Forgot to mention
This mod changes the clipping behavior to "soft clipping" instead of the previous "hard clipping" and that upgrade further increases the usable headroom.
I also employed a 0.1uf polyester (not polypro) capacitor across the speaker output terminals. This reduces the ultrasonic weapon effects so that the extra-high pitch solid-state sound isn't allowed into the speakers.
Sonic's famous T-amp has these caps too. It also prevents cascade hetrodyne of digital sources.
This is not an issue if piezo tweeters are in use because they are, internally, a capacitor.
This chipamp.com kit was a mighty difficult thing to figure out.
The power supply was sublime, but equally confusing because I have a center tap transformer. See picture for the mod.
See the double-thickess wire bridge for ground. There's an identical bridge on the output end, to prevent ground loop. That bridge on the output end is the common ground output.
This mod changes the clipping behavior to "soft clipping" instead of the previous "hard clipping" and that upgrade further increases the usable headroom.
I also employed a 0.1uf polyester (not polypro) capacitor across the speaker output terminals. This reduces the ultrasonic weapon effects so that the extra-high pitch solid-state sound isn't allowed into the speakers.
Sonic's famous T-amp has these caps too. It also prevents cascade hetrodyne of digital sources.
This is not an issue if piezo tweeters are in use because they are, internally, a capacitor.
This chipamp.com kit was a mighty difficult thing to figure out.
The power supply was sublime, but equally confusing because I have a center tap transformer. See picture for the mod.
See the double-thickess wire bridge for ground. There's an identical bridge on the output end, to prevent ground loop. That bridge on the output end is the common ground output.
Attachments
AndrewT said:
This is not unlikely.
There was no such specific help.
However:
1) The frequency of the coloration did rise out of the female voice range and onto the higher treble.
2) The power output did become twice before Spike begins his purposeful clipping.
3) The bass did increase considerably, which helped to flatten the frequency response or at least fake it nicely.
4) It does now perform a sincere imitation of LM1875--instead of its previous imitation of an angry woman with a car horn.
A specific:
It seems that a beautiful LM3875 design was put to work on the LM3886 chip without regard to tonal differences, while other components, especially resistors, were swapped for equivalents, that, likewise, weren't equivalent.
I would like to understand more thoroughly about WHY which modification did what effect.
I'm very curious about the mute circuit resistor value, of which some optional arrangement is said to lobotomize the purposeful clipping of the Spike system. How to?
P.S. I am grateful to Brian for making the kits so extraordinarily durable.
I'm using vishay s102k and panasonic fc for theses amplifiers.
I also add 200pf silver mica between legs +- directly on the chip as recommanded by the manufacturer. This remove all noises and I can't find any drawback.
Anyway they have almost the same parts and again, I'm only using mW or 1-2 watts from the amps.
I did also the test on my old kef 103.2 (not very eff.) with similar results and more power used.
AndrewT : The schematic for the non regulated is probably available on chipamp.com.
I have to draw the one for the regulated amplifier. I'll try do it this weekend.
I also add 200pf silver mica between legs +- directly on the chip as recommanded by the manufacturer. This remove all noises and I can't find any drawback.
Anyway they have almost the same parts and again, I'm only using mW or 1-2 watts from the amps.
I did also the test on my old kef 103.2 (not very eff.) with similar results and more power used.
AndrewT : The schematic for the non regulated is probably available on chipamp.com.
I have to draw the one for the regulated amplifier. I'll try do it this weekend.
There's some noticable errata in the effects of ground plane. That would show up in open wire versus PCB.
Primarily, it could have a large effect (in percentage, not volts), on DC offset.
This is audiable.
DC offset increase contributes a warmer sound.
However, the chip manufacture tolerances vary, which could make the stereo soundfield unbalanced, in addition to the warmer sound.
This was my thought when I wrote up the mod on how to get the warmer sound, and zero dc, for the PCB.
Primarily, it could have a large effect (in percentage, not volts), on DC offset.
This is audiable.
DC offset increase contributes a warmer sound.
However, the chip manufacture tolerances vary, which could make the stereo soundfield unbalanced, in addition to the warmer sound.
This was my thought when I wrote up the mod on how to get the warmer sound, and zero dc, for the PCB.
Lowthers? So, why aren't you using LM1875 or LM3875T? Those are hi fi amps. LM3886 is a television amp, albeit fairly good and powerful. I'm thinking that a big wow from a well-heatsinked and well powered LM1875--would be great stuff with the Lowthers.
Since the frequency response is flat on LM1875 and LM3875T, and the Spike protection is less aggressive on those models (*875T), I think you'd enjoy them better.
