I use a BF Fet with very high IDss so i think pushing it down to 8mA makes the current quite stabe. I could try a low noise one but i will not put much more work into that circuit at the moment. The sound is extremely pleasing right now. On par with much more expensive commercial stages and that is the essence of DIY i think. I am very bussy with the discreet INA, still working on stabilizing the feedback loop. I will try out Microcap soon to get a better handle on that kind of problems.
The Balanced Discreet INA is working stabe. I had to find a current node to implement the loop compensation correctly so i used the node where the collector of the caskode transistor and the collector of the current source meet. R7,R8,C11,C11 produce a pole at ca. 1.5usec now. Quite a bit lower then before. They roll of the open loop gain over about 60kHz. R18, R4, C13, C14 form a second pole of about 15usec. That puts the open loop rolloff of the second stage at a bit under 10kHz. The workings of the compleet stage is a bit complex but i try to explain in my own way: The cascoded input stage amplifies the signal and the gain of that stage is set to 26dB by a lokal current feedback loop around R12, R12. The open loop gain of that stage is rolled of by R7, R8, C11, C12 over 60kHz.
The output impedance of that stage is about 490 ohm and forms a shunt feedback stage with OP1 and OP2. The amplification of that stage is 33000 / 490 = 67 = 36dB. R18, R4, C13, C14 roll of the open loop gain of that stage over ca.7kHz. R1, R2, R3 provide global feedback for a closed loop gain of 37dB. So we have about 25dB of feedback.
I expect that the upper frequency limit (-3dB) of the total INA should be around 120kHz because global feedback extends the frequency response of the open loop gain stage proportional to the feedback factor. This performance though good is a little slow for my taste so i will try less agressive compensation and see how far i can extent the speed without stabilty problems. One idea is to raise R4, R18 to 100kOhm and lower C13, C14 to 68pF for example. If that circuit is stable it whould go like this. 26dB gain in the first stage, 46dB in the second stage, open loop rollof over 15kHz. We whould then have a feedback factor of 35dB now. The upper frequency limit whould then be 750kHz (-3dB) provided the Opamps are fast enough. Let´s see where i can get.
See the working circuit and the planed modifications.
P.S. I had first put in other OPs ( NE5534) to look if the stability problem was caused by the LT1115 ( that i had perhaps already fried) but that was not the case, they got warm too until i found a solution. What was interesting with the NE5534 was that they had quite low and nearly similiar offset of 6 - 7mV at the output that canceled to under 2mV in the differential stage. D.Self who is an expert on NE5534 had pre warned me that they have surprising low offset, but stiil that was good to see.
The output impedance of that stage is about 490 ohm and forms a shunt feedback stage with OP1 and OP2. The amplification of that stage is 33000 / 490 = 67 = 36dB. R18, R4, C13, C14 roll of the open loop gain of that stage over ca.7kHz. R1, R2, R3 provide global feedback for a closed loop gain of 37dB. So we have about 25dB of feedback.
I expect that the upper frequency limit (-3dB) of the total INA should be around 120kHz because global feedback extends the frequency response of the open loop gain stage proportional to the feedback factor. This performance though good is a little slow for my taste so i will try less agressive compensation and see how far i can extent the speed without stabilty problems. One idea is to raise R4, R18 to 100kOhm and lower C13, C14 to 68pF for example. If that circuit is stable it whould go like this. 26dB gain in the first stage, 46dB in the second stage, open loop rollof over 15kHz. We whould then have a feedback factor of 35dB now. The upper frequency limit whould then be 750kHz (-3dB) provided the Opamps are fast enough. Let´s see where i can get.
See the working circuit and the planed modifications.
P.S. I had first put in other OPs ( NE5534) to look if the stability problem was caused by the LT1115 ( that i had perhaps already fried) but that was not the case, they got warm too until i found a solution. What was interesting with the NE5534 was that they had quite low and nearly similiar offset of 6 - 7mV at the output that canceled to under 2mV in the differential stage. D.Self who is an expert on NE5534 had pre warned me that they have surprising low offset, but stiil that was good to see.
This circuit needs quite big 10uF coupling caps from the input to the shunt stage and i do not have much space on my proto PCB. So i build me a low distortion bipolar elcap. There is some controversy if bias on an elcap is a good thing but i trust the measurements of Bateman that found that for lowest distortion they like the least amount of bias posible. Bipolars faired better in his test too so i build my own version. It is made from four 10uF 63V elcaps. Two back to back positive and two back to back negative and then a good quality 1uF foil on top of that for high frequency bypass. I made one of my infamous little drawings and i hope you understand the concept. Some time ago i found a handy article about amplifier stability. I post it here for download.
Attachments
I put more work into the Balanced Discreet INA headamp. The goal was to see how far i can extend the bandwidth without stability issues. I am stiil not able to work in my lab in the factory because we are moving into a new building and the lab has not been setup so far. So all i have is a 40,- € multimeter and pen and paper.
