Welcome, Guest. Please login or register.

October 31, 2020, 03:57:15 PM

Login with username, password and session length

Recent Posts

 

Author Topic: 18watts peak-power amp  (Read 9227 times)

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
18watts peak-power amp
« on: September 15, 2007, 10:34:16 AM »
hey folks ,,this is the amp that i told ya about ,,again i wanna say im new to this and i designed this one hoping it works
i would like to know if there is any mistakes and problems in this design.
the output transistors are
Q1 = BD139
Q2 = BD140
Q3 = BD540
Q4 = BD539
the op amp im using is the TL074 because it has a high slew rate about 13v/uS
in the offset circuit im using 2 opamps because one cannot sink the currnet coming from the diodes alone.
i put 5 diodes in series with the 0.8k resistor to provide quisent curent of 30 mA which i think is suitable and to provide enough voltage drop between the first two base emitter transistors junctions.
the offset circuit is nessasry bcz the output voltage is 0 when the input signal down the diodes is -0.5 volts,so i offset the signal by -0.5 volts before driving the power transistors
in the tone control citcuit im using a pot that should take the values between (10k and more) and a variable cap between (10nf and more)
to give the pot the minimum value which is 10k is to boost the bass.
to give the cap the minimum value which is 10nf is to boost the treble.
the bandpass fiter i chose is the first order one because the phase response is the best.
i hope that u guys tell me what do u think about this amp and any advice would be welcome.

Good OL' BeN

teemuk

  • SSGuitar Global Mod
  • Legendary
  • ****
  • Posts: 1074
  • Chip Points: 282
    • View Profile
Re: 18watts peak-power amp
« Reply #1 on: September 15, 2007, 12:14:34 PM »
First a general suggestion: Draw the schematic in a clearer way: For example, which nodes interconnect and which ones just cross each other, where does the arrow pointing downwards from 29.5K resistor connect, does the arrow from 0.8K resistor mean a connection to 15V, etc. If the schematic is drawn poorly or illegibly people don't generally waste a lot of time in interpreting it. On the other hand, nicely and professionally drawn schematics will not be turned down at the first sight and even if they present an extremely poor circuit (that couldn't even work) they tend to be more appreciated. This ensures that you get more help, corrections to the problems (both obvious and unobvious) and create some discussion concerning your design.

Then some other suggestions concerning the power amp:

- You use a bipolar supply but an AC coupled output... It doesn't seem practical and at least you need a non-polar capacitor for that
- Where are the emitter resistors of the output stage? I know they are not neccessarily needed if no parallel transistors are used but since they help in thermal compensation I would not omit them
- You introduce way too much biasing voltage drop for that kind of output stage: The forward voltage of a Sziklai pair is one VBE so two or three diodes should be enough
- Tie the output stage inside the feedback loop, this will correct its non-linearities and thus will help to compensate crossover distortion as well.
- Hook up resistors from the Sziklai pair's driver's to the collectors of the power transistors (see attachment)
- I see no compensation against HF oscillation
- I see no Zobel network in the output
- Why on earth you use a unity gain stage to drive the power amp?
- Even more: 5M resistors in the unity gain stage! You are practically begging for noise and effects from parasitic capacitances (oscillation HF loss etc).

I don't want to put you down but your power amp design looks like it might work but only quite poorly. Why don't you use a conventional and proven topology... with some gain. (See attachment fo one example). This topology is extremely simple and if properly designed can provide very good performance and extremely low DC offset. Anyway, since you run the power amp from very low rails why even use a discrete design? Most chip amps are way more simpler to build and perform so much better than the average discrete designs.

Then the "preamp"...

- All stages are DC coupled. Is this absolutely neccessary? You can introduce a steeper hi-pass filtering if you AC couple some stages.
- No AC coupling, ground reference or overvoltage protection in the input! If you worry about DC offset then omitting the ground reference is the perfect way to make that problem worse. It says "from the preamp" - something is omitted?
- What's the use of the buffer? It seems like a completely unneccessay part of the circuit since the stage directly drives an opamp input and there likely isn't any long wiring in between. Anyway, it does no harm so I imagine it's there just to "fill" a quad-opamp.