I'd love to have a photograph of your "open wire" models that work so well.
Since the frequency response is flat on LM1875 and LM3875T, and the Spike protection is less aggressive on those models (*875T), I think you'd enjoy them better.
I'd love to have a photograph of your "open wire" models that work so well.
The frequency response is flat on the lm3886 also.
I did a comparison with friends on different speakers and systems and i'm not the only one who prefered the lm3886 over the lm3875. The lm1875 was similar to the lm3875 and lacked body for my taste.
I won't enter in a debate around the best lm chips since it's already been done many time elsewhere but I prefer the overall balance of the lm3886.
I did a comparison with friends on different speakers and systems and i'm not the only one who prefered the lm3886 over the lm3875. The lm1875 was similar to the lm3875 and lacked body for my taste.
I won't enter in a debate around the best lm chips since it's already been done many time elsewhere but I prefer the overall balance of the lm3886.
Here it is, never finished the front plates and it's housed in two very flat cisco 10base-T switchs :
Note that the second set of lm3886 are not used, they where for testing purpose only (bridged or parallele)
The heatsinks are pentium2-3 with the possibility of using the original fans on the bottom but I never had to.
The power supply (8 * 50 va 2 X 12 Vac configured in +- 24 Vac dual mono). The diodes are not visibles, they are soft recovery 6amp in 4 bridges configurations.
Note that the second set of lm3886 are not used, they where for testing purpose only (bridged or parallele)
The heatsinks are pentium2-3 with the possibility of using the original fans on the bottom but I never had to.
The power supply (8 * 50 va 2 X 12 Vac configured in +- 24 Vac dual mono). The diodes are not visibles, they are soft recovery 6amp in 4 bridges configurations.
Thank you!!
That's such a pretty job (electroncially).
You never had to use the fans because the holes underneath the heatsinks cool with venturi convection (quite professional). I'll bet they never get warm.
I like this a lot!!
What is the pre-drive?
The amp in the photos has many adventages. Excellent cooling and a careful choice of resistors is preventing the Spike governor from interfering. There seems to be some sort of pre-drive, which is good because that's a sure way to get some more control over the amp chip. The grounding scheme is preventing DC output all by itself. That thing is just so good.
Did you use the pre-drive with the LM3886 kit amp?
Did you use a similar "no loops" ground scheme?
Did you use the included unusable resistors with the kit amp, or did you use better (regardless of price) resistors instead?
Did you have the kit cooled as professionally as you did the homebrew?
Have you tried the higher fidelity models, LM1875 and LM3875?
Thanks again for the photos!!!
That's such a pretty job (electroncially).
You never had to use the fans because the holes underneath the heatsinks cool with venturi convection (quite professional). I'll bet they never get warm.
I like this a lot!!
What is the pre-drive?
The amp in the photos has many adventages. Excellent cooling and a careful choice of resistors is preventing the Spike governor from interfering. There seems to be some sort of pre-drive, which is good because that's a sure way to get some more control over the amp chip. The grounding scheme is preventing DC output all by itself. That thing is just so good.
Did you use the pre-drive with the LM3886 kit amp?
Did you use a similar "no loops" ground scheme?
Did you use the included unusable resistors with the kit amp, or did you use better (regardless of price) resistors instead?
Did you have the kit cooled as professionally as you did the homebrew?
Have you tried the higher fidelity models, LM1875 and LM3875?
Thanks again for the photos!!!
The system always consist of a bent audio (tx-102 mkIII transformers) preamplifier (or a joule electra la-100 mk3), a sony scd-1 source and the speakers are BigFunHorns with lowther dx4.
I did some comparison with tube amplifiers in this setup (homebrew 45, 2a3, commercial 6c33, 845 ...) and keep returning to gainclones.
The lm3886 kits use the same resistors/capacitors/gain, keep a good grounding scheme (probably why they have low dc output) and avel lindberg or plitron transformers.
I tried the lm3875 and lm1875, you can read my explications on my earlier posts.
I did some comparison with tube amplifiers in this setup (homebrew 45, 2a3, commercial 6c33, 845 ...) and keep returning to gainclones.
The lm3886 kits use the same resistors/capacitors/gain, keep a good grounding scheme (probably why they have low dc output) and avel lindberg or plitron transformers.
I tried the lm3875 and lm1875, you can read my explications on my earlier posts.
Sure!
The brighter amp is benefitting your application.
It would be good to measure every (every!) part of the audio path so you can purposefully do this benefit with even greater precision.
Measurements may not quite do it.