I raised the gain in the transimpedance stage by 10dB and set C13, C14 to 220pF for a time constant of 22usec. An NE5534 with a gain-badwidth of 22Mhz is just stabe under this circumstances. When i lower that timeconstant then first the offset drifts higher and then the chip gets warm and then hot when i lower that constant much. Chips like the LT1115 (GB = 40MHz) or the LT1468 (GB = 90MHz) are not stable under this condisions. I tried an OP27 ( GB = 8MHz) and that was rock solid of cause. Offset after the difference stage with the OP27 was under 1mV so a precission amp in that position is a good idea when you do not like to use a servo. With the NE5534 offset was around 5mV because of canceling in the dif stage. Instead of this old cestnuts you could use some more recent like the OPA227 that has lower distortion and settlig time. I do not expect much better measured data because the OPs are running in inverted mode and the transfer linearity used is much better then the common mode not inverted case.
A asume bandwidth of this circuit to be around 300kHz but i have to measure it to be sure.
I did also some weard experiments that shoud shock the OP swappers. I put NE5534 in one channel and OP27 in the other. The sound was just fine. I also experienced that cable differences got much smaller with this circuit ( i changed the cable from arm to headamp for that experiment). With the unbalalanced circuits the cable differences where quite obvious if not night and day. Thinking about it, it may be that shelding is an issue that did not get enough attention.
I run this high gain circuit wide in the open without case so i had some slight hum at first. It disappered when i connected one point of the groundplane over a 6.8uF cap to earth of the wall outlet. If that is a soluion that whould work in other contries i can not say but that circuit deserves a PCB and a nice case anyway.
You may ask me how that circuit sounds. Well i changed my information policy a little.
Ask me and then i reply. If you are not interested in subjective apraisal because it may influence your objective point of view, forget about that. It is just my experience and another person may have other priorities. For me the case if a circuit without global feedback sounds better then a circuit with ( at east in the case of a balanced INA) will soon settle. What i need to make now are some measurements and more listening.
P.S. : Noise is a bit higher in this circuit because of the balanced nature (3dB) and the 20 Ohm resistor. I will build a version with 4 + 4 fets and lower that resistor by using an OP with more current drive. That kind of amps have a GB of around 10Mhz so should not create any stability problems here.
Noise with the circuit right now is still low enough to be inaudible at the listening seat when i play the volume i like.
I raised the gain in the transimpedance stage by 10dB and set C13, C14 to 220pF for a time constant of 22usec. An NE5534 with a gain-badwidth of 22Mhz is just stabe under this circumstances. When i lower that timeconstant then first the offset drifts higher and then the chip gets warm and then hot when i lower that constant much. Chips like the LT1115 (GB = 40MHz) or the LT1468 (GB = 90MHz) are not stable under this condisions. I tried an OP27 ( GB = 8MHz) and that was rock solid of cause. Offset after the difference stage with the OP27 was under 1mV so a precission amp in that position is a good idea when you do not like to use a servo. With the NE5534 offset was around 5mV because of canceling in the dif stage. Instead of this old cestnuts you could use some more recent like the OPA227 that has lower distortion and settlig time. I do not expect much better measured data because the OPs are running in inverted mode and the transfer linearity used is much better then the common mode not inverted case.
A asume bandwidth of this circuit to be around 300kHz but i have to measure it to be sure.
I did also some weard experiments that shoud shock the OP swappers. I put NE5534 in one channel and OP27 in the other. The sound was just fine. I also experienced that cable differences got much smaller with this circuit ( i changed the cable from arm to headamp for that experiment). With the unbalalanced circuits the cable differences where quite obvious if not night and day. Thinking about it, it may be that shelding is an issue that did not get enough attention.
I run this high gain circuit wide in the open without case so i had some slight hum at first. It disappered when i connected one point of the groundplane over a 6.8uF cap to earth of the wall outlet. If that is a soluion that whould work in other contries i can not say but that circuit deserves a PCB and a nice case anyway.
You may ask me how that circuit sounds. Well i changed my information policy a little.
Ask me and then i reply. If you are not interested in subjective apraisal because it may influence your objective point of view, forget about that. It is just my experience and another person may have other priorities. For me the case if a circuit without global feedback sounds better then a circuit with ( at east in the case of a balanced INA) will soon settle. What i need to make now are some measurements and more listening.
P.S. : Noise is a bit higher in this circuit because of the balanced nature (3dB) and the 20 Ohm resistor. I will build a version with 4 + 4 fets and lower that resistor by using an OP with more current drive. That kind of amps have a GB of around 10Mhz so should not create any stability problems here.
Noise with the circuit right now is still low enough to be inaudible at the listening seat when i play the volume i like.
Joachim,
I apologise if I sound like a broken record, and am very interested in how these circuits sound, but am disappointed that you didn't seem to do any loading resistor comparisons with one consistent circuit. It seems you put 47k in one circuit, then changed the circuit and then tried 1k.
Have you tried 47k compared to other values in one circuit? Comments?
