@ danielwritesbac, so are you saying chip amp with Darlington pair at output will work as class a amplifier. If yes then can you post the any easy schematic with CRO output. 🙂
No,
I'm saying that a discrete parts transistor small signal section, driving chip based output devices, could make a Class AA amplifier.
For schematic, search for Circlophone Darlington with the diyaudio.com search feature located at the top of this page.
P.S.
The only IC chip involved in that IS darlington outputs (not an op-amp). Darlingtons ARE Integrated Circuit chips and they do give you a less complicated layout and an easier build. Really considerably easier. The drivers are just stuffed in there with the outputs--much less labor!!
Last edited:
A pair of little chip amplifiers with a sweet tone.
What most people like about Class A is the big bass slam from the heavy duty power circuits AND the highly palatable tone.
There's a lot of Class AB chips we could mod to sound like that, but there are two that can do it easily and naturally so that it works in most cases.
TDA2030A (not tda2030) and LM675.
This pair are made for lower gain and easily compensated with a suitable amount of gain to reach their output potential. Thus you're very likely to get the stability needed for a palatable tone with those two.
Check out their datasheets: Start! Fetch!
Those two chips can use a 12+12vac or 13+13vac transformer (and a dc power supply board). Inexpensive 25v rated capacitors can be used. The output may reach up to 16 watts per channel. For a stereo build, that 2 ampere transformer is commonplace, widely available and inexpensive.
An easy power supply prospect is a single 1004 prefabricated bridge rectifier and 3 of 3300u caps per each rail (6 caps total for the power board).
As for the power circuit on the amplifier board, actually 220u 25v per each rail OR a Parallel pair of 220u 25v per each rail (similar clarity but tamer tone plus a little more peak power), for the power decoupling tend to work well with either of those chips.
This is all really straightforward and simple, but what has really helped is that those two chips have more aggressive internal compensation so that stability/tone is more easily achieved.
The approximately 15 watts per channel can be quite loud or even too loud for a traditional living room, a kitchen, an office, or a bedroom; however, this is not enough power for a Great Room style house nor an airplane hangar. 😀
Well, if you want it sweet, and easy, TDA2030A and LM675 need a closer look.
What most people like about Class A is the big bass slam from the heavy duty power circuits AND the highly palatable tone.
There's a lot of Class AB chips we could mod to sound like that, but there are two that can do it easily and naturally so that it works in most cases.
TDA2030A (not tda2030) and LM675.
This pair are made for lower gain and easily compensated with a suitable amount of gain to reach their output potential. Thus you're very likely to get the stability needed for a palatable tone with those two.
Check out their datasheets: Start! Fetch!
Those two chips can use a 12+12vac or 13+13vac transformer (and a dc power supply board). Inexpensive 25v rated capacitors can be used. The output may reach up to 16 watts per channel. For a stereo build, that 2 ampere transformer is commonplace, widely available and inexpensive.
An easy power supply prospect is a single 1004 prefabricated bridge rectifier and 3 of 3300u caps per each rail (6 caps total for the power board).
As for the power circuit on the amplifier board, actually 220u 25v per each rail OR a Parallel pair of 220u 25v per each rail (similar clarity but tamer tone plus a little more peak power), for the power decoupling tend to work well with either of those chips.
This is all really straightforward and simple, but what has really helped is that those two chips have more aggressive internal compensation so that stability/tone is more easily achieved.
The approximately 15 watts per channel can be quite loud or even too loud for a traditional living room, a kitchen, an office, or a bedroom; however, this is not enough power for a Great Room style house nor an airplane hangar. 😀
Well, if you want it sweet, and easy, TDA2030A and LM675 need a closer look.
I did that a couple of years ago with the MiniRef - 2 versions, one with an LM1875 and another with an LM3886. Both nest the chipamp (configured as a Howland current pump) within an outer opamp-based voltage-series feedback loop, a la Walt Jung's Composite Buffer/Amp. The difference is that I don't use the chipamp as a voltage buffer, but rather as a transconductance amp (a la MyRef). Almost any opamp can be used (I've tried LME49860, AD8022, OPA2134, LT1361, NE5532 and various others) for the outer feedback loop. The best so far has been the LF06, which is my FET990-like discrete opamp.
