Contest: Linear Power Amp in a mint tin (class Aa, class AB, or class B)

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
That line looks smooth, but distortion is relevant in the 1/100 and up range, you cannot assess that visually on a scope. Actually I think I can see some Gm doubling there, it straightens out the sine curve near the crossover region so it looks more triangular.

Furthermore, say your VAS input has an impedance of 500k which is normal. Your miller cap is virtually multiplied by the voltage gain of your VAS, so your VAS gain rolloff corner is probably in the tens to hundreds of Hz. So for most of the audio band into RF, your input stage sees not what is in your scope shots, but that through a +6db/octave frequency filter. So the reality is worse. In LTSpice, you can plot this by entering D(R5) and you will probably see the crossover distortion now. Now you can compare it to an EF.
 
keantoken,

The signal is across that 1k resistor that feeds the VAS buffer, which is after where the miller cap feeds back to the LTP - the voltage at that point would look very lowpassed I'm sure, since it's a current node, but this resistor should reflect to some degree the current being fed into the VAS transistor, I think?

I'm not speaking with any authority on this, I'm by no means an expert in solid state design.
 
Here they are attached to diyaudio. It should be obvious which one is which ;)
 

Attachments

  • 9cVsi.jpg
    9cVsi.jpg
    27.7 KB · Views: 248
  • IZeLs.jpg
    IZeLs.jpg
    29.2 KB · Views: 244
  • Screen Shot 2012-11-23 at 15.32.28.png
    Screen Shot 2012-11-23 at 15.32.28.png
    61.3 KB · Views: 255
R5 will reflect the VAS current quite well. But this signal is a very tame version of the actual error signal you would see if you probed across the LTP inputs. The issue is that you are not really seeing the distortion that's there, because it is swamped out by the fundamental. 1% THD is 1/100 distortion. It is not easy to see 1% THD on a sine wave, but it's there and it matters. The D() function in LTSpice helps greatly with visually assessing distortion and tracing it to its source.

Conceptually, assessing this distortion visually helps us to mentally characterize the amp and understand it intuitively. However the problem is there is no sense of scale. The error is too small to see, yet still too large. With a better output stage, you will be able to use less transistors and less power for the same level of distortion.

I realize that for a hobbyist, economy comes second to fun, and this is a great way to enjoy one's time and learn about circuits. Each circuit is unique; from an experimenter's perspective there is not necessarily a need to compare them and fixate on one being the "best". When one does this they tend to narrow their line of sight and can become trapped in linear thinking. Finding the "best" is appropriate when there is a goal and deadline, and a clear design objective. I think you should decide whether you want to experiment or to design. Depending on your answer, this thread may not offer the type of learning experience you are after.
 
the -ve rail will inject all the rail rubbish into the -IN pin and it will get amplified to come out of the OUT PIN.

I'm aware of this - luckily in real life I was using a single rail supply so it was really going to ground. The image shows a split rail design because I quickly converted another circuit into what was on the breadboard and forgot to make it single rail.

Out of interest, if just grounding that point, how do you ensure the capacitor never gets reverse biased by any residual offset? Does it just not matter with the usual values of DC on the output?
 
I'm with Kean on this, CFP outputs produce ugly transients when switching. They are great for drivers but not for outputs unless Class A.

Bootstraps need big capacitors though - uses space, and caps don't like heat. I still think SS current source is better in this case.

But maybe discrete is more trouble than it's worth, there are many chip amp options that put all the bits inside a nice package. Even an op-amp driving a pair of external EFs might be simpler and smaller.

I'm starting to think Class D is going to be better still.
 
I like contests, but I have been involved with some that don't turn out well. It is a good idea to agree on the rules before too much work gets started. It is also good to establish the deadline date up front too.


Contest: Linear Power Amp in a mint tin (class Aa, class AB, or class B) ......I'm starting to think Class D is going to be better still.

Is this contest restricted to just class Aa, class AB, or class B, or are class D designs allowed? What about class G and class H since these are linear amps with fancy power supplies.

Amplifier is not reliant on SMD chips inside the mint tin.......We wouldn't want to hinder creativity and therefore SMD discrete parts are okay to use

So SMD discrete parts are OK, but no IC's at all, or no audio amp IC's?

Amplifier is not reliant on a digital signal processor
This implies that DSP's are not allowed. What about a dsPIC controller running the fancy power supply for a class H linear amp constructed with discrete SMD parts.