Regards, Allen
I apologise if I sound like a broken record, and am very interested in how these circuits sound, but am disappointed that you didn't seem to do any loading resistor comparisons with one consistent circuit. It seems you put 47k in one circuit, then changed the circuit and then tried 1k.
Have you tried 47k compared to other values in one circuit? Comments?
Regards, Allen
Allen, maybe you are not up to date but this thread get´s long so no wonder you missed some details. I may also not have comunicated everything i do here. I used the High Z MPP Fet Headamp consistently to listen to resistors. I have tested it with 47kOhm 1% Metalfilm, 1kOhm YAM2, 3kOhm Yam2, 1 KOhm nonmagnetic Tantal (Shinko ?), 1kOhm
Dale RS60 antimagnetic, 1kOhm Texas components Z foil, 1kOhm Caddock MK132.
Here is my ranking:
1. Z foil : best combination of detail and musicallity
2. Tantal : nearly as open as the Z foil, very elegant tonality
3. YAM2 : i could not hear a big difference between 1kOhm and 3kOhm. Very close to
Tanatal with a slightly more light tone
4: Caddock MK132: slighly bright. Came out best in an earlier version of the Fet
Headamp where it brought a wellcome opening of the sound.
5: Dale RS60 : kind of disappointing in this aplication. Loss of microdetail like tape
splices, preechos etc. Very "evarage" sound. Nothing spectacular and
moving.
6: 47kOhm Metalfilm: i only listened breefly to this combination because i prefered all
of the others.
So what i should do a comparison of the 1kOhm Z foil to a 47kOhm Z foil but i have non right now.
Sound of the Discreet INA : my finest yet. No loss of microdetail as far as i can tell. I listened to very critical records that i know in and out like "Return to Forever". The first time i heard a rattling noise in one of my speakers that proved to be of mechanical origin so low is the distortion and so high the dynamic range. I had to place that speaker slightly different on the stand to tune that problem away.
I use 2 x 47kOhm metal film for loading of this stage and use ages old OPs so improvement may be posible still.
The circuit is remarkable tolerant for PSU issues and component coices. For example i run it currently on two lead gel accus. One nearly empty at 9V and the other full at 12.6V. No problem with offset and noise or audible distortion under this circumstances.
It is really a kind of worste case szenario right now.
Dale RS60 antimagnetic, 1kOhm Texas components Z foil, 1kOhm Caddock MK132.
Here is my ranking:
1. Z foil : best combination of detail and musicallity
2. Tantal : nearly as open as the Z foil, very elegant tonality
3. YAM2 : i could not hear a big difference between 1kOhm and 3kOhm. Very close to
Tanatal with a slightly more light tone
4: Caddock MK132: slighly bright. Came out best in an earlier version of the Fet
Headamp where it brought a wellcome opening of the sound.
5: Dale RS60 : kind of disappointing in this aplication. Loss of microdetail like tape
splices, preechos etc. Very "evarage" sound. Nothing spectacular and
moving.
6: 47kOhm Metalfilm: i only listened breefly to this combination because i prefered all
of the others.
So what i should do a comparison of the 1kOhm Z foil to a 47kOhm Z foil but i have non right now.
Sound of the Discreet INA : my finest yet. No loss of microdetail as far as i can tell. I listened to very critical records that i know in and out like "Return to Forever". The first time i heard a rattling noise in one of my speakers that proved to be of mechanical origin so low is the distortion and so high the dynamic range. I had to place that speaker slightly different on the stand to tune that problem away.
I use 2 x 47kOhm metal film for loading of this stage and use ages old OPs so improvement may be posible still.
The circuit is remarkable tolerant for PSU issues and component coices. For example i run it currently on two lead gel accus. One nearly empty at 9V and the other full at 12.6V. No problem with offset and noise or audible distortion under this circumstances.
It is really a kind of worste case szenario right now.
Some may not like the coupling caps in the feedback chain so i thought about a way to DC couple. I think two simple integrators can do the job just fine. i did not build this, that has to wait until tomorrow.
While experimenting with my subwoofer i switched my poweramp on and off one time too much. When i can not get some fuses over the weekend i have to listen with my tube monoblocks. This is saturday the 13th and not fryday the 13th i thought.
Freudsche Fehlleistung included.
While experimenting with my subwoofer i switched my poweramp on and off one time too much. When i can not get some fuses over the weekend i have to listen with my tube monoblocks. This is saturday the 13th and not fryday the 13th i thought.
Freudsche Fehlleistung included.
Today i DC coupled one channel ( the one with the NE5534) to look if it works and if a run into any problems. The circuit worked fine exactly as posted before. No pumping at high level, no increase in noise.
I then did my wostcase experiment again, this time with two fully loaded 4AH lead accumulators from motorcircles. I played the DC coupled version on the left channel and the AC coupled version on the right channel. Again the sound was just fine. On stereo material it is hard to hear small differences like that anyway. Too much interesting things are going on. It certainly had no effect on the focus. Voices and instrument appeared sharply deleniated on the soundstage. Tomorrow i will DC couple the other channel and next week i will try out the same Opamps in both channels.
