Minimal Headphones Amp

[thor.zmt]

The 990 can do 200+ mA into 75 ohm.

Yup, that is what I mean with sufficient drive current.

I usually aim for 0.7A peak repetitive and sustained and expect more than that as single pulses...

Why? Planar headphones from 13 to 50 Ohm with SPL between 86dB to 100dB/V.

These headphones can hit -80dB THD acoustically, so amplifier distortion should be -90...-100dB "simple" THD at several volt into low impedance.

If you ask me, with these planar headphones available all common dynamic types are declasse, incompetetive and only use for non quality critical tasks (eg. not suitable for tracking, mixing or mastering).

Thor

 

[ruffrecords]

If you ask me, with these planar headphones available all common dynamic types are declasse, incompetetive and only use for non quality critical tasks (eg. not suitable for tracking, mixing or mastering).

Thor

What do they cost?

Cheers

Ian

 

[Tubetec]

I find a few uses for small poweramps cable of driving moderate speaker levels
I have my tube HP amp with an internal speaker , when the cans are plugged out it becomes a reasonable approximation of an average speaker found in a domestic radio , its usefull as a spot check on how the sound might present on an open baffle cabinet with a full range drive unit.

I also use it to drive spring reverb tanks , including low impedence types and it does a spectacular job of it .

I think I had a look at planar H/p recently , was it a few hundred to get a basic set , seemed like much more expensive brands used the same drive unit , only better housing/headband etc
next step up was around 500.

Recreationally I never listen on headphones , I did snap up a set of Sennheiser HD 599 recently , missed out on a pair of HD660's going for silly money too because I didnt bite fast enough , he who hesitates is lost ,as they say .

 

[thor.zmt]

Not incorrect though is it ?

The datasheet covers unity gain operation. In my books that makes it unity gain stable, if the engineer can read.

Compare an OP37 which has no external compensation.

Thor

 

[thor.zmt]

What do they cost?

Depends.

Not too long ago Fostex T50RP went for around 100-120USD. The updated "MKIII" sells at ~ 170 USD.

These may benefit from some mods, but even stock are excellent. Including the real sheapskin memory foam replacement ear pads, replacement cables, damping material for the plastic and cotton for damping and the time needed we are looking at 500 USD, most of it in "sweat equity".

My two personal favourites, the Final D8000 and Abyss Diana which weigh in at 3,800 USD and 4,500 USD respectively. Of course, that is a totally different level.

Are the Final and Abyss 10-30 times better? You all know the answer. But as sarge used to say, if you got 'em, smoke 'em...

Thor

 

[Tubetec]

No way I could fund the sugar daddy lifestyle,
well above my pay scale ,
Luck would be a fine thing , a 20 something on my arm would be nice , but the daddies are fairly protective around these parts,
Im only the new kid in town yet , not even tested the waters ,
I still manage to turn a head or two with the touch of silver in my beard and rugged good looks ,
plenty of single moms in this town too , but whos the daddy(s) ,

I took a few pints au terrace the other day up in the city , there was a jaw dropping array of women from all corners of the planet , there may be hope for hope for us yet ,
let the mountain(s) come to the man ,was it Ali said ?

 

[Newmarket]

The datasheet covers unity gain operation. In my books that makes it unity gain stable, if the engineer can read.

Compare an OP37 which has no external compensation.

Thor

A datasheet may cover it but that simply means it can be made unity gain stable. But it is not unity gain stable "as it comes". And pretty much all opamps can be made UG stable with appropriate external components so your intention here are not clear unless it is to simply antagonise.
And for clarity - I am very familiar with 5534/5532 variants; OPA 627 / 637 et al and "back in the day" had some of the first OPA134 (so early in that the labelling was in pencil) across my desk from the BB field engineer.

 

[thor.zmt]

A datasheet may cover it but that simply means it can be made unity gain stable. But it is not unity gain stable "as it comes".

Externally compensated Op-Amp allow the circuit designer to adjust compensation. In effect they make a part commonly internal accessible external.

And pretty much all opamps can be made UG stable with appropriate external components so your intention here are not clear unless it is to simply antagonise.

Using noise gain manipulation has consequences that adjusting (miller) compensation doesn't have. For one, you only get unity gain at low frequencies.

The two cannot be equated in fundamental circuit operation.

One is the "legit" way to have a part that can have enhanced performance at high gains, that is stable with unity noise gain at all frequencies.

The other is a hack with a fair bit of potential for mischief that needs care, where the part always operates at non unity noise gain at high frequencies.


So, with a 5534 you can have a circuit that directly links out to -in and is stable, with OP37 this is not possible. An OP37 will NEVER operate stable at unity gain, it will ALWAYS have to operate

Ergo,

5534 = unity gain stable (* correctly compensated as per datasheet and intentional by design)

OP37 <> unity gain stable (* and requires an extra resistor in the feedback loop and an RC to a low impedance node to allow unity gain operation at low gains but cannot offer unity gain at high frequencies).

