Very nice sounding chipamp-driven working class A amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
If you have read the thread "hum to noise problem in class A chipamp" that i started to ask about the problems i found while developing that amp, you will thing that it isn't a good thing to build. Well.. the problems are gone :).

Now it sounds very good. It's sweet and the sounds emerge from nothing and fade to nothing in a freakin' way, No 700 KHz on the output, no 3.5 MHz... Very quiet, with separated instruments and a very "black" background.

It's also the first amplifier i've heard whose bass is almost inaudible at low volumes and gets presence when increasing the volume. If you are familiar with psychoacoustics this would seem the thing to expect, but in all class B amps that i've heard the highs became prominent and agressive as the amplifier is pushed.

The solution (and revision to the schematic) to the oscillation problem was adding a cap between the inverting and noninverting inputs of the amp. It also thanks good cables and low impedance sources (it was really noisy when driven with a 5 meter unscreened power wire).

The electronics are cheap (excluding supply and heatsinks), you can get a stereo one for less than $15 (< $5 for the LM4562 (digi-key) , < $5 for the transistors and < $5 for the passives.

The TIP142 are choosen because they are very cheap and easy to find, of course adding a better transistor could give better sound and stability*.

At the moment the amp is stable without any compensating cap while not clipping, but it has a "tick" clipped region, meaning that the driver does not like to have its output clipped.

I would suggest to power it from 18.5V rails and to feed the LM4562 with 17V regulated to try to get clipping at 11 Vrms. This should give 15W of output power which seems enough if you have high efficiency loudspeakers or a family (With 89 db loudspeakers people living with me do complain).

The "key" advantadges of this amp are, apart from good sound, the ability to "tune" it by changing the few parts involve:

(1) Output transistors: These have to withstand high power dissipation and have a sufficient beta at the operating range not to place a tremendous stress on the driver opamp (2.7A center-point value with 20 mA drive current means that the opamp should be giving a maximum admisible of 10 mA at 2.7A, so a beta of 270 is the bare minimum. I don't know if there is any single transistor that meets these needs, but if exists then it would probably be a better choice than power darlingtons for speed reasons, this is a compromise between what sounds "worse", stressing the opamp or using slow darlingtons.

I don't recommend to use power mosfets because the RC filter formed by the gate stopper resistor and the gate capacitance will place a + 90 deg phase shift about 2 MHz, in the middle of the opamp banwidth (5.5 MHz with a gain of 10). This will make a power oscillator for sure and the only solution will be compensating the amp.

(2). Opamps: I go for the LM4562 because is universally accepted as sounding good. The Burr-Brown and Analog Devices ones are also fine and may suit your taste. there is a wide variety of high-fidelity opamps, all of them very good. Distortion in input stages and offset voltages depend on transistor matching and this is best done on-die, so i wouldn't be surprised if somebody reports that amp to sound better than very expensive gear.


(3) Passives: The good bias current and offset performance of these opamps makes no need for the feedback cap, the one that passes the most signal in all designs and the one that can be blamed the most for "capacitor distortion". I've found that the input cap to make slight differences, there is nothing wrong with choosing the one you like the most but it's not a good idea to obscess with it. I don't expect resistors to make absolutely any difference in the sound.

Dissipation is 50.4 peak and idle for the upper transistor and 25.2 idle, 50.4 peak for the bottom ones. This means you should use big heatsinks.

To avoid "out of current" clipping (which does not sound good) loudspeakers should not go lower than 6.5 Ohms at any range. You can make it able to go lower if you add an extra constant current sink at the bottom, having to care only about not exceeding Q7 SOAR. 4 Ohms loudspeakers are completely unusable here. I have doubts about paralleling the upper transistors since any difference in temperature would make a tremendous difference in current sharing. I would love to read advise about it, maybe a "base stopper" resistor? If paralleling is possible then maybe 4 Ohms loudspeakers would be very good as they would allow for 30W of output power with the opamp output swing.

Temperature compensation: The amp does not need temperature compensation of any kind and there is no way it can go into thermal runaway. Bias current goes lower as the programming transistors heat, but this is not much as long as they are far from the hot transistors (about 50 ma).

The design sounds very good. The only objection would be that it is not the fastest amplifier i've heard, but anyways it sounds very, very good, it has a sort of respect and politeness with the music that it's almost impossible to find in class B amps, while it is as clear as opamps can be.