Here's an example:
Empirics, dating rocks and archeology by guessing the age of discovered items by how deep or what layer.
Science, developing Radio Carbon Dating and calibrating on the previous functional system.
Science, the brand new lampshade is proven 1.531 million years old.
Empirics, the function has now become an error check to compensate for the erroneus scientific measurement.
Science, another functional system of guesswork has been added in the calibrations, to compensate for the error of millions years old lampshades.
Science, the brand new lampshade is 3 years old; however, the lamp cord is 30 billion years old.
Empirics: "Well, we're going to need a new plug."
Audio measurements are prone to the above error.
Here's how to use the 2nd calibration stage and zero equipment:
Borrow some females and/or teenagers and have them bring their own CDs.
Hide the amplifiers behind something so that they don't immediately pick out the most stylish.
Let the social music gathering begin, during which you switch the amplifier several times.
These social gatherings, historically, were the basis for calibration of audio measure. Re-create the scenerio with people who are sensitive to treble and you have cancelled out stage error without ever having needed to create it.
The last step would be to look up the medical documents for population hearing sensitivity average approximates. You can shoot for a centerline between the group described and the male adult. Doing so is a "final calibration stage."
I think that you will discover one of two possibilities:
1). The LM3886 is as bright as described.
2). The speakers or room have a reverb like effect summing at middle and low notes, so that the amp selection choice is performing a compensation.
As for myself, I hear like a teenager from one ear and a normal adult male from the other ear. With just one golden ear, there is no stereo for me. One day, I might grow up--maybe a few more decades. . .
Until then, I discovered that an EQ didn't do because of its hissing, a one-side super tweet didn't do. Piezos did remarkably well because of their 8k peaks in common, but quality models are gone from the market.
What finally did it is a Thompson center channel amp, mono, two input filter caps, stereo pot, and a full-size, full bandwidth speaker with a full range driver doing most of the work--super transparent. The source, a passive pre, has a simple splitter so that the Thompson "steals" center information before reproducing it. The Thompson likes to project soundfield behind itself, while the gainclones are forwards in stereo. That's like an orchestral configuration.
That did it for me.
I could hear clear as a bell in stereo, just like when I was young. Although measures explained the situation, it was empirics that brought a working solution.
Its funny how that goes, isn't it?
P.S. My full ranger doesn't contain a whizzer cone because that can interreact with resonances of an audio signal, that are supposed to stay concealed. Its not the the whizzer gives out more information, its that it can give out additional information that wasn't intentional.
If the LM3886's frequency response is compensated before its input stage then the above issue doesn't apply.
However, if you have leveled it like I have done to bring it to spec for the older models and if you have used the older models, there might be some reaction with a whizzer or other paper driver.
I believe we're talking Lowther?
Lowering normal amplifier effects at the predrive stage, straight on the pitches of the LM3886's boosts would have the additional effect of sending a rounded signal into the approximate frequency where paper divides and resonates--much a cancellation of that resonance without any cancellation of the audio.
If that's the case, then I'm seriously impressed.
An empiric idea is the the amount of voltage used has a large effect on the tonality of the amplifier. At a lesser voltage than I would find satisfactory for output, the LM3886 could be expected to operate at a level frequency response during low power usage. In that same situation, the LM1875, normally piping hot and brightly clear, will instead go cool, dark, and hideously masked. My recommend of it was on my experience of it, at max voltage per thermals.
So, there's quite the difference, when the voltage is low.
These empirics on power may lead to a measurable clue on power supply differences.
The brighter amp is benefitting your application.
It would be good to measure every (every!) part of the audio path so you can purposefully do this benefit with even greater precision.
Measurements may not quite do it.
Here's an example:
Empirics, dating rocks and archeology by guessing the age of discovered items by how deep or what layer.
Science, developing Radio Carbon Dating and calibrating on the previous functional system.
Science, the brand new lampshade is proven 1.531 million years old.
Empirics, the function has now become an error check to compensate for the erroneus scientific measurement.
Science, another functional system of guesswork has been added in the calibrations, to compensate for the error of millions years old lampshades.
Science, the brand new lampshade is 3 years old; however, the lamp cord is 30 billion years old.
Empirics: "Well, we're going to need a new plug."
Audio measurements are prone to the above error.
Here's how to use the 2nd calibration stage and zero equipment:
Borrow some females and/or teenagers and have them bring their own CDs.
Hide the amplifiers behind something so that they don't immediately pick out the most stylish.
Let the social music gathering begin, during which you switch the amplifier several times.
These social gatherings, historically, were the basis for calibration of audio measure. Re-create the scenerio with people who are sensitive to treble and you have cancelled out stage error without ever having needed to create it.