So when you hate servos put in caps and when you hate caps put in servos. I prefer the servo solution because an elcap can get faulty over time and a 10uF film cap is big, even a low voltage Polyester type not to mention PP or even teflon. A compact layout is then quite difficult. One theoretical advantage of the servo is that OP1 and OP2 are now running on a virtual ground and circulating ground currents can not enter PIN3 so easy any more. This beast is quite complex but sounds very transparent and works without problem. If you want the advantages of a discreet INA like good common mode rejection, very low distortion and good low level resolution i do not know which part i could spare. They all are neccessary in one or the other way.
I used Texas Instruments TL071C for servo purpose and the offset after the Diff Amp is under 2mV and rock stable. The TL051 has even better and garantied precission and other low offset OPs can be used here based on wallet and taste. I doupt if they bring a measurable benefit. This servo is very slow and should not worsen distortion in the audible range. I usually use OPA134 in that position to satisfy my soul to avoid conflict with the advanced user and even the very expensive OPA627 has been witnessed in this position. Anyway, the circuit erodes component coices and PSU issues extremely effective. Well i used paralled 1% PS-HX from LCR in the 75usec RIAA for good channel matching and stabilty. The new Röderstein 1% MKP´s work as well here. Both are easy to get at Farnell or RS Components and are not expensive for my taste. I have not tried Teflon in this position yet but have specified them in one of my commercial stages.
All the listening tests of the different Headamps where all done with the capacitive RIAA. I will post that circuit again for the compleet picture to what i am listenimg right now. Tha RIAA stage is run double momo on two LC Audio low noise regulators with 29mH coils in series and generous general and distributed decoupling. The INA runs wide in the open, dead bug style with my triple ground plane system and one point of the groundplane connected with a 6.8uF PP cap to earth of the wall outlet. That made a huge difference. I use 1m Wire World Silver Eclipse between tonearm and Headamp and 0.5m Platinum Eclipse between Headamp and RIAA. As i said before i could not hear much difference between my chaos fractal cable and the Wire World this time. I use the Wire World now because my cable is kind of fragile and i do not want to destroy it during stage swapping.
By the way, the Opamp in the capacitive RIAA is an LT1468 and the servo is an AD711 i think (maybe OPA134). I mounted the servo upside down so i can not see it. Maybe with a small mirror. The 100nF cap is a Vishey/Röderstein MKP1837 1%, the caps in the servo are NOS yellow Röderstein MKT. That simple circuit sounds very pleasing but is in no way the end of the story.
So what more can be done ?
I wait for the MU metal encapsulated precission coils from Souther.
When i have them i will try the inductive RIIA again and will see if i can erode the last trace of hum.
A fully ballanced RIAA with Balanced ouput could be developped. I have posted circuits like that before (without the balanced output).
That could make sense in my system because i have a 5m run of Koax Video cable between the RIAA stage and my preamp
I then did my wostcase experiment again, this time with two fully loaded 4AH lead accumulators from motorcircles. I played the DC coupled version on the left channel and the AC coupled version on the right channel. Again the sound was just fine. On stereo material it is hard to hear small differences like that anyway. Too much interesting things are going on. It certainly had no effect on the focus. Voices and instrument appeared sharply deleniated on the soundstage. Tomorrow i will DC couple the other channel and next week i will try out the same Opamps in both channels.
So when you hate servos put in caps and when you hate caps put in servos. I prefer the servo solution because an elcap can get faulty over time and a 10uF film cap is big, even a low voltage Polyester type not to mention PP or even teflon. A compact layout is then quite difficult. One theoretical advantage of the servo is that OP1 and OP2 are now running on a virtual ground and circulating ground currents can not enter PIN3 so easy any more. This beast is quite complex but sounds very transparent and works without problem. If you want the advantages of a discreet INA like good common mode rejection, very low distortion and good low level resolution i do not know which part i could spare. They all are neccessary in one or the other way.
I used Texas Instruments TL071C for servo purpose and the offset after the Diff Amp is under 2mV and rock stable. The TL051 has even better and garantied precission and other low offset OPs can be used here based on wallet and taste. I doupt if they bring a measurable benefit. This servo is very slow and should not worsen distortion in the audible range. I usually use OPA134 in that position to satisfy my soul to avoid conflict with the advanced user and even the very expensive OPA627 has been witnessed in this position. Anyway, the circuit erodes component coices and PSU issues extremely effective. Well i used paralled 1% PS-HX from LCR in the 75usec RIAA for good channel matching and stabilty. The new Röderstein 1% MKP´s work as well here. Both are easy to get at Farnell or RS Components and are not expensive for my taste. I have not tried Teflon in this position yet but have specified them in one of my commercial stages.
