At around the time when the 5534 was considered SOA i really wanted to build a simple single box phono preamp as alternative to my three-box valve unit. The initial attempt to follow an app note and some magazine articles proved more than discouraging sound wise. So, back to the drawing board, which in my case meant: choose a simple gain stage, adjust gain for unity and compare against straight wire. Ok, it's usually not that simple as you have to take into account driving impedances and cables, but allows at least to discard unusable topologies.
The non-inverting 5534 (gain set to 10 and adjusted with an input pot to unity) proved to be such an 'unusable' topology. It simply changed the character of sound way too much. To the borderline of very distorted subjectively. Amusing, when you think how many 'high end' components used quite a few of those.
The invering topology, especially when care is taken that both inputs are fed from the same impedances proved an incredible improvement: it even contributed some 'depth' to the sound field and it's sins were mostly of ommission: slightly softened micro and macro dynamics, reduced stereo images.
Eventually i used inverting gain stages to build a 3-stage phono preamp with split riaa between the stages (why am seeing so few split riaa stages when they play so much better?) fully dc coupled (the dc needed a trim every few days...). I built this with 6 discrete regulators and eventually upgraded to six separate NiCd power supplies for and extremely pleasant and musical sound with amazing rhythm and timing. It was however rather noisy as the resistors in the split riaa were both part of the load of one stage and setting the gain of the next. Maybe a more modern opamp with higher current capabilty will be more suitable.
peter
The non-inverting 5534 (gain set to 10 and adjusted with an input pot to unity) proved to be such an 'unusable' topology. It simply changed the character of sound way too much. To the borderline of very distorted subjectively. Amusing, when you think how many 'high end' components used quite a few of those.
The invering topology, especially when care is taken that both inputs are fed from the same impedances proved an incredible improvement: it even contributed some 'depth' to the sound field and it's sins were mostly of ommission: slightly softened micro and macro dynamics, reduced stereo images.
Eventually i used inverting gain stages to build a 3-stage phono preamp with split riaa between the stages (why am seeing so few split riaa stages when they play so much better?) fully dc coupled (the dc needed a trim every few days...). I built this with 6 discrete regulators and eventually upgraded to six separate NiCd power supplies for and extremely pleasant and musical sound with amazing rhythm and timing. It was however rather noisy as the resistors in the split riaa were both part of the load of one stage and setting the gain of the next. Maybe a more modern opamp with higher current capabilty will be more suitable.
peter
My experience is similar to Nelson's. In many cases, it appears that the sonic issue with global feedback is largely one of phase margin, stability and ringing/settling behavior - across a variety of loads and in the presence of whatever RF contamination you may expect in modern society.
I have observed that many circuits each appear to have an optimal range of global feedback for good sound, and empirically this seems to depend on the physical structure as well as device choice and topology.
When I change the closed-loop gain, I will usually also adjust the open-loop gain to maintain a similar amount of loop feedback.
regards, jonathan carr
I have observed that many circuits each appear to have an optimal range of global feedback for good sound, and empirically this seems to depend on the physical structure as well as device choice and topology.
When I change the closed-loop gain, I will usually also adjust the open-loop gain to maintain a similar amount of loop feedback.
regards, jonathan carr
Many single opamps are equipped with compensation nodes that you can use to create interesting hybrid circuits.
For example, pins 1 & 8 allow direct access to the second-stage differential bases of the much-unloved NE5534, and you can splice in something like a discrete JFET front end (short the normal input pins 2 & 3 to the minus voltage rails). Now you have complete flexibility in setting the front-end operating currents, degenerative feedback, cascoding and so on.
Various opamps may be equipped with different styles of compensation nodes that will allow you to design various types of hybrid circuits.
As another example, see the file attachment that I posted in this thread.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6328
Believe me, you can have a lot of freedom when designing with opamps. So have fun!
regards, jonathan carr
For example, pins 1 & 8 allow direct access to the second-stage differential bases of the much-unloved NE5534, and you can splice in something like a discrete JFET front end (short the normal input pins 2 & 3 to the minus voltage rails). Now you have complete flexibility in setting the front-end operating currents, degenerative feedback, cascoding and so on.
