I have played a few tricks on my wife with swpping amps and speakers around.
In my living room I have come to like horns for both mid and high, in most other rooms in my house I remain a dedicated ribbon fan.
I am yet to try a Heil though.
Amps, I can easily tell chip amp (detailed but harsh) vs the new 5.1 (very mushy and lacking detail) vs the old school SS (low on detail but good power) vs what I consider my favorite, a Marantz which really is very detailed, but never harsh.
I prolly dont get clipping either though. I may have never heard it cos I dont run em loud.
Cant wait for the tube+3886 kit to get here.
Cool.
Srinath.
In my living room I have come to like horns for both mid and high, in most other rooms in my house I remain a dedicated ribbon fan.
I am yet to try a Heil though.
Amps, I can easily tell chip amp (detailed but harsh) vs the new 5.1 (very mushy and lacking detail) vs the old school SS (low on detail but good power) vs what I consider my favorite, a Marantz which really is very detailed, but never harsh.
I prolly dont get clipping either though. I may have never heard it cos I dont run em loud.
Cant wait for the tube+3886 kit to get here.
Cool.
Srinath.
The LME49810, LME49811, LME49830 support up to +/-100 V rails. That's enough for about 625 W RMS into 8 ohms or 2.5 kW if you run two channels bridged. is that really not enough for you?!
Sure, if you're going for max power, you'll have to go discrete. But if you are shooting for something of reasonable power that offers good performance (i.e. low THD, low noise) the integrated amps is probably where you should throw your money.
I think you are getting things confused here. Typical semiconductor processes offer various devices. In many processes, the high voltage devices are a separate module that the circuit designer (in collaboration with the rest of the design team) can request to have enabled. Once enabled, the wafers go through a manufacturing flow where the high-voltage devices are built. Mixing low-voltage and high-voltage devices is common and if you can build transistors, you can certainly build logic.
Take a look at attached equivalent circuit schematic for the LM3886. Notice how the "logic" function of the MUTE pin is implemented with an analog mux and a couple of diff pairs? Nobody says you have to use a process optimized for high-speed CMOS to create a simple logic function.
The high-speed CMOS processes generally suck for anything analog due to the small device size, and thus, higher 1/f noise. Besides, even analog CMOS circuits never use the minimum device size. Why bother with a $1M mask set for a 45 nm CMOS process if you're going to use 2 um channel length devices to get the noise down...
~Tom
Attachments
Anybody who puts "easy" and "kW" in the same sentence should set out to build one of those systems to see exactly how "easy" it is...
~Tom
Not sure if this is being exhibited as a "chip amp" but it is not. It's what's called a hybrid ie a tiny printed thick/thin film substrate with discrete chips, caps, and resistors using wire bonds from semi-chips to substrate and from the substrate to outside pins. LOL
My proprietary high gain 3886 chip amp competes (according to many others who have listened) well at the $3000 level for a stereo amp. Has 225dB gain at DC and 125dB gain at 10kHz. Vanishingly small distortion. Also features a second order DC servo loop and absolute stability with no output coil or RC network to make it safe or stable. Will drive any load with the correct power supply. Have used them below one ohm on low voltage power supply and worked flawlessly.
AV=+20 or 26dB gain with feedback. Yes 100dB of feedback at 10kHz. "If your going to use feedback more is better so most is best." quote of John Gordon Iverson. In that vain this amp has a ton of feedback and gain. Works far better than it has any right to and is stable into any kind of load. I could barely get it to ring part of one cycle driving .1µF or .5µF load with 20kHz square wave and no resistance. Full power bandwidth is 110kHz.
Hi All
This subject always interested me. How can you 'hear' the amp clipping? What to listen to and how to determine that your amp clips?
Thanks
Consensus ,
Chip amps good , detailed but harsh , F5 or discrete type amp better ? chip plus discrete best ?
From different posters on many different forums, this seems to be the opinions of those that voice such , chip amps seem to lack bass and top end detail , good detail in the middle , less dynamics when compared to a good discrete amp. Of course there are those that enjoy the chip-amp sound , again audio being audio everything is system (speaker dependent) limited bandwidth speakers tend to be there best match and true enuff associates and friends who rave over their chip amps use small limited bandwidth speakers with a few exceptions of course.
