Calcs for op amp loading vs THD?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
In any case, I don't know what this pissing contest has to do with the OP.
I mean, we can talk about headphones too. There’s a really great example of “what can this op amp drive under X conditions” in one of the application papers for the INA1620, talking about the difference between single and parallel drive.

But if i can nail down a (conservative) rule of thumb for the OPA210, my next target is the LME49724. And that’s a different kettle of fish because I don’t quite understand the way loading works in a differential context. Does each side have to drive the entire load, alternating in a push-pull fashion (600, 600)? Or do you calc it as half the load, all the time (300, 300)? Or do you treat the center of the load (aka the common mode point) like a passive summing point and call it 1.5x of half the load, per side, all the time (450, 450)?
 
Last edited:
You misinterpreted my defition of "Vsupply". Vsupply does not refer to the output of the circuit at whatever point. Vsupply refers to the voltage of the power supply which is generally a fixed voltage. Power supply voltage is going to define what power can be dissipated in a power amplifier application.

Putting Rbuildout inside or outside the FB loop has no effect on making "effective power less dependant on headphone impedance".

The only difference is the gain of the amp itself. With Rbuildout outside the FB loop, the 47R does not impact gain. With Rbuildout inside the FB loop, the gain is higher because of the voltage divider created by the 47R and Rphones.

But again, the listener does know or care what the gain of the amplifier is.

But the power supply rails do not affect the power delivered to the phones. Assuming that the rails have enough voltage and current to allow proper operation of the amplifier. Anything more is just headroom. If you run out of headroom then, of course, you have a problem but that is not the discussion here.
Wrt amplifier gain - the gain of the op amp with the 47R inside the FB loop - a "zero impedance" circuit" is higher. But the gain of the circuit as a whole remains unity (assuming unity gain design for simplicity)
Abbey's maths is correct. Thanks Mr Ohm 🙂
 
You guys have wandered off well into the weeds...

The OP's question about THD wrt loading, distortion increases somewhat linearly with falling impedance up to current limiting then is similar to clipping or saturation.

Distortion with a NF circuit is a function of open loop distortion reduced by loop gain margin (ratio of closed loop gain vs open loop gain). Since most amplifiers exhibit falling open loop gain with rising frequency, that suggest that distortion will rise with frequency also.

Some circuit topologies are less linear than others, so output distotion will begin to increase well before voltage saturation. Pretty much all amplifiers will exhibit increasing distortion just before saturation. Different industries or market segments tolerate different measurement standards for specifying max power. Many guitar amplifiers specify power at 3% thd while clean hifi amplifiers are more inclined to spec 1% or 0.1% THD. Allowing 3% distortion can make a big enough difference on paper to influence customers.

Of course I have a Peavey story about this too but maybe for another time...

JR
 
I say bring on the Peavey tale 🙂
OK by invitation... Last century I was involved in Peavey's effort to enter fixed install (background) sound products with AA (Architectural Acoustics brand). Hartley's logic was that since he was selling truckloads of 75W guitar amps, why not take those same power transformers and use them to make 75W fixed install amps. The oops that we didn't realize until well into the product development process was that guitar amps that make 75W with 3% distortion, do not make 75W clean for the fixed instal market :rolleyes: (they checked, and re-checked because Peavey was the f'n new guy in that market. )

Since we had already assigned model numbers and publicized these as 75W amps we had no choice but to tool up new slightly higher voltage transformers just for the fixed instal market, completely trashing the economy of scale expected from piggy backing on our existing guitar amp transformer volume. :cry:

JR
 
Anyone care to do a brief ELI5 on differential loading? I don’t think it makes a difference whether it’s an FDA or not…
 
What data am I bringing in to a rough estimation of whether an amp can drive a particular load at a particular output level without veering into distortion? And what is the math thereafter? Looking at any data sheet, you have the RMS vs. THD+N plot, the short-circuit current, and what else? Without expanding the question to in-loop buffering or parallel amps (or negative impedance for that matter!), maybe a good place to start would be “I see that OPAXYZ with a 65mA short circuit current can output 7V into 600R at .XXX% THD+N, but what can it output into 400R without the THD+N curve taking off?”

You can use some very basic "rule of thumb".

Assuming no current or voltage limits, once loading exceeds the limit for Class A Operation, every halving of load impedance will appx. double HD (all harmonics).

So if you have an Op-Amp rated to have current enough to drive +22dBU (10V) into 600 Ohm with reserve (> 50mA short circuit current recommended) but is only rated for 2kOhm for THD, the 600 Ohm THD will be around 3 times or 10dB higher.