I would love to read from anybody who builds it, mine does work. if you have any problems ask here. If somebody wants to develop a board is of course invited to do and all credit goes to him (orcad makes me want to dissapear). The sound of the amp is worth the effort*.

*I belive that the schematic is so simple that it has to have been posted before. All credit about how it sounds goes to national semiconductor.

I would add that i've found there is not much innovation on the forum, almost all threads refer to people having problems soldering a bunch of parts to a board. Nothing bad about helping people get good sound, but i would also love to read from people trying different things.
 

Attachments

  • 5w to 15w chipamp.pdf
    14.7 KB · Views: 1,661
Originally posted by eketeheSounds great..
I'll learn the schematic, n check if parts are available,hope will have the sensational sounds.

Brgds.
Eka

Nice to hear that you have interest on it. The opamp on the schematic has nothing to do with the opamps i reccomend, i put it here because it had the same pinout than all hi-fi dual opamps so i could try to develop a board. The output transistors are a "baseline choice", they are as common as the 2N3055 and you should not have any problems finding it. If you have a preference for other darlingtons or want to make a discrete darlington with couple of transistors and a couple of resistors you can of course do as long as they withstand the dissipation (better if 3A 40V is below the secondary breakdown region, otherwise you may get an amplifier that explodes when it is powered unloaded.). (*)

I would also suggest to place the opamp in a socket. Then you will have an easy time changing it and you can first try to run the amp with a very cheap one (TL072 will do) and then upgrade to the one you like the most.

(*) Forgot to say that the upper transistor dissipation goes 100.8W peak, 50.4 idle and average in the "No load" condition. The conclusion is that the amp should run unloaded only for offset checking (should be arround 10-40 mV for most opamps)!
 
Member
Joined 2002
Paid Member
Hi ionomolo

Nice that you provided the schematic.

I have 3 questions:

1. You draw the op. amp. power supply lines as +18V and -12V. Is it correct?

2. There are two constant current sinks in the schematic. Do you think that these are enough to drive lower than 6 Ohm speakers?

3. The dissipation you are quoting for the "bottom" transistors is for one CCS or for the two CCSs?

Regards
George
 
Originally posted by gpapag
1. You draw the op. amp. power supply lines as +18V and -12V. Is it correct?

No. The -12V is a mistake, i'm feeding it +/- 12V from scientific power supplies because i'm not listening loud (Yeah, krell's way, dynamic adjust of class A bias, too bad it has to be done by hand).

Originally posted by gpapag
2. There are two constant current sinks in the schematic. Do you think that these are enough to drive lower than 6 Ohm speakers?

It depends on how loud you plan to play it. if you want 10Vrms (10^2/6 = 16W) you need a current of 14.1/Zmin where Zmin is not 6 Ohms but rather the minimum impedance of your loudspeakers (btw the only manufacturer who openly quotes that is B&W, i'm not saying that others hide it but that it is not as openly exposed). Each sink takes something between 1.4 and 1.32 A depending on the temperature of the programming resistor. If it takes less then it means that the amp is overheating. The two current sinks in the design will give the minimum current needed for 16W in 6 Ohms. If loudspeakers dip a bit then it will clip.

Originally posted by gpapag
3. The dissipation you are quoting for the "bottom" transistors is for one CCS or for the two CCSs?

The dissipation is for each transistor.

EDIT: Powering the ompamp from lower rails as suggested 17V vs 18.5 is not only to get the maximum power from it but also to prevent "terrible clipping".
 
I post revision 2 with some errors corrected. The caps C7 and C8 are imprescindible. It also corrects some minor mistakes in the original version.

The cap C12 makes the amplifier more stable but it might not be requiered. The idea is that decouples placed near the active parts should be small since these decouples are to kill high frequency garbage. If you place a big cap there it will load with sharp pulses due to rectifier spikes and will probably inductively couple a buzz and make the clean ground less clean. I have no buzz in my design so i let it there.
 

Attachments

  • rev2.pdf
    15.1 KB · Views: 655
help with opamp and transistor choice

Well, i found that adding the C7/C8 caps improoved the amplifier performace while trying random mods to make thething more stable, but credit to this comes to AndrewT who suggested me to add these much before i started to try that mods. Sorry for not having mentioned him early.

