New Posts  All Forums:Forum Nav:

The Opamp thread - Page 265

post #3961 of 4752

You don't become an EE overnight basically. You may wanna have a look at this document, especially regarding AD797: http://e2e.ti.com/support/amplifiers/precision_amplifiers/w/design_notes/upgrading-op-amps-in-audio-equipment.aspx

 

That's AD797 in the ONE: http://www.head-fi.org/t/542563/asus-essence-one-headphone-amp-dac-cebit-2011/810#post_8300531

 

Gluing a giant heatsink on top of your DAC is *not* a practical solution ^^

 

Anyway, I believe we've beaten this subject to death.....blind careless rollers can keep rolling mindlessly if they like, nothing wrong w/ that beerchug.gif

post #3962 of 4752

You cant blame the datasheet for that, there is a huge difference between using 1 opamp per phase in a PCB designed for it with a proper ground plane, well compensated and driving the rather benign high impedance load of an amplifier input (the dac before it has nothing to do with it) and stacking 4 per channel in a PCB not designed for it, without supporting components and driving a headphone load which is many orders of magnitude harder than the high impedance of the amp. worlds apart.

 

the first rule you need to learn is ohms law, its at the heart of almost any design. its a common mistake around here and there is a certain amount of misinformation surrounding it, that a high impedance load is harder to drive than a low impedance load. In fact the opposite is true, provided you have the voltage swing in the power supply plus losses, a very high impedance draws very little current and hardly loads the amplifier at all.

 

A low impedance load requires a LOT more current and is far harder to drive in a linear fashion, to the point many amps will actually go into current limiting, roll off the high current bass lines and distort. it creates much more heat and stresses the chip. the lower the load impedance, the closer to a short circuit it becomes.

 

think of it like connecting a resistor across the output of your amp, think of this as the headphone load, (impedance and resistance are similar) in fact for load testing this is exactly what many designers will do if they dont own an electronic load. so the less resistance in that resistor, the more the amp wants to push current across, lower and lower and lower until you have zero impedance/resistance, in other words a short circuit, almost pure current; we all know a short circuit isnt an easy load on an amplifier is it? not good

 

probably the first thing that would give if you kept using it like you were, is the caps surrounding it, you can bet they didnt use 105c caps, more like 85c max, because they presume someone will cease operation if the amp is too hot to hold..... audiophile approved caps are rarely 105c, so they will eventually dry up and fail, if you are lucky they will fail closed and short the amp, maybe trigger a protection circuit, if they fail open you could see the full power supply rail DC at the headphones

 

but I did not say anything of the sort, roll to your hearts content, but keep your eyes open and if something doesnt seem right, chances are it ISNT right

 

would you drive a Ferrari around Le mans in 1st gear, revving the crap out of it just because it was making a fantastic noise?


Edited by qusp - 6/12/12 at 9:25am
post #3963 of 4752

That reminds me of me raving about the AD797B as I/V and LT1028AC as buffer on the HD2.......killer SQ considering that it's using the AK4396 "miracle DAC" w/ two discrete clocks for 44.1/48kHz multiples happy_face1.gif

 

Then one day out of the blue, AD797 completely refused to work anymore, and then LT1028 shortly afterwards...AD797 would make a very loud whining noise and whatever 1028C or AC would give some very high pitched sound for a second and then become dead silent =/

 

The board worked fine w/ UGS opamps, mind you!

 

I poisoned quite a lot of ppl to try this set up and one of them came up w/ a dirty fix involving resistors in order to lower the gain of the buffer to -1 hence making LT1028 happy. It was a success apparently! But all boards died a few months later duh ^^


Edited by leeperry - 6/12/12 at 10:02am
post #3964 of 4752

qusp and leepery,

 

Thanks again for your comments.

 

For the record, I wasn't blaming the data sheet for my decision to try the AD797s in the buffer stage - I was blaming my inability to interpret the datasheets - and the inability of the vast majority of people who roll op-amps.  