Have you built or simulated this circuit? What are your own impressions about it? There is a very close resemblence to this design: http://sound.westhost.com/project76.htm so checking it out might provide some ideas...

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
Re: 18watts peak-power amp
« Reply #2 on: September 15, 2007, 07:51:38 PM »
well thanks man for being concerned in the design
first of all im sorry about the drawing,,ill make sure to make better drawing next times
second about the circuit ,,i havnt built the circuit but i simulated it over proteus. it works fine there .
about the diodes .... in the proteus software when using only 2 or 3 diodes the provide only around 0.5 voltage drop between the two bases .....is that enough to open up the 2 transistor base-emitter junctions to create some quiescent current?
by using 5 diodes i can have a voltage drop of 1 volts when the input signal is -0.5 and so i have 0.5 volts on each base emitter junction which is enough to create a quiscent current of 30mA .... thats what i see by proteus ,,, :(
and yes i mean by the downward arrow -15 and the uppward one +15.
i thought that this is very common.....
what are the vaules of the emitters resistors should be?
,,and about the unity gain stage. it is just to offset the signal by -0.5 volts,and i chose the 5 M cuz i didnt know that it would cause that much noise ,,thnx for the information..
again about the resistors to the collector of the transistor pairs ,ive seen schematics that include them and some they dont,, still not sure what s the main benefit of them.
thanks again for the schematic u included and do u know any source that explains that feedback u told me about? to include the output transistors in the feedback loop of the opamp?
Good OL' BeN

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
Re: 18watts peak-power amp
« Reply #3 on: September 15, 2007, 08:00:50 PM »
sorry again there is a mistake in the second resistor value in the bandpass filter
instead of 5k it is 11k to provide a gain of 12
Good OL' BeN

teemuk

  • SSGuitar Global Mod
  • Legendary
  • ****
  • Posts: 1074
  • Chip Points: 282
    • View Profile
Re: 18watts peak-power amp
« Reply #4 on: September 16, 2007, 02:46:01 PM »
well thanks man for being concerned in the design

My pleasure, that's what these kind of forums are for.

Quote
second about the circuit ,,i havnt built the circuit but i simulated it over proteus. it works fine there .

Are you sure about that? My simulation with LTspice tells me that the output devices dissipate over 100 watts at idle...

Quote
about the diodes .... in the proteus software when using only 2 or 3 diodes the provide only around 0.5 voltage drop between the two bases .....is that enough to open up the 2 transistor base-emitter junctions to create some quiescent current?

Yes, it should be: Essentially you need to create bias voltage potential that is about the same as total forward voltage of the output transistors. Consider the compound (Sziklai) pair as one transistor: It has a forward voltage of only one VBE (there is only one "diode" junction between input and load). For a darlington pair the amount is two. Since the output has a total of two compound transistors you need about 2 x VBE. That's the same as two diodes.

Quote
what are the vaules of the emitters resistors should be?

Quite small: The smaller they are the less they have effect. However, too high values will limit the current too much. Typical values are between 0.22 - 1 ohm, so values like 0.33 or 0.47 ohms are quite common.

Quote
,,and about the unity gain stage. it is just to offset the signal by -0.5 volts,and i chose the 5 M cuz i didnt know that it would cause that much noise ,,thnx for the information..

If you worry about the offset you have other means to compensate it more effectively: Try AC coupling the input and setting unity gain for DC for example. In my opinion, the inverting configuration is quite poor for an output stage since all practical methods to compensate the offset are difficult to implement. Besides that, it has narrower bandwidth and poorer noise performance.

Quote
again about the resistors to the collector of the transistor pairs ,ive seen schematics that include them and some they dont,, still not sure what s the main benefit of them.

In Sziklai pair they set gain and omitting them often leads to a situation where output stage draws gigantic amounts of current. In Darlington configuration they increase the speed of the transistor. I suggest you simulate their effects.

