I downloaded the datasheet of the opa827 and had a look. Comparing to opa627, gain bandwidth product is increased from 16MHz to 22Mhz; open loop gain is increased by 6-10dB. Settling time is the same. However, PSRR for the negative rail is decreased approximately from 78dB to 62dB at 20kHz, a 16dB deduction. That is a lot.
I wonder if anybody has done any subjective comparisons between opa627 and opa827? Which sounds better?
What I gain from Andrew's post is not just about the RC constant. In the case of 68pF the RC Constant is about 0.4uS. In the case of 100pF is about 0.6uS. I am not totally convinced yet the time delay at such a high frequency is audible, if so, no loudspeaker crossover would work.
What Andrew's post did was to remind me to look for something else, parasitics, for example.
I was thinking that, adding the 68pF hence a very low impedance to ground at high frequencies to both +/- inputs of the opamp is effectively adding a 34pF between the +/- inputs of the opamp. This may upset the input bias of the opamps, or not? Is this the reason that the 68p caused subjective sound degradation? I am a novice so experts please throw in your answers here.
But then, I have a 220R "gate stopper" before the 68p at the inputs of the opamps, which should be enough to eliminate that effect, or not? 😕
I recall that in Rod Elliot's project (I am digging it out now and am getting the text here): "Earlier boards included RF interference immunity by adding a small capacitor between the two inputs of U1 (the space for this cap can be seen in the PCB photo, which is of an early board). This has now been removed, as it caused more problems than it solved - in particular, opamp oscillation with some devices."
What Andrew's post did was to remind me to look for something else, parasitics, for example.
I was thinking that, adding the 68pF hence a very low impedance to ground at high frequencies to both +/- inputs of the opamp is effectively adding a 34pF between the +/- inputs of the opamp. This may upset the input bias of the opamps, or not? Is this the reason that the 68p caused subjective sound degradation? I am a novice so experts please throw in your answers here.
But then, I have a 220R "gate stopper" before the 68p at the inputs of the opamps, which should be enough to eliminate that effect, or not? 😕
I recall that in Rod Elliot's project (I am digging it out now and am getting the text here): "Earlier boards included RF interference immunity by adding a small capacitor between the two inputs of U1 (the space for this cap can be seen in the PCB photo, which is of an early board). This has now been removed, as it caused more problems than it solved - in particular, opamp oscillation with some devices."
Last edited:
My low resolution oscilliscope (20MHz 10mV per grid) did not show me any noise within the limits of what it can display. I would not see anything below 500uV, because at 500uV it would appear as a thick line. What I see is a thin line probably less than 1 grid / 20.
In that case, can the opamp still oscillate or even be saturated with HF noise, only because the frequency is so high that the oscilliscope failed to catch and display it?
Last edited:
Morning Bill 🙂 well it is here (7.00am) and so many posts since last night.
Easy one first... can opamps oscillate and you wouldn't see it on a 20 mhz 'scope ?
Not normally... a 20mhz scope will often "show" signals considerably above it's -3db point anyway, although at reduced amplitude. My 'sope is 100mhz but I often use it at the reduced 25mhz (switchable) bandwidth simply to get a finer trace (less internally generated noise).
Curiously Doug Self has mentioned that the NE5532 and NE5534 can oscillate at HF and that this can go unnoticed. The cause is poor decoupling in use. I always recommend a small (0.1uf or so) cap soldered directly across the supply pins.
Easy one first... can opamps oscillate and you wouldn't see it on a 20 mhz 'scope ?
Not normally... a 20mhz scope will often "show" signals considerably above it's -3db point anyway, although at reduced amplitude. My 'sope is 100mhz but I often use it at the reduced 25mhz (switchable) bandwidth simply to get a finer trace (less internally generated noise).
Curiously Doug Self has mentioned that the NE5532 and NE5534 can oscillate at HF and that this can go unnoticed. The cause is poor decoupling in use. I always recommend a small (0.1uf or so) cap soldered directly across the supply pins.
I think only you can answer this (is it audible or not question)... what I suggest is rigging up a small in line network perhaps from CD to amp of say your 68pf cap and various resistors feeding it... or better a pot say 10 k.
Construct a two channel version with dual pot... you only need a bit of point to point wiring... and listen at various settings of the pot. Try also a 100 pf etc.
Now a word of caution 🙂 it's so so easy to "think" you hear an effect... it's something I have had to caution myself on many times.
So each time you listen get someone else to turn the pot to it's new position before you listen.
It's a rough and ready test but might help.
A practical issue is that this arrangement depends on the impedance the network it's feeding into to a small extent as that will alter the overall level to.