Audibly, I believe my MiniRef LM3886 version with the LF06 discrete opamp comfortably beats the MyRef Rev C, while the LM1875 version probably also exceeds it comfortably, but at lower power output of about 20-25W (which may be an issue if you have low-sensitivity speakers). The best audible version of the MiniRef LM1875 is presently offered by PreSapian on this forum, who has done a bunch of mods to allow the use of an AD797 as the outer-loop opamp (which is otherwise notoriously difficult to stabilize).
For more details, search on this forum for "MiniRef".
Edit: For the MiniRef 1875, any of LM675, TDA2030/2030A/2040/2050/2050A are drop-in replacements for the chipamp, as long as the rail voltages are adjusted accordingly. A test prototype with the TDA2050A has been on my to-do list for while, but meanwhile ST discontinued the Pentawatt chipamps. I can still get several of the ST chipamps from my local ST distributor for very reasonable prices, which will last until his stock runs out.
Last edited:
Thank you. This is really great detailed information. However, it seems that you missed what I said. I'm sorry I didn't say it clear enough. That bit you quoted had "TDA7293 in Slave Mode" which, according to the datasheet (see modular) has killed the voltage amplifier. That is a high current buffer. There's no voltage amp at all. In that case, we could shove it with a Class A small signal Amplifier and attach global negative feedback so the buffer wouldn't go bonkers. The only source of amplification in that case, is Class A. Therefore the sonic signature should be similar to a Class A amplifier.
However, half of the main source of concern for any amplifier is the output devices, so there is a fair chance that I may be mistaken.
Edit:
There is no chance for that plan to beat your LM1875 based MyRef except for increased output power that would be a truly insignificant amount of decibels, 3.5db, if measured at the speaker. That statement is true with any 8 ohm speaker. However, at heavier loads like a 4 ohm speaker. . . this plan is on the loose. 😀
However, half of the main source of concern for any amplifier is the output devices, so there is a fair chance that I may be mistaken.
Edit:
There is no chance for that plan to beat your LM1875 based MyRef except for increased output power that would be a truly insignificant amount of decibels, 3.5db, if measured at the speaker. That statement is true with any 8 ohm speaker. However, at heavier loads like a 4 ohm speaker. . . this plan is on the loose. 😀
Last edited:
I admit to being unfamiliar with the TDA729x-family in slave mode. The feasible options for the output device are: Unity-gain buffer (i.e. current amplification only, a la Walt Jung), Transconductance or VI converter (a la Mauro Penasa), or both current and voltage gain (a la Technics New Class A and several others).
If you have a naturally-stable unity-gain buffer like the BUF6xx, LME49600, etc., then the nested unity-gain buffer is the easiest to stabilize and probably the way to go. Mauro's genius in the MyRef was to take a chipamp (LM3886) which is not unity-gain stable, and still manage to comfortably stabilize the global loop and make it sound good. I went a bit further in the MiniRef by combining Mauro's and Walt's topologies and verifying that it could be still be stabilized with a wide variety of outer-loop opamps, while sounding very good (the sonics are predominantly that of the outer-loop opamp, which could of course be biased to Class-A).
Whether the TDA7293 can be run stably in this application is a matter for testing/prototyping (in the absence of an official SPICE model for the TDA7293), but I have no reason to believe that it can't be done. The MiniRef uses a sufficiently generic compensation schema (a modified DeBoo integrator) which should work with a wide variety of output chipamp devices - all that's needed is generic opamp functionality for the output device: high-impedance inverting and non-inverting inputs, and a low-impedance high current output.
The choice of the output device is largely a matter of personal preference - there are many who swear by the sonics of the LM1875, while others prefer the higher power and dynamics of the LM3886. The MOSFET outputs of the TDA7293 could well be another dimension in this matrix.
That is good. But there is something more fun at the cost of more labor. It is difficult to do an LM1875 Parallel amplifier since it does not agree copacetically to this use. However, that output is astounding. It sounds horribly expensive but it doesn't cost a lot. When you hear that, you can't un-hear it. It does answer the question of what an LM1875 might sound like if it had appropriate output linearity and 44 watts capacity.
Mosfet are fun. The imaging is crap but the tone fantastic. The practicality for mosfet is really very loud playback so you can crank it up really loud, pretend you're at a concert and it not hurt your ears instantly. But for other use it is not an ultimate. When indoors, the usefulness is pretty much limited to short bursts of enthusiasm. It does that well. But it is not especially good for multi-function. Unfortunately, the placid tone mosfet party blaster probably won't be doing high end imaging nor analytic. So, if that is your main system, then you will probably require a second system for the other types of listening.