 

In other words, a datasheet is of little value to a guy like me.  I know it must be hard for the more knowledgeable people to dumb down to my level, but seriously, easily half of what you write, goes right over my head, and yes, I know that it takes a lot of time and reading to come up to speed on this stuff.  I'm trying. I want to learn.  tongue.gif

 

Even when I wrote that I was going to smoke my Cuban cigars, I just meant that I will probably continue to roll op-amps WITHOUT the knowledge and test equipment I lack - because it's fun and I CAN hear differences.  I didn't mean that I would continue to use the AD797s as buffers with the case getting so hot.  Believe me, I respect your warnings and do not want to fry my LCD-2s.

 

Meanwhile, I decided to send an e-mail to iBasso, asking them about use of AD797s in the buffer stage and the overheating problem.  (Note that iBasso ships the PB2 with AD797s included their rolling kit.)

 

Here's my e-mail to iBasso:

 

Quote:
Date: Tue, 12 Jun 2012 00:17:50 +0000 (GMT)
To: iBasso Audio <service@ibasso.com>
Subject: Questions about use of AD797 in buffer stage of PB2
 
Hi!
 
When I use AD797 in the buffer stage of my PB2 (with OPA1612s in the
voltage gain stage), the PB2 gets very warm - almost hot.  I have to open
the lid to keep it cool.
 
I really like the sound I get with this combination, though - OPA1612 with
AD797 as buffers.
 
The AD797 datasheet says they can be used as buffers, but I don't know if
that's true with the PB2.
 
I do not hear any oscillations when listening to music, but I do not have
an oscilliscope with which to test for oscillations at high frequenicies.
 
I have not added any bypass capacitors.
 
Questions:
 
1) If I operate the PB2 with the lid removed, to keep the amplifier
cool, do you think it is it safe to operate the PB2 with OPA1612 in L/R
and AD797 as buffers?  (No damage to amplifier or headphones?)
 
2) Do you think the AD797 are stable (no oscillations) when used without
bypass capacitors in the voltage gain stage?
 
3) Do you think the AD797 are stable (no oscillations) when used without
bypass capacitors in the buffer stage?
 
Thank you!
 
Mike

 

And here is their response:

 

 

Quote:
Date: Tue, 12 Jun 2012 21:44:12 +0800
From: iBasso Audio <service@ibasso.com>
Subject: Re: Questions about use of AD797 in buffer stage of PB2
 
Hi,
 
Before I answer your questions, I have to state that
 
1, The AD797 produce heat during operation.
 
2, I am not sure what "bypass capacitor" you are mentioning.  In our
understanding, buffer stage doesnt need bypass capacitors.
 
1) If I operate the PB2 with the lid removed, to keep the amplifier
cool, do you think it is it safe to operate the PB2 with OPA1612 in L/R
and AD797 as buffers? (No damage to amplifier or headphones?)
 
-----Can you measure the temperature. We dont know how hot it is.
 
2) Do you think the AD797 are stable (no oscillations) when used without
bypass capacitors in the voltage gain stage?
 
------Yes.
 
3) Do you think the AD797 are stable (no oscillations) when used without
bypass capacitors in the buffer stage?
 
------Yes. However, the output current of the AD797 is only 50mA. It isnt
an ideal buffer.
 
Sincerely
iBasso Audio

 

 

 

So, according to iBasso, it's normal for the AD797 to run hot, but they want to know how hot, even with the lid off.  They don't believe I need to use bypass capacitors on the AD797s in either the voltage gain stage or the buffer stage - iBasso believes the AD797 will be stable at either location - but it sounds is if they're saying the AD797 isn't an ideal buffer because its output current is only 50 mA.  (Which doesn't bother me, as I'm using LCD-2, not something current hungry like the HE-6.)

 

Now, I guess I have to buy one of those those laser thermometers, if I want to continue this exchange with iBasso.  Is there a better way (a right way) to measure the temperature of the AD797s?  

 

Can anyone think of any other questions I should ask iBasso?

 

Does anyone think this guy at iBasso might be steering me wrong?

 

Thanks!

 

Mike

post #3965 of 4752
Quote:
Originally Posted by leeperry View Post

That reminds me of me raving about the AD797B as I/V and LT1028AC as buffer on the HD2.......killer SQ considering that it's using the AK4396 "miracle DAC" w/ two discrete clocks for 44.1/48kHz multiples happy_face1.gif

 

Then one day out of the blue, AD797 completely refused to work anymore, and then LT1028 shortly afterwards...AD797 would make a very loud whining noise and whatever 1028C or AC would give some very high pitched sound for a second and then become dead silent =/

 

The board worked fine w/ UGS opamps, mind you!