Quote
thanks again for the schematic u included and do u know any source that explains that feedback u told me about? to include the output transistors in the feedback loop of the opamp?

That's pretty simple: Instead of taking the feedback from the output of the operational (voltage) amplifier, take it from the output of the push-pull "buffer" (current amplifier) which follows it. Since feedback tries to correct the errors in the signal it will now compensate the non-linearity of the output stage. Also, it will try to correct the DC offset that may drift a lot if the output stage is not included to the feedback loop.

BTW, I drawed the schematic with haste so it has some errors: Basically the driver's emitters should connect the output through resistors. Omitting them sets a very high gain (collector resistance/Remitter) so the output stage will dissipate a lot of power: Not good.

I took a liberty of "tweaking" (or basically re-designing) your design to create another circuit that seemed to work at least decently (see attachment). I suggest you A-B it against your design and see what differences you get in

- DC offset
- Maximum output voltage swing
- Quiescent current and power dissipation of the output transistors
- Linearity
- Stability.

Don't worry about the output transistors being different, this is just an example. Anyway, some explaining is in order: the two bootstraps should provide a more constant current load (for the opamps) than plain resistors so the output voltage swing ability is effectively increased. You see also that only two diodes are used, yet the output stage hardly suffers from crossover distortion, which the biasing is intented to eliminate in the first place. The input is connected "symmetrically" between the diodes and bootstraps are also symmetric. Output is fitted with zobel that tries to compensate the rising inductance of a speaker load (you hardly see any effect of this with simulator unless you model the speaker).

This is a fairly basic design, similar to ones that are used, for example, in countless reverb driver and headphone amplifiers. Anyway, simulating and tweaking this circuit - in order to see how different component values or omitting some of the stuff completely affects the operation - should teach you a lot about power amp design. I suggest you also try replacing the inverting unity gain stage configuration with that bandpass filter of yours (tie the output to feedback loop). It will then work as a high-power bandpass filter: This is the topology that 99,9% of modern SS power amps use. (Most of them just have a "discrete" opamp). Hook the circuit's input to ground with 11K resistor (same values as the one in feedback loop) and see how varying this value will affect the DC offset.

Anyway, this post has gotten too long already and I'm not trying to tell you how you should design your amp. I hope you will take these as mere suggestions that (I hope) will help you learn more about power amp design.

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
Re: 18watts peak-power amp
« Reply #5 on: September 16, 2007, 04:14:25 PM »
well thanks man you did a great job to the amp but i still have some points thats not so clear to me.
the diodes ,,,again,,the BE junction of the BD139 is not the same diode as the 1N4007.
i did some simulation to the circuit u attached,for example if the opamp out is at 0 volts,the amount of current thats gonna pass the first diode is about 10mA and this current introduces a 0.2 v voltage drop across the 1N4007 diode which is clearly pretty far from opening up the BD139 BE junction cuz its totally a different diode so there is clearly no quiescent current at all!
the BE junction of the BD139 introduces a voltage drop of 0.61 volts when running 10mA through it, while the 1N4007 introduces only 0.2 volts thats basically why i put 5 diodes in my design.
again,, werent you against using the 5mega resistors???
and also in the schematic u draw u havnt included a lowpass filter of 20000hz
is it that the zobel configuration is enough?
one thing more ,whats the bootstrap circuit for? i would really welcome a little explanation about it.
again really thanks for the great job uv done to the circuit and yes your modification really helped me understand more about ss amps.thanks alot.
how long have ya been in building amps?
did u build the schematic u included me?
does it work fine?
Good OL' BeN

teemuk

  • SSGuitar Global Mod
  • Legendary
  • ****
  • Posts: 1074
  • Chip Points: 282
    • View Profile
Re: 18watts peak-power amp
« Reply #6 on: September 16, 2007, 05:39:23 PM »
According to datasheet, 1N4007 should have a forward voltage of about 600 mV at 10 mA (my sim shows 7 mA through diodes at idle). Anyway, you are right in that you should match the diode's properties to those of the driver transistors. Yet, I still think that 2 or 3 diodes should be all that you need; there will not be 10 mA over BE junction because the base current is only a fraction of the bias current (through the bootstraps and diodes).

