10 W Class AB amp: Adapting Single Supply to Dual Supply

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Dear diyaudio forum,
I have some experience in electronics (took several circuits/electronics courses in college), but this is my first attempt at building a modest power amp: 10W class AB.

I found a nice, fairly simple design in "Transistor Circuit Techniques" by GJ Ritchie. His design shows a single supply (+25 V). I am thinking of how to adapt this for a dual supply design (+/- 12.5 V).

I've attached the schematics of the original (single supply design), as well as the one showing my proposed modifications for the dual supply version. I wonder if anyone can comment whether my modest changes will be sufficient. (Somehow I have this nagging feeling, there is more I need to do
icon11.gif
). Briefly, here are the changes I've made:

For proper biasing, I think I might need to change the ratio R6/R5 to be closer to 1. It seems like I ought to have R5 connected directly to +12.5 rail, but I can't do that AND have the negative feedback (set by R5 and R4). Hmmm...

Would very much appreciate any comments/suggestions/feedback you may have. Would love to know if/how I've made an error here!

Thanks a lot. Jon E

- Top/bottom rails now at +/- 12.5
-removed CL (1000u), since dc offset should be set to 0v.
-input is now shown referred to ground
-inserted power line filter caps (100 nF)
-Zobel network connects between output and ground.
 

Attachments

  • singleSupply10WclassABamp.jpg
    singleSupply10WclassABamp.jpg
    236.9 KB · Views: 926
  • dualSupplyAmp.jpg
    dualSupplyAmp.jpg
    242.5 KB · Views: 854
Last edited:
Administrator
Joined 2007
Paid Member
The big big problem is drift... you will never achieve a stable "zero volt" output.
Also switch on/off thumps with all those time constants... easily fixed with a relay delay.

It would be easy to add a DC servo (one FET opamp).

Any reason for wanting dual rail supplies ?
 
I see that you are doing some effort to really understand how thw holy thing works, so I believe you deserve some answer. BTW I am not an expert, so take my advice with a grain of salt..
So, there is nothing really wrong on your circuit at a first look. I would only recomend to join the Vbe trimmer pins central and top, so in case of contact failure you do not have too much bias current. The compensation capacitor (22pF) it looks to be at the lower limit for stability. You may start with a safer value, i.e 100 pF.
BUT
Your amplifier, while correct seems missing something... an input stage. I mean the input impedance that you show might be too low and you probably need an input stage to correct for this. This is usually done with a differential amplifier, a couple of BJTs in its simplest form. Then you will probably have to hang the feedback there, but thre are a lot of choices. Good luck
effebi
 
Mooly and effebi: thanks very much for your rapid replies!

Before I address the specific points you made, I ought to say:

I plan to build a pre-amp based around a Baxandall tone control with a TL072 op amp. (See attached. jpg). I've already built this as shown, hooked up an ipod for input, ran the output into my stereo receiver, works great. I plan to cascade that to another inverting op-amp with volume control to make the whole "pre-amp" section non-inverting, a la project 97 on Rod Elliot's website (sound.westhost.com/project97.htm). The output of the it will drive my power amp (in theory anyway
icon10.gif
).

With regard to your specific questions.

Mooly: I didn't say previously, but I do plan to use a relay to avoid thumping. Already bought one of those (at somewhat relatively great expense, well 10 bucks anyhow).

Can you say a word about the DC servo FET you mention? I'm not quite sure I follow, other than to guess that my source (ipod, say) should drive the FET, which essentially drives TR6 in my power amp?

I was after using dual rails because it seemed easier to use my home-built power supply for powering both the op-amps in the preamp with +/- Vcc AND the power amp. (I built a power supply from a 12-0-12 center-tapped transformer. Power supply just uses standard full-bridge configuration with filter caps, working just fine.) I realize it is possible use a single power to power op-amps. As a wee lil lad in basic circuits courses I always used a dual supply, so that's what I was comfortable with. Maybe time to push my safety zone a bit, eh?

Also, I realize I could "skip over" the center tap and use the full +24 V for my power amp (well, actually ~24*sqrt(2)). But wouldn't that leave me with an input to my pre-amp referenced to the center tap ("ground") and my power amp referenced essentially to 12*sqrt(2) below ground. That seemed like trouble to me...maybe not, if the blocking caps just block the dc offsets anyway.

Thoughts on this?

effebi: Indeed, I agree about the compensation cap. Other designs I looked at used 220 pF, so I'll very likely opt for something a bit larger.

Good point, too, about using a differential amp as an input stage. I had seen a design that implements this in Martin Jones' "Practical Introduction to Electronic Circuits" (love that book, btw). I guess I could always switch and use that topology instead. But, if possible, I would like to figure out (if for nothing more than my own edification) the proper way to make the amp currently under question work off a dual supply.

Thanks again for all your comments. Please let me know if you have further thoughts.