You could always rig up a buffer ? 🙂
Remember the signal generator produces square waves with extremely fast rise times... nothing in audio whether off CD or SACD will ever aproach these times.
Again a "practical" experiment. Feed a squarewave into 16bit AD convertor running at 44.1khz sample rate and then feed that back to a DA convertor. I tried this with a Minidisc recorder which has similar specs to CD regarding bandwidth etc and above about 6 khz that squarewave is reduced to a triangular wave... which is exactly what you would expect... it's just a bit of a reality shock to see it.
Again on a practical note, most power amps have an RC filter at the input for exactly this reason, to limit the rise time of any applied signals to below the slew rate of the amp to minimise TIM etc. The time constants of these networks is often much greater than your 1k + 68pf.
Phase shift is generally considered to be inaudible... certainly at the values we are talking about. As you say, think of crossovers or even tone control networks etc.
Capacitors and sound. Hmmm... I am not a fan of so called "boutique" parts. I know many swear by them... I am not a believer.
Just though of another thing (obvious really). The network of 1k and 68pf feeding an opamp. I haven't got that particular device to hand, but although it's a highly specified device, I'll bet it introduces more HF roll off than the RC network 🙂 when used at modest gain or even as a buffer.
One reason I always test with squarewaves, as the photos way way back show is that opamps are influenced by stray circuit capacitance etc and it's nice to actually see for real what is happening and to be able to compensate and tweak to get the best results with no obvious over or undershoot.
Also remember that 99% of recorded material has been through far far more bandwidth limiting filters on it's way to your amp.
Last edited:
RF getting into circuitry based on opamps, unless you are near a strong source of RF I can't really see that being an issue tbh.
Is compensation required when the input is devoid of HF ?
That's a good question... and probably opinons will differ.
Lets be clear on the term "compensation", as I use it loosely.
Certain opamps need compensating as a matter of course depending on the gain they are configured for. A common example is the NE5534 which is not unity gain stable. So that IC has two pins where a small cap can be added. Without, and the device oscillates. The NE5532 is unity gain stable, and that is because it as that cap inbuilt... a consequene of which is that the slew rate suffers and falls to around 9v/us.
I guess that's the proper usage for "compensation"
Adding a small cap across the global feedback resistor of an opamp is slightly different, and from a practical point of view can be used to tweak the squarewave response to eliminate ringing or over and undershoot as I think I showed to Andrea way back. Again, it's very dependant on the exact application as to whether it's needed or not.
Have a look at this,
http://www.diyfactory.com/data/NE5532 c.pdf
I can also fully recommend Doug Selfs new book, Small Signal Audio Design.
It covers a lot of the questions you are asking and is written in a very easy going but very comprehensive style... well worth a read.
The Audio Power Amplifier Design Handbook
Thanks, Mooly. That is quite a bit of information to diggest, and I will do so slowly.
After all those discussions, I did one easy experiment. I moved the RF shunt in front of the 20k pot. For convenience the caps are soldered at the input RCA jack. This should completely eliminate the possibility of adding additional parasitics to the opamp because of the RF shunt, or upsetting the bias of the opamps. I did some calculations and found 1nF before the 12k // 8k divider gives a -3dB point at 600kHz, at 16Mhz (gain bandwidth of the opa627 at unity gain) it is -46dB down. In theory, this is completely inaudible.
Put the probe of the scope at the output (no input signal), I saw nothing.
Here is the same result: I prefer the sound without the 1n RF shunt. Without it, subjectively the sound is more "open" and "vivid".
There may be multiple possible reasons for this. I guess the primary suspect is psychological. It is possible that my ears have been trained getting used to such sound / distortions therefore unconsciously make judgments based on the distorted patterns stored in my memory.
Joachim presented 3 buffers. First one (simple two jFETs) has high even order distortions. Second one with low level of high frequency distortions. The third one (opamp buffer) with low distortions difficult to measure.
Joachim liked the second one. As far as I know not many people like the third one - the sound of opamps (implemention plays a big part though), while there is a large following for the first one with higher distortions. If you don't believe me on this, start a new thread to collect statistics on the preferances on the simple jFETs and the opamp buffer. I think the simple jFETs will totally outnumber the opamps.