Really simple and very different.

True Unity Gain Stability is very different from faking noise gain.

QED.

Thor

 

[thor.zmt]

I took a few pints au terrace the other day up in the city , there was a jaw dropping array of women from all corners of the planet , there may be hope for hope for us yet ,
let the mountain(s) come to the man ,was it Ali said ?

Sounds good, I'm just a beach bum in South East Asia. You might be surprised what's possible here within your pay scale...

Thor

 

[Newmarket]

Externally compensated Op-Amp allow the circuit designer to adjust compensation. In effect they make a part commonly internal accessible external.

Yes. I know.

Using noise gain manipulation has consequences that adjusting (miller) compensation doesn't have. For one, you only get unity gain at low frequencies.

The two cannot be equated in fundamental circuit operation.

Yes. I know.

One is the "legit" way to have a part that can have enhanced performance at high gains, that is stable with unity noise gain at all frequencies.

The other is a hack with a fair bit of potential for mischief that needs care, where the part always operates at non unity noise gain at high frequencies.

Usage case obviously needs to be taken into account for any solution.

So, with a 5534 you can have a circuit that directly links out to -in and is stable, with OP37 this is not possible. An OP37 will NEVER operate stable at unity gain, it will ALWAYS have to operate

Are you missing the end of a sentence there ?

Ergo,

5534 = unity gain stable (* correctly compensated as per datasheet and intentional by design)

OP37 <> unity gain stable (* and requires an extra resistor in the feedback loop and an RC to a low impedance node to allow unity gain operation at low gains but cannot offer unity gain at high frequencies).

Really simple and very different.

True Unity Gain Stability is very different from faking noise gain.

QED.

Thor

Not really "QED" though is it ? Given that you have to caveat the 5534 case.
It's fine and I would prefer the compensation approach in general over using noise gain to achieve unity gain stability for signal.
But it remains the case that a 5534 itself is not unity gain stable.

 

[thor.zmt]

Agree with many of your comments but perhaps not your conclusions.

Up to you. I base them on my best understanding the results of MY testing.

We did ABC tests rather than ABX cos you get 'statistical significance' a lot faster.
...
And BTW, for those who have never tried to do DBLTs properly, it is VERY $$$$

There is a lot of room for debate, especially the underlying reasons for the excessively high frequency of failing to correctly reject the Null hypothesis aka "false negative" [or type 2 / type B statistical error] which has as much to do with applied psychology as it has with statics.

(TL;DR the test was insufficiently blind to remove bias and the bias acted as strong randomising agent which together with statistics that pace an emphasis on avoiding "false positives" [type 1/ type A statistical error] jointly obscured actually audible but small differences in fidelity impairments. Masking of fidelity impairments by poor quality transducers may also have played a role.)

It does however not belong into this thread. I'm happy to debate this to death if needed in a separate thread you are welcome to start. Tag me in, I'll reply as time allows.

Thor

 

[thor.zmt]

Are you missing the end of a sentence there ?

Yes, my lysdexia strikes again.

"I mean it will always operate with non-unity noise gain at sone frequencies to be stable!"

It's fine and I would prefer the compensation approach in general over using noise gain to achieve unity gain stability for signal.

Yup.

But it remains the case that a 5534 itself is not unity gain stable.

But it is designed for unity gain stable operation if applied according to the datasheet, in the very way the OP37 cannot. And it's direct dual version 5532 has internal unity gain compensation.

My "qualification" was especially for professional nitpickers, to pre-stall a comment on the external compensation being needed.

The uA709 is "unity gain stable" according to the datasheet, if the external compensation is set correctly. In fact, without external compensation it is not stable at almost any gain.

Again, QED.

At this time we are splitting already splitted hairs.

Or is that Hares?



Thor

 

[JohnRoberts]

You won that "straw man" argument....

Some of us are old enough to remember the ua741. The first unity gain op amp with compensation cap integrated onto the die. This opened up simplistic op amp design to a whole generation of circuit designers who didn't grok compensation.

JR

 

[thor.zmt]

Some of us are old enough to remember the ua741.

Some of us even remember the uA709 (first practical monolithic IC Op-Amp), LM101 and Siemens TAA761 & co on top of uA741. And the uA741 was really useless for audio.

Thor

 

[Newmarket]

Some of us even remember the uA709 (first practical monolithic IC Op-Amp), LM101 and Siemens TAA761 & co on top of uA741. And the uA741 was really useless for audio.

Thor

For hifi and pro audio yes.
I recall that a well liked line of guitar amps were stuffed full of them. Of course there's not much above a few kHz going on.

 

[JohnRoberts]

Some of us even remember the uA709 (first practical monolithic IC Op-Amp), LM101 and Siemens TAA761 & co on top of uA741. And the uA741 was really useless for audio.