The last step would be to look up the medical documents for population hearing sensitivity average approximates. You can shoot for a centerline between the group described and the male adult. Doing so is a "final calibration stage."
I think that you will discover one of two possibilities:
1). The LM3886 is as bright as described.
2). The speakers or room have a reverb like effect summing at middle and low notes, so that the amp selection choice is performing a compensation.
As for myself, I hear like a teenager from one ear and a normal adult male from the other ear. With just one golden ear, there is no stereo for me. One day, I might grow up--maybe a few more decades. . .
Until then, I discovered that an EQ didn't do because of its hissing, a one-side super tweet didn't do. Piezos did remarkably well because of their 8k peaks in common, but quality models are gone from the market.
What finally did it is a Thompson center channel amp, mono, two input filter caps, stereo pot, and a full-size, full bandwidth speaker with a full range driver doing most of the work--super transparent. The source, a passive pre, has a simple splitter so that the Thompson "steals" center information before reproducing it. The Thompson likes to project soundfield behind itself, while the gainclones are forwards in stereo. That's like an orchestral configuration.
That did it for me.
I could hear clear as a bell in stereo, just like when I was young. Although measures explained the situation, it was empirics that brought a working solution.
Its funny how that goes, isn't it?
P.S. My full ranger doesn't contain a whizzer cone because that can interreact with resonances of an audio signal, that are supposed to stay concealed. Its not the the whizzer gives out more information, its that it can give out additional information that wasn't intentional.
If the LM3886's frequency response is compensated before its input stage then the above issue doesn't apply.
However, if you have leveled it like I have done to bring it to spec for the older models and if you have used the older models, there might be some reaction with a whizzer or other paper driver.
I believe we're talking Lowther?
Lowering normal amplifier effects at the predrive stage, straight on the pitches of the LM3886's boosts would have the additional effect of sending a rounded signal into the approximate frequency where paper divides and resonates--much a cancellation of that resonance without any cancellation of the audio.
If that's the case, then I'm seriously impressed.
An empiric idea is the the amount of voltage used has a large effect on the tonality of the amplifier. At a lesser voltage than I would find satisfactory for output, the LM3886 could be expected to operate at a level frequency response during low power usage. In that same situation, the LM1875, normally piping hot and brightly clear, will instead go cool, dark, and hideously masked. My recommend of it was on my experience of it, at max voltage per thermals.
So, there's quite the difference, when the voltage is low.
These empirics on power may lead to a measurable clue on power supply differences.
Power supply goodness!!
I'm just loving that power supply!! Five times 2200!! What a fantastic idea! So, this is a divisional value that is blocking a certain "chord" from going into the audio? And, there's small caps of certain values. Would these all be shunting 5ths and 2nds?
So, I'm very interested in your power supply that shunts/snubs specific divisionals instead of all of them. I'm just so curious on how/why it works. I can't find out for myself because I don't have any drivers as sensitive as the Lowthers.
I'm just loving that power supply!! Five times 2200!! What a fantastic idea! So, this is a divisional value that is blocking a certain "chord" from going into the audio? And, there's small caps of certain values. Would these all be shunting 5ths and 2nds?
So, I'm very interested in your power supply that shunts/snubs specific divisionals instead of all of them. I'm just so curious on how/why it works. I can't find out for myself because I don't have any drivers as sensitive as the Lowthers.
Like I said, this amp as moved to different installations always with great results so I wouldn't tend for any synergistic reactions between it and the lowther whizzer ... I even tried it on a pair of b&w dm70 and mid horn of altec vot also with great results
The opposite is also true, I often have amplifiers at home for checkup and try them on my speakers and would easily point to any coloration/distortion from the gainclone by comparison.
Other amplifiers as played well here like atma-sphere (but noisy), david berning (also noisy), my very good old mcintosh mc-2100, some tube amplifiers ...
Next week I may be able to assemble and test regulation pcbs I've build to be used on a lm3886 chipamp kit (between the two pcb of the kit) using always the sames components (lt1083, panasonic fc, s102k, mur860 ...). This should give me a good idea of the regulation effects by comparison to another similar kit unregulated.
I will post my results here.
The opposite is also true, I often have amplifiers at home for checkup and try them on my speakers and would easily point to any coloration/distortion from the gainclone by comparison.
Other amplifiers as played well here like atma-sphere (but noisy), david berning (also noisy), my very good old mcintosh mc-2100, some tube amplifiers ...