All the listening tests of the different Headamps where all done with the capacitive RIAA. I will post that circuit again for the compleet picture to what i am listenimg right now. Tha RIAA stage is run double momo on two LC Audio low noise regulators with 29mH coils in series and generous general and distributed decoupling. The INA runs wide in the open, dead bug style with my triple ground plane system and one point of the groundplane connected with a 6.8uF PP cap to earth of the wall outlet. That made a huge difference. I use 1m Wire World Silver Eclipse between tonearm and Headamp and 0.5m Platinum Eclipse between Headamp and RIAA. As i said before i could not hear much difference between my chaos fractal cable and the Wire World this time. I use the Wire World now because my cable is kind of fragile and i do not want to destroy it during stage swapping.
By the way, the Opamp in the capacitive RIAA is an LT1468 and the servo is an AD711 i think (maybe OPA134). I mounted the servo upside down so i can not see it. Maybe with a small mirror. The 100nF cap is a Vishey/Röderstein MKP1837 1%, the caps in the servo are NOS yellow Röderstein MKT. That simple circuit sounds very pleasing but is in no way the end of the story.
So what more can be done ?
I wait for the MU metal encapsulated precission coils from Souther.
When i have them i will try the inductive RIIA again and will see if i can erode the last trace of hum.
A fully ballanced RIAA with Balanced ouput could be developped. I have posted circuits like that before (without the balanced output).
That could make sense in my system because i have a 5m run of Koax Video cable between the RIAA stage and my preamp
Hi Joachim,
it'd be so cool if you could provide a photo of that discrete ina input stage with triple groundplane and deadbug-opamp style!
Rüdiger
it'd be so cool if you could provide a photo of that discrete ina input stage with triple groundplane and deadbug-opamp style!
Rüdiger
It certainly looks quite jammed !
My digital camera was stolen and my son lost his in vacation.
Anyway, i wil buy a new one, and soon i will be back in the factory where we have professional posibilizies.
My digital camera was stolen and my son lost his in vacation.
Anyway, i wil buy a new one, and soon i will be back in the factory where we have professional posibilizies.
In search of low noise
Today i DC coupled the other channel and everything worked fine.
Listening revealed that there was not much of a difference between the AC and DC version. I think the DC version is a little bit more transparent and the AC version a little bit more sweet and forgiving. No big deal. I can recomend both.
Yesterday i started to work on my subwoofer because at insane levels there was a slight hum comming out of the woofer only and interestingly not out the main speakers.
I have a quite sofisticated wall power outlet scheme and tried to trace down the hum problem by trying out different power connections. I have 3 independent lines. One line with particular low impedance and filtering. Wherever i took the power from there was no difference. Then i tryed to isolate the earth connection. Again no success.
I usually piggy bag my woofer high level after the poweramp, so i tried to connect the woofer low level from the preamp. I wanted to try that anyway because it is one amplifier less in the chain. Again no succes with the hum but the sound got more clear and transparent in the bass and also a bit lighter in tone. I readjusted the sub volume to my usual level by listening to familiar records. I do only simple acoustic measurements when i set up a woofer. Usually with pink noise. Records vary in amount of bass tremendoesly so even when i find a well measuring setup, i tune by ear afterwards.
Thinking about the problem for a while it struck me that i had used coaxial unbalanced cable from my tonearm so far. Technically it works because my arm has a seperate earth but much of the advantage of a balanced stage to suppress hum is lost that way.
So i made a truly balanced cable in the usual style. Twisted pair with seperate shield.
I made it very short (0.3m) and used Spiral Groove Litzwire with Eichmann Bullet Plugs.
After i inserted the cable and put the shield to the central mass point the hum in the sub was totally gone !
Also the hiss in the speakers sounded different. Before i had heard high frequency components that sounded a bit like muted fire crackers. Now it was just plain old white hiss with a much more benign character. The channel with the NE sounded even more soft so i decided not to compare the NE5534 with the OP27 any more but use NE5534 on both channels. I am running out of OPs, so i have to buy some next week.
The sound after that major change was better in any way. Especially dynamic range was increased and the bass much cleaner and faster.
I do not claim that the magic properties of the used wire and termination made the difference. No, this time it was only good engineering praxis.
In search of low noise:
The circuit i posted here has a voltage noise of ca. 1.45nV/qHz. I a balanced stage the voltage gain is higher but the input transistors are in series for a noise disadvantage of 3dB. A stage with an integrated INA163 would need a gain resistor of ca. 80 Ohms for the same gain of 36dB. Noise under this circumstances is around 2nV/qHz so some 3dB worse. A lot of expensive phonostages up to 3000,-€ use that chip and people are still not complaining about noise as far as i know.
1.45nV/qHz is not a world record and i started to think how that circuit could be improved.
The NE5534 can drive a 600 Ohm resistor to over 10V without much distortion so i lowered the feedback resistors to 340 Ohm.
50mV at the input whould drive the resistors to 1.7V and asuming that the NEs can put out 20mA of clean power there is still some headroom to 6.8V. The compleet stage doubles that at the output so there is still ample overload margin in theory.