Transistors, power amp chips, and tubes are welcome in the contest
.

A true tube amp is not a contender here. Two 9 watt OPT's won't go into a mint tin. Some type of hybrid design may be possible, but highly unlikely.

I haven't designed a solid state audio amp in about 30 years. I would be interested if the deadline date was far enough out so that I have time to work on it. Sometimes I don't have time to look at this stuff for weeks. I'm wondering if some of the power supply modulation stuff I was doing at work for super efficient 1GHz RF amps could be modified for audio. It would be different.
 
What about a TDA1521
http://www.datasheetcatalog.org/datasheet/philips/TDA1521Q.pdf
Very few components. Room for a bottom fan with ventilation holes trough the bottom (Allowed?) each side. External symmetrical supply.
Yes, that's all okay. Auto-speed fan might be good. Philips has the reputation for very cool running class B car chips. I don't have any temperature information on the "compact stereo" chip that you've mentioned (because I haven't bought that chip). It is specifically designed to work within a small stereo and minimize heatsink size/costs. That seems good. It is a likely candidate.

Heat versus mute mode circuit: Sometimes the Philips datasheet advises the mute be hooked up directly to a power tap; however, that's wrong. When in doubt, try a 10k to 20k resistor (whatever resistor value is just enough to light an LED to dim). There's no need to get the inbuilt solid state switch fire hot with bias, even though that thing is amazingly sturdy.

Heat versus input load: With Philips single rail applications, input loads are typically from input side of the input cap to ground via 10k. Or you can simply use a 10k pot. In that case, I'd add to the load with a picofareds cap and a 250K resistor, which is just enough so that momentary contact failure of a dirty pot doesn't make extremely loud buzzes. Stiff input loads lower the heat.

If your Philips chip has "distortion detector" you simply "power it up" via 10k (nearby figures work) and that pin "grounds out" during clips and overloads. This can be very useful for some form of simplistic clipping reduction circuit added at input. Like an ordinary soft clipper, the options include mildly (in)effective or terrible sound. In this case, "mild" works brilliantly and can decrease unnecessary workload somewhat.

Heat versus power circuit: Large caps mounted directly on chips from Philips and ST is likely to inspire much heat and dull sound. With the Philips chips, 22u min to 220u max is the likely range of capacitance at/onto the chip and then cable to the DC power jack where you can locate larger capacitance. Given the very short distance, I would try Cat5 from chip to DC jack. *If single ended (non-BTL) the speaker return (speaker negative) would have to go to the DC jack, in this example, since that is the location of a cap large enough to deal with it. Although single rail is recommendable for ease, if you do happen to use a traditional linear split rail power supply, try a CRC and snub an ordinary bridge rectifier with 4 of 10n polyester dip caps to knock off power noise extras that would otherwise cause heat in the amplifier.

An example with the TDA8561Q (BTL)
Mute mode to V+ via 10K to 22K
DDD to V+ via 10K
Input side of the input caps to ground via 10K resistors or stereo pot
RF filtering for input
22u onto the chip from V+ to ground
2200u onto the DC jack from V+ to ground
$9 laptop pack plugged into the DC jack.
Could it be easier? I think not. :)

I've been exploring chips to spot likely candidates:

ST
TDA7292
TDA7265B
TDA7265 reportedly popular in hi-fi bi-amp
TDA7269
TDA7576 Fet
TDA7296 Fet, can be under-volted for cool running
TDA7295 Fet, can be under-volted for cool running

Philips/NXP
TDA1521Q
TDA1554Q can run cool
TDA8561Q can run cool
TDA1560Q (qualifies, but just barely)
TDA1562Q

All of these look possible, but I haven't tried them all.
With the more powerful chips, I suggest reducing the power supply voltage to lower the heat.
 