Basically, the important question is: Do you have crossover distortion at low amplitudes when only two diodes are used? My simulation does not show any visible amount of it.

The reason why I used 5Meg resistors in the (example) circuit is that you had them in the original desin of the (unity gain) output stage. The schematic should show how the feedback loop is tied around the circuit and (plus few other "tricks"), therefore I didn't want to confuse with any other resistance values. I didn't draw any of the filters of your circuit either - just the unity gain/current amp part.

The Zobel only linearizes the inductive speaker load and helps to prevent oscillation. It should have no audible filtering qualities.

I haven't build the attached schematic - only simulated it with SPICE. The circuit is essentially a result from about half an hour "tweaking" with LtSpice - just for the purpose of showing an example. So... it's not proven and tested design. I wouldn't build a unity gain power amp - or even a discrete power amp for less than 30 W of (RMS) output power. Even more, I think few hours of tweaking that circuit would even result into a better circuit. It's just for educational purposes.

I have been dealing with amplifier circuitry (researching & developing, repairing, building etc.) for few years. I'm for the most part self-educated in electronics and my main occupation at the moment has nothing to do with amplifiers - or even electronics.

The bootstrap circuit is essentially a constant current source: The AC voltage swing over the "bottom" resistors is kept constant because the node in between the two resistors is referenced to the output with a capacitor. (The output has about equal voltage swing as the input). I have attached a simplified schematic that should show the difference of operation between a simple resistor load and a bootstrapped load. Notice how the capacitor can shift the voltage potential at the junction of the resistors to a higher potential than the actual supply voltage is. This extends the voltage swing of the output stage because the transistor bases will not run out of current when input signal's amplitude is close to voltage potential of the supply rail.

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
Re: 18watts peak-power amp
« Reply #7 on: September 17, 2007, 04:19:57 AM »
Quote
According to datasheet, 1N4007 should have a forward voltage of about 600 mV at 10 mA (my sim shows 7 mA through diodes at idle).
lookslike my proteus is using another type of 1N4007 ,,!!! possible?
i got really shocked on this. i spent hours to modify the design because of this.
Quote
I wouldn't build a unity gain power amp - or even a discrete power amp for less than 30 W of (RMS) output power.
...whats wrong with that? is it not worth the effort?
Quote
I have been dealing with amplifier circuitry (researching & developing, repairing, building etc.) for few years. I'm for the most part self-educated in electronics
thats Awsome,why dont u consider a career in amps! lookslike u have a big load of info about em.

also seems like the bootstrap circuit is some heaven for amp design.
anyways
im really considering to build a first amp
what schematic u think i should go for?
an opamp or the differential amp configuration for an input stage?
what power transistors u prefer for the output?
and about cutting high frequencies to protect the speaker,,is it a must?
should the -3db freq be at 20000hz?
how much phase shift is allowed for the sound signal to be heard without any distortion?
i also did read some stuff about ground loops,,here in the ssguitar and somewhere in epanorama ,its really somthing to be concernd about,
whats the best wiring u think to get rid of it? i liked the star one,lookslike its pretty neat.
again thank u and im sorry if im asking too much,its just im new to this.
Good OL' BeN

teemuk

  • SSGuitar Global Mod
  • Legendary
  • ****
  • Posts: 1074
  • Chip Points: 282
    • View Profile
Re: 18watts peak-power amp
« Reply #8 on: September 17, 2007, 09:05:44 AM »
I don't quite follow you on that HF speaker protection issue. Guitar speakers mainly blow because too much power is fed to them, not because of a high frequency signal. Cutting high frequency signals is done to prevent oscillation, which often happens way above our hearing range. The 3dB point should be set above 20 kHz (maybe to 30 - 40 kHz) because otherwise the filter will affect on audible frequencies as well. It's another issue if you want to remove "harshness" after some clipping stages. Also, a typical speaker works like a low-pass filtered circuit with -3dB at about 5kHz. However, this inherent filter cannot prevent oscillation.

