Maker Pro
Maker Pro

TDA2030A Headphone Amp Gain too high

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
I have a TDA2030A based amplifier that should make a decent headphone amp except that its gain, at about 32 dB, is way too high.

It would be easy to change the gain setting resistors or even put in a switch to change the gain from say 10 to 15 to 20 dB for proper Headphone use. But there is a problem.

The TDA2030A data sheet says the chip needs to run at a gain in excess of 24 dB to maintain stability. That is too high for low impedance high sensitivity headphones.

A resistor could be placed in series with the output. But that destroys damping factor and sound quality. That is not an acceptable solution.

Meier Audio uses the TDA2030A in their Corda Brick. It has switchable gain of -1 to +14 dB. So how do they do that with the TDA2030A?

Anybody got a working circuit they will share that gets the TDA2030A down to a stable switchable gain of 10-15-20 dB?

Suggestions?

Thanks
 

davenn

Moderator
Sep 5, 2009
14,254
Joined
Sep 5, 2009
Messages
14,254
apart from that .... you don't REALLY want an 18Watt amplifier feeding a headphone speaker right next to your ear do you ??
A sure way to achieve complete deafness

You really only need something ~ 0.5 - 1.5 W.

Dave
 

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
Where do you see that minimum gain in the datasheet.

I am looking at this one:

http://www.st.com/st-web-ui/static/active/en/resource/technical/document/datasheet/CD00000129.pdf

Bob

On your URL it is at page 19. Look for part 6, Application recommendation, Footnote 1.

Supposedly, below that gain the amp may become unstable with certain loads. I am looking to power headphones at around 32 Ohms impedance, maybe a little less on the bottom up to 300 Ohms at the top.
 

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
apart from that .... you don't REALLY want an 18Watt amplifier feeding a headphone speaker right next to your ear do you ??
A sure way to achieve complete deafness

You really only need something ~ 0.5 - 1.5 W.

Dave

Not really Dave,

Thanks for your concern. But I have maintained my hearing acuity quite well for over 70 years and do not want to lose it now.

I am looking to supply sufficient clean current to the HPs for very high quality reproduction while minding my volume level. That can be done more easily with lower gain. A little power amp chip like the TDA2030A should fit the bill quite nicely

Most all 8 pin OPAs used to drive HPs cannot supply the current necessary for really high quality sound with low impedance HPs even if operated at +-15V. All I need to do is swing less than 2V PP to handle 32 Ohm phones and if resultant amp is powered by +-9V or more it may also handle 300 Ohm phones too. But I need to knock down the gain and still maintain stability.

Do you know how to do that?

Thanks
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
Use a chip that doesn't have the gain limitation.

Bob
 

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
You can just put a passive attenuator (two-resistor voltage divider) on the input. But I definitely would not drive headphones directly from a high-power amp. A fault could easily destroy them, and my headphones are not cheap. Also the background noise, which would be inaudible from a speaker, will be very noticeable with headphones.

Google headphone amplifier design. There are lots of designs available; some simple, some complicated; some using only op-amps, some hybrid and some entirely discrete; some making all kinds of claims about their sound quality. headwize.com has a selection of DIY designs.

Also look at Digikey in their audio amplifiers section; there are lots of ICs designed for use as headphone amplifiers. National Semiconductor (now Texas Instruments) had a range called the "Boomer" series for headphones and small speakers. There are lots of options.
 

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
You can just put a passive attenuator (two-resistor voltage divider) on the input. But I definitely would not drive headphones directly from a high-power amp. A fault could easily destroy them, and my headphones are not cheap. Also the background noise, which would be inaudible from a speaker, will be very noticeable with headphones.

Google headphone amplifier design. There are lots of designs available; some simple, some complicated; some using only op-amps, some hybrid and some entirely discrete; some making all kinds of claims about their sound quality. headwize.com has a selection of DIY designs.

Also look at Digikey in their audio amplifiers section; there are lots of ICs designed for use as headphone amplifiers. National Semiconductor (now Texas Instruments) had a range called the "Boomer" series for headphones and small speakers. There are lots of options.

Thanks Kris,

Took your advice and reviewed the web literature. For discrete circuits it seems there is an overwhelming favorite. It is the Diamond buffer. I cannot see why it is favored over the common Complementary Symmetry PP topology. Perhaps you can explain.

The schematic below shows two common, simple but functional versions of common buffers. The one on the left is a Complementary Symmetry Push-Pull (CSPP) buffer and the one on the right is a Diamond buffer.

