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.
To promote efficient linear (non-switching) amplifiers, here's "Linear power amp in a mint tin" contest. :D
Transistors, power amp chips, and tubes are welcome in the contest.

Here's suggested amplifier parameters:
At least 9W@10%THD (7.5W@1%THD) per channel to 8 ohms stereo amplifier
OR at least 20W@10%THD (16W@1%THD) to 8 ohms monobloc Parallel amplifier
Not hot running (not hot enough to harm capacitors)
Power amp that has speaker jacks, and input jack
Amplifier small signal is analog (doesn't convert signal to digital)
Not Class D

External power supplies, SMPS and Linear are okay
Extras for "integrated power amp" functions are okay
Small rubber feet added to the bottom of the mint tin is okay
Reasonably decorous vent holes are okay--ventilation is good
Reduced linearity at pitches below 44Hz is okay if done for headroom.
Linearity is not required for power output above 9W@10%THD per channel in the stereo amplifier.
Linearity is not required for power output above 20W@10%THD in the monophonic parallel amplifier.

Prizes are announced at the end of the contest. There are multiple judges, not selected until the end of the contest. A possible alternative judging method is by poll. The goal of the contest is a useful efficient linear (non-switching) amplifier that works well inside a mint tin and promotes efficient linear audio. Advanced linear classes, like AaH, AH, ABH, AaG, AG, ABG, etc, must do sine, square, triangle and other common tests at least as nicely as Class AB.

For judging, I suggest this:
1). It is built (there are no votes for non-existent entries)
2). It is a linear (non-switching) power amp that doesn't overheat
3). Creativity is encouraged
 
Last edited:
I think I could reconfigure this BTL 45 watt amp for stereo using horizontal mount ICs and lower profile components. The board would mount up to a heatsink with impeller type fan. The board in the pic is only 1-3/4" x 2-1/4".

I'm not sure if I could get it below 3/4" thickness (or even an 1") and not have thermal issues.

An externally hosted image should be here but it was not working when we last tested it.


Some of the car stereo power amp ICs need as few as three film caps to function, although I like to add RF filtering on the input. This may be another option. Power supply would be simpler as well.
 
Power is limited not only by space but by choice of outputs. Powerful outputs are too slow to be used at low bias. Something with FT>20 at 10mA, and Hfe linear to over 1A, in a TO-220 or TO-126 package. EF output would cause massive crossover distortion unless there was some sort of low-bias non-switching scheme. Another possibility is common-collector outputs.
 
Interesting challenge - looks like fun !

I believe we'll have to make the tin the basis of the heatsink and use an external PSU.

Big space users are capacitors so using few as possible is good.

I don't know anything about chip amps and I assume the solution is expected to be a chip amp given where you posted this challenge. But discrete amps can be pretty small.

Go Class AB for best tradeoff of sound quality against power dissipation. TO-220 LAT FETs for outputs - eliminates need for drivers and bias generator which cuts further down on parts. Single device input, feeds single device VAS. Load the VAS with a CCS (not a bootstrap that needs a large cap but a single JEFT CCS perhaps).
 
Bigun said:
Interesting challenge - looks like fun! I believe we'll have to make the tin the basis of the heatsink and use an external PSU. . . .
I don't know anything about chip amps and I assume the solution is expected to be a chip amp given where you posted this challenge. But discrete amps can be pretty small.
Thank you. Discrete amplifiers are welcome! Thanks for your input. Handmade amplifiers can be fine tuned for low heat. Tubes, Transistors, Chips and Hybrids are all welcome! A chip amp has the bias sealed inside and can be selected at time of purchase by either choosing a cool running chip or undervolting. And all of these options do exist.

So, I can't imagine any way to predict what or who will win.
But, here's some ideas. . .

