Not using too excessive bandwidth for diy audio.
.
My ironi aside,
John, in what way complicated ?
.. what is it you find so complicated
.. or do you just don't know what to say on this matter
You can do better than this, man, I sincerly hope
Maybe you are unsure what to say, to argue,
as not using too excessive bandwidth for diy audio is contradictionary
to your current beliefs & obsession in circuitry.
(john is ultra fast, ultra slewrate audio advocate, as far as i know)
Lineup
.
how enlightning, don't you think fellasjohn curl said:
My ironi aside,
John, in what way complicated ?
.. what is it you find so complicated
.. or do you just don't know what to say on this matter
You can do better than this, man, I sincerly hope
Maybe you are unsure what to say, to argue,
as not using too excessive bandwidth for diy audio is contradictionary
to your current beliefs & obsession in circuitry.
(john is ultra fast, ultra slewrate audio advocate, as far as i know)
Lineup
unclejed613 said:
I couldn't tell you if it is oscillating or not. I have used both the THS4131 and OPA1632 in a handful of different circuits and they all get hot. I use proper decoupling, so if all the circuits I built have oscillation problems then I don't know how to remedy it. Even if they are oscillating, they definitely do not sound "bad" at all; quite the opposite.
These IC's were obviously made to run hot or they wouldn't be available in a POWERpad package.
if the oscillation frequency is high enough, and at a somewhat low level, it might not impact the sound a whole lot, but could still make the chip draw more current. if you're using an op amp in an SMT package, it will get hotter than it's DIP counterpart, since there's less physical mass and surface area to dissipate heat. if it has a thermal pad on it, then i would heat sink it.
the only reason i mentioned GPB wasn't to distract from the original question about diff inputs/outputs, but to answer the question that came up later about the op amp running hot.
actually that wasn't even the original question.... the original question was whether monolithic op amps could measure up to discrete op amps, which in many ways they are actually better than discretes, but discretes also have some things in their favor, so it really comes down to what characteristics are more important. if you want simplicity, low offset, small footprint, high CMRR, PSRR, and open loop gain, use a monolithic. if you want total control over the open loop gain, slew rate, compensation, output stage type, and other methods of optimization, go with a discrete.
the only reason i mentioned GPB wasn't to distract from the original question about diff inputs/outputs, but to answer the question that came up later about the op amp running hot.
actually that wasn't even the original question.... the original question was whether monolithic op amps could measure up to discrete op amps, which in many ways they are actually better than discretes, but discretes also have some things in their favor, so it really comes down to what characteristics are more important. if you want simplicity, low offset, small footprint, high CMRR, PSRR, and open loop gain, use a monolithic. if you want total control over the open loop gain, slew rate, compensation, output stage type, and other methods of optimization, go with a discrete.
I looked at the THS4130/1.
Its a current feedback op amp, so very wide bandwidth which is fairly independent of loop gain
Standing current is 13.5mA a t c. 25C. This is 0.39W on +-15V supplies
On the 2 packages I looked at in the data sheet, the junction to case temperature was in the region of 40-54 degrees. There is a very small package with 260C Junct to ambient - probably only usefull on the very low supply rails
So, in normal use, this device will get quite warm, even into a high impedance load. If you are driving a heavy load (e.g. 600 Ohms + cable capacitance, it will get even hotter - so this is why it comes with with a header pad at the bottom of the chip.
The frequency response does peak out at about 30 or 40 Mhz - I dont know that this is a problem for audio. With a CFA topology, you usually cannot put capacitors around the feedback elements - it causes instability. Best way to get a 'flat' response if this indeed is a concern (as I said its out at 40Mhz), then a simple RC amplifier in front of the amp is what is needed - and for CFA this is the usual recomendation.
I think this device is designed primarily for high speed video and communications applications driving typically an A-D converter. It would not be my ideal choice for a balanced amp for audio - probably better to go for a VFA Instrumentation amp - I think AD do some quite nice ones and LT as well. You can certainly ge t better noise performance as well.
BTW, this parts Iq also increases with temperature, so in most cases I would expect the quiescent current to quite quickly go to the max level of c. 16mA - check out the power vs ambient temperature graphs in the data sheet.
Its a current feedback op amp, so very wide bandwidth which is fairly independent of loop gain
Standing current is 13.5mA a t c. 25C. This is 0.39W on +-15V supplies
On the 2 packages I looked at in the data sheet, the junction to case temperature was in the region of 40-54 degrees. There is a very small package with 260C Junct to ambient - probably only usefull on the very low supply rails
So, in normal use, this device will get quite warm, even into a high impedance load. If you are driving a heavy load (e.g. 600 Ohms + cable capacitance, it will get even hotter - so this is why it comes with with a header pad at the bottom of the chip.