 

I poisoned quite a lot of ppl to try this set up and one of them came up w/ a dirty fix involving resistors in order to lower the gain of the buffer to -1 hence making LT1028 happy. It was a success apparently! But all boards died a few months later duh ^^

oh god, Lol I have a long list of opamps I was about to buy for my STX that you used to roll a few years ago, especially the LT ones. good thing I trimmed it down lol. I'll still come ask for confirmation here before I buy anyway. I love the AD797 in my STX with my AKG K242HD, it really does sound good, but the treble makes you bleed, probably from oscillation. So I took it out. Now running THS4032, its the most neutral, suits my Denons and the AKG's without the harsh treble and an overall smooth signature. But they are also very high bandwidth, and correct me if I'm wrong but that generally spells a recipe for oscillation? arg Tough crap, not taking them out, they sound too good.

post #3966 of 4752
Quote:
Originally Posted by qusp View Post
I wouldnt trust all the numbers on the front page. for the GBW numbers they give

110 MHz gain bandwidth (G = 1000)

8 MHz bandwidth (G = 10)

which doesnt make any sense, negative feedback as a process uses up gain to extend bandwidth and vice versa, feedback used to increase voltage gain will LOWER bandwidth, so the bandwidth for G=10 above should be higher, not lower than the bandwidth for G=1000. i'm not the first to mention this error either.

That's not an error, that statement is correct, Analog Devices staff are not so silly.

110/10=11MHz so 8MHz is jus OK since no opamp is ideal.

 

Quote:
Originally Posted by leeperry View Post
That reminds me of me raving about the AD797B as I/V and LT1028AC as buffer on the HD2...

LT1028 shouldn't be used as buffer without further compensating: see datasheet page 7 - figure "Gain, Phase vs. Frequency" (top middle).

The phase curve crosses 0° sooner than gain curve crosses 0dB. For stable operation there should be a phase margin of ~45°.

post #3967 of 4752

I have a question. A stupid question but humour me.
Opamps sound different in different conditions right, so if they sound great in non-ideal conditions and oscillate and start damaging the circuit, and then you compensate them correctly and all the components are happy, could it sound vastly different and even bad?

post #3968 of 4752
Quote:
Originally Posted by AmarokCZ View Post

That's not an error, that statement is correct, Analog Devices staff are not so silly.

110/10=11MHz so 8MHz is jus OK since no opamp is ideal.

 

LT1028 shouldn't be used as buffer without further compensating: see datasheet page 7 - figure "Gain, Phase vs. Frequency" (top middle).

The phase curve crosses 0° sooner than gain curve crosses 0dB. For stable operation there should be a phase margin of ~45°.

so you claim that lower gain has lower bandwidth? that certainly goes against every other datasheet or reference ive read, or build manuals i've used, thats the point, gain uses bandwidth or do you think they are not quoting gain bandwidth in the second line? it didnt make sense to me that an error would stand for this long either, but I found references to this going back 6 years including some confusion by the designer of the lisa III it also doesnt match the rest of the datasheet.

 

i'm open to all of us being confused though, but I do wish there was some consistency in datasheets between manufacturers

 

 

first link that came up, at ecircuit center, so there is something in the way the DS is layed out, or the way i'm interpretting it thats odd

 

Quote:
Quote:

GAIN-BANDWIDTH-PRODUCT

The term fu is also called the the Gain-Bandwidth-Product (GBP). Why call it this? If we rearrange the above equation, we get

fu = GBP = GN x fc

Notice, that the product of gain GN and bandwidth fc is constant and bounded by GBP (fu)!

What does this mean? You can't arbitrarily set the gain and bandwidth for a given op amp. Increase the gain GN, and the bandwidth fc will drop to keep GBP constant. Alternatively, if you need a higher bandwidth, then you must choose a lower gain. If you need both higher gain and bandwidth, you're out of luck with this device. You need to pick an op amp with a higher GBP (fu) on its data sheet.

