads in the hundreds of ohms would be impossible to drive to any significance


edit: forgot to respond to the bottom - I meant each op-amp bridged (its 2 channels to 1), and one op-amp for the L headphone, the other for R

 

[solderdude]

the output impedance is just 0.785 ohms - is it possible with such a low output impedance?

2 resistors of 1.5ohm each would give such an output R.
The resistors can be that low when the DC offset is low enough.
The O2 even uses 1ohm resistors.

Since there are only 2 it seems it'd have to be the master slave.

Can still be any of the topologies I mentioned.

 

[egellings]

Why do so many people have to state "can't hear the difference", "solved problem" whenever well spec'd stuff is being discussed? It's redundant and annoying. If your attitude is that home audio is a "solved problem", why read or write about it? It didn't add to the conversation, right?

I do agree with you that this op-amp + DAC combo is very transparent. It's not perfect though. A lot of people say -100 is good enough. "solved problem" etc. etc. But that's at full power and being considered on its own! Put it in series with the other components of an entire audio system, say 7 other steps, all "good enough" (-100) as well.. Then use a couple of them (pre and power amps) at less than full power, add some efficient speakers to the mix, treat the room properly and use some good source material with a high dynamic range, and "good enough" isn't good enough anymore (that perfect -100 ain't near-100 anymore...)

What's annoying are claims of audible differences when deviations are so minute that the human ear cannot possibly detect them, no matter the gold content of the ears. Once it's good enough, getting better doesn't get you anything. Why order a drop forge when all you need is a tack hammer? For doing scientific research, then I can see where ultra-low distortion and DC to light frequency response could be desirable. For living room sound enjoyment, not so much. As for adding to the conversation, this reply is a legitimate one. If you believe in your heart of hearts that you really need such over-the-top specifications, then go for it. It will most certainly not cause any harm whatsoever, except for, perhaps, the wallet. Needless over-specification is misguided.

 

[mike7877]

2 resistors of 1.5ohm each would give such an output R.
The resistors can be as low when the DC offset is low enough.


Can be any of the topologies.

Ohhh, ok.

Do either of them result in signal degradation? The THD+n is 2 or 3dB worse on the HP out vs the RCA

 

[mike7877]

What's annoying are claims of audible differences when deviations are so minute that the human ear cannot possibly detect them, no matter the gold content of the ears. Once it's good enough, getting better doesn't get you anything. Why order a drop forge when all you need is a tack hammer? For doing scientific research, then I can see where ultra-low distortion and DC to light frequency response could be desirable. For living room sound enjoyment, not so much. As for adding to the conversation, this reply is a legitimate one. If you believe in your heart of hearts that you really need such over-the-top specifications, then go for it. It will most certainly not cause any harm whatsoever, except for, perhaps, the wallet. Needless over-specification is misguided.

I can't count the pixels on my 1080p S22 or my 1440p S9, but I can see the difference the extra pixels make on the sharpness of text. Food for thought