The frequency response does peak out at about 30 or 40 Mhz - I dont know that this is a problem for audio. With a CFA topology, you usually cannot put capacitors around the feedback elements - it causes instability. Best way to get a 'flat' response if this indeed is a concern (as I said its out at 40Mhz), then a simple RC amplifier in front of the amp is what is needed - and for CFA this is the usual recomendation.
I think this device is designed primarily for high speed video and communications applications driving typically an A-D converter. It would not be my ideal choice for a balanced amp for audio - probably better to go for a VFA Instrumentation amp - I think AD do some quite nice ones and LT as well. You can certainly ge t better noise performance as well.
BTW, this parts Iq also increases with temperature, so in most cases I would expect the quiescent current to quite quickly go to the max level of c. 16mA - check out the power vs ambient temperature graphs in the data sheet.
Please understand that I could have used that same argument 40 years ago. In fact, I believed it myself. All that I had to do then was wait until better IC's were produced. It has now been 32 years since I first measured and 31 years since I A-B listener tested the 5534. It just isn't as good as a properly designed fet input discrete op amp. That is the point that I am making and typical objective measurements won't tell you why, either. I sincerely wish that they would. It is NOT TIM either, the 5534 is fast enough, it is something else. Perhaps very low levels of higher order harmonic distortion, or perhaps a high frequency oscillation (a favorite theory of the late Dean Jensen) or it could be PIM. I just don't know for sure.
I will say that we used to modify selected 5534's with a fet input stage and sold them to Dave Wilson for his WAMM speaker system. Dave could hear the difference, and maybe some of you could too.
I will say that we used to modify selected 5534's with a fet input stage and sold them to Dave Wilson for his WAMM speaker system. Dave could hear the difference, and maybe some of you could too.
Hello John,
Your comments on the 5534/2 are interesting. Used them in a design a good few years back now. The prototype used TL071/2 and was great but in the final build I used the NE's.
Now a particular piece of music I know well - It was a Mozart piano concerto actually - didn't sound right, in fact the whole preamp didn't sound as good as the prototype. A particular bar in the music, and I could put the CD on repeat to check had some "extra" sound in there as the piano and strings faded out. It was like a "false string" on a piano almost, very unpleasant sounding. It was the OpAmps. Changing back to TL072's and harmony was restored. It was a real eye ( should that be ear ) opener.
Your comments on the 5534/2 are interesting. Used them in a design a good few years back now. The prototype used TL071/2 and was great but in the final build I used the NE's.
Now a particular piece of music I know well - It was a Mozart piano concerto actually - didn't sound right, in fact the whole preamp didn't sound as good as the prototype. A particular bar in the music, and I could put the CD on repeat to check had some "extra" sound in there as the piano and strings faded out. It was like a "false string" on a piano almost, very unpleasant sounding. It was the OpAmps. Changing back to TL072's and harmony was restored. It was a real eye ( should that be ear ) opener.
john curl said:
Interestingly the Ambrosia preamp made by James B. is using an AD797 with FETs upfront.
http://ampzilla2000.com/Ambrosia Superior Audio Review.htm
These are valid points.
I also see a lot of people swapping op-amps in and out of circuits (e.g. trying one type in place of another) when in actual fact its not quite as easy as that.
For example, don't put a non-unity gain stable op-amp in a circuit where the gain is less than the manufacturers recommendation. You will get instability issues - and og course that leads to 'glassy sound', edgy sound and a whole lot of other problems
Some op-amp circuits drive stuff in the outside world directly - i.e. op-amp output directly to the output connector. Bad practice since you can never be sure about the capacity load you are looking into, always isolate the op-amp output with a low value resistor - 100 Ohms is good in most cases (but check it in practice!)
For power supply decoupling and layout dressing, similar comments apply.
Another example of bad practice in op-amps (applies to discrete ccts as well) is where the feedback devider is placed a long way from the -ve op-amp input. I've seen this in quite a few cases in my designer days. This is a recipe for noise, oscillation and a whole lot of other ills. Because IC op-amps tend to have high loop gains, its especially bad - you cannot get away with it.
On the input side, also be careful about noise and EMI ingress - devices behave differently because the input topologies and structures are not the same from one type to another.
Bottom line is, a competent designer who chooses to use an op-amp will optimise the circuit for that op-amp, so just changing the device, unless you are aware of all the small detials, is to invite trouble.
I also see a lot of people swapping op-amps in and out of circuits (e.g. trying one type in place of another) when in actual fact its not quite as easy as that.
For example, don't put a non-unity gain stable op-amp in a circuit where the gain is less than the manufacturers recommendation. You will get instability issues - and og course that leads to 'glassy sound', edgy sound and a whole lot of other problems
Some op-amp circuits drive stuff in the outside world directly - i.e. op-amp output directly to the output connector. Bad practice since you can never be sure about the capacity load you are looking into, always isolate the op-amp output with a low value resistor - 100 Ohms is good in most cases (but check it in practice!)
For power supply decoupling and layout dressing, similar comments apply.