 


Edited by qusp - 6/12/12 at 11:55am
post #3969 of 4752

Qusp,

 

Quote:
Originally Posted by qusp View Post

zilch0md, get yourself a cheap ebay thermal camera, or heat sensor and find out exactly what temp they are running, add some for the thermal impedance of the opamp package to opamp die and check the datasheet for SOA (safe operating area) wrt temp.

 

[snip]

 

if you can get the temp right down, not just of the case, but the opamp die itself, you may not kill the amp, but as is you will be shortening the life of the chips and all of the caps surrounding them. is the amp DC coupled? (no caps on the output) if it is, if they fail open or less than gracefully, you could feed the DC rail voltage directly to the headphones

 

You wrote this four days ago - well ahead of my e-mail to iBasso.  Cheers!

 

What is "a heat sensor" (in the context quoted)?

 

Thanks,

 

Mike

post #3970 of 4752

Thanks for the interesting posts I have to catch up on, some quick questions...

 

 

Originally Posted by qusp View Post

 

[/] the blinkered chase for high slew rate and wide bandwidth at all costs is not all that relevant for our purposes

 

You mean like horse blinkers?  The OPA2111KP seems pretty popular and I like it, IIRC it has a pretty low slew rate.

 

 

Originally Posted by leeperry View Post

You don't become an EE overnight basically. You may wanna have a look at this document, especially regarding AD797: http://e2e.ti.com/support/amplifiers/precision_amplifiers/w/design_notes/upgrading-op-amps-in-audio-equipment.aspx

 

Interesting list, how come the 1611 and 827 aren't very popular?

 

 

Originally Posted by leeperry View Post

That reminds me of me raving about the AD797B as I/V and LT1028AC as buffer on the HD2.......killer SQ considering that it's using the AK4396 "miracle DAC" w/ two discrete clocks for 44.1/48kHz multiples happy_face1.gif

 

What's the HD2?  I sortof like the sound of AK4396 -> LT1028...

post #3971 of 4752
Quote:
Originally Posted by WiR3D View Post

I have a question. A stupid question but humour me.
Opamps sound different in different conditions right, so if they sound great in non-ideal conditions and oscillate and start damaging the circuit, and then you compensate them correctly and all the components are happy, could it sound vastly different and even bad?

Depends what is bad for you, because everyone has different taste, many people likes intentionally induced distortion (and instable amp is more distorting)

Sure thing is that this condition is bad for the circuit and it's against basic philosophy.

 

Quote:

Originally Posted by qusp View Post
so you claim that lower gain has lower bandwidth? that certainly goes against every other datasheet or reference ive read, or build manuals i've used, thats the point, gain uses bandwidth or do you think they are not quoting gain bandwidth in the second line?

Yes, lower gain certainly doesn't have lower bandwith.

On first line there is GBP at x1000 gain = 110MHz

Second line is just -3dB bandwith (not GBP) at x10 gain which is stated as 8MHz that means the GBP would be 80MHz.

 

It surely is odd and confusing since the first line is GBP and second line isn't, but it's not an error.

post #3972 of 4752
Quote:
Originally Posted by zilch0md View Post

Qusp,

 

 

You wrote this four days ago - well ahead of my e-mail to iBasso.  Cheers!

 

What is "a heat sensor" (in the context quoted)?

 

Thanks,

 

Mike

 

yeah, I did didnt I hehe. ibasso's claim that buffers dont need bypass caps and that multiple AD797 used as a buffer is a good idea, with the only limit being the current, is confusing for such a technology company, i'd say the person who replied only looked as far as the line that says its unity gain stable and may not have actually used the chip.

 

the fact its in a FB loop doesnt mean its all of a sudden another opamp, we have demonstrated that it is proven mathematically to be unstable. Even chips designed to be buffers should have the rails bypassed/decoupled individually, they should also really have current sharing resistors. When all connected in parallel, there will be a tendency for some to hog the current more than others.

 

there is a reason Walt Jung stopped recommending AD797 for his super regulators, even though when stable its still pretty hard to beat even today. its plain to me they are not happy and they should not even be considered for this application in the first place, so continuing along this track even now seems ridiculous to me

 

as for the sensor, I mean not a camera, just an infrared heat sensor/thermometer rather than a camera if you want to save a bit. I cant find the link to the cheap thermal cameras I had, most on ebay are OTT for this, but the above will do

Quote:
Originally Posted by kiteki View Post

Thanks for the interesting posts I have to catch up on, some quick questions...