Various opamps may be equipped with different styles of compensation nodes that will allow you to design various types of hybrid circuits.
As another example, see the file attachment that I posted in this thread.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6328
Believe me, you can have a lot of freedom when designing with opamps. So have fun!
regards, jonathan carr
Why not invering mode?
One reason is that unless you use a low input bias curent type (read......JFET input), there will be offset errors as drive impedance is changed. Such as when you plug a source into it.
The Eagle series of amps from Electron-Kinetics used an inverting op-amp in its input stage.
One reason is that unless you use a low input bias curent type (read......JFET input), there will be offset errors as drive impedance is changed. Such as when you plug a source into it.
The Eagle series of amps from Electron-Kinetics used an inverting op-amp in its input stage.
offset errors
But isn't the bias current very low and the driving inpedance small?
What about AC coupling, very often used?
What about bias current on the positve input?
`I have asked three questions, and that is enough,'
Said his father; `don't give yourself airs!
Do you think I can listen all day to such stuff?
Be off, or I'll kick you down stairs!'
Alice
But isn't the bias current very low and the driving inpedance small?
What about AC coupling, very often used?
What about bias current on the positve input?
`I have asked three questions, and that is enough,'
Said his father; `don't give yourself airs!
Do you think I can listen all day to such stuff?
Be off, or I'll kick you down stairs!'
Alice
I am sure it's possible to tweak the performance of an opamp by messing with its internal architecture, but is it worth it? I mean sound-wise. The little good that op-amps have going for them the way i see it is convenience, few passive components to worry about and physically short path for the feedback and grounding. If you build a discrete input stage and an output buffer (and not mess the phase margins in the process) what's the point of using an op-amp in the first place?
peter
peter
Re: offset errors
depends on what you mean by "low" and "small". for a FET or JFET input opamp, yes the currents are almost negligible, but with bipolar inputs they can be a factor. also, low driving impedance is not always guaranteed. if i'm using a high-impedance pot before the opamp, it can be fairly high (tens of kOhms).
i'm considering how to implement the LM1875 power amp IC. i'd like to use it in inverting mode, but that will necessitate a low driving impedance before it for best noise/distortion performance. so i may need to use a small-signal opamp as a preamp stage before, maybe a AD8610 or something, also in inverting mode.
Alice Liddell said:
depends on what you mean by "low" and "small". for a FET or JFET input opamp, yes the currents are almost negligible, but with bipolar inputs they can be a factor. also, low driving impedance is not always guaranteed. if i'm using a high-impedance pot before the opamp, it can be fairly high (tens of kOhms).
i'm considering how to implement the LM1875 power amp IC. i'd like to use it in inverting mode, but that will necessitate a low driving impedance before it for best noise/distortion performance. so i may need to use a small-signal opamp as a preamp stage before, maybe a AD8610 or something, also in inverting mode.
>I am sure it's possible to tweak the performance of an opamp by messing with its internal architecture, but is it worth it? I mean sound-wise.<
Why not find out for yourself? You will be able to experiment with a variety of composite topologies that would be a serious headache to implement in discrete. Maybe you will come up with something unique and interesting. And besides, isn't direct hands-on experience much more educational and rewarding than what you may hear of from someone else or read about?
In my own case, a number of ideas that are now used in my products were originally "sketched out" using hybrid and composite IC implementations. When I felt that the outcome warranted further investigation, I subsequently transferred the ideas to discrete implementations.
>The little good that op-amps have going for them the way i see it is convenience, few passive components to worry about and physically short path for the feedback and grounding.<
The Jeff Rowlands Model 10, 11, 12, the American Hybrid Technology phono stage, the AudioPhysic Strada line preamp, and other "high-end" audio products are based on opamps. These may not be the very best audio components in the world, but I daresay that intelligently implemented opamps are sonically preferable to ineptly implemented tubes or discrete solid-state.