That circuit has an equivalent voltage noise of 1.2nV/qHz. Not a huge improvement but i rebuild the stage with that modification to prove the concept. It works without detectable problem. The trimmers have to readjusted because the Fets draw a bit more current under tis conditions. Around 6mA each.
The next option is to double the input transistors. 4 plus 4 is the maximum i have seen so i will try that tomorrow. The current sources have to be slighly modified to be able to supply the expected 24mA of current.
That circuit should have around 1nV/qHz of noise.
The next step is to lower the values of the feedback resistors as far as posible.
One posibility is to use shunt feedback amps with higher current drive but i decided otherwise. I will use a second set of NE5534 as paralled buffers that will double the current drive. The circuit looks a bit exotic and unusual, especially with the aditional integrators but elemnts of it are well documented and it should work, I can now reduce the feedback resistor to 5 Ohm for a total of ca. 0.8nV/qHz.
From now on improvemenzs will be very small. Even if i whould reduce the resistors to 1 Ohm this circuit can not go under 0.6nV/qHz in theory. I praxis this amounts to less then 1dB.
Today i DC coupled the other channel and everything worked fine.
Listening revealed that there was not much of a difference between the AC and DC version. I think the DC version is a little bit more transparent and the AC version a little bit more sweet and forgiving. No big deal. I can recomend both.
Yesterday i started to work on my subwoofer because at insane levels there was a slight hum comming out of the woofer only and interestingly not out the main speakers.
I have a quite sofisticated wall power outlet scheme and tried to trace down the hum problem by trying out different power connections. I have 3 independent lines. One line with particular low impedance and filtering. Wherever i took the power from there was no difference. Then i tryed to isolate the earth connection. Again no success.
I usually piggy bag my woofer high level after the poweramp, so i tried to connect the woofer low level from the preamp. I wanted to try that anyway because it is one amplifier less in the chain. Again no succes with the hum but the sound got more clear and transparent in the bass and also a bit lighter in tone. I readjusted the sub volume to my usual level by listening to familiar records. I do only simple acoustic measurements when i set up a woofer. Usually with pink noise. Records vary in amount of bass tremendoesly so even when i find a well measuring setup, i tune by ear afterwards.
Thinking about the problem for a while it struck me that i had used coaxial unbalanced cable from my tonearm so far. Technically it works because my arm has a seperate earth but much of the advantage of a balanced stage to suppress hum is lost that way.
So i made a truly balanced cable in the usual style. Twisted pair with seperate shield.
I made it very short (0.3m) and used Spiral Groove Litzwire with Eichmann Bullet Plugs.
After i inserted the cable and put the shield to the central mass point the hum in the sub was totally gone !
Also the hiss in the speakers sounded different. Before i had heard high frequency components that sounded a bit like muted fire crackers. Now it was just plain old white hiss with a much more benign character. The channel with the NE sounded even more soft so i decided not to compare the NE5534 with the OP27 any more but use NE5534 on both channels. I am running out of OPs, so i have to buy some next week.
The sound after that major change was better in any way. Especially dynamic range was increased and the bass much cleaner and faster.
I do not claim that the magic properties of the used wire and termination made the difference. No, this time it was only good engineering praxis.
In search of low noise:
The circuit i posted here has a voltage noise of ca. 1.45nV/qHz. I a balanced stage the voltage gain is higher but the input transistors are in series for a noise disadvantage of 3dB. A stage with an integrated INA163 would need a gain resistor of ca. 80 Ohms for the same gain of 36dB. Noise under this circumstances is around 2nV/qHz so some 3dB worse. A lot of expensive phonostages up to 3000,-€ use that chip and people are still not complaining about noise as far as i know.
1.45nV/qHz is not a world record and i started to think how that circuit could be improved.
The NE5534 can drive a 600 Ohm resistor to over 10V without much distortion so i lowered the feedback resistors to 340 Ohm.
50mV at the input whould drive the resistors to 1.7V and asuming that the NEs can put out 20mA of clean power there is still some headroom to 6.8V. The compleet stage doubles that at the output so there is still ample overload margin in theory.
That circuit has an equivalent voltage noise of 1.2nV/qHz. Not a huge improvement but i rebuild the stage with that modification to prove the concept. It works without detectable problem. The trimmers have to readjusted because the Fets draw a bit more current under tis conditions. Around 6mA each.
The next option is to double the input transistors. 4 plus 4 is the maximum i have seen so i will try that tomorrow. The current sources have to be slighly modified to be able to supply the expected 24mA of current.
That circuit should have around 1nV/qHz of noise.
The next step is to lower the values of the feedback resistors as far as posible.
One posibility is to use shunt feedback amps with higher current drive but i decided otherwise. I will use a second set of NE5534 as paralled buffers that will double the current drive. The circuit looks a bit exotic and unusual, especially with the aditional integrators but elemnts of it are well documented and it should work, I can now reduce the feedback resistor to 5 Ohm for a total of ca. 0.8nV/qHz.