Anyone care to do a brief ELI5 on differential loading? I don’t think it makes a difference whether it’s an FDA or not…

Simple, a FDA with differential loading will cancel even harmonics, mainly H2. So you get less THD.

At the same time you get twice the voltage, which if driving the same load will double the current and quadruple the power which will in effect half the load seen by the individual Output.

So distortion and H2/4/6/8... will be around doubled with the output level also doubled, presuming enough current capacity.

On the other hand, if driving the same voltage into the same load with an FDA each half will see half the impedance load but also half the voltage, so HD will be neutral.

Now note all the above is "rule of thumb" and not reliable, it will work most of the time.

Past that...

TI now has "dedicated" drop in replacements for the common audio op-amp's that are fully characterised with 600 Ohm loads at a competitive price, I see no reason not to use OPA1678 (USD 0.25 in 1Ku -> NE5532, TLx2 etc.) and OPA1679 (USD 0.4 in 1kU -> TLx4 etc.) for most positions instead of the "common audio op-amp's".

For specific applications that need lower noise or higher current are best accommodated the same way they are if using the common Op-Amp's, using external discrete components. Looking at Rupert Neve's later IC based designs is a good source for relevant ideas.

Thor
 
a FDA with differential loading will cancel even harmonics, mainly H2. So you get less THD.

By ~6dB, compared to single ended drive? In other words, when you drive a line output with an FDA and lift the cold at a single ended receiver, do 2/4/6/8 harmonics double?

When you drive a line output with an equivalent dual op amp and lift the cold at a single ended receiver, is it the same H2/4/6/8 result as with the FDA, or is there something particular to the CM cancellation of an FDA that is more/less aggravating to H2 et al?

if driving the same voltage into the same load with an FDA each half will see half the impedance load but also half the voltage, so HD will be neutral.

So…an FDA which drives 28dBu/20V into 600R is outputting 33.3mA. Thus each side is driving 22dBu/10V into 600R and is outputting 16.7mA? Or are you saying that each side is driving 1200R and outputting 8.33mA?

Maybe a dumb question; I assume it’s the former. I might have gotten thrown off by the quadrupling comment. But either is better than each side driving 300R! LMK if I need to sketch this.

I am totally ignoring the load contribution of Rf here (or Rf+Rg on one side of a single ended to differential FDA application), but I would phrase the question the same if it were a differential voltage follower at the line output.
 
Last edited:
TI now has "dedicated" drop in replacements for the common audio op-amp's that are fully characterised with 600 Ohm loads at a competitive price, I see no reason not to use OPA1678 (USD 0.25 in 1Ku -> NE5532, TLx2 etc.) and OPA1679 (USD 0.4 in 1kU -> TLx4 etc.) for most positions instead of the "common audio op-amp's".
Yes. Lotta people using those parts now. Even the OPA1692 specs better than similar to the 5532, and that’s a micropower amp. I stopped referencing the 5532 when the LM4562 got affordable, and OPA2134 was already out and about when I was getting started, so no need to wrestle with TL0 or TLE really. I’m liking OPA2210 and INA1620 for all my low impedance shenanigans, the one caveat with the 2210 being 7V before clipping. Wish I could find a 10V part with a Vos and Ib like that (super beta only, i suppose). OPA2209 is pretty good, and is available as a quad. The 4625-2 is also a good option but power and cost are a factor there.

Looking at Rupert Neve's later IC based designs is a good source for relevant ideas.

Focusrite? Any particular schematic spring to mind?
 
Last edited:
By ~6dB, compared to single ended drive?

Yes.

In other words, when you drive a line output with an FDA and lift the cold at a single ended receiver, do 2/4/6/8 harmonics double?

No, you are still running Single-Ended. So nothing changes and no cancellation of even harmonics happens.

Use the output differential, now 10-20dB cancellation of even order harmonics can be expected, AT THE COST of increased odd order harmonics.

When you drive a line output with an equivalent dual op amp and lift the cold at a single ended receiver, is it the same H2/4/6/8 result as with the FDA, or is there something particular to the CM cancellation of an FDA that is more/less aggravating to H2 et al?

If you make an FDA from two Op-Amps, it operates as FDA.

If you make a balanced out from a dual Op-Amp (say voltage follower and inverter) you get some degree of eblven order cancellation, but the noise penalty is large.

Such balanced outputs are usually worse than a simple SE out. If the input is balanced, using a single ended impedance balanced drive is IMNSHO preferred to "balanced", less noise and distortion and 50% lower cost.