The cap in each circuit is as suggested by Jung and in my simms that cap does a great job at reducing distortion in Class A mode. If either circuit will spend considerable time in Class B, an additional cap can be placed from the top of R5/R105 to the bottom of R6/R106. This will improve distortion in the Class B region. I do not want to go there so I have not included those caps.

It seems to me that the CSPP is better suited for discrete construction. Thermal feedback in the CSPP is between identical transistors, NPN to NPN, PNP to PNP. They are easy to match for diodic equivalency if you wish and of course the bias of the drivers is set by the diode junctions of identical transistors that are thermally connected.

The Diamond on the other hand has its output transistors biased by unlike transistors, NPN to PNP and PNP to NPN. These are much harder to match if you wish. Also, I have seen Diamond circuits on the net where the thermal feedback is NPN to NPN and PNP to PNP which makes no sense to me because the thermal feedback is not directed to the actual transistor that provides bias to a given output. Indeed, I have seen cases where the current in the bias transistors significantly exceeds the current in the outputs (a bit much if you ask me even for a current mirror) and each transistor is on a separate HS with no thermal feedback at all.

Am I missing something? What is it that makes the Diamond circuit favored over the CSPP? Is it higher and stiffer input impedance? Or is it the name?

I have simmed both circuits and both result in all distortion products down more than 120 dB with outputs biased at 25 mA with the buffer inside a feedback loop. The Diamond is a couple dB better though. As far as I can see they are virtually identical. With 25 mA Ic in the outputs, and a supply of +- 9V or more, it seems both circuits will supply sufficient current and voltage to fully power headphones with impedances from 30 to 300 ohms before crossing to Class B mode with low impedance phones. I know that a sim beggs reality but I am comparing apples to apples.

So, my two questions:

1. Why the Diamond buffer over the CSPP?

2. Shouldn't the temp compensation in a discrete Diamond be from bias transistor to output transistor NPN-PNP for proper temp stability?

Well, here are the example circuits and thanks for any help you can give.

Buffer%2520Comparison.PNG
 
Last edited:

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
That's very interesting. You've obviously done some homework!

I had never seen the "diamond buffer" design before, and I'm glad to know about it now. I think it's quite elegant, but I don't know enough about transistor characteristics and audio circuit design to be able to tell you much about the differences between them. I think you probably know more than me.

I certainly do see the reason for C1 and C101, especially C101. As well as providing a strong coupling between the bases, which will stabilise the current flow in the otuput transistors, it will help to keep the voltage drops across R102 and R103 constant. This was a slight concern because of the voltage divider effect between R101/102 and R103/104, which would cause the voltage across R102 or R103 to drop as the circuit voltages approached the positive or negative rails.

Edit: But I've just realised that that wouldn't be a problem anyway, because the output stage bias is determined by the sum of the four voltages, and when the voltage across one resistor drops, the voltage across the other will increase.

Other versions of the diamond buffer I've seen have used current sources for R101 and R104 to keep the voltage drops across R102 and R103 stable, but C101 will have the same effect, as long as it's large enough to cover the lowest frequencies that the amp will have to reproduce.

As for thermal stability, it's true that in the standard circuit does use an NPN to counter the NPN output transistor's offset and a PNP to counter the PNP's offset, whereas the diamond buffer does the opposite, but the total offset for the input/bias transistors still cancels the total offset in the output transistors, so the quiescent current should be stabilised; it's just the DC offset that may not be, and this can be fixed by applying negative feedback - unless you want to avoid an overall feedback loop. Even without correction, the output offset voltage shouldn't be very much.

As for connecting a capacitor between the emitters of the output transistors, I agree that shouldn't be very important; the emitter resistors are already pretty low in value. I would be interested to know if there's any quantifiable improvement in performance with them added though. I've never seen them used in the standard circuit, nor in the diamond buffer circuits that I found with a Google search.

Your standard circuit is not exactly how I usually see it - instead of a fixed bias using base-to-collector-tied transistors, versions I've seen use a single transistor with a trimpot across it, with the wiper connected to the base. (Or the safer version, a resistor from base to collector and a trimpot as a variable resistor from base to emitter). I guess that arrangement would give more thermal negative feedback, but I'm not sure. Have you analysed the differences? Did you have a reason for using the two transistors that way?

As far as offset tracking goes, I'm not sure how successful it would be in any case, considering that the input and output transistors will not normally be the same types. Also thermal resistance will mean that the output transistor junctions will always be hotter than the input transistor junctions. For a headphone amp, none of this should be very important anyway, I think. Would you agree?