Since i have built that amp i get hooked to music much more than i used to do with my previous reference (LME49810 + MOS), but today i decided to do some A/B comparison and found that the LME49810 had some sort of special transparency that the Class-A one didn't have. The LME+MOS one is a fast cold amp, while this is a warm polite one, but there is something i like in having a female singer shouting at my ears.

Any idea on which opamp/output transistor could give a faster sound? I've been looking at spice simulations and it seems that the pole created by the gate stopper resistor and the cgs could stil allow the amplifier to be stable. Should i go mosfet or there are fast nice bjt output transistors there?

Could the burr-brown OPA627/637 be a good choice?
 
Hi Ionomolo,
The idea to build this class A is coming since i already have pair of LM4562 in hand and wait for a good idea,

I just find all parts, and being noticed from a friend that class A amp is not for beginners.

well... how hard it will be?
what kind of PSU i will need for stereo ?
if TIP 142 is too late for LM4562, is LM833 or TL072 will suitable?

i am a newbie :D

need your honesty sir... will i able to do this?

its ok with all parts i have buy, i can make a BJT with its pair TIP147

Brgds.
 
I was going to pm you to warn about some of the issues regarding class-A amps as i've noticed that your aren't among the oldest-registered members.

Well, the true begginers where those who built the first amps, they had nobody to ask! And they did build class A amps because they are the conceptually-easier.

About the opamp, i suggest to try with the TL072 and then replace it with the LM4562 once you know it won't explode. The output transistors aren't the best you can find, but they will do the job.

There are only four things you must know before buliding it:

The transformer must be able to supply continuous > 3 A for mono and > 6 A for stereo.

The heatsinks should be very big. Running the amp without heatsinks will immediately destroy output transistors and loudspeakers.

You cannot feed it higher than rated voltage. I suggest you to start with anything arround 12-14 Vdc and a single ccs (you will notice that there are two identical sub-circuits connected to the speaker output. Start building it with a single one). This will allow for 6-8W of output power, will require a much less expensive supply (1.5/2A will do) and will let you know if you want to invest in the big heatsinks and transformers requiered for maximum output.

I suggest you to post or send me by pm a picture of the transformers and heatsinks you plan to use.

The amp does not have an incredible amount of parts so it should be an easy build (it took 30 minutes for me to do a mono one with all parts over the bench).

The amp is not very powerful so you need to use efficient speakers and not listen to outrageous levels.

At the moment i'm trying to design a faster output stage (either bjt ccs-loaded sziklai or mosfet ccs-loaded sziklai) but this will be a hell to get stable, the "slowness" of the output transistors plays for your part if you are new to this since it will not requiere the extreme care with layout that faster parts need.

I belive you will like it because it's a very "politically correct" amp. Nothing prominent, warm, a bit of "tubey midrange" and low distortion, which means it's not adding so much of its own to make anything not liking its character hate it.

The picture attached is from the first prototype i did (6W powered from +/- 12V with a single ccs.). I belive you shouldn't have any problems building it.

Somebody has done anything releated to ccs-loaded sziklai (replacing the usual 100R with a ccs arround a couple of mA's) *?

*This simulates to give a sub 0.01 % 1KHz THD open-loop close to the clipping voltage due to the immense amount of local feedback it has.
 
Sorry, i've just noticed i forgot to attach the picture.

EDIT: The psu i'm using is unregulated snubberized "classical" from chipamps. The voltage can never exceed +/- 18V because the opamp, but each unnecesary volt turns in a lot of heat. You should only make sure that it can hold those currents.
I may add a choke (pi-filtered) one later, but don't mess up with that if you are new to this.
 

Attachments

  • dsc_0320.jpg
    dsc_0320.jpg
    50.6 KB · Views: 2,600
Dear Ionomolo,
i'll start with a single ccs,
maybe a bit slow development, theres some thing i need to learn first, than bother you with some newbie questions.

Here attached are the parts.
i got some of shown heatsink. may will consider to invest toroidal transformer and better parts for upgrading later.
the shown toroidals are only 15-0-15 20VA and 15-0-15 2A which only enough for buffer power. 2 x 5A 12-0-12 E1, a lot of resistors with vary watt. all the cheapest.

The 12 - 14vdc maybe a problem since 9 or 10vac /min 3A transformer is rarely found here. is it bad to have a different power for the opam? ( regulated ) so that we able to have about 16VDC for TRs from other bigger power.
what is U3A?
have you try MJ5003? can it be other option? ( i dont know what you guys call for TR with that UFO lump, we call 'em jengkol ).
i have OCL amp using that UFO 5003/5004, they resist to higher heat and voltage than 2955/3055, higher the bias and lower the voltage was make the sound nicer. that a bit same tech with this class A don't you think.