Later this week I will have the pleasure to hear Mono bloc's Bel Canto's playing maggies vs my threshold S500 on the same Maggies, so i will have a reference to see how well this type of amplifier performs. The owner believes it to be better than his Krell KSA200 and the threshold is much better sounding than the KSA200 IMO so it will be a good shootout.
I'm not married to any type of topology , I'm interested in only good sound,
If chip is the way, I'm in , if not , i will say so ...
regards,
In my limited experience - extremely speaker dependent, especially in bass and dynamics. Don't quite see how adding hundreds of db extra open loop gain can really correct for this - there just seems to be insufficient current from single chips.
Offtopic: last week upgraded my old S300 with Rifa main caps, polyprops in place of polyesters in output zobel and feedback decoupling and Vampire binding posts/jacks. The improvement is beyond all my expectations as i've never really liked the S300 before. Can a chip amp sound as powerful @1W? I seriously doubt.
Interesting analog , i too have been tempted to upgrade the s500 PSU caps and diodes and after your comments , I'm leaning that way now more than ever.
When we do the comparison it will be between stock S500 and chip amp
"Bel canto" .
I went from all 1950s and 60s Fisher and Scott amps to a Bryston 4B to home built LM3886 amps.
The only thing I can say to people who claim discrete is better is, good luck matching your transistors even remotely as close as those in a chipamp. And to people who claim they can be matched close enough, I guess you can argue any point so use what makes you happy. My fisher and scott amps sounded great, the Bryston 4B sounded great and gave me bragging rights. The LM3886 amps give me great sound and simplicity. And since bragging rights are useless when it comes to getting the job done, I'm sticking to my lm3886's
Build and use what makes YOU happy. Thats what its about.
The only thing I can say to people who claim discrete is better is, good luck matching your transistors even remotely as close as those in a chipamp. And to people who claim they can be matched close enough, I guess you can argue any point so use what makes you happy. My fisher and scott amps sounded great, the Bryston 4B sounded great and gave me bragging rights. The LM3886 amps give me great sound and simplicity. And since bragging rights are useless when it comes to getting the job done, I'm sticking to my lm3886's
Build and use what makes YOU happy. Thats what its about.
Last edited:
Any suggestions about the diodes? I have no experience with 35A+ diodes. Another thing to do is replace the aluminium heatsink which doubles as a ground bus with copper. Not sure how audible this will be but i'll do it out of principle
And lastly - a separate transformer/rectifier/(regulator?) for the input/driver board.
Tom, the LME49* are drivers not chipamps - they replace input stage and VAS. I'm sure you know it.
So, I wouldnt exactly compare to the 3886 &C.
Originally Posted by analog_sa
"In my limited experience - extremely speaker dependent, especially in bass and dynamics. Don't quite see how adding hundreds of db extra open loop gain can really correct for this - there just seems to be insufficient current from single chips."
Feedback makes an amplifier more linear. More feedback is more linear. Also having the amp stable with that much feedback makes it drive bad loads a lot better. I use no compensation at all on the output like an inductor or RC snubber to calm the amp from oscillation. It is like big rock stable. It also means the dominant pole gets moved out to a very high frequency so out of band signals do not cause foldback distortion of the worst kind. Comparing the very well built straight 3886 to the "high gain composite amp" version on the bench and in listening test was not even fair. One sounds like the typical halfway decent amp the the other sounds like music and forget the amp part.
Also the extra gain means higher damping at all frequencies which translates directly into more control over the transducers. This reduces the effects you mention of the speakers making the difference. This damping also makes the amp run hotter because it is absorbing more back EMF from the voice coil.
As far as current I am not certain what you mean. The 3886 with a good supply will put out 11 amps which translates into an easy 7.5 amps RMS so that is 450 watts into 8 ohms and 225 into 4 if it could swing the voltage. That is plenty of current. So much current that I run 2 3886 balanced output to increase the voltage swing making 160 watt amps out of a pair of them (±32 volt supplies) which takes less than 5 amps RMS into 8 ohms.
Further I run the very same amp on ±15 volts and drive loads down to half an ohm. Most amps puke at that kind of load but this is the easy way to do it. It will direct drive a ribbon just fine if the power supply voltage is set appropriately.