...
Also, I have seen Diamond circuits on the net where the thermal feedback is NPN to NPN and PNP to PNP which makes no sense to me because the thermal feedback is not directed to the actual transistor that provides bias to a given output.
Do you mean that the NPN driver and NPN output share a heatsink, and the PNP driver and PNP output share a different heatsink?

If so, that should be alright for the reason I mentioned before - if you're not worried about DC offsets, only about stabilising the output circuit current, then it doesn't matter which driver is thermally linked to which output transistor.

Indeed, I have seen cases where the current in the bias transistors significantly exceeds the current in the outputs (a bit much if you ask me even for a current mirror) and each transistor is on a separate HS with no thermal feedback at all.
That's interesting. How is that supposed to work? I guess the self-heating in the drivers is deliberately made higher than the self-heating in the output transistors, to effectively give more than 100% negative thermal feedback, on the assumption that the conditions that cause the output transistors to get hot will also cause the drivers to get hot.

I don't think that's true, and in any case, most of the heating in the output transistors will be due to load current, not quiescent current, so unless the load has a high impedance, the output transistors will still dissipate more power than the drivers.

Does that make sense? What are your thoughts?

Does the design you describe have a proper circuit descripiton that covers the rationale behind it? If so, can you post a link to it?

Am I missing something? What is it that makes the Diamond circuit favored over the CSPP? Is it higher and stiffer input impedance? Or is it the name?
I think it could be the high input impedance, or the inherent elegance. Could also be the name :) And if you're missing something, I'm missing it too!

I have simmed both circuits and both result in all distortion products down more than 120 dB with outputs biased at 25 mA with the buffer inside a feedback loop. The Diamond is a couple dB better though. As far as I can see they are virtually identical. With 25 mA Ic in the outputs, and a supply of +- 9V or more, it seems both circuits will supply sufficient current and voltage to fully power headphones with impedances from 30 to 300 ohms before crossing to Class B mode with low impedance phones.
That sounds fair enough.

Do you have some 300 ohm headphones?

I know that a sim beggs reality but I am comparing apples to apples.
Yes, I agree.

I've tried to answer your two questions. I'm very interested to hear your responses.
 
Last edited:

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
That's very interesting. You've obviously done some homework!

Yup! But I am not trained in the field. So much is over my head.

I had never seen the "diamond buffer" design before, and I'm glad to know about it now. I think it's quite elegant, but I don't know enough about transistor characteristics and audio circuit design to be able to tell you much about the differences between them. I think you probably know more than me.

First time I saw the circuit was in Audio Amatuer magazine over 20 years ago. Jung wrote an article about it ref the LH0002 buffer. You might check that datasheet. It is on line.

Also you might check this more modern article by Jung. It is a much more complex version than the original. http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_2.pdf

I certainly do see the reason for C1 and C101, .....

Yeah, they work fine for the purpose.

Other versions of the diamond buffer I've seen have used current sources for R101 and R104 to keep the voltage drops across R102 and R103 stable, but C101 will have the same effect, as long as it's large enough to cover the lowest frequencies that the amp will have to reproduce.

For a simple HPA I wanted to keep it simple as possible and still get the job done. So I wanted to avoid the current sources and make it more like the original diamond buffers.

As for thermal stability, it's true that in the standard circuit does use an NPN to counter the NPN output transistor's offset and a PNP to counter the PNP's offset, whereas the diamond buffer does the opposite, but the total offset for the input/bias transistors still cancels the total offset in the output transistors, so the quiescent current should be stabilised; it's just the DC offset that may not be, and this can be fixed by applying negative feedback - unless you want to avoid an overall feedback loop. Even without correction, the output offset voltage shouldn't be very much.

Three points. DC stability is very important if you are biasing in Class A. You may want to operate the buffer outside a feedback loop. Thermal runaway is not controlled without the bias trans and output trans in thermal contact. So why not get it stable in all respects before deciding to apply feedback or not?

As for connecting a capacitor between the emitters of the output transistors, I agree that shouldn't be very important; the emitter resistors are already pretty low in value. I would be interested to know if there's any quantifiable improvement in performance with them added though. I've never seen them used in the standard circuit, nor in the diamond buffer circuits that I found with a Google search.

The second cap seems to have little effect in Class A region. But does help at crossover and beyond into Class B. Found it here: http://www.tubecad.com/2012/09/blog0244.htm

Your standard circuit is not exactly how I usually see it - instead of a fixed bias using base-to-collector-tied transistors, versions I've seen use a single transistor with a trimpot across it, with the wiper connected to the base. (Or the safer version, a resistor from base to collector and a trimpot as a variable resistor from base to emitter). I guess that arrangement would give more thermal negative feedback, but I'm not sure. Have you analysed the differences? Did you have a reason for using the two transistors that way?

