I've been working for one year in gainclones and now I have two designs (will post schematics soon). I have noticed that there's a huge difference in their sound.
The problem now is that i can't guess which way to go, partly because there is a compromise between good-sounding and fidelity and partly because when someone has an amplifier that does something really well it's difficult to change it for another which does that thing worse, despite doing other things better.
The first of my current designs (yes, i did that first) was one that used a huge nfb (noninverting gain of 11, 10K/1K) with no cap on its feedback loop. I dealed with bias-current induced offset trough using a LM4562 buffer. The buffer is AC-coupled (1uF/47K with the source and DC-coupled to the LM3886) 0.22u caps are attached to all supply pins. The buffer is powered trough a pair 7812/7912 + 470 uF/rail bypassed with 1uF MKT. The board also includes 470uF/rail+snubber to get rid of residual cable inductances. PSU is unregulated snubberized +/-32V 22000 uf/rail bypassed by 1uF mkt on their screws. It also has two 1uF mkt at the pins of the rectifier. It's fan cooled, the fan is connected to the main supply trough a 7812 regulator + 6000uF + "big" inductor + 4000 uF and is absolutely unlistenable with the ear attached to the speaker and no music playing. It does sound full, authoritative and tubey.
The second board is a "classic gainclone" (less parts better aproach, the first has lot of parts). It uses the same PSU. 6x4 cm veroboard, 2200 uF on-board reservoir, ac-coupled trough 1uF/47K, lower nfb (47K/1K) + 100uF offset-removing cap. 0.22 uF mkt soldered to all power legs.
Why I don't like my first design? It's rich and warm but its slow and instrument separation is horrible. It can't be used to play movies trough my speakers (they have a tendency towards being boomy, but this amplifier exagerates it up to a point that sometimes it becomes difficult to understand dialogues in films) and it has a curious harshness that at some passages may become unnoticeable. Female vocals are recessed and mixed with the band up to a point they loose all the magic.
The second design is much faster and cleaner. Vocals are in front of the stage. At first I thought it was more analytical, but after a while i noticed that it was exagerately bright. Bass is recessed up to a point that hurts classical music making it sound more like a tv-ad than like a concert. Rock and electronic music tend to have an excess of bass and aren't affected so much. I would define it's sound as being very tv-ish.
I would ask if someone knows how to improve any of the two designs. The second without it's drawbacks would be the best since it's much cheaper. But I think that using a cap in the feedback loop will always induce bass rollof and some phase-shift. Is the 20K/1K a "magic value"? I've read that the original gaincard was operating at higher gain (slightly above 30 dB, which it's not far from my 33.6 dB gainclone). My first boards used the 20K/1K setup, but they were flawed in other ways and when I did new ones without hum, buzz and noise i could not resist to change other parameters.
Is instrument separation/clarity/female vocal sharpness and warmth/avoiding brightness an excluding choice?
The problem now is that i can't guess which way to go, partly because there is a compromise between good-sounding and fidelity and partly because when someone has an amplifier that does something really well it's difficult to change it for another which does that thing worse, despite doing other things better.
The first of my current designs (yes, i did that first) was one that used a huge nfb (noninverting gain of 11, 10K/1K) with no cap on its feedback loop. I dealed with bias-current induced offset trough using a LM4562 buffer. The buffer is AC-coupled (1uF/47K with the source and DC-coupled to the LM3886) 0.22u caps are attached to all supply pins. The buffer is powered trough a pair 7812/7912 + 470 uF/rail bypassed with 1uF MKT. The board also includes 470uF/rail+snubber to get rid of residual cable inductances. PSU is unregulated snubberized +/-32V 22000 uf/rail bypassed by 1uF mkt on their screws. It also has two 1uF mkt at the pins of the rectifier. It's fan cooled, the fan is connected to the main supply trough a 7812 regulator + 6000uF + "big" inductor + 4000 uF and is absolutely unlistenable with the ear attached to the speaker and no music playing. It does sound full, authoritative and tubey.
The second board is a "classic gainclone" (less parts better aproach, the first has lot of parts). It uses the same PSU. 6x4 cm veroboard, 2200 uF on-board reservoir, ac-coupled trough 1uF/47K, lower nfb (47K/1K) + 100uF offset-removing cap. 0.22 uF mkt soldered to all power legs.
