Has anybody tried paralleling opamps to drive headphones like a project I saw on headwize? just curious. jer
Mooly,
I have just spent an hour reading this thread. Interesting discussions. I really appreciate your posts especially those showing your measurements with and without compensation.
I have a number of opamp stages in my line level XO/EQ for a 4 way active speaker. Some of them have small gains (up to 6dB) and I have spent some time trying to work out the values of the compensation caps.
But a discovery today made me wonder if I should install those compensation caps at all.
I could hear some "Zzzz..." in my system when playing certain CDs. This morning I put in 1k and 68p (to ground) after a 20k volumn pot and in front of an opamp buffer (opa627) to shunt RF to ground. I checked this with LTSpice and was sure that there was no frequency rolloff and phase shift within the audio bandwidth. In theory, this 1k and 68p should not degrade the sound but only reduce high frequency noise hence enhance the sound
The result was a bit of a shock to me. The sound seemed to be coloured by the 68p capacitor (Wima FKP2, polypropylene) and some music details were lost. I then replaced the 68p MKP with 100p polystyrene. The colouration was reduced but still there. I did not hear any reduction of the noise "Zzzz...". Those "Zzzz..." may come from the recordings, as for good recordings I don't have the noise at all.
I then removed the 1k (leaving a 200R there) and 100p. The sound was obviously back to be good, and the noise level appeared to be the same.
It appears shunting RF in front of the opamp does not reduce noise in my case, while adding the RC filter coloured the sound, due to possible distortion introduced by the capacitor.
I think that perhaps effective shielding has prevented RF from getting into the signal path. The power supply ripples are below 100uV to be measurable by ordinary equipment and has an output impedance below 1mR. In that case, if no high frequency signal gets into the opamp, would the compensation be necessary?
My worry is that adding a feedback capacitor for compensation would introduce similar (or worse) effect like the 1k 68p in front of the opamp. I understand that if RF is present then compensation is a must. But what if there is no RF in the signal, for example, the opamps used after the LP filters?
Your advice will be very much appreciated.
Best regards,
Bill
I have just spent an hour reading this thread. Interesting discussions. I really appreciate your posts especially those showing your measurements with and without compensation.
I have a number of opamp stages in my line level XO/EQ for a 4 way active speaker. Some of them have small gains (up to 6dB) and I have spent some time trying to work out the values of the compensation caps.
But a discovery today made me wonder if I should install those compensation caps at all.
I could hear some "Zzzz..." in my system when playing certain CDs. This morning I put in 1k and 68p (to ground) after a 20k volumn pot and in front of an opamp buffer (opa627) to shunt RF to ground. I checked this with LTSpice and was sure that there was no frequency rolloff and phase shift within the audio bandwidth. In theory, this 1k and 68p should not degrade the sound but only reduce high frequency noise hence enhance the sound
The result was a bit of a shock to me. The sound seemed to be coloured by the 68p capacitor (Wima FKP2, polypropylene) and some music details were lost. I then replaced the 68p MKP with 100p polystyrene. The colouration was reduced but still there. I did not hear any reduction of the noise "Zzzz...". Those "Zzzz..." may come from the recordings, as for good recordings I don't have the noise at all.
I then removed the 1k (leaving a 200R there) and 100p. The sound was obviously back to be good, and the noise level appeared to be the same.
It appears shunting RF in front of the opamp does not reduce noise in my case, while adding the RC filter coloured the sound, due to possible distortion introduced by the capacitor.
I think that perhaps effective shielding has prevented RF from getting into the signal path. The power supply ripples are below 100uV to be measurable by ordinary equipment and has an output impedance below 1mR. In that case, if no high frequency signal gets into the opamp, would the compensation be necessary?
My worry is that adding a feedback capacitor for compensation would introduce similar (or worse) effect like the 1k 68p in front of the opamp. I understand that if RF is present then compensation is a must. But what if there is no RF in the signal, for example, the opamps used after the LP filters?
Your advice will be very much appreciated.
Best regards,
Bill
Last edited:
Something going on here that I cannot explain.
Rs = 200r
Added resistor = 1k
RF cap = 68pF.
RC time constant = [1k0+200r] * 68p = 0.08us This should be completely inaudible.
Rs=200r
no added resistor
RF cap = 100pF
RC time constant = 200 * 100p = 0.02us. This too should be completely inaudible. But it is two octaves higher than the 68p+1k0.
68pF is very low for an MKP. Are you sure is is 68pF. Can you measure it, or confirm in some other way.
Rs = 200r
Added resistor = 1k
RF cap = 68pF.
RC time constant = [1k0+200r] * 68p = 0.08us This should be completely inaudible.
Rs=200r
no added resistor
RF cap = 100pF
RC time constant = 200 * 100p = 0.02us. This too should be completely inaudible. But it is two octaves higher than the 68p+1k0.
68pF is very low for an MKP. Are you sure is is 68pF. Can you measure it, or confirm in some other way.
I agree with peranders, I've modded a V-DAC a while ago and there was a huge thread on Head-Fi (Musical Fidelity V-DAC Owners? - Head-Fi: Covering Headphones, Earphones and Portable Audio) about what was best. In the end I tried a lot of combinations as well and what I found was that a cheap combo (LME49740NA + LME49860NA) sounds a LOT better then a more expensive one (OPA4227PA + LME49720HA)...