However, a parallel BJT amplifier comes a lot closer to performing the multiple functions with only one system.
Last edited:
If the goal is to remove class B distortion. You need to run the output in Class A. I have been doing some simulations and its not to difficult to do. LM3886 in single ended Class A works in my simulation.
Yeah, probably that works. It will need enough voltage for 32W and then can't do more than 19W before setting off the Spike protection involving a hard clipper (which is quite loud). However, if used at less than 19W (max), it may work. Reference that in Class AB mode, spike goes off with epic distortion at 45 watts. However, when using only one of the outputs, we may be able to do half of that. *maybe*
Since this is for indoor use, where 19W is awfully loud (its surely enough), this plan just might work.
Edit: Question: Did you use a 2 transistor bias?
Edit2: This post has an error that whatever goes to the heatsink will eat up the output power budget and is not available to the speaker, therefore the above output figures are overly optimistic if the op amp is biased heavily.
Last edited:
Class AB without crossover distortion.
I have tried this and can report that there are many inconveniences, including re-tuning the whole thing, especially the power circuit, and there's the inconvenience of hand-matching the resistors via digital ohmmeter, which is an epic workout while also re-tuning the whole thing simultaneously. However, John did tell us the truth and indeed it does work delightfully well. His book doesn't mention the amount of labor for it, which is at least obnoxious; however, the phrases where he ventured to tell us the audio advantages of it are accurate.
Real engineers never brag on their own stuff with observations, but the fact that a big chunk of the book was dedicated to paralleling is a battleship sized clue; and, when engineers go on and on and on about one thing like that, we had better pay attention.
I did try it with LM1875.
And it certainly isn't easy.
Yes, a heck of a lot of nasty goes into the ballast resistors and Not into your ears. Crossover distortion is the least of the problems wiped out by that approach, but perhaps it was the easiest to measure. After a dreadful workout, then you're in for an audio treat.
Even with chips, the tone is perfectly level--no hard spots!!!
And, you know, the labor might not be too bad but merely relocated, because after you get the main stuff to work, the remainder of the fine tuning is greatly simplified and nearly effortless. . . because it has worked.
Perhaps I am simply reporting that Class A is not the only way to wipe out crossover distortion from Class AB amplifiers.
John Linsley Hood in "The Art of Linear Electronics" says make a parallel amplifier and then the crossing distortion goes into the ballast resistors, not into the speaker (therefore, not into your ears).
I have tried this and can report that there are many inconveniences, including re-tuning the whole thing, especially the power circuit, and there's the inconvenience of hand-matching the resistors via digital ohmmeter, which is an epic workout while also re-tuning the whole thing simultaneously. However, John did tell us the truth and indeed it does work delightfully well. His book doesn't mention the amount of labor for it, which is at least obnoxious; however, the phrases where he ventured to tell us the audio advantages of it are accurate.
Real engineers never brag on their own stuff with observations, but the fact that a big chunk of the book was dedicated to paralleling is a battleship sized clue; and, when engineers go on and on and on about one thing like that, we had better pay attention.
I did try it with LM1875.
And it certainly isn't easy.
Yes, a heck of a lot of nasty goes into the ballast resistors and Not into your ears. Crossover distortion is the least of the problems wiped out by that approach, but perhaps it was the easiest to measure. After a dreadful workout, then you're in for an audio treat.
Even with chips, the tone is perfectly level--no hard spots!!!
And, you know, the labor might not be too bad but merely relocated, because after you get the main stuff to work, the remainder of the fine tuning is greatly simplified and nearly effortless. . . because it has worked.
Perhaps I am simply reporting that Class A is not the only way to wipe out crossover distortion from Class AB amplifiers.
You can find it here. Still testing the sims but it is relatively efficient for Class A. Half the voltage and half the current so 1/4 of normal maximum power.
http://www.diyaudio.com/forums/chip-amps/253834-single-ended-class-chip-amp.html#post3874201
http://www.diyaudio.com/forums/chip-amps/253834-single-ended-class-chip-amp.html#post3874201
It is not so useful to mention this, but since we mentioned everything else, here goes. . .
Class A is traditionally highly reliant on a clean power source. Much of the talk on Class A sound is output transistors, output tubes, output transformers and output bias; however, there is an input piece.