Of course, there may be technical reasons that the RF Shunt may indeed somehow makes it sound worse, just something I have not learned. I recall that my Marantz SA11 SACD player has 3 output filter switches to control the high frequency rolloffs. The frequencies are much higher than the audioband width. If they have no impact on sound, Marantz would not have given choices to 3 selectable filters. I played with the filters and found some subtle differences. Some other people reported bigger differences than me. They have better ears than me. 🙂
Regards,
Bill
After all those discussions, I did one easy experiment. I moved the RF shunt in front of the 20k pot. For convenience the caps are soldered at the input RCA jack. This should completely eliminate the possibility of adding additional parasitics to the opamp because of the RF shunt, or upsetting the bias of the opamps. I did some calculations and found 1nF before the 12k // 8k divider gives a -3dB point at 600kHz, at 16Mhz (gain bandwidth of the opa627 at unity gain) it is -46dB down. In theory, this is completely inaudible.
Put the probe of the scope at the output (no input signal), I saw nothing.
Here is the same result: I prefer the sound without the 1n RF shunt. Without it, subjectively the sound is more "open" and "vivid".
There may be multiple possible reasons for this. I guess the primary suspect is psychological. It is possible that my ears have been trained getting used to such sound / distortions therefore unconsciously make judgments based on the distorted patterns stored in my memory.
Joachim presented 3 buffers. First one (simple two jFETs) has high even order distortions. Second one with low level of high frequency distortions. The third one (opamp buffer) with low distortions difficult to measure.
Joachim liked the second one. As far as I know not many people like the third one - the sound of opamps (implemention plays a big part though), while there is a large following for the first one with higher distortions. If you don't believe me on this, start a new thread to collect statistics on the preferances on the simple jFETs and the opamp buffer. I think the simple jFETs will totally outnumber the opamps.
Of course, there may be technical reasons that the RF Shunt may indeed somehow makes it sound worse, just something I have not learned. I recall that my Marantz SA11 SACD player has 3 output filter switches to control the high frequency rolloffs. The frequencies are much higher than the audioband width. If they have no impact on sound, Marantz would not have given choices to 3 selectable filters. I played with the filters and found some subtle differences. Some other people reported bigger differences than me. They have better ears than me. 🙂
Regards,
Bill
Last edited:
As for the negative feedback cap, I think I will spend the time to rebuild (point to point on veroboard) my opa627 buffer to have 6dB gain, try a few values of the feedback cap and see if it makes any subjective difference.
I don't have a signal generator that can generate square waves, and worse, I don't have the time now. I am planning to complete the build of the XO/EQ within a couple of weeks to wrap up my HiFi project and make some serious commietment to work (for a living), and may come back to the hobby in a year or two's time.
I don't have a signal generator that can generate square waves, and worse, I don't have the time now. I am planning to complete the build of the XO/EQ within a couple of weeks to wrap up my HiFi project and make some serious commietment to work (for a living), and may come back to the hobby in a year or two's time.
It seems I did not make my answer clear enough.
We are adding a 1k0 resistor and/or adding a 68pF/100pF cap to a system that already has an RF filter built into it.
The roll-off frequency of this parasitic RF filter is determined by the source resistance/impedance (Rs) of each filter.
The Pot has a source resistance for the downstream side of 8k//12k1 = 4k8
The parasitic capacitance (Cpara) after that could be 5pF or 10pF or 20pF.
The parasitic RF filter is effectively Rs & Cpara.
Before adding the extra components and assumed Cpara=10pF we have RC=0.05us. This will not be audible as an audio frequency filter.
Adding a 1000pF (=1nF) at the RCA input is ludicrous. Try 47pF at the RCA socket and come back with your audibility results.
Now add on 1k0, RC becomes 0.06us, still inaudible.
Now add 68pF alone, RC becomes 0.37us, some may argue this is just approaching audibility.
Now add 1k0 & 68pF, RC becomes 0.45us, even closer to audibility.
Replace the stacked film low inductance 68pF PP with a rolled foil medium inductance 100pF PS and the filter characteristics will be changed. This may be audible. In addition the RC becomes 0.58us. This slope is modified by the inherent inductance of the polystyrene.
I am not talking about swapping boutique /exotic parts for technically competent components. I am talking plain and simple engineering.
Last edited:
It seems I did not make my answer clear enough.
We are adding a 1k0 resistor and/or adding a 68pF/100pF cap to a system that already has an RF filter built into it.
The roll-off frequency of this parasitic RF filter is determined by the source resistance/impedance (Rs) of each filter.
The Pot has a source resistance for the downstream side of 8k//12k1 = 4k8
The parasitic capacitance (Cpara) after that could be 5pF or 10pF or 20pF.
The parasitic RF filter is effectively Rs & Cpara.
Before adding the extra components and assumed Cpara=10pF we have RC=0.05us. This will not be audible as an audio frequency filter.