Hi Bill,
I'm just trying to visualise your set up. You have the wiper of the volume pot connected to this R/C network you mention (1k +68pf). If that's not correct perhaps you could draw what you have. That network just forms a filter, it's not compensating any opamp as you describe it.
Anyway, as Andrew says at, that should be totally inaudible. The effect of increasing the value of the cap is to roll the high frequencies off earlier... buteven a 100 pf on a 20k pot wiper is still inaudible.
Can you describe what you mean by "coloured" the sound. Do you mean loss of treble ?
Another wild thought is that the opamp it's feeding is actually oscillating ( the opa627 is a wide bandwidth device) and needs care in use and application. Perhaps the addition of your cap is causing some unexpected change.
Have you an oscilloscope ? that you can actually do tests on the set up with ?
I'm just trying to visualise your set up. You have the wiper of the volume pot connected to this R/C network you mention (1k +68pf). If that's not correct perhaps you could draw what you have. That network just forms a filter, it's not compensating any opamp as you describe it.
Anyway, as Andrew says at, that should be totally inaudible. The effect of increasing the value of the cap is to roll the high frequencies off earlier... buteven a 100 pf on a 20k pot wiper is still inaudible.
Can you describe what you mean by "coloured" the sound. Do you mean loss of treble ?
Another wild thought is that the opamp it's feeding is actually oscillating ( the opa627 is a wide bandwidth device) and needs care in use and application. Perhaps the addition of your cap is causing some unexpected change.
Have you an oscilloscope ? that you can actually do tests on the set up with ?
Andrew and Mooly,
Thanks for your reply. Your help would be very much appreciated.
I am completing my HiFi system and my line level XO/EQ is the very last stage I must do, the rest are already done. Below is a photo of my system. It is designed as 4 way U-frame fully active, but I have run out of time now so I am making it 3 way first and will come back to it perhaps in a year's time and make it 4 way. I must complete this system within 2 to 2.5 weeks because I have just got some new work commitments.
Thanks for your reply. Your help would be very much appreciated.
I am completing my HiFi system and my line level XO/EQ is the very last stage I must do, the rest are already done. Below is a photo of my system. It is designed as 4 way U-frame fully active, but I have run out of time now so I am making it 3 way first and will come back to it perhaps in a year's time and make it 4 way. I must complete this system within 2 to 2.5 weeks because I have just got some new work commitments.
An externally hosted image should be here but it was not working when we last tested it.
Below is the circuit for my test set up.
I had the same thought as you. I believed that the 1k and 68p should have no negative impact on the sound. But subjectively it sounds obviously better without them.
It is for this reason that I think perhaps there is no RF getting into the signal so no compensation caps may be needed in the real filter stages.
By the way, the schematic and the actual built circuit has a difference. The circuit has a 10R between the signal ground and the circuit common, and 0.1uF from the signal ground to chassis.
I had the same thought as you. I believed that the 1k and 68p should have no negative impact on the sound. But subjectively it sounds obviously better without them.
It is for this reason that I think perhaps there is no RF getting into the signal so no compensation caps may be needed in the real filter stages.
By the way, the schematic and the actual built circuit has a difference. The circuit has a 10R between the signal ground and the circuit common, and 0.1uF from the signal ground to chassis.
An externally hosted image should be here but it was not working when we last tested it.
Before I go and cut the board and build it, since I am an novice and lucky to have you experts here, it would be appreciated if you could have a look at my circuit designs and see if you can spot any obvious faults.
This is for the tweeter. The filter transfer function combined with the mounted driver response form an acoustic LR4.
This is for the tweeter. The filter transfer function combined with the mounted driver response form an acoustic LR4.
An externally hosted image should be here but it was not working when we last tested it.
This is the band pass filter for the midrange drivers. A price to pay for having U-frame baffle is the ballooning. It takes 4 opamps to do 2 notch filters and 1 boost filter.
An externally hosted image should be here but it was not working when we last tested it.
Last, this is for the woofer.
An externally hosted image should be here but it was not working when we last tested it.
The discreet circuits i tryed that sound "better" them OPamps all had a measure of second harmonic or a series of exponentially falling distortion at low level thrown in.
Actualy when you use CD as your main source that can sound quite spectacular.
Cd is not famous for accurate treble response and the "harmonic" distortion of some circuits restores the missing overtones to a certain degree.
If you like i can show you 3 circuits of a buffer that can be build quite cheep that have the 3 different characters i talk about.
Actualy when you use CD as your main source that can sound quite spectacular.
Cd is not famous for accurate treble response and the "harmonic" distortion of some circuits restores the missing overtones to a certain degree.
If you like i can show you 3 circuits of a buffer that can be build quite cheep that have the 3 different characters i talk about.
I am sure. It is the Wima FKP2. I checked the label many times and it is 68p. I got them from Farnell: Browse for Products | Farnell Australia
It is rather difficult to describe "coloured", same difficulty with describing "sweet". It is all subjective. With "coloured", it is more like listening to a HiFi. Without the "colour", it is more like sounding the real thing. I am familiar with natural acoustic music.