Brgds.
Eka
 

Attachments

  • img_3750.jpg
    img_3750.jpg
    47.2 KB · Views: 2,423
I'm not a fan of toroidal transformers because they have wider bandwith and let all the garbage from the ac line to get into the amplifier. Mine has two blocks, one containing the psu and the other containing the amplifiers, so emf is not much of an issue.

The 0.5R resistor should be at least 3W! a 1/4 one will burn immediately after power-on.

The 12V-0-12V transformer is ok (a bit on the high-voltage side). assuming 0.7V drop on the diodes it will give 16V.

The transistors should have a beta of at least 300. I belive that the ones suggested in Douglas Shelf book would do a great job here.

The power of the opamp may be the same of the output stage or a bit lower. The reason for choosing lower voltage is that the opamp would not be able to drive the output transistor into hard clipping and this will prevent massive carrier storage at the base (*).

I belive that the heatsink is a bit on the small size for that voltage (assuming you heatsink each TIP142 transistor on a different heatsink, if you plan to place both on the same heatsink it is definitely insufficient). The "other" transistor in the CCS (i belive it is a bc5xx in my current ones) should be placed as far as possible (without creating too long wires) from the big heatsinks (it is in the center of the board in my current prototype). because if it's close to the heatsinks it will thermally track the output stage temperature and make the current go lower than expected (One problem with push-pull amps is that they do the opposite, go higher currents when they heat, and this is dangerous).

Conclusion:

( a ) 12-0-12 5A transformer is fine.
( b ) Heatsinks are bit small.
( c ) 0.5R needs to be rated at least 3W
( d ) EXTRA: Place decoupling transistors close to the power pin they are decoupling (that applies to all amplifiers either class A, B, C, D and to all other circuits that aren't amplifiers).
( e ) If you wire it properly it should just work like any other amplifier.

A bjt transistor operates the following way: It is like a diode, current goes from the emmiter to the base, but the carriers at the base take a time to recombine and during this time most of they are sucked to the collector region because the base is a very thin layer. If the collector hasn't enough voltage to attract them all the carriers will be stored at the base and this will make the transistor take much more time to recover from clipping. The beta of the transistor is the ratio between sucked carriers and recombined carriers.

Apart from being a class A amp, this design is single-ended, which may give a bit more "SET sound". The opamp is also single ended because the sinker transistor does nothing unless the amp goes into hard clipping.

EDIT 1: Remember to place the anti-popcorn-sound transistor found in REV2.

EDIT: You can ask all the questions you want, i will reply you quickly if i have time. In fact you are doing me a favour because you are preventing the thread to fall apart from the spot due to threads asking and asking again about a standard non-inverting amp.
 
I would start without regulation at the opamp, since 12-0-12 will give 16 V rectified. That's a bit on the high side for a fist prototype but it should work. If you regulate the opamp at 12V this will probably decrease output swing to 10-11Vpk (6.25), while in the unregulated case you may get 14Vpk (12W)

I said that regulating the opamp would probably give better performance because there is a lot of people here saying so. Mine is unregulated.

There are some threads about which regulators work the best. 78XX are said to be the low end, if i'm right jung's superregs (discretes) are said to be very good.

Anyways this is an extra tweak and it's not needed for a first prototype. If i had to tweak something this would be the output transistor.

Now i would like to work a bit on the output stage. I have to choose between integrated darlingtons (MN2488 that are less linear than some competitors but with a nice 55MHz transition frequency) or doing a CFP/EF with sustained beta devices (maybe MJL4281A or 2SC3281). Any suggestion?
 
To improve the output power capability;
May we use NE5534 for each channel (so it's DC offser can be easily adjustable to zero) and feed it with symmetrical 20v and 24v for the output stage,
Then we could get 13v swing on NE5534's output, that makes 21,5W into 8R (or 28W into 6R)..
So in this case we will need ~2A bias for the output stage (for 8R). How can we adjust the bias current to that value?
And the second question;
May we use MOSFETs for the output? (Like IRFP240 etc).. I have an experience with MOSFETs very like to this circuit. And with this way, we could run the opamp in class A mode, because theorically we wont sink any current from the opamps output!
What do you say?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.