Why I don't like my first design? It's rich and warm but its slow and instrument separation is horrible. It can't be used to play movies trough my speakers (they have a tendency towards being boomy, but this amplifier exagerates it up to a point that sometimes it becomes difficult to understand dialogues in films) and it has a curious harshness that at some passages may become unnoticeable. Female vocals are recessed and mixed with the band up to a point they loose all the magic.
The second design is much faster and cleaner. Vocals are in front of the stage. At first I thought it was more analytical, but after a while i noticed that it was exagerately bright. Bass is recessed up to a point that hurts classical music making it sound more like a tv-ad than like a concert. Rock and electronic music tend to have an excess of bass and aren't affected so much. I would define it's sound as being very tv-ish.
I would ask if someone knows how to improve any of the two designs. The second without it's drawbacks would be the best since it's much cheaper. But I think that using a cap in the feedback loop will always induce bass rollof and some phase-shift. Is the 20K/1K a "magic value"? I've read that the original gaincard was operating at higher gain (slightly above 30 dB, which it's not far from my 33.6 dB gainclone). My first boards used the 20K/1K setup, but they were flawed in other ways and when I did new ones without hum, buzz and noise i could not resist to change other parameters.
Is instrument separation/clarity/female vocal sharpness and warmth/avoiding brightness an excluding choice?
Welcome to the wonderful world of hi-fi compromise! 
You haven't mentioned important factors like what speakers are being used, and whether you use an active pre amp or buffer.
The 'trick' to GC success is in using an active stage before them, and using efficient, easy-to-drive speakers. If you don't, then you will have to accept certain compromises.
It is a fine line between having enough power, and retaining that finesse that gives the GC its 'magic'. One solution is to bi-amp (or tri-amp) your speakers, possibly using a different design for each driver.
You haven't mentioned important factors like what speakers are being used, and whether you use an active pre amp or buffer.
The 'trick' to GC success is in using an active stage before them, and using efficient, easy-to-drive speakers. If you don't, then you will have to accept certain compromises.
It is a fine line between having enough power, and retaining that finesse that gives the GC its 'magic'. One solution is to bi-amp (or tri-amp) your speakers, possibly using a different design for each driver.
I havn't mentioned the speakers because they are quite bad. I knew nothing about "high end" before starting to 'boost' my laptop speakers. When I bought that $6 ic i had no idea that it would make me plan to invest in 1k$ loudspeakers.
Despite not sounding very good they are very revealing of what's driving them, probably due to crossover issues. Impedance drop shall not be a problem since both gainclones nicely drive 4 ohms car speakers that go well below 2 ohms at dc.
The sound differences between them are not releated to the speakers, although exagerated by them, because they are also noticeable with my sennheiser headphones, both direct-driven and with series-parallel dummy loads.
I guess that harshness and sounding like if the amplifier is forced to do something it doesn't want to do of the buffered one is releated to excessive NFB. I choose 10K/1K because its used in BR100 in the national AN-1192 and it shows less distortion than the PA100/single ended ones.
The sound of the second reminds me of one I did six months ago that used a low voltage tube as preamplifier (12AE6). It did not sound well because ac-coupling network sinked too much current from the tube, and this rolled off the bass. I think that the slightly rolled-off bass/bright sound has nothing to do with the cap in the feedback loop because it doesn't change very much if I short it (assuming 1V offset for 5 seconds won't kill my cheap speakers). I did the buffered one because it has less than 5 mV offset and no ugly 100uF caps on the signal path.
I direct-drive them with various sources that range from a m-audio card to a relatively nice-sounding pioneer cd player.
This afternoon i will try to reduce nfb on the first model trough paralleling 1K resistor with the one already installed. I hope that less impedance in the feedback loop will lead to a more musical sound without increasing noise, and that's possible because it does not have any cap so there is no compromise between phase shift and bass rollof. Maybe some of the harsness it's due to being at the edge of stability (in the datasheet it says it's stable at a gain over 10).