 

 

 

You mean like horse blinkers?  The OPA2111KP seems pretty popular and I like it, IIRC it has a pretty low slew rate.

 

 

 

Interesting list, how come the 1611 and 827 aren't very popular?

 

 

 

What's the HD2?  I sortof like the sound of AK4396 -> LT1028...

 

yes Horse blinkers, its a figure of speech, ie. pushing ahead at speed without looking around. The problem is, often massive slew rate is chased while thinking of it somehow being meaningful to having a responsive amp at audio frequencies, wide bandwidth and high slew rate are good indicators that the chip will be sensitive to layout and other conditions. These types will often need a lot of NFB and heavy compensation to be stable at low audio bandwidth, low gain applications, so any advantage can easily be lost.

 

if you like LT1028 in this application and arent able to check performance, try buffering or following it with a discrete follower, I like mosFETs, but some of the modern bipolars and definitely the power jfets are superb. I dont know why its not more popular either, personally I love opa827 and IMO its one of if not the best voltage feedback opamp for Dac IV stages, because of its stupidly low noise, high impedance jfet input stage and incredible AC performance, VERY low input bias current, which makes it great for the input of DC coupled circuits.

 

the opa1611 wont be very good without a buffer, has bugger all current, same with the opa1641, great chip and I love jfet opamps, but its like they forgot the output stage on this one, simply pissweak output power, so would need a buffer. I keep coming back to the lme49990 and I do like the 827 a lot, just dont have enough of them to throw around and its not a cheap one

 

IV stages are a bit of a different case, here I would champion the use of a very fast current feedback opamp designed for video mostly, as the glitch energy of delta sigma type dacs in particular is at very high speed, but you cant just drop current FB chips in place of voltage FB.

Quote:
Originally Posted by AmarokCZ View Post

Yes, lower gain certainly doesn't have lower bandwith.

On first line there is GBP at x1000 gain = 110MHz

Second line is just -3dB bandwith (not GBP) at x10 gain which is stated as 8MHz that means the GBP would be 80MHz.

 

It surely is odd and confusing since the first line is GBP and second line isn't, but it's not an error.

 

yeah youre right, the numbers make sense with that in mind, but its sure an odd way to present the info and has fooled smarter men than me. I just glanced straight past that as both being GBP until looking at it again when writing that reply above


Edited by qusp - 6/13/12 at 11:56am
post #3973 of 4752

qusp,

 

Quote:
Originally Posted by qusp View Post

yeah, I did didnt I hehe. ibasso's claim that buffers dont need bypass caps and that multiple AD797 used as a buffer is a good idea, with the only limit being the current, is confusing for such a technology company, i'd say the person who replied only looked as far as the line that says its unity gain stable and may not have actually used the chip.

 

the fact its in a FB loop doesnt mean its all of a sudden another opamp, we have demonstrated that it is proven mathematically to be unstable. Even chips designed to be buffers should have the rails bypassed/decoupled individually, they should also really have current sharing resistors. When all connected in parallel, there will be a tendency for some to hog the current more than others.

 

there is a reason Walt Jung stopped recommending AD797 for his super regulators, even though when stable its still pretty hard to beat even today. its plain to me they are not happy and they should not even be considered for this application in the first place, so continuing along this track even now seems ridiculous to me

 

as for the sensor, I mean not a camera, just an infrared heat sensor/thermometer rather than a camera if you want to save a bit. I cant find the link to the cheap thermal cameras I had, most on ebay are OTT for this, but the above will do

 

[snip]

 

 

Just today, I received another e-mail from iBasso in response to my asking if the PB2 is DC coupled:

 

Quote:
Date: Thu, 14 Jun 2012 01:00:25 +0800
From: iBasso Audio <service@ibasso.com>
Subject: Re: Questions about use of AD797 in buffer stage of PB2
----------------------------------------
 
Hi,
 
The PB2 doesnt have DC coupling cap. It doesnt need it at all.
 
We dont use cap in the signal path (output).
 
Sincerely
iBasso Audio

 

Forgive me, but I don't understand this well enough to even tell if what iBasso has written here conflicts with what you've written above.   Please comment - I trust you!