From now on improvemenzs will be very small. Even if i whould reduce the resistors to 1 Ohm this circuit can not go under 0.6nV/qHz in theory. I praxis this amounts to less then 1dB.
I am listening to the 5.0 version of the INA right now and doubling of he input transistors gave a bigger subjective improvement then going from 20 to 10 Ohm in the feedback resistors. If you allow me some speculation, i asume that noise from a plain resistor is less objectionable then noise from a FET. Equivalent input noise measured in nV/qHz does not tell about the quality of the noise. I have to make some measurements to sustain that notion of cause.
I am very pleased with the sound and dynamic range is now so high and distortion is so low that i can hear at very advanced volume without fatique.
The only problem is that my speakers give up before i give up.
I am using small 2-way monitors with 13cm woofers and 25mm dome tweeters plus an active subwoofer with 140W Hypex class D amp. That woofer stands in the middle between the main spaekers close to the frontwall and i am sitting close to the backwall.
Under theese circumstances the woofer (25cm long throw metalcone) has ample roomgain and no problems with subsonic because it summs the outputs of both channels. Disturbance from warped records cancel in that setup.
The problem is cone movement in the monitors and now that i supply the woofer from the preamp, i thought how i could releave the speakers a bit from heavy basswork.
I have lend my TDS subsonic filter to a friend, so i needed another quick solution.
The poweramp supplies the monitors only so i placed a little cap in front of the poweramp. It rolls of with 6dB octave under 100Hz and has 20dB damping at 10 Hz the aerea where the arm-cartridge resonance is. Really slow movements of the cones because of warped records are even more suppressed that way.
I had build little filters that way over 25 years ago with good succes.
There was a commercial version for TMR Standart speakers and Omtec class a monoblocks. The filter between woofer and midrange of the TMR was 6dB so i substituted the passive filter with little modules that could be inserted at the input of the poweramps. The gain in soundquality was quite substancial and the combination got popular in Germany.
The sound i am getting now after that modification took a big leap forward.
I can listen now extremely loud without bottoming the small woofers.
The sound gained in clarity and precission and was somehow "sweeter" without loss of information. I attribute that to lowering of mechanical distortion.
Whatever few losses that little cap may have ( i used a 33nF Röderstein MKP1937 to be specific) is swapped by the huge gain in quality because of less mechanical stress.
Somewhere in a chain of DC coupled high gain stages i recomend a safety cap anyway and this is the perfect place because it serves a double purpose. "Zwei Fliegen mit einer Klappe"
Just for fun i had disconnected one of the supply voltages of a DC coupled stage at full volume and you can imagine what nasty things happened next.
I am very pleased with the sound and dynamic range is now so high and distortion is so low that i can hear at very advanced volume without fatique.
The only problem is that my speakers give up before i give up.
I am using small 2-way monitors with 13cm woofers and 25mm dome tweeters plus an active subwoofer with 140W Hypex class D amp. That woofer stands in the middle between the main spaekers close to the frontwall and i am sitting close to the backwall.
Under theese circumstances the woofer (25cm long throw metalcone) has ample roomgain and no problems with subsonic because it summs the outputs of both channels. Disturbance from warped records cancel in that setup.
The problem is cone movement in the monitors and now that i supply the woofer from the preamp, i thought how i could releave the speakers a bit from heavy basswork.
I have lend my TDS subsonic filter to a friend, so i needed another quick solution.
The poweramp supplies the monitors only so i placed a little cap in front of the poweramp. It rolls of with 6dB octave under 100Hz and has 20dB damping at 10 Hz the aerea where the arm-cartridge resonance is. Really slow movements of the cones because of warped records are even more suppressed that way.
I had build little filters that way over 25 years ago with good succes.
There was a commercial version for TMR Standart speakers and Omtec class a monoblocks. The filter between woofer and midrange of the TMR was 6dB so i substituted the passive filter with little modules that could be inserted at the input of the poweramps. The gain in soundquality was quite substancial and the combination got popular in Germany.
The sound i am getting now after that modification took a big leap forward.
I can listen now extremely loud without bottoming the small woofers.
The sound gained in clarity and precission and was somehow "sweeter" without loss of information. I attribute that to lowering of mechanical distortion.
Whatever few losses that little cap may have ( i used a 33nF Röderstein MKP1937 to be specific) is swapped by the huge gain in quality because of less mechanical stress.
Somewhere in a chain of DC coupled high gain stages i recomend a safety cap anyway and this is the perfect place because it serves a double purpose. "Zwei Fliegen mit einer Klappe"
Just for fun i had disconnected one of the supply voltages of a DC coupled stage at full volume and you can imagine what nasty things happened next.