I had a pair of quite complicated ideas for the next two. One is about paralleling one Class-A biased LM3886 with an unparalleled one and seeking how to make the class A one supply-sink only the current difference between the "correct output" and the Class-B output I would call it class A+B. The other is about putting an LM3886 in the feedback loop of a LM4562, but setting the LM3886 at a voltage gain of 10 instead of the voltage unity-gain of the Mauro Penasa one, because LM3886 does its best at about 4Vrms output (Also think about biasing it into class A). Does somebody know wich is the best way to go? I would also ask if somebody has listened to LME49810 driving mosfets (I have 20 IRFP240/IRPF9240 that may behave well with enough bias current, despite being quite dangerous since they experiment termal runaway because temperature coeficient is positive at normal operating conditions and only becomes negative at huge currents).
Despite not sounding very good they are very revealing of what's driving them, probably due to crossover issues. Impedance drop shall not be a problem since both gainclones nicely drive 4 ohms car speakers that go well below 2 ohms at dc.
The sound differences between them are not releated to the speakers, although exagerated by them, because they are also noticeable with my sennheiser headphones, both direct-driven and with series-parallel dummy loads.
I guess that harshness and sounding like if the amplifier is forced to do something it doesn't want to do of the buffered one is releated to excessive NFB. I choose 10K/1K because its used in BR100 in the national AN-1192 and it shows less distortion than the PA100/single ended ones.
The sound of the second reminds me of one I did six months ago that used a low voltage tube as preamplifier (12AE6). It did not sound well because ac-coupling network sinked too much current from the tube, and this rolled off the bass. I think that the slightly rolled-off bass/bright sound has nothing to do with the cap in the feedback loop because it doesn't change very much if I short it (assuming 1V offset for 5 seconds won't kill my cheap speakers). I did the buffered one because it has less than 5 mV offset and no ugly 100uF caps on the signal path.
I direct-drive them with various sources that range from a m-audio card to a relatively nice-sounding pioneer cd player.
This afternoon i will try to reduce nfb on the first model trough paralleling 1K resistor with the one already installed. I hope that less impedance in the feedback loop will lead to a more musical sound without increasing noise, and that's possible because it does not have any cap so there is no compromise between phase shift and bass rollof. Maybe some of the harsness it's due to being at the edge of stability (in the datasheet it says it's stable at a gain over 10).
I had a pair of quite complicated ideas for the next two. One is about paralleling one Class-A biased LM3886 with an unparalleled one and seeking how to make the class A one supply-sink only the current difference between the "correct output" and the Class-B output I would call it class A+B. The other is about putting an LM3886 in the feedback loop of a LM4562, but setting the LM3886 at a voltage gain of 10 instead of the voltage unity-gain of the Mauro Penasa one, because LM3886 does its best at about 4Vrms output (Also think about biasing it into class A). Does somebody know wich is the best way to go? I would also ask if somebody has listened to LME49810 driving mosfets (I have 20 IRFP240/IRPF9240 that may behave well with enough bias current, despite being quite dangerous since they experiment termal runaway because temperature coeficient is positive at normal operating conditions and only becomes negative at huge currents).
Hey,
With your "classic", I'm not sure why you would be getting 1V of DC offset at the outputs without the cap. I typically get between 20 to 60 mV on non-inverting GC's I've built.
Have heard that the "100uF offset-removing cap" can have an impact on the sound. If this a standard electrolytic it may add an etched/grainy/bright sound in this position-and a few seconds of listening with the cap bypassed may not be enough to really hear the difference. The gain used on the two of your amps are kind of at the far ends of the spectrum. The original Gaincard, I believe, used 20k/680. It may be a good experiment, as you mentioned, to try different gain settings on the amps.
I'd be interested in seeing the schemo's.
Scott
With your "classic", I'm not sure why you would be getting 1V of DC offset at the outputs without the cap. I typically get between 20 to 60 mV on non-inverting GC's I've built.
Have heard that the "100uF offset-removing cap" can have an impact on the sound. If this a standard electrolytic it may add an etched/grainy/bright sound in this position-and a few seconds of listening with the cap bypassed may not be enough to really hear the difference. The gain used on the two of your amps are kind of at the far ends of the spectrum. The original Gaincard, I believe, used 20k/680. It may be a good experiment, as you mentioned, to try different gain settings on the amps.