 

Meanwhile, I've taken the plunge and ordered one of these - just so that I can answer iBasso's outstanding question:  "How hot are the AD797s?"  

 

Raytek_MiniTemp_MT6_-_Same_as_Fluke_62.jpg

 

Raytek MT6 Non-contact MiniTemp Infrared Thermometer 

 

         Temperature range: -30º to 500ºC

 

(This is identical to the Fluke 62, but sells for about half as much.)

 

I'm not hell-bent on using the AD797s, but I am intensely curious as to what iBasso will say when I report the temperature, and I figure I can find other reasons for this tool - like telling from a distance whether or not my wife is having a hot flash!  tongue.gif  

 

No, seriously, I intend to use it for comparing op-amp running temps to the operating ranges given in their datasheets.  Doh!

 

I'm currently using HA5002s in the buffer stage (instead of AD797s or dummy buffers), along with the OPA1612s in L/R - and everything is running cool.   (How cool?  I will soon know...)

 

Thanks again for all your tutelage,

 

Mike

post #3974 of 4752
Originally Posted by qusp View Post

 

yes Horse blinkers, its a figure of speech, ie. pushing ahead at speed without looking around. The problem is, often massive slew rate is chased while thinking of it somehow being meaningful to having a responsive amp at audio frequencies, wide bandwidth and high slew rate are good indicators that the chip will be sensitive to layout and other conditions. These types will often need a lot of NFB and heavy compensation to be stable at low audio bandwidth, low gain applications, so any advantage can easily be lost.

 

if you like LT1028 in this application and arent able to check performance, try buffering or following it with a discrete follower, I like mosFETs, but some of the modern bipolars and definitely the power jfets are superb. I dont know why its not more popular either, personally I love opa827 and IMO its one of if not the best voltage feedback opamp for Dac IV stages, because of its stupidly low noise, high impedance jfet input stage and incredible AC performance, VERY low input bias current, which makes it great for the input of DC coupled circuits.

 

the opa1611 wont be very good without a buffer, has bugger all current, same with the opa1641, great chip and I love jfet opamps, but its like they forgot the output stage on this one, simply pissweak output power, so would need a buffer. I keep coming back to the lme49990 and I do like the 827 a lot, just dont have enough of them to throw around and its not a cheap one

 

IV stages are a bit of a different case, here I would champion the use of a very fast current feedback opamp designed for video mostly, as the glitch energy of delta sigma type dacs in particular is at very high speed, but you cant just drop current FB chips in place of voltage FB.

 

What's a discrete follower?  What are the intricate differences between jfet, difet, mosfet?  When driving headphones directly, do op-amps have different output impedance levels?  I can't find them in the datasheets.

post #3975 of 4752
Quote:
Originally Posted by zilch0md View Post

qusp,

 

 

Just today, I received another e-mail from iBasso in response to my asking if the PB2 is DC coupled:

 

 

Forgive me, but I don't understand this well enough to even tell if what iBasso has written here conflicts with what you've written above.   Please comment - I trust you!

 

Meanwhile, I've taken the plunge and ordered one of these - just so that I can answer iBasso's outstanding question:  "How hot are the AD797s?"  

 

.....snip

 

Thanks again for all your tutelage,

 

Mike

ok mate, no worries, I reckon i'd be pretty curious too by this stage. Being DC coupled is a good thing for audio, particularly headphones as output caps tend to need to be large to allow driving low impedance headphones without rolling off the bass. they form a high pass filter at the output in combination with the headphone impedance and the amps output impedance (I will call impedance Z from now on). Electrolytics are not so linear and have lower damping factor (means high output Z) so the amp does not have quite as good electro-mechanical control of the driver, not such an issue with headphones some would say, due to being physically small, but I disagree.

 

DC coupling is not without its caveats though, the output caps help to protect the headphones from DC at the output if the amp, so the fact its a DC coupled amp, means although audio performance is good, if there is a failure condition in the amp that creates DC your headphones will see all of it.

 

indeed the temp sensor is handy for all sorts of things and most are pretty cheaply priced, so you didnt buy the fluke? would have been risky on ebay, fluke gear on the bay is often fake, as its a well known big brand.

New Posts  All Forums:Forum Nav:
  Return Home