To reduce the noise of the Discreet INA further it is neccesary to reduce the feadback resistor to as low a value as posible. My solution is to add a buffer to double the current drive. I thought that putting it in the shunt feedback loop is not neccesarry and may create stability problems. Simply adding it at the output whould work as well i think. The buffer has 100% of feedback and is fast and low distortion. It is still in the global loop anyway. So i have redrawn the shematic slightly. I raised the decoupling resistors slightly and adjusten the 170 ohm resistors to 150 Ohm to compensate for the aditional impedance.
I will build this first and when it performes well, i will keep it that way.
I will build this first and when it performes well, i will keep it that way.
I had a funny idea. The input stage of the INA could be transfered into a Grounded Gate Low Z version. I have seen that done with tubes in an old VLT book by David Manley.
If it works here i do not know yet. I may try it later in the day.
Big elcaps are nessesary at the input but there is very low bias at that point and they work like a bipolar cap in this aplication so distortion from this source should be low.
They can be bypassed with a foil of say 1uF.
I have only 2200uF caps of good quality around so provided they work on a 10 Ohm feedback resistor they form a highpass with -1.8dB at 20 Hz. I can compensate for that in my sub. It has an equaliser that works in that low range.
The caps are Nichicon CL with very low ESR. They have an organic dielectric and have some leakage current so they may not be ideal here. I know at least one cap from Elna that is optimised for low leakage audio work, so that one should do a bit better. Anyway, leakage should cancel somewhat in that arrangement.
If you want to go low and have no provision to adjust here are some calculations what whould happen provided the feedback resistor is 10 Ohm. If a 5 Ohm is used the frequency doubles.
4700 uF : 20Hz -0.4dB 6.6Hz -3dB
10000 uF : 20Hz -0.1dB 3Hz -3dB
A measue of subsonic filtering could be intruduces this way and the sound can actually improve because of that.
If it works here i do not know yet. I may try it later in the day.
Big elcaps are nessesary at the input but there is very low bias at that point and they work like a bipolar cap in this aplication so distortion from this source should be low.
They can be bypassed with a foil of say 1uF.
I have only 2200uF caps of good quality around so provided they work on a 10 Ohm feedback resistor they form a highpass with -1.8dB at 20 Hz. I can compensate for that in my sub. It has an equaliser that works in that low range.
The caps are Nichicon CL with very low ESR. They have an organic dielectric and have some leakage current so they may not be ideal here. I know at least one cap from Elna that is optimised for low leakage audio work, so that one should do a bit better. Anyway, leakage should cancel somewhat in that arrangement.
If you want to go low and have no provision to adjust here are some calculations what whould happen provided the feedback resistor is 10 Ohm. If a 5 Ohm is used the frequency doubles.
4700 uF : 20Hz -0.4dB 6.6Hz -3dB
10000 uF : 20Hz -0.1dB 3Hz -3dB
A measue of subsonic filtering could be intruduces this way and the sound can actually improve because of that.
i prommised to be more specific in the future so i looked up the spec sheet of the Nichicon CL and compared it to the Elna RLB that is optimised for audio work:
Nichicon CL : Leakage current: 0.01CV or 3uA
Tan Delta: 0.26%
ERS: 0.06 Ohm @ 100kHz
Ripple current: 1190mA
Elna RLB: Leakage current: 0.006CV or 0.5uA
Ripple current: 1400mA
The CL is not that bad in any way and has a very small case.
The Elna is exeptional concerning low leakage but is somewhat bigger.
I compared the 6.3V versions.
Nichicon CL : Leakage current: 0.01CV or 3uA
Tan Delta: 0.26%
ERS: 0.06 Ohm @ 100kHz
Ripple current: 1190mA
Elna RLB: Leakage current: 0.006CV or 0.5uA
Ripple current: 1400mA
The CL is not that bad in any way and has a very small case.
The Elna is exeptional concerning low leakage but is somewhat bigger.
I compared the 6.3V versions.
I ocured to me that noise could be lower in this arrangement because the cartridge impedance is now in parallel with the feedback resistor. The disadvantage is that the NE5534 have to drive (indirectly) a 3.75 Ohm resistor. That lowers dynamic range somewhat and makes the gain higher. The votage gain is now 182 x so 45dB. When i asume that the NEs can drive a 340 Ohm resistor to 5V for a total of 10V after the dif amp. 10 / 180 = 55mV so asuming that the cartridge has 0.5mV output at medium cutting level there is still 40dB overload margin. This headamp has huge dynamic
range !
range !
Actually the RIAA curve boosts the treble by 20dB so overload margin in the treble is somewaht reduced. Music has not much content here but scratches and blemishes could give high transient output. A clean and well preserved record is a good starting point or the gain could be reduced and the feedback driven with higher current.
The 75usec part of the RIAA could be implemented in the INA feedback loop to cure that issue too.
Anyway, having measured the voltage that develops over the 340 Ohm resistors while actually playing records i can asure you that this does not happen often in reality.
The 75usec part of the RIAA could be implemented in the INA feedback loop to cure that issue too.
Anyway, having measured the voltage that develops over the 340 Ohm resistors while actually playing records i can asure you that this does not happen often in reality.