I'd be interested in seeing the schemo's.
Scott
Just remember signal voltage x gain should be well below the PSU rails to avoid clipping, also these chips DO have a hard time driveing 4ohm loads...and with 4 ohm loads, i hope you drive these off a relatively low voltage...
On the other side of the spectrum, do not go below, or even too near to the minimum gain figure expressed in the datasheet...
On the other side of the spectrum, do not go below, or even too near to the minimum gain figure expressed in the datasheet...
These are the schematics from my gainclones. Please tell me if you found that they take too much to load since they are aprox 700Kb.
I think i won't mess up with class A until i get the LME49810. I plan to use it with IRPF240/IRPF9240 power mosfets. Hope that generous bias will lead to good results.
An externally hosted image should be here but it was not working when we last tested it.
I think i won't mess up with class A until i get the LME49810. I plan to use it with IRPF240/IRPF9240 power mosfets. Hope that generous bias will lead to good results.
Hi ionomolo
Something seems wrong with your second circuit: the 47k should be on the other side of the 1uF input cap.
A way to improve the sound of the first circuit is to supply the 4562 with proper power. The 78(9)xx are really quite unsuitable for high quality applications (IMO). Nevermind that lots of cheapskate manufacturers use them; it's not because they're good. I have experimented with upgrading the LM4562 regulation from 317/337 to Jung superregs and the difference is dramatic.
A regulator + choke for the fan? You really must like that little fellow.
Something seems wrong with your second circuit: the 47k should be on the other side of the 1uF input cap.
A way to improve the sound of the first circuit is to supply the 4562 with proper power. The 78(9)xx are really quite unsuitable for high quality applications (IMO). Nevermind that lots of cheapskate manufacturers use them; it's not because they're good. I have experimented with upgrading the LM4562 regulation from 317/337 to Jung superregs and the difference is dramatic.
A regulator + choke for the fan? You really must like that little fellow.
Thanks for your advice analog_sa, i havn't thought that the regulator would make a difference in the sound. The capacitor is in the prooper place on the board, I did a mistake while drawing the schematic and will correct in a day or two.
The choke+reg+big caps are to prevent noise from the fan being feed to the amplifier, since i'm planning to use a tube preamp based on the "pete millet hybrid amp" from the headphone community, which does have a much lower psrr than most audio opamps. I havn't found that with this setup the fan is inaudible even with the worst psrr possible.
Today I'll order some LME49810 to give them a try since they look very promising.
The choke+reg+big caps are to prevent noise from the fan being feed to the amplifier, since i'm planning to use a tube preamp based on the "pete millet hybrid amp" from the headphone community, which does have a much lower psrr than most audio opamps. I havn't found that with this setup the fan is inaudible even with the worst psrr possible.
Today I'll order some LME49810 to give them a try since they look very promising.
Now I've received the LME49810 i'm planning to do a mosfet class AB amplifier. Has somebody listened to deeply biased AB amplifiers (1 A or more idle current? 10W in class A). It seems that it must have the advantages both of class A (very good reproduction of quiet passages) and class B (loudness when necessary).
How would that sound?
How would that sound?
gainclone sound vs speaker efficiency
Someone pointed out that gainclones require high efficiency speakers to behave their best, but looking at the graphs from the datasheet it seems that when low listening levels are required they would benefit from low efficiency ones since disortion before clipping point descends monotonically.
Since i consider buying new speakers soon i would ask if someone has lisened to a chip amp with both high and low efficiency speakers and can tell how they sound.
Someone pointed out that gainclones require high efficiency speakers to behave their best, but looking at the graphs from the datasheet it seems that when low listening levels are required they would benefit from low efficiency ones since disortion before clipping point descends monotonically.
Since i consider buying new speakers soon i would ask if someone has lisened to a chip amp with both high and low efficiency speakers and can tell how they sound.
ionomolo,
The GamuT D200 I guess would qualify as a very good sounding amp and it is a A/B amp biased at 10W or so class A.