Another example of bad practice in op-amps (applies to discrete ccts as well) is where the feedback devider is placed a long way from the -ve op-amp input. I've seen this in quite a few cases in my designer days. This is a recipe for noise, oscillation and a whole lot of other ills. Because IC op-amps tend to have high loop gains, its especially bad - you cannot get away with it.
On the input side, also be careful about noise and EMI ingress - devices behave differently because the input topologies and structures are not the same from one type to another.
Bottom line is, a competent designer who chooses to use an op-amp will optimise the circuit for that op-amp, so just changing the device, unless you are aware of all the small detials, is to invite trouble.
Mooly, might be because the inputs remain at or near gound (virtual earth) that you do get a different sonic signature compared to non-inverting where the inputs (inv and non-inv) follow the input signal. But same issues apply to discrete op-amps as well I have to add.
Note layout on inverting types also critical - put the junction of the feedback resistors (upper and lower leg) as close as phyicially possible to the inv input.
As a general rule (inverting and non-inv configuration) this becomes more important with higher open loop gain op amps. With low loop gain amplifiers as I mentioned above, you can get away with it - 1uV of noise pickup x 20db = 20uV. On an op amp with 120dB of loop gain anything picked up on the trace between the feedback resistor junction and the inv input is amplified by a huge margin . . . . ugly, very ugly.
Another example of why op-amps can get bad press.
Power supply decoupling. Most op-amps have great PSRR but it falls off quickly once you get to a few KHz. On small signal amplifiers I like to rely on the op-amp PSSR at frequenceis up to a few KHz, and after that, on RC filtering to help at the higher frequencies. 22 Ohms and 100uF will give over 40db at 10KhZ and over 60dB at 100KHz. Typical volt drop across the 22 Ohm resistors is <0.15V. If th eop-amp is driving the outside world, and the load is not accurately defined, you have to be careful though.
Note layout on inverting types also critical - put the junction of the feedback resistors (upper and lower leg) as close as phyicially possible to the inv input.
As a general rule (inverting and non-inv configuration) this becomes more important with higher open loop gain op amps. With low loop gain amplifiers as I mentioned above, you can get away with it - 1uV of noise pickup x 20db = 20uV. On an op amp with 120dB of loop gain anything picked up on the trace between the feedback resistor junction and the inv input is amplified by a huge margin . . . . ugly, very ugly.
Another example of why op-amps can get bad press.
Power supply decoupling. Most op-amps have great PSRR but it falls off quickly once you get to a few KHz. On small signal amplifiers I like to rely on the op-amp PSSR at frequenceis up to a few KHz, and after that, on RC filtering to help at the higher frequencies. 22 Ohms and 100uF will give over 40db at 10KhZ and over 60dB at 100KHz. Typical volt drop across the 22 Ohm resistors is <0.15V. If th eop-amp is driving the outside world, and the load is not accurately defined, you have to be careful though.
Dear Mr. Curl. I hope this is not a supid question. Although I read a lot of technical material I am not formally trained in semi-conductor technology so I am a bit self-conscious of my lack of knowledge. But I thought this thread may be the place to ask about the importance of "settling time" as an objective measure of our subjective response to circuits. This is not an observation I dreamt up myself but one that a couple of Englishmen raised 35 yrs ago. I own a 10W Class-A amp (John Linsely-Hood) and he had argued for many years that minimising settling time was a desirable goal for "clean", pleasant sounding amps.
I notice that some of the IC op-amps that are favoured by people on this forum such as the AD825 do have v.low settling times. Do the relatively more simple discrete op-amps with lower component counts score in this area?
Any thoughts?
I notice that some of the IC op-amps that are favoured by people on this forum such as the AD825 do have v.low settling times. Do the relatively more simple discrete op-amps with lower component counts score in this area?
Any thoughts?
Hi Jonathan,
If you are selecting an OpAmp for I/V conversion such as directly from a DAC in a CD player settling time does become an important parameter.
I guess it's all a matter of degree really in power amps. The signals that we test with ( fast rise time squarewaves for ex ) bare no relationship to a real music signal.
In fact I can't recall seeing a power amp with what you might call a poor settling time -- as distinct from "ringing" which is a different problem.
If you are selecting an OpAmp for I/V conversion such as directly from a DAC in a CD player settling time does become an important parameter.
I guess it's all a matter of degree really in power amps. The signals that we test with ( fast rise time squarewaves for ex ) bare no relationship to a real music signal.
In fact I can't recall seeing a power amp with what you might call a poor settling time -- as distinct from "ringing" which is a different problem.
Thanks for the responses. As I said, I'm not a wizz at this stuff but I think JLH said that he favoured clean sqaure waves and low settling times because they implied either less internal phase change or greater stability margins and that this translated into a more attractive sound.......This was also acknowledging that the 'ringing' might well be supersonic......but I'll stop now as I am out of my depth at this point. Jonathan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.