As with most other technology, opamps are only a tool. How they are used and how well they are used is up to the sensibilities and capabilities of the individual designer.
regards, jonathan carr
Why not find out for yourself? You will be able to experiment with a variety of composite topologies that would be a serious headache to implement in discrete. Maybe you will come up with something unique and interesting. And besides, isn't direct hands-on experience much more educational and rewarding than what you may hear of from someone else or read about?
In my own case, a number of ideas that are now used in my products were originally "sketched out" using hybrid and composite IC implementations. When I felt that the outcome warranted further investigation, I subsequently transferred the ideas to discrete implementations.
>The little good that op-amps have going for them the way i see it is convenience, few passive components to worry about and physically short path for the feedback and grounding.<
The Jeff Rowlands Model 10, 11, 12, the American Hybrid Technology phono stage, the AudioPhysic Strada line preamp, and other "high-end" audio products are based on opamps. These may not be the very best audio components in the world, but I daresay that intelligently implemented opamps are sonically preferable to ineptly implemented tubes or discrete solid-state.
As with most other technology, opamps are only a tool. How they are used and how well they are used is up to the sensibilities and capabilities of the individual designer.
regards, jonathan carr
In reference to noise, the technique of throwing away open
loop gain by resistors from input to ground, usually the
noise and offset due to bias currents are not much degraded
because the source impedance is going down with the
open loop gain. Input device voltage noise does increase,
however.
In any case this means that this approach works about
equally well with bipolar and Jfet.
loop gain by resistors from input to ground, usually the
noise and offset due to bias currents are not much degraded
because the source impedance is going down with the
open loop gain. Input device voltage noise does increase,
however.
In any case this means that this approach works about
equally well with bipolar and Jfet.
opamp inverting
Nelson,
You are right in your earlier post on opamps often being on the edge of oscillation. Especially for opamps that are supposed to be unity gain stable, the manufacturer tries to squeeze the last drop of gain-bandwidth out of it.
As Jonathan says, creative use of compensation often can get you a bit further.
If you want to throw away exta open loop gain by putting a resistor from the inverting input to ground (in the inverting stage case) be aware that you increase the noise gain of the circuit equally. If you decrease the open loop gain by 20dB, you get 20dB extra noise from the opamp input stage, plus 20dB less THD reduction because of less loop gain.
If you don't want to mess with the external compensation, I would get an opamp that is not so hot, but in the end gives better overall results. The above is the reason that I turned away from the 797, for instance.
Jan Didden
Nelson,
You are right in your earlier post on opamps often being on the edge of oscillation. Especially for opamps that are supposed to be unity gain stable, the manufacturer tries to squeeze the last drop of gain-bandwidth out of it.
As Jonathan says, creative use of compensation often can get you a bit further.
If you want to throw away exta open loop gain by putting a resistor from the inverting input to ground (in the inverting stage case) be aware that you increase the noise gain of the circuit equally. If you decrease the open loop gain by 20dB, you get 20dB extra noise from the opamp input stage, plus 20dB less THD reduction because of less loop gain.
If you don't want to mess with the external compensation, I would get an opamp that is not so hot, but in the end gives better overall results. The above is the reason that I turned away from the 797, for instance.
Jan Didden
Re: offset errors
Alice,
The bias current can be very small, but if the
IP resistor is large enough and the OPA closed
loop gain large enough, then output offset
can be significant.
There are ways to get around this by balancing
impedances of + & - inputs so the offset cancels.
I have done this in balanced pre amps with gain
of 40dB (100) and managed to get offset to 1mV
or less. In this case the balanced topology also helps.
As you stated, generally there is no offset
due to bias current into driving impedance which
is usually low enough.
Having said all of the above, using a precision
FET OPA such as OPA627 is offset heaven. I have
driven these with megs and still were hard pressed
to measure any offset.