So…an FDA which drives 28dBu/20V into 600R is outputting 33.3mA.

Correct.
Thus each side is driving 22dBu/10V into 600R and is outputting 16.7mA?

No, each side is driving 22dBu/10V into 300R and is outputting ~33.3mA.

I am totally ignoring the load contribution of Rf here (or Rf+Rg on one side of a single ended to differential FDA application), but I would phrase the question the same if it were a differential voltage follower at the line output.

It is usually not a good idea to ignore the loading from the feedback network.

Especially with FDA we usually need low resistance networks due to both noise gain and resistor noise.

On the other hand, a classic balanced compound feedback pair input circuit (aka "British microphone preamp") followed by an FDA made from (say) OPA1679 using the Birt circuit can be extremely low noise, distortion and high gain & bandwidth, exceeding any "canned" microphone preamps.

If we add an external Class A buffer using a highish beta PNP (go ahead, ask why PNP) & CCS with (say) 50mA IQ per side we can also drive +28dBu into 600 Ohm differential at low HD.

Even better, the input could be made direct coupled with phantom power, for a DC coupled Microphone preamp in to out, with a "flying cascoded" frontend and DC servos.

It will even sound great for a "clean" mic pre. I'm surprised nobody ever did this, it's so obvious!

Thor
 
@thor.zmt I’m down to get into capacitorless mic pre and impedance balanced output discussion (and hey who doesn’t love a discrete output buffer - cooler than the BUF634A for sure), but i wanna make sure I understand the basic assumptions re: differential heavy load line driving with minimum practical resistances. Here is a visual. Yea/nay?

A81E0DB6-F29C-4DCB-811D-7D0F9BFEE325.jpeg

(Cf for the Birt inverter might want to be an order of magnitude smaller for speed purposes, also I think it’s driving 3K at the summing junction, so maybe a little less load on and current out of that amp)

(also want to note the regular 1.3dB loss on the receiving end here)
 
Last edited:
@thor.zmt I’m down to get into capacitorless mic pre and impedance balanced output discussion (and hey who doesn’t love a discrete output buffer - cooler than the BUF634A for sure), but i wanna make sure I understand the basic assumptions re: differential heavy load line driving with minimum practical resistances. Here is a visual. Yea/nay?

View attachment 106558

The Birt Circuit would normally use larger value resistors in the feedback loop, this would lower the output current. Otherwise correct.

Also note, using the + input in the Birt circuit you can create a common mode offset. Adding Servo to that allows a "CM" input equivalent to the FDA.

Next would be noise and distortion analysis.

Also, how about adding an Impedance Balanced SE Output (simple follower) and a follower + inverter output?

Say NE5532 for all of them.

Thor
 
using the + input in the Birt circuit you can create a common mode offset. Adding Servo to that allows a "CM" input equivalent to the FDA.

Next would be noise and distortion analysis.

Also, how about adding an Impedance Balanced SE Output (simple follower) and a follower + inverter output?

that’s a neat Birt idea, i’ll draw it just to see how it works. i can see value with the servo method. also i’ll see about configuring it with a trim across the centers of the feedback resistances…

you said it yourself, an impedance balanced output is always gonna beat a dual drive output in terms of noise floor

i’d be more interested in analyzing this circuit below than the standard buffer + inverter output. i believe the noise penalty is 3dB before factoring in resistor values. requires a low Ib amp of course. it has Rod Elliot origins — his zero phase lag solution, but stripped back to essentials and set up for least load on the preceding stage (even an inverter with 1K Rf can drive it). i’m sure someone else has thought of it somewhere, but i haven’t seen it.

A3006F34-6949-47BC-80A0-1F4C82A56414.jpeg
 
Last edited:
i’d be more interested in analyzing this circuit below than the standard buffer + inverter output. i believe the noise penalty is 3dB before factoring in resistor values.

View attachment 106611

Hmm, upper Op-Amp noise gain = 6dB (gain also 6dB and attenuation on the input by 6dB, lower Op-Amp noise gain = 6dB.

Thus noise (ignoring resistor noise) is 9dB worse than a simple voltage follower.

I'd do follower + inverter. Follower noise gain 0dB, inverter noise gain 6dB, circuit gain 6dB, so noise stays lower at around the 6dB of the inverter.

I think we could find a way to lower noise in the inverter, with a noise servo, to the same level as the follower. Then we may end up with something closer to Birt.