Yes, simplicity. I want to avoid constant current sources and keep parts count down. The original CSPP uses one resistor and a diode for each output trans and has been used in countless buffers and power amps. You can change the resistor value to get more drive to the output but it becomes a trade off of ability of the front end to supply voltage vs current. By doing it my way you can essentially set the current of the driver and then adjust the voltage to the output base as well without starving the driver of either voltage swing or current to feed the output.

As far as offset tracking goes, I'm not sure how successful it would be in any case, considering that the input and output transistors will not normally be the same types. Also thermal resistance will mean that the output transistor junctions will always be hotter than the input transistor junctions. For a headphone amp, none of this should be very important anyway, I think. Would you agree?

I am not an engineer. Indeed, that is one of the questions I was seeking guidance on. But it seems to me that the transistors do not need to be at the same temp before thermal feedback for the feedback to be beneficial. For examp. say the two transistors were very different in Tj when separated and then they were thermally connected and heatsunk. The cooler trans (driver) would be heated up and deliver less drive while the hotter trans would cool and dissipate less. At some point, homeostasis would occur and thermal runaway would be held in check. I assume that would be at a temp in excess of the driver trans Tj when on its own. It might take some fiddling to get the output to bias properly in Class A, but thermal runaway should be avoided. At least that is the way it seems to me.


Do you mean that the NPN driver and NPN output share a heatsink, and the PNP driver and PNP output share a different heatsink?

That is correct. I do not have a URL for that cause it made no sense to me. But here is one where the PNPs are thermally connected and the NPNs are thermally connected (no sinks). It too makes no sense to me: http://phonoclone.com/diy-sapp.html

If so, that should be alright for the reason I mentioned before - if you're not worried about DC offsets, only about stabilising the output circuit current, then it doesn't matter which driver is thermally linked to which output transistor.

DC offsets are important if you want to DC couple and avoid large and expensive film caps.

That's interesting. How is that supposed to work? I guess the self-heating in the drivers is deliberately made higher than the self-heating in the output transistors, to effectively give more than 100% negative thermal feedback, on the assumption that the conditions that cause the output transistors to get hot will also cause the drivers to get hot.

I don't think that's true, and in any case, most of the heating in the output transistors will be due to load current, not quiescent current, so unless the load has a high impedance, the output transistors will still dissipate more power than the drivers.

Does that make sense? What are your thoughts?

Does the design you describe have a proper circuit descripiton that covers the rationale behind it? If so, can you post a link to it?

See the URL above. Sorry, I do not understand it. But I believe that in Class A, as the outputs deliver current the trans dissipate less heat cause the current is going to the load and each output trans is cycling through the amount of current through it as the frequency it is reproducing varies. I think that in Class B, the peak dissipation is around 65% of max output power. I am not sure though.

Do you have some 300 ohm headphones?

No, mine are 32 Ohm. But the current required for lower impedance phones is higher and that is where these little guys tend to fall down if they are capable of swinging the necessary V for higher impedance phones. Indeed, that is one of the problems in trying to cover 30 ohms to 300 ohms. If you have the voltage capability to handle 300 and the current capability to handle 30 total possible power at low impedance is dangerously high without some sort of current limitation.

I've tried to answer your two questions. I'm very interested to hear your responses.

Well, I have responded above. But my real response is that perhaps it is time to buy the bricks to build a house rather than make the bricks to build a house.

By that I mean that perhaps I should go the OPA route rather than discrete.

I started this thread by asking how to cut back the gain on a 2030. It was suggested that a power amp should not be used for phones. I disagree. Phones need a power amp. But it should be a little one with limited current capability. The buffers we are discussing are in effect the same as a power amp output stage. In fact, they can supply enough current to fry headphones and your ears. So even these circuits need to be used with care.

A possible solution is to use OPAs that current limit. For example the venerable 5532 current limits at about 40 mA. But the current limit is hard and it gets ugly long before it limits. So paralleling two or three of them might do the trick. Worst case the phones will see 80-120 mA and that should be survivable by low impedance phones (not your ears).

Here are two app notes re paralleling:

http://www.ti.com/lit/an/sboa031/sboa031.pdf

http://www.intersil.com/content/dam/Intersil/documents/an11/an1111.pdf

And see this by a real guru on amplifiers; https://www.google.com/url?sa=t&rct...=BxQlcijnGvHFWlCcMphBag&bvm=bv.55139894,d.eWU

Those are not the only way. See this article by an engineer on a mission. Read the whole page including the measurements on the bottom. His writing style is not the best, but I bet you will end up reading all the pages.

http://nwavguy.blogspot.com/2011/07/o2-headphone-amp.html

That might well be the way to go. In fact I am almost fininshed with a PCB for that topology.