Put the subjective matters aside, I do have a scope, abeit a low resolution one - maximum 20MHz and 10mV per grid. If I see a thicker line I assume it is under 500uV, etc. It is an old scope and I am not even sure if both channels are working, I have used only one. I also have a signal generator but I have never used it. A friend gave both these equipments to me only a few months ago. Because now I have to complete the system within 2-2.5 weeks and I wish I could have the time to investigate them.
I don't know if my "calculations" for the compensation capacitors are correct for the tweeter, midrange and woofer filters. I modelled it in LTSpice with simple RC filter (the feedback resistor and the parallelled capacitor) and tried to get the corner frequency at around 200kHz so that the bandwidth is maximally limited but without upsetting the frequency and phase response under 20kHz. I could be very wrong here. Please correct me if I am wrong.
Regards,
Bill
Last edited:
That sounds very interesting. Sometimes I have the same feeling. Certain types of higher frequency noise compensates for the lack of high frequency harmonics in some recordings, making them sound a lot more exciting than they should. However, long term listening may be tiring. Eventually, when comparing to a system that is low distortion, low noise, you suddenly find that the later is a lot better and you then set your new goal.
It is interesting to see how you achieve those different characters of sound with a simple buffer.
Regards,
Bill
This buffer adds a measure of low level 2nd harmonic. Actually the Fet is a Linear System
LSK389C. Idss selected 2SK170BL can be used but the LSK389C has more class A and no worry about offset . It sounds spectacular with CD.
Tomorrow i will post a buffer that does the exponential falling trick but on a very low level plus a OPamp buffer that has no measuarable distortion. You can compare the sound and draw your own conclusion.
LSK389C. Idss selected 2SK170BL can be used but the LSK389C has more class A and no worry about offset . It sounds spectacular with CD.
Tomorrow i will post a buffer that does the exponential falling trick but on a very low level plus a OPamp buffer that has no measuarable distortion. You can compare the sound and draw your own conclusion.
Yes, I have built that with 2sk170BL and compared it to opa627. They sound different but I personally preferred opa627. They were driven by a very clean low impedance shunt regulator with no ripples shown on my scope.
If we see it from LTSpiece, the RC filter only works above audio bandwidth, so it seems not to have any sonic effect.
But Jung did some capacitor measurements and found that capacitor distortions not only affect the pass band, but to similar degree to the stop band! That may be the answer.
Last edited:
If I read this right in those networks for the tweeter, mid, and woofer, and please give some leeway to my very inexperienced opinion in this matter, but do I read that each network is, at its heart, a 12db Butterwoth 2nd order electrical filter? Wouldn't that provide a 6db boost at the crossover points, thus seeming to need more notch filter that could otherwise be overcome with a 24db 4th order LR filter and the resulting 3db of boost at the crossover points, along with less driver excursion? The tweeter especially? I don't kniow enough about phase and subsequent eq sections.
I'm still learning this active filter stuff.
I'm still learning this active filter stuff.
Last edited:
If I read this right in those networks for the tweeter, mid, and woofer, and please give some leeway to my very inexperienced opinion in this matter, but do I read that each network is, at its heart, a 12db Butterwoth 2nd order electrical filter? Wouldn't that provide a 6db boost at the crossover points, thus seeming to need more notch filter that could otherwise be overcome with a 24db 4th order LR filter and the resulting 3db of boost at the crossover points, along with less driver excursion? The tweeter especially? I don't kniow enough about phase and subsequent eq sections.
I'm still learning this active filter stuff.
The filter transfer functions do not conform to any text book electrical order. Fortunately we are dealing with minimal phase system here so it is the combined response (of the electrical response of the above XO and the natural acoustic response, i.e. response measured when the driver mounted on the loudspeaker baffle) that matters. A shallen key topology may yield a 2nd order response, not necessarily Butterworth, as different R and C values will yield different Q. In any case, the final acoustic responses of the speaker drivers combined with the above filters yield near perfect LR4 response, and the frequency response is almost ruler flat with +/-1.5dB max from 180Hz to 20kHz.
I would suggest that the filter functions be discussed in a spearate multi-way thread. I would really like to get the experts' opinions on how to work with opamps (opa627 and lm4562) in question. Let us focus on opamp circuits, compensation, bandwidth limiting, transient response, etc. which is the centre theme of this thread.
Last edited:
Bill... hopefully I'll come back to this later today. You have posted a lot of info... and I have to confess I don't use LTSpice etc.
Your filters and integrating them with your speakers really are a topic on their own.
Having seen your circuit and the 68pf cap I can't really believe that the effect could be audible at all... so I want to just try something if I have chance.
If you have the 'scope you really should use it to confirm all is well and that there is no unexpected HF oscillation anywhere.
Your filters and integrating them with your speakers really are a topic on their own.
Having seen your circuit and the 68pf cap I can't really believe that the effect could be audible at all... so I want to just try something if I have chance.
If you have the 'scope you really should use it to confirm all is well and that there is no unexpected HF oscillation anywhere.