Best regards. Jon E
 

Attachments

  • Baxandall_tonecontrol.jpg
    Baxandall_tonecontrol.jpg
    61 KB · Views: 607
Administrator
Joined 2007
Paid Member
The DC servo is a widely used technique even in amps with differential inputs. It uses an opamp (must be a FET type like the TLO71 etc) connected as an integrator. The integrator removes all "audio" from the output due to a long time constant and the output of the opamp is used to bias the input transistor of the main amp. It's a really simple circuit, one opamp and either one or two resistors and caps depending on whether the action from the servo is inverting or not.

Your amp is "inverting" so the servo needs to be "non inverting" such that a drift positive in the output from the amp causes the error voltage to be positive also. The servo should hold the DC offset to around 1mv or less with no drift whatsoever.

Your amp is based on the classic "Lin" design and can be elaborated on and I'm a huge fan of the single ended input stage for sonic reasons.

Your 12 0 12 volt transformer will give DC rails of around 17-0-17 which I think your aware of.
 
Can't tell you how much I appreciate you al sharing your time and expertise in these forums. I know that basic questions from newbies (like the ones I am asking) can cause a seasoned pro to roll the eyes, but have been most patient--I am already learning a lot, and again, many thanks!

hitsware: Definitely see your point about the input being referenced to neg. supply instead of ground. I was hoping I could get away with this because (and I failed to mention earlier), my power supply actually has LM317/318 (adjustable) regulators, so the supply line should be fairly quiet, very low ripple.

mooly: I'm following just fine with the non-inverting integrator and the concept behind the dc servo. With that...here comes a (probably) stupid question. Indeed my amp is inverting. So...why doesn't the amp already do what the dc servo is gonna help it do? In other words, say voltage at output goes up. Negative feedback loop with have high voltage applied at top, so base voltage of TR6 goes up, turns on that transistor more, decreases voltage applied to the output transistors, which should act to counter/decrease the dc voltage at the output. Maybe the magic has to do with the frequency response--that a dc servo can actually do its job at dc, but the feedback loop itself cannot. (If that's true, I am confused, because, its just a dc feedback loop, ya?)

Definitely gotcha on the +/-17 rails (rms vs p-p). One other quick stupid question, because all of the sudden I find myself lacking any degree of confidence in my abilities: If were to use the single supply version of the circuit, and I want to supply the full + 25V to the top rail: couldn't I just set the regulators to output say +/- 12.5V, then treat -12.5 as if it is the reference potential--i.e., ground--and treat the +12.5 V as being +25V above ground? Perhaps the ground rail will be slightly noisier than doing it the proper way (which isn't much trouble at all...)

Could always just start constructing and see how good--or bad--things really are, but kinda nice to understand all the issues first!
icon7.gif

Cheers.
Jon
 
Administrator
Joined 2007
Paid Member
There's a lot to be said for experimenting... it's how we all learn.

Single supply. If you wanted to use that, and there's absolutely nothing wrong what so ever in doing that, then all you need is one regulator in the + rail. The ground rail doesn't need any form of regulator in it. Also this amp would work with voltages much higher than +25 as long as caps and semiconductors were suitably rated. Power amps are generally better run off an unregulated supply as a general rule. As to hum and noise. The bigest factor is the grounding layout... and this is where so many builds fall down, and yet it's so easy once actually think about it. We can come back to that again.

DC feedback. I can understand your reasoning :)
The problem is that there is no absolute fixed reference DC wise on a single ended input like this. The bias is "fixed" yet the output drifts like crazy. It's because the transistor parameters change wih temperature. The "fixed" base emitter volt drop of a transistor actually follows a precise curve with temperature. The small DC bias current taken by TR6 actually varies with this temperature change and that change alters the bias voltage on the two resistors slightly. Also the DC gain of the transistors rises with temperature.
So although it's fairly stable, it's not stable enough to use for a DC coupled design.

The DC servo has ground (zero volts) as its reference to compare the output too, and so generate an error voltage. If we set the reference voltage to any other value then the amp output would sit at that value too.

Even amps with a true differential input stage suffer small DC offsets due to imbalance and mismatched transistors etc. Hold your finger on one of the pair to warm it and the output shifts dramatically. In an IC (opamp) the parts are tighly matched and all in thermal contact with one another.
 
Ah! I get the problem with dc stability now. I had thought of temperature effects, and initially thought that feeding back to TR6 would cure all ills, but now I definitely see the error of my ways
icon10.gif
having not thought properly about the issue.

All good on the single supply--get what you are saying about using just one regulator in the + rail. I think I better get my self a LM338 (5A max) instead of LM317 (1.5A) or I might be very sad when the music (err, the voltage regulator) stops! Or maybe I ought to ditch the regulators, as you suggest. I've read multiple perspectives on regulators vs non. Most audiophiles seem to prefer non-or are forced to do w/o b/c the currents are high. For a fairly low current application like mine, it seems the main objection is poor regulation at the higher audible frequencies is the main issue. Curious to know more about your take, why they are typically avoided in low power (current) applications.

Ground: to the best of my knowledge a star ground layout usually does well. Thoughts on that?