Of course simpler is easier. So I left out the battery option and protection circuit. But how about on-off transients? I am hoping that switching the DC side of the supply while leaving the AC side always active might help. Also, having a snubber across the DC switch might help moderate pops and clicks. Got any simple ideas that do not involve relays, complex protective circuits, etc. that will minimize on-off thumps and clicks?

Thanks,its been fun so far.
 
Last edited:

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
First time I saw the circuit was in Audio Amatuer magazine over 20 years ago. Jung wrote an article about it ref the LH0002 buffer. You might check that datasheet. It is on line.
Is that AN227 "Applications of Wide-band Buffer Amplifiers"? I just had a quick look at that one. It doesn't help me much.
Also you might check this more modern article by Jung. It is a much more complex version than the original. http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_2.pdf
Yes, I read that one already. It doesn't show the capacitor between the bases of the output transistors! Nor does AN227.

That's disappointing, because that capacitor is very helpful - it allows the drive stage to actually provide significant base current to the output transistors. Without it, the driver transistors can only suck current away from their respective output transistors; with the capacitor, each driver transistor is an active current source for the base of the opposite output transistor. I'm not surprised that the capacitor makes a significant improvement, especially when operating into a significant load.
Three points. DC stability is very important if you are biasing in Class A. You may want to operate the buffer outside a feedback loop. Thermal runaway is not controlled without the bias trans and output trans in thermal contact. So why not get it stable in all respects before deciding to apply feedback or not?
Agreed, and I wasn't suggesting no thermal connection between the drivers and the output transistors. I was saying that any thermal connection - from top output to top driver and bottom output to bottom driver, or vice versa - should prevent thermal runaway. And I agree that it's helpful to have a circuit that starts off clean without feedback - that's why I suggested that option, and the dual emitter follower design is the obvious choice for a low-distortion buffer.

I'm not sure I agree with you about whether a DC output offset is a major problem though. I wonder if 10 mV or 20 mV of DC offset will cause any actual problems with headphones - apart from a click when you plug them in and unplug them.

Assuming it really IS a problem, surely a DC blocking capacitor of a few hundred uF isn't a big issue?

And if you can't have the blocking capacitor, you could stabilise the DC output voltage by applying DC-only feedback without losing the apparently magical audiophile qualities of a no-feedback signal path...

The second cap seems to have little effect in Class A region. But does help at crossover and beyond into Class B. Found it here: http://www.tubecad.com/2012/09/blog0244.htm
That's an interesting blog. Nothing much new for the basic diamond buffer, but he does show the use of one or two Schottky diodes to increase the output stage bias and current. Have you considered that idea? I'm not suggesting it, just wondering.

He mentions that adding the base-to-base capacitor improves distortion from 1% to 0.1%. Those figures are far higher than the claims for the LH0002. Perhaps that's because the LH0002 is trimmed in some way? What distortion figures do you get with your SPICE analysis?
I am not an engineer. Indeed, that is one of the questions I was seeking guidance on. But it seems to me that the transistors do not need to be at the same temp before thermal feedback for the feedback to be beneficial. For examp. say the two transistors were very different in Tj when separated and then they were thermally connected and heatsunk. The cooler trans (driver) would be heated up and deliver less drive while the hotter trans would cool and dissipate less. At some point, homeostasis would occur and thermal runaway would be held in check. I assume that would be at a temp in excess of the driver trans Tj when on its own. It might take some fiddling to get the output to bias properly in Class A, but thermal runaway should be avoided. At least that is the way it seems to me.
Yes of course, you're right.
That is correct. I do not have a URL for that cause it made no sense to me. But here is one where the PNPs are thermally connected and the NPNs are thermally connected (no sinks). It too makes no sense to me: http://phonoclone.com/diy-sapp.html
I wouldn't take much notice of that site. I read the text. He talks about avoiding IC regulators in the power supply, and recommends a zener regulator, R-C filter, and emitter follower and says "It is efficient, effective, and sounds very good." Even before I read that, I was concluding that his explanations are not very rational. So he has little credibility with me. The nail in the coffin is the lack of even the base-to-base capacitor.
DC offsets are important if you want to DC couple and avoid large and expensive film caps.
I answered this before, but... film caps? Are electrolytics "dirty" in some way? Do they colour the sound? Sorry to sound cynical, but I've learnt enough about perception in the brain to know that all of the claims of so-called golden-eared audiophiles can be very convincingly explained as psychological, and unrelated to reality. These people often make claims like "vinyl sounds better" or "digital sounds bad" which really beggar belief.