Last edited:
I like contests, but I have been involved with some that don't turn out well. It is a good idea to agree on the rules before too much work gets started.
There aren't rules and there aren't disqualifications; however, there are suggested parameters, votes, and competitors. Good design follows the suggested parameters and is most likely to win. The parameters should not be changed to any large extent; however we would not want to hinder creativity. The parameters cannot be altered after the first working entry that meets or betters the parameters, is built and displayed with a photo attachment.
Is this contest restricted to just class Aa, class AB, or class B, . . . What about class G and class H since these are linear amps with fancy power supplies . . .or are class D designs allowed?
Creativity is greatly encouraged. But, if you want to win, I would suggest at least 9W per channel non-switching stereo amplifier. If you want to do Class H, it seems recommendable to have the first 9W per channel in Class B. If you do "something else" that matter is up to the voters to discern it. Therefore I suggest that you avoid class D in any contest for non-switching amplifiers.
This implies that DSP's are not allowed. What about a dsPIC controller running the fancy power supply for a class H linear amp constructed with discrete SMD parts.
Let's check the weighting scale:
1). It exists (there are no votes for non-existent entries)
2). It promotes efficient linear (non-switching) power amp design
3). Creativity is encouraged (not hindered)
Regardless, it is entirely up to the voters to discern the details. Therefore, I suggest that your entry needs to be recognizable as doing "linear audio" for at the first 9 watts per channel stereo. You have creative freedom with the power supply. If you have additional input, please hurry, since the parameters are thoroughly locked at the first existent working entry that meets the parameters.

Like under-volting a more powerful chip amplifier drops the heat considerably, of course automated under-volting power circuit is one possible way to improve linear audio amp efficiency; however, those who elegantly make the amp run cooler regardless of simple power circuit may be even more competitive. I cannot predict which option would win. Can you do both?
A true tube amp is not a contender here. Two 9 watt OPT's won't go into a mint tin. Some type of hybrid design may be possible, but highly unlikely.
I believe it to be possible for Nuvistor and Subminiature-8 (and others) to be used for small signal input. These were used inside tiny radios, and so at least the possibility exists. In modern times, a tube can be used at the input to remove noise that annoys the ear, and in that task one tube can outperform equalizers and tone controls. This "integrated power amp function" is a creative possibility, but it isn't required.

Likewise, integrated receiver in a mint tin is another possible creative option, as is remote control. And unlikely never means impossible. :) Amplifier topology, power supply, integrated tuners, simple elegance and audiophile market harmonics are all much different skill sets. So, there are many different ways to surpass the minimum parameters. People with vastly different skill sets can all compete and it is up to the voters to sort that.
It is also good to establish the deadline date up front too.. . . I haven't designed a solid state audio amp in about 30 years. I would be interested if the deadline date was far enough out so that I have time to work on it. Sometimes I don't have time to look at this stuff for weeks. I'm wondering if some of the power supply modulation stuff I was doing at work for super efficient 1GHz RF amps could be modified for audio. It would be different.
Only existing amplifiers get votes. Creativity is encouraged. I was trying to decide between Christmas and New Year's. What do you suggest?

P.S.
I recently received an awesome construction tip. He said to begin the casework first before all the topology pondering starts. Then your hard work produces a finished amplifier instead of only a tangle.
 
shorted dc protection?

luckily in real life I was using a single rail supply so it was really going to ground. The image shows a split rail design because I quickly converted another circuit into what was on the breadboard and forgot to make it single rail. Out of interest, if just grounding that point, how do you ensure the capacitor never gets reverse biased by any residual offset? Does it just not matter with the usual values of DC on the output?
With a cool-running non-switching amplifier (this particular contest), and some vent holes in the enclosure, then I think you can use whatever capacitors you need.
Small DC (<100 mV) offset is not a problem. A diode can be placed in parallel to protect it for abnormal high-reverse DC.
But, that fix can cause worse problems. . .
In a Split Rail example, and perhaps the NFB cap with diode protection applied, the third order harmonics with diode misfire in practical use, plus the possibility of bouncing offset are all possible annoyances. I believe that a possible alleviant is a bat diode chain or a germanium diode chain. The bat diodes low capacitance means low noise, and the germanium (and led) diode's contribution is mainly 2nd order which isn't nearly as annoying as a standard silicon diode. These made-for-signal BAT/germanium types of approximately 0.3v switch-on can be chained 3 in series for approximately 0.9v switch-on, and the higher switch-on voltage helps to decrease the occurrence of bouncing offset.

The noise possibilities are the same as a very low current soft clipper. Hopefully the low noise diodes have suitable current tolerance or perhaps we should just buy a better cap?

For a single rail amplifier diode clipping a signal cap can huge surges of DC offset emitted into the speaker or an exploded amplifier. The cause of failure would have been a mystery except that I tried it several times, and it makes a lot of smoke.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.