One last question if I may before I begin to actually start construction: I do want to have an op-amp based tone control in my circuit, and I have two power supplies at my disposal (both ones I built, same 12 0 12 transformer, and can be reconfigured easily). So couple of options.
1. TLC072 is specifically made to work with single rail--in fact the datasheet shows an example of using it with Baxandall tone control powered by single supply! So I could just go that route--very tempting.

2. However, I had planned to use a TL072 running off dual supply (for no other reason than familiarity). In this case, let's say I use one power supply as a single supply (0 30V) for the power amp, the other one as a dual supply (-15 0 15V) for the tone control, connect the 0V terminals of both to reference them to each other. If I connect output of tone control into input of power amp, now they share the same reference potential, so a) hopefully any dc offset is minimal between tone control and power amp, and b) if there is a dc offset, it will just get blocked by the input cap (C3) in my diagram. Do I have that right? Or am I missing something obvious (yet again!).

Thanks x 23674684 again, Mooly. I hope this is the last of my questions. When I'm done, I'll let you know how it went, whether good tunes be pumping or not.

Jon E
 
...
But, if possible, I would like to figure out (if for nothing more than my own edification) the proper way to make the amp currently under question work off a dual supply.
...Best regards. Jon E

the input referred to the negative rail is pretty much makes this a single supply design - anything you do "convert" it to dual supply will just be awkward and have no advantage over single supply operation
some signal is going to require C coupling to a supply rail without more transistors - and DC coupling the output is asking for trouble

to make best use of dual supply you simply want another circuit topology

the JLH Class A amp is similar power, single ended input with low parts count

The Class-A Amplifier Site

a consequnce of these single eneded designs from the dawn of transistor power amp design is that they are based on minimizing the number of then expensive transistors - at the expense of increased offset and bias stability problems
they are actually harder to fully understand and make "work right" than using a few more transistors to simplify the input/output reference and bias conditions

additonally the dual supply uses more power rectifiers and reservoir caps - again less expensive today


there is a reason the diff pair input is used nearly universally today - it does several things "right" and small signal transistors cost pennies apiece

the diff pair is much more linear than using the single ended inputs - due to the cancellation of the biggest exponential Vbe error term when biased with balanced collector current

the same Vbe cancellation also gives good DC offset cancellation

and the input has very high PSRR and can be trivaly referenced to the same center supply gnd as the output
 
Last edited:
Administrator
Joined 2007
Paid Member
Star grounding... if only it were that easy. Have a read this as this and see if you follow the reasoning... it really starts around post #14
http://www.diyaudio.com/forums/solid-state/101321-3-stage-lin-topology-nfb-tappings.html

As far as the PSU goes the biggest problem is just 100/120 hz ripple from the bridge making itself audible due to the less than perfect PSRR (power supply rejection ratio) of the (any) amp. Amps vary a lot in this... I suspect this is mediocre at best on that score, but that doesn't mean it's actually audibly a problem.
A problem with regulators is that they need a minimum input to work properly, the amp when delivering much current can easily pull the rail below that threshold if the reg voltage hasn't much headroom. The action of the reg is lost, noise and riple come through and the outout rail suddenly dips. Allow for that by having a higher input and lower regulated output and the dissipation in the reg is high... it gets hot... and it robs you of precious headroom. Probably the best compromise is to run the driver stages of a regulated supply and the outputs of the unregulated.

When it comes to opamps (or any other circuit) remember that ground is just an arbitrary reference point. All the opamps we are discussing are equally at home on single or dual supplies... yes some have features that give them an edge in low voltage single rail work but for these circuits any of the "audio" opamps would work just as well whether on single or dual supplies.

Just quickly looking at the tone control circuit you posted... first you must use a FET opamp as any tiny bias current flowing into/outof the opamp - terminal would cause the pots to be noisy when turned (crackly). All you need do for a single rail is bias the + input terminal to around half supply (I would use a 15 v zener +cap and say 10k) to generate a stable reference. The input must be AC coupled, and the output too which it is with C3 in the power amp. Positive end C3 to go to the opamp as that will be at 15 volts.

To me that is much better than using multiple PSU's. Don't be put off by AC coupling.
 
HI

THere are a couple of problems with your circuit which others have suggested work-arounds but to be clear here are the issues:

1. The input is offset from ground by a large voltage, being the difference between -Vee (or -Vcc whatever you call the negative rail) and the Vbe of the input transistor, meaning, about -10V. You CAN circumvent this but charging of the input capacitor has to be allowed before the amplifier is stable.

2. Drift of the transistors as the temperature changes will lead to possible offset voltages which have to be eliminated ("DC servo") or adjusted with a potentiometer.

Most people would use a couple of PNP input transistors with one referenced to ground and the other for the feedback to hold the output at DC zero. THis is not what you intended, but is more widely used for the reasons above.

In the original AC coupled circuit the only real issues I see are that the output capacitor is small for the load (4ohms). I recommend increasing the capacitance which would give you a better low frequency response (e.g. 22mF).

John
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.