I can't really recommend you a schematic since I hardly know what type of music you play, what features you need, how much complexity in the circuit you "tolerate" etc. If you need a practical low-power amp (is there such thing?) I suggest you make the preamp versatile and nice sounding, then build the power amp section based on a chip. There must be dozens of schematics picturing such a design around the internet. For example, try Dean Markley, Rickenbacker, Fender or Gibson site for schematics. Those are bound to give you some ideas and inspiration.

For low power I would use an opamp.... or rather build just an IC-based power amp because those perform so much better than simple discrete designs and even include stuff like short circuit and thermal protection. With all the effort spent to design something even comparable it just wouldn't make sense to build something as unpractical as a low power SS amp.

For anything that needs a higher output swing than the supply rail of opamps is I would use the discrete differential (plus the required voltage amplifier stage that follows it). There are also numerous ways to increase the output swing of an opamp past the rails or use it as a differential but that's another story. There are also special opamps designed for driving power output stages with higher voltage swing.

High power chip amps are pretty good but tend to have some problems: For example, they try to dissipate the heat away from a very small area. A discrete design can spread the heat along the sink more evenly. In my opinion, discrete stuff is more fun and lends itself to mods and repair a lot more easily than those ICs. However, it will be huge project to design and build a discrete power amp from the scratch (and gets harder and harder the more the output power increases). It's usually demanding enough to rob all enthusiasm away so I really can't recommend it as a first project.

About component selection: I prefer stuff (transistors) that I can find from the local shops selling electronics components. i.e. BD139/140 (drivers) and TIP142/147 (output devices). This of course depends a lot on the circuit you are intending to build. Those are pretty good for a 60 W amp but overkill for 10 - 20 watts. I tend to avoid all obscure transistors since there is a great chance that thay can't be found locally and ordering would get too expensive (this is usually the case with FETs, MOSFETs and alike). Plus I always buy more than needed in case I blow stuff up. Other reason is that I never know when the stores run out of supply of certain component - and I like to have stuff around for other projects that come up. You usually can't have enough capacitors from 4.7 - 100 uF range and small signal all-purpose transistors like BC546 and BC556. I always tend to need that stuff when working with electronic projects.

The grounding is an issue I wrote quite a long post here. A REAL star ground is too damn difficult to implement in an effective manner but luckily that's not even needed. Anyway, I'm not going to rephrase myself on that issue. If you understand the concept of ground current flow then implementing an effective grounding scheme should not be too difficult. :tu:

benzer

  • Master
  • ***
  • Posts: 63
  • Chip Points: 0
    • View Profile
Re: 18watts peak-power amp
« Reply #9 on: September 17, 2007, 02:10:01 PM »
Quote
I can't really recommend you a schematic since I hardly know what type of music you play, what features you need, how much complexity in the circuit you "tolerate"
well i play country music and the blues
about the complexity of the circuit i just need a practical one containing all the circuits it should has,im doin the fifth year of electronics engineering so i can handle it if there is a little explanation,
although as a first build i would recommend a simple shot.
Quote
I suggest you make the preamp versatile and nice sounding, then build the power amp section based on a chip.
what do u exactly mean versatile? what can it has more than a tone control and amplification?
and im really considering desiging with discrete elements in the perpose of educating myself ,,i dont want chips.
its true what u say about that its a hard shot and maybe someone would quite but i think ill just work my ass off for it
thanks for noting about the websites lookslike there are loads of schematics over there,,but are they reliable to build?
Quote
TIP142/147 (output devices). This of course depends a lot on the circuit you are intending to build. Those are pretty good for a 60 W amp but overkill for 10 - 20 watts.
why overkill? what could go wrong if they dont use their full range of amplification?
anyways,, how many amps have ya built? whats the best shot for country music? ,,,maybe i should try to get the sound of tubes..


Good OL' BeN