I hink I read about a guy that had high sensitivity spekaers and indeed the chipamp was a poor match due to crossoverdistortion which tend to be a low level problem that is highlighted with high such a speaker.
OTOH I have listened to a LM3886 amp in a blind test with my Sennheiser HD600 and could not detect if the signal had passed thru the chip amp. IOW the sound was identical when the music was looped via the soundcard and looped via the soundcard+amp. The level was aprox. 1w or maybe a little less.
/Peter
The GamuT D200 I guess would qualify as a very good sounding amp and it is a A/B amp biased at 10W or so class A.
I hink I read about a guy that had high sensitivity spekaers and indeed the chipamp was a poor match due to crossoverdistortion which tend to be a low level problem that is highlighted with high such a speaker.
OTOH I have listened to a LM3886 amp in a blind test with my Sennheiser HD600 and could not detect if the signal had passed thru the chip amp. IOW the sound was identical when the music was looped via the soundcard and looped via the soundcard+amp. The level was aprox. 1w or maybe a little less.
/Peter
Re: gainclone sound vs speaker efficiency
Lowther PM6A got me started with my first gainclone. Very dynamic, lightning-fast sound, really quite impressive. Tannoy HPD 15" produced very balanced sound with good dynamics and excellent bass.
Sonus Faber Signum and ProAc Studio 150 were both quite unlistenable: pale, uninteresting sound (even with high-capacitance PS), anaemic, ill-defined bass. They were both around 88db and not particulaly challenging loads, even to low power tube amps.
Doesn't seem to be a simple issue of efficiency.
ionomolo said:
Lowther PM6A got me started with my first gainclone. Very dynamic, lightning-fast sound, really quite impressive. Tannoy HPD 15" produced very balanced sound with good dynamics and excellent bass.
Sonus Faber Signum and ProAc Studio 150 were both quite unlistenable: pale, uninteresting sound (even with high-capacitance PS), anaemic, ill-defined bass. They were both around 88db and not particulaly challenging loads, even to low power tube amps.
Doesn't seem to be a simple issue of efficiency.
Recently added a 12BH7 (low voltage [30V B+ from 50V V+] ccs loaded) pre to my gainclone and sounds great.
I'm between vienna acoustics haydn grand and tannoy sensys dc2. Somebody has listened to them?
I'm quite skeptical about single-driver speakers like the lowthers, specially regarding midrange magic.
I'm between vienna acoustics haydn grand and tannoy sensys dc2. Somebody has listened to them?
I'm quite skeptical about single-driver speakers like the lowthers, specially regarding midrange magic.
Here's a hint to help evaluate these things. Get yourself a decent pair of headphones. I'll leave the definition of decent to you, but they don't have to be terribly expensive. I use an old pair of Sony MDR-V4, and sometimes even the somewhat dead sounding old Koss Pro4AA. The advantage of headphones is a lack of crossover and the ability to hear very subtle problems. Now, don't hook these direct. Use a divider pad so the amp operates at a more realistic level, as if you were driving speakers. I use 25 ohms and 1 ohm. You can scale that to more heavily load the amp if you want, but make sure the resistors are sufficiently high wattage. 8 ohms and 0.33 ohms might be good. Scaling the resistors will let you see how the amp performs at high and low levels, and with 4 and 8 ohm loads.
I haven't built a gain-clone, but have built enough similar stuff to suggest that layout can have a large amount to do with the harmonic content of the signal you get out. Even though you're not doing discrete design, you might want to get hold of Doug Self's book on audio amp design. It will provide some inspiration for future designs, and describes how to properly set up power and grounds.
I haven't built a gain-clone, but have built enough similar stuff to suggest that layout can have a large amount to do with the harmonic content of the signal you get out. Even though you're not doing discrete design, you might want to get hold of Doug Self's book on audio amp design. It will provide some inspiration for future designs, and describes how to properly set up power and grounds.
Conrad Hoffman said:
Nothing against headphones but this setup won't even give a hint of what an amp will sound into real crossovers. Or any type of speakers for that matter.
Headphones provide such a dramatically different view of most audiophile concepts, such as soundstaging, that i would hesitate to draw any conclusions based on can's listening alone. Even of source components.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.