As I've said before, any of these claims COULD be supported by properly conducted double-blind testing, but the people who make them seem to be allergic to the whole idea. So I have to try to apply skepticism and refuse to accept their claims until they can demonstrate their validity. This would include the idea that electrolytics are somehow bad for sound quality.

[... snip ...]

No, mine are 32 Ohm. But the current required for lower impedance phones is higher and that is where these little guys tend to fall down if they are capable of swinging the necessary V for higher impedance phones. Indeed, that is one of the problems in trying to cover 30 ohms to 300 ohms. If you have the voltage capability to handle 300 and the current capability to handle 30 total possible power at low impedance is dangerously high without some sort of current limitation.
I agree. Perhaps a headphone amplifier could be designed to measure the actual load resistance (weak DC current into load; measure voltage across it) and adjust its characteristics accordingly... Using some kind of very soft clipping to limit the output power without colouring the sound unnecessarily the way a compressor/limiter would... sounds like an interesting project that I would never be bothered with!
Well, I have responded above. But my real response is that perhaps it is time to buy the bricks to build a house rather than make the bricks to build a house. By that I mean that perhaps I should go the OPA route rather than discrete.
Oh, that would be a shame! I like the diamond buffer idea, with the base-to-base capacitor. By all means, use an op-amp in the input stage, and within the feedback loop, but don't you want to use the diamond buffer as well?

I started this thread by asking how to cut back the gain on a 2030. It was suggested that a power amp should not be used for phones. I disagree. Phones need a power amp. But it should be a little one with limited current capability. The buffers we are discussing are in effect the same as a power amp output stage.
No, I definitely disagree. A headphone amplifier is not a power amplifier. It's part way between a signal amplifier and a power amplifier.
In fact, they can supply enough current to fry headphones and your ears. So even these circuits need to be used with care.
Sure. But that's not the only reason I said you shouldn't use a power amplifier with headphones. My first reason was that an amplifier designed to driver a loudspeaker will produce too much noise for direct connection to headphones. That and the fact that a power amplifier could literally melt the voice coils of a pair of headphones and set them on fire! And the fact that it's simply huge overkill and totally inappropriate! So I stand firmly by my opinion that you should not drive headphones from a power amp!
A possible solution is to use OPAs that current limit. For example the venerable 5532 current limits at about 40 mA. But the current limit is hard and it gets ugly long before it limits. So paralleling two or three of them might do the trick. Worst case the phones will see 80-120 mA and that should be survivable by low impedance phones (not your ears).
Yeah, I might do that for simplicity, but I would rather have a clean output stage and do the limiting in a controlled way by soft clipping the voltage rather than current limiting.

I don't see the point in using a whole lot of op-amps in parallel when you could just use a standard output stage such as the two we've been discussing. It has some curiosity value but that's all, IMO.
Those are not the only way. See this article by an engineer on a mission. Read the whole page including the measurements on the bottom. His writing style is not the best, but I bet you will end up reading all the pages.
http://nwavguy.blogspot.com/2011/07/o2-headphone-amp.html
That might well be the way to go. In fact I am almost fininshed with a PCB for that topology.
Good luck with that!
Of course simpler is easier. So I left out the battery option and protection circuit. But how about on-off transients? I am hoping that switching the DC side of the supply while leaving the AC side always active might help. Also, having a snubber across the DC switch might help moderate pops and clicks. Got any simple ideas that do not involve relays, complex protective circuits, etc. that will minimize on-off thumps and clicks?
Yeah. Unplug the headphones when you're turning it ON and OFF!

Not really. Perhaps you could arrange for the supply rails to come up gradually using an R-C circuit driving an emitter follower in series with each supply rail. That might work assuming the op-amps don't do anything strange while their supply voltages are low.
 

oneoldude

Jun 16, 2013
14
Joined
Jun 16, 2013
Messages
14
I am going to try to respond without all the quotes. Here goes.

There is very little information available re: the use of the base to base capacitor. I only once saw a comment by Jung that said it was helpful. But caps are not typically used internally in OPAs due to difficulty of manufacture, etc. The only other reference I found was in that Magic Diamonds blog. So I simmed it and found it worked very well in both circuits.

I am an objectivist. No magical qualities about anything. But some things do sound different than others. And when they do, measurements should tell you why. If there are no measurable differences, then I suspect the subjective conclusions more than properly conducted measurements. Actually both need to be verified.