Terry
Alice Liddell said:
Alice,
The bias current can be very small, but if the
IP resistor is large enough and the OPA closed
loop gain large enough, then output offset
can be significant.
There are ways to get around this by balancing
impedances of + & - inputs so the offset cancels.
I have done this in balanced pre amps with gain
of 40dB (100) and managed to get offset to 1mV
or less. In this case the balanced topology also helps.
As you stated, generally there is no offset
due to bias current into driving impedance which
is usually low enough.
Having said all of the above, using a precision
FET OPA such as OPA627 is offset heaven. I have
driven these with megs and still were hard pressed
to measure any offset.
Terry
Stability
Jonathan and Jan,
It is not quite clear to me whether you mean only that we must
make sure that the op amp is stable under worst-case load
conditions (which we do want, of course) or if you are asking
for something more? That is, do you also mean that the sound
could benefit from improving a phase margin that is already
sufficient for stability?
Jonathan and Jan,
It is not quite clear to me whether you mean only that we must
make sure that the op amp is stable under worst-case load
conditions (which we do want, of course) or if you are asking
for something more? That is, do you also mean that the sound
could benefit from improving a phase margin that is already
sufficient for stability?
Anything you can do to reduce noise at the front end of an amp will give you better returns than an equivalent reduction at the back end. Noise from the front end is multiplied by all the stages later (and added to, of course). A reduction in noise at the front end keeps this snowballing effect from occuring. Reduction at the back end, after the fact, will work but not be as efficient.
As an admittedly oversimplified view of things, assume an arbitrary number of identical gain stages. The noise figure of the circuit is grossly determined by the noise figure of the very first stage. Let's say that each stage amplifies by a factor of 100 (40dB), a not unrealistic number for solid state circuits, although it's stretching things for tubes. If there's, say, 1mV of noise injected per stage, then the noise from the first stage is amplified by the second stage to become 100mV. The second stage then contributes its 1mV of noise, so now we stand at 101mV of total noise. You can see that the first stage's noise swamps that of all following stages. If you cut that first 1mV of noise in half, you've only got 51mV of noise coming out of the second stage.
It's easier just not to put it in there in the first place. Trying to get it out later is a royal pain. Not to mention more subtle factors like the noise intermodulating with the signal, etc.
Grey
As an admittedly oversimplified view of things, assume an arbitrary number of identical gain stages. The noise figure of the circuit is grossly determined by the noise figure of the very first stage. Let's say that each stage amplifies by a factor of 100 (40dB), a not unrealistic number for solid state circuits, although it's stretching things for tubes. If there's, say, 1mV of noise injected per stage, then the noise from the first stage is amplified by the second stage to become 100mV. The second stage then contributes its 1mV of noise, so now we stand at 101mV of total noise. You can see that the first stage's noise swamps that of all following stages. If you cut that first 1mV of noise in half, you've only got 51mV of noise coming out of the second stage.
It's easier just not to put it in there in the first place. Trying to get it out later is a royal pain. Not to mention more subtle factors like the noise intermodulating with the signal, etc.
Grey
GRollins said:
Grey, I must correct you little bit. Hope you don't mind.
You can't add noise and distortion in the way you describe.
Total noise will be SQR(100^2+1^2) = 100,004999875 (*)
Conclusion: It's even more important to have a low noise first stage. All of the noise is generated there for most amps.
(*) What is this type of addition called? Is it algebraical addition?
dorkus
If you want to have high input resistance AND high gain with an inverting OP-AMP circuit, then you can avoid high-value feedback resistors by a simple trick: Instead of using a simple feedback resistor you can use a voltage divider between output and feedback resistor.
Per
What you are talking of is geometrical addition (it's at least called like that in German).
Regards
Charles
If you want to have high input resistance AND high gain with an inverting OP-AMP circuit, then you can avoid high-value feedback resistors by a simple trick: Instead of using a simple feedback resistor you can use a voltage divider between output and feedback resistor.
Per
What you are talking of is geometrical addition (it's at least called like that in German).
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.