Thor

The 709s were expensive hand grenades that would blow up if you looked at them sideways. I recall seeing them infrequently back in the late 60s (because of >$50 ea cost).

I read a technical paper in the IEEE journals back then about the significance of the ua741. The 0.5V/uSec was not exactly hifi fast and input noise was not notably low, but it was easy enough for junior engineers to design working circuits all by themselves. The 741 and 1458 duals were quite popular back then. I preferred LM301s for higher performance audio paths as long as I could use them inverting with feed forward compensation.

Slew rate is all relative. Its V/uSec so lower voltage rails can tolerate slower slew rate op amps. The entire -10 dBV community mostly survived slower slew rate devices thanks to modest operating audio signal voltages.

I wrote a piece for my old "audio mythology" column back in the 80s comparing +4 dBu audio paths to -10 dBV audio paths... The -10dBV paths scored respectably well, delivering more headroom than the more expensive "professional" +4 dBu paths.

JR

PS; By the late 70s inexpensive fast opamps like TL07x and NE553x made the op amp selection process much easier.

 

[thor.zmt]

The 709s were expensive hand grenades that would blow up if you looked at them sideways. I recall seeing them infrequently back in the late 60s (because of >$50 ea cost).

In East Germany we had a clone called A709 that was widely used into the 80's in non-audio applications. We had an A741 as well (uA741 copy). We also has an A07x (TL07x copy) range and A08x range.

And we had a range of singles to quad's of the TAA761, which was used for audio, where it was slow enough. The external compensation on the 709 meant it was useful in some situations but the Class B output needed to be biased into Class A. TL07x/8x also could use being biased.

Even well into the 80's we stuck to discrete circuitry for audio as most Op-Amp's we had were not great. We never got 5532/34 copies.

I read a technical paper in the IEEE journals back then about the significance of the ua741. The 0.5V/uSec was not exactly hifi fast and input noise was not notably low, but it was easy enough for junior engineers to design working circuits all by themselves.

0.25V/uS actually, on the original version. If there was any gain, the uA709 (A709 east german) was a better choice.

PS; By the late 70s inexpensive fast opamps like TL07x and NE553x made the op amp selection process much easier.

Especially 553x were a significant step forward, TL07X less so.

Better use an external J-Fet pair with a 553x if J-Fet inputs are needed, especially if running at much higher gain than unity.

Thor

 

[JohnRoberts]

Especially 553x were a significant step forward, TL07X less so.

Better use an external J-Fet pair with a 553x if J-Fet inputs are needed, especially if running at much higher gain than unity.

Thor

I used thousands of TL07x before working at Peavey, and truckloads after. IMO the TL07x is still respectable today when properly applied. The 553x had lower ein voltage but higher ein current so circuit impedance matters. Again used properly those old soldiers are still OK today.

The way I tend to look at them is that they both were way faster than needed for audio.

JR

 

[thor.zmt]

I used thousands of TL07x before working at Peavey, and truckloads after.

Very noisy. Low output stage Iq. I have also used them but usually not for audio.

IMO the TL07x is still respectable today when properly applied.

Try OPA1678/79 instead.

The price difference is marginal (25 cent in 1kU directly from TI), noise approaches 5532 with a Fet Input, (T)HD into 2k is very low, 9V/uS Slew Rate, 16MHz GBWP, there is no phase-reversal when clipping, output current is +/-35mA, output is "rail to rail" and input common mode range is very wide.

As Audio-Op-Amp it basically replaces NE5532 and TL072/074 with one much superior part. For very low input noise external low noise pairs can be added, where higher current or loads below 600R are needed a simple discrete buffer can handle that.

The 553x had lower ein voltage but higher ein current so circuit impedance matters. Again used properly those old soldiers are still OK today.

Depends on the definition of OK. I think the TL072 is ok as DC servo, but given I stock OPA1678/79 anyway it easier to just these and they have less offset.

The way I tend to look at them is that they both were way faster than needed for audio.

That is very conditional. Once we get to digital audio and other modern stuff, this may not be true any longer, though technically speaking the ultrasonic noise from DS modulators or the very high edge rate current steps from multi-bit DAC are not "audio", but they exist in parallel and are mixed with the audio signal.

Thor

 

[ruffrecords]

Some of us even remember the uA709 (first practical monolithic IC Op-Amp), LM101 and Siemens TAA761 & co on top of uA741. And the uA741 was really useless for audio.

Thor

That we do. Way back in 1969 I started my undergraduate apprenticeship with British Aircraft Corporation (now BAe) in Stevenage. One of the first things I got wo work on was a chopper stabilised dc amplifier that was chock full of 709s. I remember it was also common practice to improve their performance by fitting one of more discrete transistors at the front end.

Cheers

Ian