If it is not obvious, Birt has the inverter in the feedback loop of the first Op-Amp but noise from this becomes common mode (as does DC offset on the positive in of the inverter). This lowers differtial noise of the Birt Circuit to close to single op-amp levels.

But honestly, I fail to see why we need an active balanced output. It has no benefit whatsoever, other than that it simulates a center tapped (or floating) transformer input.

I can see why commercial designs need to offer that (compatibly with the one or two oddball pieces that needed it).

For DIY? I'd agressively optimised for performance in my own system. In that case a noise optimised single Op-Amp balanced in and impedance balanced out materially outperform industry standard solutions.

For commercial gear I'd include a switch to select active balanced if needed, but default to impedance balanced.

I also like P68 non-standard Phantom Power and Capacitor microphones getting the 60V capsule bias from 68V Phantom power. The mic's loose a bit performance compared to P68 when running on P48, but work well enough. Getting 48V instead for "P48 max" mic's is also easy.

Thor
 
Hmm, upper Op-Amp noise gain = 6dB (gain also 6dB and attenuation on the input by 6dB, lower Op-Amp noise gain = 6dB.

Thus noise (ignoring resistor noise) is 9dB worse than a simple voltage follower.

I don’t know why I was thinking that the outputs cancel by more when they are summed back to single ended at the receiver.

ohhhhhh i know what it was. different topology:

with a double balanced output made from two equal-value diff amps, a SE or differential signal comes out differential and 6dB higher. but an equal value diff amp has 3dB noise gain (plus johnson details). so gain is commensurate with noise gain. do tell me if i have that wrong.

you can make a double balanced output with an INA1620. the load on the preceding stage is 6??R, so you need a buffer (a fader buffer, for instance). resistors in that package average 965R.

965R || 350R = 257R

10V / 257R = 39mA
39mA x 10V = 390mW…nope

7V / 257R = 27.2mA
27.2mA x 7 = 190mW…close

6.2V / 257R = 24.1mA
24.1mA x 6.2V = 150mW…bingo, full power at 1.6MHz BW. 24dBu differential into 600R with 50R buildouts.

But into a 5K differential load, it’ll do 28dBu. Just that dual amp and two 100p caps, no resistors to speak of.

I'd do follower + inverter. Follower noise gain 0dB, inverter noise gain 6dB, circuit gain 6dB, so noise stays lower at around the 6dB of the inverter.

I think we could find a way to lower noise in the inverter, with a noise servo, to the same level as the follower. Then we may end up with something closer to Birt.

I’d like to see what you mean on the back of a napkin.

In your other Birt suggestion, were you saying to feed signal to the inverting input and DC servo the non-inverting input via the existing voltage divider, or were you saying to drive the circuit differentially and come directly in to the inverting input with a high-value resistor off the servo output? Where are you drawing the DC sample from?

if you wouldn’t mind showing these two Birt-based circuits it would be helpful…

If it is not obvious, Birt has the inverter in the feedback loop of the first Op-Amp but noise from this becomes common mode (as does DC offset on the positive in of the inverter). This lowers differtial noise of the Birt Circuit to close to single op-amp levels.
Nope, not obvious. Also, referencing the typical higher value resistors for inverter Rf and Rg, wouldn’t you induce additional noise if you use, say 5K/5K rather than 2K/2K as drawn?

But honestly, I fail to see why we need an active balanced output. It has no benefit whatsoever, other than that it simulates a center tapped (or floating) transformer input.
That I would agree with. People generally know enough to lift a cold when need be.
I can see why commercial designs need to offer that (compatibly with the one or two oddball pieces that needed it).
I think it’s pretty important to provide an input and output that can at least accommodate a converter running at 24dBu = 0dBFS. That’s why I’m okay with the 2210 — 25dBu across the hot and cold. There are a few amps that can put out 11 or 11.5Vrms on 18V dual supplies, but there’s always some caveat from what I’ve seen.

I’m willing to bet there are broadcast and educational clients who don’t buy anything that doesn’t have capabilities that an impedance balanced I/O can not provide.

Anyways, I don’t wanna get too in the weeds here. Optimizing a Birt output for low noise with acceptable drive would be pretty on topic tho.
 
Last edited:
But into a 5K differential load, it’ll do 28dBu. Just that dual amp and two 100p caps, no resistors to speak of.

Well, not EXACTLY a Dual Op-Amp.

I’d like to see what you mean on the back of a napkin.

It uses Error extraction from inverting input (which will among others show the noise).
The error is amplified inverting (say with a low noise CFP) by say 1,000 times and injected by 49R9 * 1,000 into the output. I had a quick stab in TINA-TI but could't quite get it...

