Say Hello To "Sparky" the Parallel Chipamp

[lgreen]

Here is Sparky-- perhaps otherwise known as SPARCie or SPARCy or just SPARC for short.

A Brian GT based parallel LM4780 GC put into an old "lunchbox" case for a Sun SPARC IPC. You really cannot tell, but this amp is quite small, about the size of two mouser/digikey catalogs stacked one on the other.

Left Perspective, Rear View. Looks kind of like a Goldmund. Same binding posts too.

Lunchbox Open, Another Lunchbox view

Fan Power Supply

Isolated "live" Heatsinks

Boards with Thermal Switch The thermal switch flips when the heat on the heat sinks gets to 45C, which causes the red LED to illuminate (visible through the floppy disk drive opening). The airflow comes in the floppy disk drive opening and then flows out from the rear fan. There are also some small holes on the side that might let in a trickle of air. I suppose the LED is a warning that things are getting hot and to turn up the fans.

I tried to keep as much original stuff as I could but this was limited to some rear panel connectors (unused) and the light guide for the front panel. I also used all the original power connectors and the cage for the PS, which I had to cut down. Had I remembered that the floppy drive had a small opening for a light I would have tried to match up the red LED to it in a better way.

This is using a 330 VA transformer, rails at + 35 and -35 (incl. heatsink) and another PS for the fan. The fan speed can be adjusted from the rear panel by flipping the switch, and turning the potentiometer.

The power supply is snubberized just like the Brian GT schematic, and I did it point-to-point. And I threw in some Black Gate N type 4.7uF caps from a prior GC that I had removed a long time ago. This is using 2 (15,000 uF) caps on the base and the rooftop deck houses 8 (15,000 uF) caps, so I've got 75,000 uF per rail plus 4.7 + 220 on board the boards. I used the largest XFO and most capacitors that would fit. DC offset is less than 50mV.

This is not designed to be a world beater ideal amp, I just wanted to make something that looked neat and worked to get my motivation and confidence up after blowing up MOSFETs on my Aleph X caused me to give up DIY for a few months.

There will soon (a day or two) be a page up on my website (parttimeprojects, below), and power measurements.

[Mad_K]

Very cool Sorry about your Aleph-X. Looks like your're back in business

[preiter]

Nice! I looked at some of those old Sparc cases and thought about using one as a chassis. The sparc disc drives look good for a small amp.

[davidlzimmer]

VERY NICE!

Having just finnished one, I can certainly apreciate the planning involved.

I'd like to hear more about the thermal switch if you get the time.

[lgreen]

Mad_K- thanks. The pain from that has not yet dissipated, may be awhile before I go back to that project.

Prietier- the cases are plastic but note, there is a highly conductive (0.0 ohm) coating in the inside so be careful.

davidlzimmer- thanks for asking. The thermal switch is A Stancor "snap action" thermostat that I got at www.mouser.com. Digikey calls these bistable switches, so thanks to Jacco I was able to figure this out. The switches either open or close at a preset temperature, and go the other way once the temperature lowers by a certain amount. Usually, us DIYers put the AC inlet to the transformer primary through a normally closed switch of this type, and if the sink gets too hot (say 60 C) the switch opens protecting the amp. The switch closes once the temp drops to say 50 C and we are automatically protected.

Here I used this to simply connect an LED to indicate heat. The chassis was too complicated to run the AC through the switch, since the AC comes in from the top plate.

With no fan the sinks indicate 45C after about 15 min playing music at moderate levels, sinks don't get very warm if the music level is very low. With fan spinning slowly, I have been below 45C for 2 CDs. Thus, while this "parallel" design is able to get very hot, its heat is very highly related to output level.

These temp sensors make a clicking sound like a relay when they actuate. Kind of neat.

FYI- the web page for this is up, measurements over the weekend will be posted.
See LM4780 Web Page

[davidlzimmer]

Thanks! I saved that info. My next project will be a pair of parralled LM4780s. Will definitly concider a thermal switch and fans.

[lgreen]

OK, been checking out Sparky on the scope. I drove this to clipping into paralleled 100 watt 5 ohm 10% resistors. Hmmm....I also measured 2.5 ohms with the DMM. So I'm fairly certain its in the neighborhood of a 2.5 ohm load.


About 25 volts peak, 50 peak to peak, driving 1 channel only.
Clipping, 10V/Div, 2Khz, 2.5 ohms


So what does this give me for 2.5 ohms, about

P = ((25*.707)^2)/2.5 = 124 WPC RMS?

(at 5 ohms I got about 30 vk peak, 60 peak to peak at clipping, 89 watts?)

[davidlzimmer]

Just wondering. Is driving to clipping the excepted method of figuring PEAK output?

Do you know what the input level was. (or did I miss that)

Also, I thought there was a built in clip prtection in these chips.

[lgreen]

Originally posted by davidlzimmer
Just wondering. Is driving to clipping the excepted method of figuring PEAK output?

I would think that this would be maximum continuous output vs. peak since we are dealing with continuous stress to the amp versus "bursts" of signals. When you see measurements of peak power its usually a short burst followed by silence, followed by a short burst, maybe 10mSec tone and 100 mS silence or something lke that. You need a special signal/ automated setup for this.

Do you know what the input level was. (or did I miss that)

Didn't keep track of it. Note that my portable sony cd player could not achieve sufficient levels to drive this to clipping.

Also, I thought there was a built in clip prtection in these chips.

No, the protection is for safety by shutting down based on operating temperature. There is nothing in there to detect electrical characterisics such as distortion or clipping... I'm sure if it clips long enough we will see this circuitry engage but I did not run it at clipping very long.

[davidlzimmer]

Ok. So did you test for output just below clipping? Seems to me, that would be the figure you would want to know. The maximum undistorted output. Or is doing it this way close enough?

Sorry if I'm being a pest.

But, I'm learning! ''