Over the years I have conducted blind and double blind comparisons of lots of stuff. The only differences I ever found were objectively explainable. The only constant I found was that there was a bit more resolution on the very top end with big SOTA pure Class A power amps. Probably due to residual effects of crossover distortion and TIA in those AB designs of 20 and 30 years ago. But it was only observable in instantaneous comparisons and not worth the cost of going to SOTA Class A mega buck designs.

Today things might be even better for higher quality consumer designs vs mega buck SOTA designs. Also, my hearing is not as good as it was 30 years ago so I am out of the game

Clearly, a DC feedback design will reduce offset. And I have no problem with feedback. It is a great janitor. But in some cases you do not want a closed loop from amp to buffer and back. For example, in a low noise HPA, one topology that works very well is to use a gain stage in front of the vol pot with the pot feeding the buffer. In that instance, you cannot close the loop with the pot in between. So you want the best buffer you can muster for that topology.

The use of a Schottky diode providing an additional fixed drop seems to be the addition of a non-necessary component. It is somewhat like using a more complex current source vs a resistor. Yes the current source is stiffer, but is it really necessary? If, the two resistor scheme can set current and voltage sufficiently well to get the job done, why bother with greater complexity unless dynamic distortion increases in the real world?

According to my LTSpice simulations, the distortion figures for both circuits are well below 1% when properly configured. More like .001%. Real world measurements may well be worse. If you read the LH0002 datasheet, it specs the original simple buffer circuit at 0.1%.

Here is the datasheet: http://datasheet.elcodis.com/pdf/16/64/166478/lh0002cn.pdf

That chip is over 20 years old and of course has no B-B cap. But such a cap could be implemented from pin 6 to pin 10 on that chip. I have never seen it discussed though. I have a couple LH0002s on hand. May give them a shot!

You ask about the simms. A picture is worth a thousand words. Here are four of them. BTW check the scale on the left. They are not all scaled the same.

My CSPP without the cap:

https://lh4.googleusercontent.com/-...LY/LabS5-2Q2eA/s1152/CSPP%20without%20cap.PNG

My Diamond without the cap:

https://lh4.googleusercontent.com/-...tcFseLiA66w/s1152/Diamond%20without%20cap.PNG

My CSPP with the cap:

https://lh3.googleusercontent.com/-...LU/KROrnF2_LJc/s1152/Diamond%20with%20Cap.PNG

My Diamond with the cap:

https://lh3.googleusercontent.com/-...AAALQ/G9prXfERlF0/s1152/CSPP%20with%20Cap.PNG

Pretty impressive, what? The Diamond is arguably a bit better. But why bother? I think the inherent more natural thermal feedback of the CSPP may well render it more useable by a DIYer. Not being an engineer, I have asked, "Why the Diamond?" I may well be missing something.

In a proper setting, electrolytics are as good as they need be. But, to be used properly they need to be biased properly. They work very well when a polarizing voltage is applied as intended. That keeps the dielectric properly charged. For example: At the output of a single supply amplifier with a positive supply, the voltage will ideally be positive at one half the supply voltage. There, an electrolytic is just fine. But in an amp with a bipolar supply, the output should ideally be at zero volts. If so, there is no polarizing voltage for the electrolytic. In real life the output may be slightly positive or slightly negative so it is uncertain how to hook up the electrolytic. One could use back to back electrolytics, but of necessity one of them will be reverse biased. These uncertainties may exist at both the input and output of a circuit. Hence, the desire to use a film cap, or no cap at all. It is not a question resolved by the magical sonic qualities of film caps.

I see you have been bitten by the Diamond bug. It is beautifully symmetric, new to you, and has interesting possibilities. Something like that girl at the pub. But the purpose here is to put together a high quality HPA in the easiest and most inexpensive and efficient manner. Using the latest technology in the simplest way suggests extraordinarily low distortion with sufficient current and both thermal and current limitation in inexpensive OPA packages. Isn't that really the way to go? Especially since the CSPP and Diamond are more complex and more expensive?

BTW, on the O2 site, did you scroll down and look at the measurements? Objectivism at its finest. VERY IMPRESSIVE RESULTS!!!

Here is the URL again: http://nwavguy.blogspot.com/2011/07/...phone-amp.html
 

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
There is very little information available re: the use of the base to base capacitor. I only once saw a comment by Jung that said it was helpful. But caps are not typically used internally in OPAs due to difficulty of manufacture, etc.
Of course! The kind of value that's useful in audio applications is not difficult to manufacture on silicon; it's impossible!