The following seem good for useful and efficient:
Vent holes through the bottom and sides are okay, but not required. Drillings in the vertical portion of the lid are okay, but not required. Heat spreader bar is okay if inside the mint tin. Rubber feet and rubber-coated feet are okay and good for functional air intake vents in the bottom of the mint tin, and you're sure to notice that some input and output jacks have insulators that exceed by about 1/8"; so, some little feet on the bottom of the mint tin can help resolve that problem too. External power supply is okay. Headroom and current management accessory circuits are okay but not required. Integrated amplifier accessory parts, such as volume knob, tuner, etc. . . are okay but not required since the contest is for a linear power amp. Input jack(s), power jack, and speaker jacks seem to be required for function.

What is unlikely or less likely to win the contest:
Too hot for longevity
External fans/heatsink sticking out
Non-decorous vents in the top of the mint tin and/or extreme mutilation.
*The point and the fun challenge of the contest is efficient linear design that actually runs cool enough to work within the constraints of a mint tin.
Powerful outputs are too slow to be used at low bias.
Too much bias = swamping clumsy with bias = inefficient = too hot? Is that why this (attachment #2) is terrible? Apparently, some chips aren't suited for mint tins. :)

The chip shown in the attachment wasn't "plan a" anyway. :D

Since I can't vote for myself, I decided to try a few different chip options to learn what works and it wasn't this:
 

Attachments

  • MintTin.jpg
    MintTin.jpg
    66.3 KB · Views: 1,040
  • MintTin-STK459.jpg
    MintTin-STK459.jpg
    47.1 KB · Views: 1,982
do I read you right, you see using SMD as cheating? its against the rules to use SMD?
My apologies for the language error. That part was worded badly, but the idea of it was to block prefabs hastily shoved into mint tins. The #1 goal is promoting efficient linear design. We wouldn't want to hinder creativity and therefore SMD discrete parts are okay to use. I expect that any judges will be looking for successful efficiency as a primary concern. The size of the parts is not important if you can close the lid. :)
Can I use a fan? =)
Of course. However, some of your competition is probably going to be efficient design that doesn't require a non-decorous fan sticking out. So, if you want to win, try making a cool running amplifier. :) Just remember that the target goal is an efficient linear power amp.
 
Last edited:
Too much bias = swamping clumsy with bias = inefficient = too hot? Is that why this (attachment #2) is terrible? Apparently, some chips aren't suited for mint tins. :)

That's basically it. Our outputs need to be Hfe linear to 2A to drive a 6R load to 9W. What outputs are sufficient? Where is a 2A transistor with Ft>20 at 10mA... Perhaps we could find something in the <80V Vcemax region...

If we knew the thermal resistance to ambient of the altoids tin we could calculate our minimum bias. I would stand the tin up on its side with vertical vents, this way convection is maximized, and more surface area is in contact with the air.

How about someone rig up a 1W CCS with a DN2540 and test the temperature of the altoids tin after 30 minutes? This would get us the thermal info we need. Does an altoids tin have enough heatsinking power to run 9W?

I think it needs at least a +-15V PSU (30V single rail). It looks like an external PSU is mandatory unless there is magic 10W joule-thief, and a way to filter it well enough for audio.
 
I fear a micro fan to be compulsory.
Can you do an internal fan with decorous vents and thermo speed control?
Some of the car stereo power amp ICs need as few as three film caps to function, although I like to add RF filtering on the input. This may be another option. Power supply would be simpler as well.
That's a good idea. I don't actually want to win, but do want to raise the bar slightly, so perhaps I'll do something like a TDA1554Q with a $9 laptop cord for power. Those parts are somewhere in the desk and might as well be employed to make a nifty little Christmas present.