Next week I may be able to assemble and test regulation pcbs I've build to be used on a lm3886 chipamp kit (between the two pcb of the kit) using always the sames components (lt1083, panasonic fc, s102k, mur860 ...). This should give me a good idea of the regulation effects by comparison to another similar kit unregulated.
I will post my results here.
danielwritesbac :
I'd like to know what, technically speaking, would make the lm3875 lm1875 a higher fidelity chip over the lm3886 ?
The gbp is not very relevant; we do not need high gain and have to limit the bandwidth in order to protect the chip from rf.
The slew rate spec is better on the lm3886. Ok, technically audio application at 20khz doesn't ask for that big of a slew rate but for sacd or LP application, why not considering the higher slew rate better.
The spike protection, are you referring to the immunity to rf and electrical interference or about clipping behaviour?
The lm3886 deal more easily with low impedance load and no speaker is flat in this area except maybe a resistive load like a magnepan. I've been able to push my old kef reference 103.2 harder with the lm3386 than the lm3875 before distortion ocure.
Once limited in bandwidth I have absolutely no electric or rf interferences from the lm3886 and it were the same story with the lm3875 with proper implementation.
I heard about the "sound destructive" muting circuit, which first get me away from this chip until I tried it.
franksmith said:
I fixed that. Actually, Andrew fixed that. I just used equal parts soldering iron and desoldering iron until we arrived at values that pleased us both.
See the last chart on the lm1875 thoughts thread.
Um, I've left a little clue in the center of that chart. That and input filter choices is going to do any amount of "body" (audio presentation both forwards and rearwards of the cones?) that you'd like from the little chip. I think you'll be surprised, or considering the power of the Lowthers, maybe even terrified.
For Lowthers and many speakers, the output zobel is a problem. I haven't figured out that last part yet, and I'd sure like your views on it with such sensitive equipment as Lowthers.
Thank you!!!
I'm waking up an old thread but I forgot to give feedback when I've finally build the second regulated gainclone (mars 2008).
It's using a complete chipamp kit but I've implemented the lt1083 regulation on two boards and wire it between the chipamp psu board and the lm3886 boards.
It's housed in a old dynaco st-120 amplifier and using a pair of avel lindberg 300va 25+25 transformer. The original heatsink are used.
The parts selected around the lm3886 are the sames used for my first homebrew ciscoclone. The gain is the same also and I lowered the reserve capacitors close to the lm3886 to 47uf.
Each PSU board (1 per channel) use 30 000uf panasonic electrolytic but are not of fc series. The diodes are the same also.
I called it the Dynaclone.
The sound from the dynaclone is very similar to the ciscoclone. I think I prefer the sound of the point to point one but I doubt I would be able to find it in a blind test. It doesn't heat and the small original dynaco heatsink are enough.
I also use a 5v drop at the regulator and the operating voltage is around 1v higher than the ciscoclone.
Dc offset at the output is 4-5 mv which if very similar.
My conclusion of all this is that I prefer the sound of regulated gainclones over unregulated ones. It seem to open the sound, bring more definition, better control of the bass and lower the noise floor (but not by much if I compare with or without in this amplifier).
Keep in mind that 4X lt1083 (adjustable) raise the total cost of construction and may sound ridiculous considering the price paid for a pair of lm3886.
Here are pictures of the regulation boards and the amp finished.
It's using a complete chipamp kit but I've implemented the lt1083 regulation on two boards and wire it between the chipamp psu board and the lm3886 boards.
It's housed in a old dynaco st-120 amplifier and using a pair of avel lindberg 300va 25+25 transformer. The original heatsink are used.
The parts selected around the lm3886 are the sames used for my first homebrew ciscoclone. The gain is the same also and I lowered the reserve capacitors close to the lm3886 to 47uf.
Each PSU board (1 per channel) use 30 000uf panasonic electrolytic but are not of fc series. The diodes are the same also.
I called it the Dynaclone.
The sound from the dynaclone is very similar to the ciscoclone. I think I prefer the sound of the point to point one but I doubt I would be able to find it in a blind test. It doesn't heat and the small original dynaco heatsink are enough.
I also use a 5v drop at the regulator and the operating voltage is around 1v higher than the ciscoclone.
Dc offset at the output is 4-5 mv which if very similar.
My conclusion of all this is that I prefer the sound of regulated gainclones over unregulated ones. It seem to open the sound, bring more definition, better control of the bass and lower the noise floor (but not by much if I compare with or without in this amplifier).
Keep in mind that 4X lt1083 (adjustable) raise the total cost of construction and may sound ridiculous considering the price paid for a pair of lm3886.
Here are pictures of the regulation boards and the amp finished.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.