In your other Birt suggestion, were you saying to feed signal to the inverting input and DC servo the non-inverting input via the existing voltage divider, or were you saying to drive the circuit differentially and come directly in to the inverting input with a high-value resistor off the servo output? Where are you drawing the DC sample from?

Look at the detailed inside block diagrams of the TI FDA's, you will see how it's done.

if you wouldn’t mind showing these two Birt-based circuits it would be helpful…

Ok, tomorrow if I find time...

Nope, not obvious. Also, referencing the typical higher value resistors for inverter Rf and Rg, wouldn’t you induce additional noise if you use, say 5K/5K rather than 2K/2K as drawn?

Yes, but the noise (and DC) of the inverter becomes common mode. If this matters depends if the output is taken SE (then noise goes up) or BAL (common mode noise is suppressed).

That I would agree with. People generally know enough to lift a cold when need be.

Yup. Here a proposal of an active circuit that will do +28dBu into 600Ohm, can be made with a pair of NE5532 and has acceptable performance (AND a +22dBu limited passive option).

1679154531105.png
Output noise is dBu and the active balanced circuit is less than 2dB noisier compared to pure inverted output, but 10dB worse than the "passive" option (all re 0dBu).

Thor
 
It uses Error extraction from inverting input (which will among others show the noise).
The error is amplified inverting (say with a low noise CFP) by say 1,000 times and injected by 49R9 * 1,000 into the output. I had a quick stab in TINA-TI but could't quite get it...

Ok, I got it. Needed to go noise gain ~ 100 & fettle the gain set resistor value for the noise Amp.

1679169412583.png

Around 3.5dB noise reduction overall. Should also reduce HD.

If looking only at the inverter noise, this can be reduced by nearly 6dB.

I may not consider it worthwhile, but it is quite a few dB less noise than even using double double Op-Amp's.

The circuit one post above might be also better with the inverters driven from the direct input, but 1.1k Load vs near infinite load on source are a consideration.

Thor
 
Last edited:
The circuit one post above might be also better with the inverters driven from the direct input, but 1.1k Load vs near infinite load on source are a consideration.

Ok, and now "what if we use a more modern, low noise Op-Amp?". Using two OPA1612 at 4 USD each when buying a reel of 3,000 (and nearly 7 USD each in small quantities) will give -119dBu noise and +28dBu without significant distortion (presuming TI Op-Amp models are accurate enough:

1679208647822.png

But if we use more NE5532 & noise reduction, what can we get?

1679209391532.png

I think for mass production the "cheap Op-Amp plus noise reduction" makes sense and performs well enough (THD may be able to be tuned lower by another 20dB I don't want to spend the time).
If modifying commercial gear in hand or building new, I thing using "expensive op-amp's" makes perfect sense.

Thor
 
THD may be able to be tuned lower by another 20dB I don't want to spend the time.

Ok, tried after all, cascoding the noise reduction amplifiers in the 4-Op-Amp version kills the excess THD:
1679212606316.png
Now we are very close to "expensive op-amp" with less than 1 USD BOM (assuming full reels of all components).

Thor
 
Last edited:
@thor.zmt this is really impressive. I went “wow, -115!” a few posts back and “these concepts should get put together,” and then you did it. This is publishable research.

Trying to avoid asking any dumb questions here. I don’t have a strong background in discretes, so i’m having a hard time following the way the noise servos work without all the values. Are they just not showing up?

Do I understand correctly that the same concept could be applied to any inverter, and that values are pretty widely scalable as long as the injection ratio matches the Sziklai pair gain? If the noise sample comes from the inverting input of a diff amp that is being relied upon for CMR, does the non-inverting input need to see a similar effect as is caused by the base of the transistor (i.e. a dummy transistor or high value resistor to ground)?

I saw that you spec’d your original transistor choices and then they went away and am wondering what the factors are in parts choice as you figure this out.

It would be pretty cool to see this architecture surrounding a solitary LME49724 in a diff-to-diff context. Not a big difference between 2.1nV/rtHz and the 1.1nV/rtHz of the OPA1612, and it’s on the cheap side. Also note that the 1.1nV/rtHz (!) OPA1633 is in preview; there must be a model available.

I might have to buy a Windows machine just to run TINA. This makes me want to try the same tactic with an all-2K double balanced output, and with a variety of differential line trim circuits I’ve been playing with. It’s always the resistor values that make +/-12 to +/-15dB way harder to do elegantly than +/-6dB to +/-10dB.
 
Last edited:
Back
Top