This explains everything you could not see so I hope now you can "see."
"In my limited experience - extremely speaker dependent, especially in bass and dynamics. Don't quite see how adding hundreds of db extra open loop gain can really correct for this - there just seems to be insufficient current from single chips."
Feedback makes an amplifier more linear. More feedback is more linear. Also having the amp stable with that much feedback makes it drive bad loads a lot better. I use no compensation at all on the output like an inductor or RC snubber to calm the amp from oscillation. It is like big rock stable. It also means the dominant pole gets moved out to a very high frequency so out of band signals do not cause foldback distortion of the worst kind. Comparing the very well built straight 3886 to the "high gain composite amp" version on the bench and in listening test was not even fair. One sounds like the typical halfway decent amp the the other sounds like music and forget the amp part.
Also the extra gain means higher damping at all frequencies which translates directly into more control over the transducers. This reduces the effects you mention of the speakers making the difference. This damping also makes the amp run hotter because it is absorbing more back EMF from the voice coil.
As far as current I am not certain what you mean. The 3886 with a good supply will put out 11 amps which translates into an easy 7.5 amps RMS so that is 450 watts into 8 ohms and 225 into 4 if it could swing the voltage. That is plenty of current. So much current that I run 2 3886 balanced output to increase the voltage swing making 160 watt amps out of a pair of them (±32 volt supplies) which takes less than 5 amps RMS into 8 ohms.
Further I run the very same amp on ±15 volts and drive loads down to half an ohm. Most amps puke at that kind of load but this is the easy way to do it. It will direct drive a ribbon just fine if the power supply voltage is set appropriately.
This explains everything you could not see so I hope now you can "see."
What ...
OK so there were tube amps, which gave way to solid state, which gave way to IC which gave way to chip. Now you're saying there is a different one ... man have mercy on me, I am confused enough.
So FET is also a chip right. MOSFET, JFET etc etc.
Scary.
Cool.
Srinath.
OK so there were tube amps, which gave way to solid state, which gave way to IC which gave way to chip. Now you're saying there is a different one ... man have mercy on me, I am confused enough.
So FET is also a chip right. MOSFET, JFET etc etc.
Scary.
Cool.
Srinath.
Depends on exactly how you define "chip amp". I'm sure the discrete crowd would lynch me if I called it a discrete design. But I do agree that the LME498xx family is somewhere in the gray area between discrete and chip amp.
In terms of complexity of design, the LME498xx series is only marginally more complex than the LM3886 -- especially when compared to the complexity of a discrete design. I've done all of them...
~Tom
Good luck. Matching discretes on a curve tracer (been there, done that) can yield matching on one parameter to within a few percent if you are patient and have a large sample size to choose from. Resistor matching is the same -- though, one could get within 2 % by buying +/-1 % tolerance resistors. Note, however, that if there is any thermal gradient across the PCB, the tempco of the resistors will throw that matching right out the window unless you have some very clever PCB design (and most designers don't).
Precision analog semiconductors (not to be confused with the run of the mill, tiny feature size CMOS logic chips) fabricated using processes optimized for analog products (National's VIP50 process for example) will yield far better matching between semiconductors than you can ever achieve by manually matching a bag of discrete devices. The resistor matching is just unreal. At the same time, the tempco of precision resistors will be controlled in the process. Typically, there will be at least one resistor type with "zero" tempco. NiCr, TiW, etc.... Great materials. Furthermore, the circuits can be designed to take advantage of the matching, i.e. circuit performance depends on the matching between components not the absolute value of the components (which is much more difficult to control in ICs). By careful chip layout, the matching between parts can be improved. For example a differential pair may be made up of, say, 2x8 devices placed on a common centroid for better matching. The routing to the transistors can be designed to balance the parasitics of each side of the diff pair. Hastings, "Art of Analog Layout" is a good place to start if you are more curious about this topic.
In a discrete circuit, you are usually better off by getting precision components and taking advantage of the precision in their absolute value rather than designing your circuit to be dependent on the matching between parts. Whoever came up with the idea of emitter degeneration resistors deserves a big hug!
Anyway... Back to work. Oh, and good luck with your knock-our-socks-off discrete design...
~Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.