I'm thinking of trying a heat spreader bar thermo-pasted and pop-riveted to the bottom (inside) of the mint tin and vent holes drilled through the bottom of the mint tin. And, then little rubber coated feet elevate and allow those vents to work. Well, that should let some cool air in, and then the remaining task seems to be determining some fairly decorous way to let the warm air out.
keantoken said:
Does an altoids tin have enough heatsinking power to run 9W?
A 9w+9w (10%thd, 8R load) stereo amp of Class D or hyper-efficient Class B will run with a mint tin as the heatsink. After all, some Linear plug packs compete with switchmode plug packs for Class 5 efficiency.
So, it is possible.
If the efficiency percentage of 18 watts total system power goes into the speakers, the remaining percentage heats the mint tin. I think that the main problem is ventilation versus capacitor longevity.
keantoken said:
I think it needs at least a +-15V PSU (30V single rail). It looks like an external PSU. . .
External PSU looks good. If you could arrange for lower loss at the amplifier then it would run cooler and could use lower voltage.
keantoken said:
Where is a 2A transistor with Ft>20 at 10mA... Perhaps we could find something in the <80V Vcemax region...
Possibly Toshiba BJT? Maybe a Sanyo or Fairchild BJT? But, have you considered a FET design?
 
Last edited:
FETS are a good idea, and you can bootstrap the drivers to get full rail voltage swing. However at low bias currents the large transconductance will cause massive capacitive current spikes and I don't think that will sound good. BJTs can be biased low and still have reasonably low crossover distortion. I don't know, maybe there is a suitable FET pair.

The only pair I can think of for this is the 2SC4883A/A1859A from Sanken. Paralleling outputs would be a good idea but 4 outputs is already plenty to fit in an altoids tin with room to spare for lytics and such. FET input would allow us to save space on film caps.

Low-bias amplifier design is an art.
 
The problem with a CFP output stage is you have to keep the drivers from turning off. It can be done, but by that time there is little advantage over an ordinary EF.

I have finally found a page on what current dumping is really about:

Current Dumping Article

A current-dumping output stage is by nature a low-gain stage, which means we will have to make up for the gain in other stages, possibly making the amp more complex than it needs to be. Furthermore, the preceding stage will have to have its own quiescent in order to drive in the "off" region, making the whole arrangement somewhat redundant.

In terms of performance, I think an ordinary EF output will work best, .82R emitter resistors gives us 15mA output quiescent, making for 450mW output idle consumption. Can't go too low or emitter resistors cut into voltage swing. Current-dumping may get us lower but is it necessary to sacrifice the performance?
 
The problem with a CFP output stage is you have to keep the drivers from turning off. It can be done, but by that time there is little advantage over an ordinary EF.

Is this a problem? It seems fairly common to run CFP structures (including the driver) in Class-AB. You also have the advantage of the driver transistor having some control over Iq instead of relying on the tempurature dependant Vbe drop of the finals
 
Okay, but what is controlling the driver transistor's Iq? Variations in the driver's Iq will be multiplied in the output's Iq. Depending on the implementation this can be many times worse for a CFP. In a CFP the outputs' Hfe tempco is added to the driver tempco, and Hfe tempco is unique to each transistor type. All this must be designed around. Typically a CFP tempco is larger than an EF tempco for this reason, and much more tedious to compensate.

A CFP output stage could be used, but it would be a tedious hassle, a liability, and will be marginally better than an EF at best.

It would certainly be interesting, in the sense of a learning experience, but personally I think I've learnt most of this one already.

If you have seen all these issues overcome, then I would be eager to discuss the schematics you provide.
 
Okay, but what is controlling the driver transistor's Iq?

The way I understand it, the output of the CFP goes through an "emitter" resistor to the load, and the driver transistor samples the Iq via its voltage drop. Obviously the temperature will vary this to a degree, but importantly it's not thermally linked to the current drawn by the output transistors. If the output stage decides to draw more current, the driver will turn it off more. I'd imagine there's also a degree of self compensation - if the driver warms up and conducts more, the output transistor will conduct more in turn thus increasing Iq and turning off the driver in rough proportion.

If you parallel output devices then of course you would have "real" emitter resistors from emitter to rail on the big transistors.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.