That is (I assumed) the reason that the base connections were both brought out to pins on the LH0002 - for connection of an external capacitor!
The only other reference I found was in that Magic Diamonds blog. So I simmed it and found it worked very well in both circuits.
Yes, and for very good reasons!
[... snip ...]
Clearly, a DC feedback design will reduce offset. And I have no problem with feedback. It is a great janitor. But in some cases you do not want a closed loop from amp to buffer and back. For example, in a low noise HPA, one topology that works very well is to use a gain stage in front of the vol pot with the pot feeding the buffer. In that instance, you cannot close the loop with the pot in between. So you want the best buffer you can muster for that topology.
... or apply the feedback after the volume pot, i.e. a separate feedback loop for the output stage.
The use of a Schottky diode providing an additional fixed drop seems to be the addition of a non-necessary component. It is somewhat like using a more complex current source vs a resistor. Yes the current source is stiffer, but is it really necessary? If, the two resistor scheme can set current and voltage sufficiently well to get the job done, why bother with greater complexity unless dynamic distortion increases in the real world?
The reason for the Schottky diodes is to increase the quiescent current in the output stage. Assuming that the driver base-emitter drop roughly cancels the output transistor base-emitter drop, and you want to run the output transistors in something closer to class A, you need a source of voltage drop to correspond to the voltage drop across the emitter resistor on each side of the output. Then you set the output stage quiescent current by increasing those emitter resistors.

Of course increasing the output stage emitter resistors increases the output impedance; it might be better to increase them slightly, and add only part of a Schottky diode voltage drop on each side, using a trimpot or voltage divider to tap off part of the voltage drop.

I saw another idea on the diamond blog, I think it was. That's a single Schottky diode between the bases of the drivers. Of course that adds half a Schottky diode drop as an offset, which would have to be removed with feedback.

All just ideas to increase the quiescent output stage current.

According to my LTSpice simulations, the distortion figures for both circuits are well below 1% when properly configured. More like .001%. Real world measurements may well be worse. If you read the LH0002 datasheet, it specs the original simple buffer circuit at 0.1%.
That's without the base-to-base capacitor though, right? Even then I'm surprised - I mean, it's only emitter followers! Where is the distortion coming from? (Assuming it's not crossover distortion.) But then I've never really delved into distortion in transistor circuits so my gut feeling is probably wrong.
That chip is over 20 years old and of course has no B-B cap. But such a cap could be implemented from pin 6 to pin 10 on that chip. I have never seen it discussed though. I have a couple LH0002s on hand. May give them a shot!
I'd be interested to see how they affect the distortion figures.
You ask about the simms. A picture is worth a thousand words. Here are four of them.

Pretty impressive, what? The Diamond is arguably a bit better. But why bother? I think the inherent more natural thermal feedback of the CSPP may well render it more useable by a DIYer. Not being an engineer, I have asked, "Why the Diamond?" I may well be missing something.
As I said, if you've missed something, I'm missing it too :)
[...]But in an amp with a bipolar supply, the output should ideally be at zero volts. If so, there is no polarizing voltage for the electrolytic. In real life the output may be slightly positive or slightly negative so it is uncertain how to hook up the electrolytic. One could use back to back electrolytics, but of necessity one of them will be reverse biased. These uncertainties may exist at both the input and output of a circuit. Hence, the desire to use a film cap, or no cap at all. It is not a question resolved by the magical sonic qualities of film caps.
AFAIK electrolytics don't mind a bit of reverse bias.
I see you have been bitten by the Diamond bug. It is beautifully symmetric, new to you, and has interesting possibilities. Something like that girl at the pub. But the purpose here is to put together a high quality HPA in the easiest and most inexpensive and efficient manner. Using the latest technology in the simplest way suggests extraordinarily low distortion with sufficient current and both thermal and current limitation in inexpensive OPA packages. Isn't that really the way to go? Especially since the CSPP and Diamond are more complex and more expensive?
LOL :)
Well, your own simulations show that the diamond buffer with base-to-base cap performs better than the standard complementary push-pull design in terms of distortion. And I don't see how the diamond buffer has a problem with thermal stability.

I said before that I think it would be better to use soft clipping to limit the voltage rather than the current, but that's not fair. I think it depends on how the current limiting is implemented.
BTW, on the O2 site, did you scroll down and look at the measurements? Objectivism at its finest. VERY IMPRESSIVE RESULTS!!!
Sure! There are so many graphs that I lost interest the first time, but those sure are some impressively low noise and distortion figures. Mr. NwAvGuy does seem to know his stuff - actually I think he's a bit obsessive! I can't fault anything he has said; he knows a lot more about this stuff than I do.
 
Top