Now add on 1k0, RC becomes 0.06us, still inaudible.
Now add 68pF alone, RC becomes 0.37us, some may argue this is just approaching audibility.
Now add 1k0 & 68pF, RC becomes 0.45us, even closer to audibility.
Replace the stacked film low inductance 68pF PP with a rolled foil medium inductance 100pF PS and the filter characteristics will be changed. This may be audible. In addition the RC becomes 0.58us. This slope is modified by the inherent inductance of the polystyrene.
I am not talking about swapping boutique /exotic parts for technically competent components. I am talking plain and simple engineering.
Adding 1000pF at the RCA socket is ludicrous. Try 47pF and come back and tell us the audibility result. The reason for such a low value at the input is that a 2pole filter is being created using the inductance of the preceding interconnects to create an LC RF filter.
The RF filter is not intended to alter and/or modify the audio signal.
It is there solely to attenuate the RF interference and thus allow the audio circuit to do it's job properly.
High gain circuits are especially sensitive to power supply hash, a lot of care is needed in the layout of phono preamps. You might want to check (by examination of the pcb) if the designer has used separate grounds for power and signal - one very common approach is to decouple the opamp supplies direct to the combined signal/power ground. My money would be on there being no distinction on that cheap unit😛
Andrew,
I guess we have moved on from this. The roll-off is not an issue.
See below. Right from the CD output to the opamp input. I have not included parasitic inductance and resistance. But you know they won't affect the result of the roll-off as much. I have made the capacitance to the extreme.
Regards,
Bill
🙂
I guess we have moved on from this. The roll-off is not an issue.
See below. Right from the CD output to the opamp input. I have not included parasitic inductance and resistance. But you know they won't affect the result of the roll-off as much. I have made the capacitance to the extreme.
Regards,
Bill
🙂
An externally hosted image should be here but it was not working when we last tested it.
Hi Bill,
You are losing me on your descriptions of your experiments 🙂
You moved the RF shunt (68pf and 1k) to the front of the pot... that's fine, however the filter response is now determined more by what feeds the pot. I'm looking at your circuit in post 1707.
If you feed it from the wiper of the volume control, then the source resistance alters as the pot is turned, this in turn has a small effect on the reponse of the filter. In other words the response is different for every setting of the pot... but the effect is slight with the values used.
Feed the filter from the constant impedance of the CD player output and the response is "fixed".
no, you haven't. You are still reporting anomalous results from your tests because you have not understood why an RF filter does it job and why it uses particular component/circuit values to achieve an inaudible attenuation of RF.
Post 1754.
Are R1, R2 and C1 inside the source CDP?
Last edited:
Hmmm... where's the 1nf come from ?
I'm looking at your circuit and you feed the 20 k pot from the CD player output which will be at a fairly low Z.
Hi Alexg !
You are on the right track. Pannasonic FC is a good replacement and the sound may get a bit smoother over time.
0.47 uF seems a bit small to me.
What is the input impedance of your next ( line ? ) stage ?
For 47 kOhm i whould recommend something around 2uF.
The RIAA curve in the V-LPS should be accurate. Those guys are not stupid but may as you say have used affordable components to keep the price down.
You are on the right track. Pannasonic FC is a good replacement and the sound may get a bit smoother over time.
0.47 uF seems a bit small to me.
What is the input impedance of your next ( line ? ) stage ?
For 47 kOhm i whould recommend something around 2uF.
The RIAA curve in the V-LPS should be accurate. Those guys are not stupid but may as you say have used affordable components to keep the price down.
A$nd you have just posted a new circuit 🙂
1nf is too high to use on the output of an opamp (in practice).
Andrew keeps mentioning parasitic capacitance etc.
That's one reason I mentioned earlier that this was a real world measurement taking into account the 'scope probe capacitance. I nearly posted another picture of the result at HF using a 1 to 1 probe to show you.
When you are talking about small cap values (pf range) then circuit layout and stray capacitance starts to become more and more important... and you can even make small caps on the PCB using close running print, not just caps but inductors too, and this is often done in RF work.
Back to audio...
1nf is too high to use on the output of an opamp (in practice).
Andrew keeps mentioning parasitic capacitance etc.
That's one reason I mentioned earlier that this was a real world measurement taking into account the 'scope probe capacitance. I nearly posted another picture of the result at HF using a 1 to 1 probe to show you.
When you are talking about small cap values (pf range) then circuit layout and stray capacitance starts to become more and more important... and you can even make small caps on the PCB using close running print, not just caps but inductors too, and this is often done in RF work.
Back to audio...