That is the singleton input.
She absolutely requires clean power and that's an inconvenience.
The options for chips with singleton input are TDA2003 and TDA2009.
These are not very powerful, but they can be bridged for more power.
Even so, almost all discrete options with singleton input will outperform those chips.
What they actually have in common is a very lovely tone with low listening fatigue.
When given a deluxe power supply like a tracking pre-regulator (chips) or a discrete reg, it is very typical for the singleton input amplifier to image like a champ and have a highly practical tone.
I think that the appropriate use for those two chips is for driving a 3" Full range equipped with a bsc/notch and deployed for desktop speakers.
Notes:
With only one transistor for input, of course the singleton is a Class A input.
This sort of input comes with the necessity of much extra effort at the power circuit.
At end result, the efforts have made all of the other desirable parameters for audio more easily achieved.
An NTP input is a more convenient alternative that may have a similar performance.
Class A is traditionally highly reliant on a clean power source. Much of the talk on Class A sound is output transistors, output tubes, output transformers and output bias; however, there is an input piece.
That is the singleton input.
She absolutely requires clean power and that's an inconvenience.
The options for chips with singleton input are TDA2003 and TDA2009.
These are not very powerful, but they can be bridged for more power.
Even so, almost all discrete options with singleton input will outperform those chips.
What they actually have in common is a very lovely tone with low listening fatigue.
When given a deluxe power supply like a tracking pre-regulator (chips) or a discrete reg, it is very typical for the singleton input amplifier to image like a champ and have a highly practical tone.
I think that the appropriate use for those two chips is for driving a 3" Full range equipped with a bsc/notch and deployed for desktop speakers.
Notes:
With only one transistor for input, of course the singleton is a Class A input.
This sort of input comes with the necessity of much extra effort at the power circuit.
At end result, the efforts have made all of the other desirable parameters for audio more easily achieved.
An NTP input is a more convenient alternative that may have a similar performance.
Is that a non-inverting LM3886 on the datasheet values? If so, I don't think that the experiment will hurt the performance at all, because there's nothing worse. We could try inverting mode instead, which would be a lot better. But if we weirdly use non-inverting mode for no good reason, then we could use 2k2 vs 56k or 2k2 vs 68k for the gain divider. It will still image like a champ and the tone will be a lot better. Still frightful, but yet a lot better than the datasheet. 😀
Mark, sorry about that. Generally, the class A idea seems alright, but I worry about device linearity when highly loaded.
Member
Joined 2009
Paid Member
Class A doesn't have to be expensive. The JLH 10W is very forgiving, allowing use of different parts. You can scrounge or make your own heatsink - just needs a large enough area of metal, and you can always use a CPU heatsink + fan assembly. Run it off a couple of car or motorcycle batteries, or a SMPS from a computer etc. Build it for 5W instead of 10W if needed. There are plenty of tube amps that operate in Class A and put out 1W on a good day, for SET fans a 5W output is pretty strong.
The feedback loop will remove any linearity issues as it would with normal operation.
On op-amps you might do worse than read Bruce Putzey's article "The F-word or, why there is no such thing as too much feedback" as it has a very clear outline of the issues associated with op-amps (and the differential/integrator/output amp design)
FWIW the JLH always ends up back in my system - I suspect because it doesn't have a differential input stage.
Steve, What do you want to know? Chip amp designs that sound like class A or operating chipamps in class A or the importance of the feedback loop to the workings of a chipamp.
What I think sounds like Class A, is Parallel LM1875.
Usually that will need somewhat excessive ballast, so firstly, there's no chance of negative output impedance, and secondly, you get a miniscule touch of current drive.
Also, driving an 8 ohm speaker with Parallel LM1875 makes for very high linearity output or in other words, current headroom.
This, of course is not exactly Class A sound, but parallel op-amps of any size tend to have their own very relaxing sonic signature, probably because they're not stressed--they really are relaxed that way.
Usually that will need somewhat excessive ballast, so firstly, there's no chance of negative output impedance, and secondly, you get a miniscule touch of current drive.
Also, driving an 8 ohm speaker with Parallel LM1875 makes for very high linearity output or in other words, current headroom.
This, of course is not exactly Class A sound, but parallel op-amps of any size tend to have their own very relaxing sonic signature, probably because they're not stressed--they really are relaxed that way.
- Status
- Not open for further replies.