ExtremA, class-A strikes back?

[pinnocchio]

Would be a pity if you have to abandon. I have build an ExtremA as well and it works great and sounds awesome. I followed the original BOM and haven't been able to do the "update"/correction. Had some problems when powering up the amp but that was solved with a small modification (that I posted here on diyaudio). Not able to do measurements for you since I am stationed abroad for a few years and did not bring my audio equipment.

Scanspeakman

I built mine with the newer changes that Sander had suggested... Then I guess those could be the issue since they were only simulated but never tested live. I will try to revert to the original values in the BoM and see if it works or not.

I'm glad to hear from someone that has one and really enjoys it. I've put so much time, money in the creation of this amplifier (the heavily modified chassis and transformers being the most expensive part) that I want to at least give it a fair chance before I flush the project.

Thanks for the encouragements.

Do

BTW, couldn't find your post about how you solved your issue.

 

[scanspeakman]

This is build thread:

http://www.diyaudio.com/forums/soli...-extrema-100watt-class-power-amplifier-3.html

I was under the impression I posted the modification on diyaudio but I cant find it. Here a link to a dutch diy audio website with the modification that solved the strange start up behaviour. A diode in series with one of the CRD's. This issue only occurs in the first version of the PCB.

forum.zelfbouwaudio.nl • Toon onderwerp - Nog een ExtremA

Scanspeakman

 

[pinnocchio]

This is build thread:

http://www.diyaudio.com/forums/soli...-extrema-100watt-class-power-amplifier-3.html

I was under the impression I posted the modification on diyaudio but I cant find it. Here a link to a dutch diy audio website with the modification that solved the strange start up behaviour. A diode in series with one of the CRD's. This issue only occurs in the first version of the PCB.

forum.zelfbouwaudio.nl • Toon onderwerp - Nog een ExtremA

Scanspeakman

Thanks for the information!

I got a free scope from a friend, it is a Iwatsu SS-5711 4 channels 100MHz. It is in pretty good condition. Don't know how good it is as a scope but I'm pretty sure it'll do the job properly.

Ciao!
Do

 

[pinnocchio]

Ok, I'm getting somewhere! I've removed one of the four 1 ohm emitter resistor of each quads. Result is 59 watts in Class A and much less heat but those emitter resistors are still getting pretty hot... So I've ordered some 5 watts 1 ohm non-inductive resistors. This is the last test before I scrap everything.

Do

 

[jacco vermeulen]

I've removed one of the four 1 ohm emitter resistor of each quads.

Single way to lower the quiescent current level, even wrote you so.

(ftr: I've warned at several threads/forums on this design that heat/dissipation is the biggest hurdle to take. I can do bearing and gearbox friction/dissipation calculations on the back of my hand, hvac calculations, even to what extent the sun will heat a single grain of silica, any thermal aspect of audio electronics is kiddy stuff. Those for who it's not, are wiser to read threads thoroughly before commencing)

 

[Jay]

(ftr: I've warned at several threads/forums on this design that heat/dissipation is the biggest hurdle to take. I can do bearing and gearbox friction/dissipation calculations on the back of my hand, hvac calculations, even to what extent the sun will heat a single grain of silica, any thermal aspect of audio electronics is kiddy stuff. Those for who it's not, are wiser to read threads thoroughly before commencing)

Jacco, when fixing a PS3 YLOD I found out that there are expensive kinds of thermal compound. Gold colored, injection method. Is this compound relevant or suitable for class A amplifiers?

 

[jacco vermeulen]

relevant

Nope

 

[pinnocchio]

Hi Jacco,

This is the reason why I did the change but keep in mind that I only found this thread way after I had started the project, before it was to late to back off ... I usually read threads before starting a project. If I had seen it, I would have never invested that heavily but would have done some other project.

That said, your explanations about thermal conductivity made total sense to me.

Thanks
Do

 

[jacco vermeulen]

So, don't let a stupid resistor discourage you.

 

[AndrewT]

You could extract all the expensive components and recycle them into another amplifier.
The PCB and resistors probably only account for less than 5% of the amplifier cost.

 

[Jay]

Nope

Okay, thanks. That's what I thought.

 

[SSassen]

pinnocchio. this amp has never been proven in the real world. the designer will have nothing to do with it. sander is long gone and wan't answer.
make the most out of it. drop the design and use the parts for something else.

I'm sorry? For whatever reason my subscription to this topic got removed, hence I got no notification emails sent. Rest assured I'm willing and able to help with debugging.

@Do,
Not sure why you didn't email me directly to tackle these issues you're seeing? You've done so previously.

 

[SSassen]

I have been looking on it. Bias current is ~5A giving 130Watt to dissipate in every sanken transistor. What where they thinking of!????

By shorting D11 and D12 it comes down to ~3.5A

By changing D5 and D8 to a schottky type (BAT42) it comes down to ~2.5A.

But Maybe R32,R33,R47 and R48 should be reduced to 470R when using BAT42 + shorting D11 and D12. But leave the 1K in the circuit for a start.

Bias current will drift all over the place as it the change with a rate of 2.3mV/K per diode/transistor...

So in original configuration it will start out at 5A and when the inside of the case reaches 50 degrees Celcius the bias will be reduced to ~4A.

Something that would boost an change in sound signature over time...

Wrong! Bias is 2.5A per transistor, which yields about 70W per Sanken device, which is about as high as we figured we could push it whilst keeping a close eye on reliability. As per the original article efficiency when driven at 100W/8R is around 40%.

 

[SSassen]

Ok, I'm getting somewhere! I've removed one of the four 1 ohm emitter resistor of each quads. Result is 59 watts in Class A and much less heat but those emitter resistors are still getting pretty hot... So I've ordered some 5 watts 1 ohm non-inductive resistors. This is the last test before I scrap everything.

Do

I see you're using an alu plate screwed to the heatsink. I hope there's thermal paste in between it and the heatsink, otherwise thermal resistance will be poor.

Also, did you use the recommended parts on the output stage, as due to its simplicity it is near impossible to not arrive at the proper bias current. The bias current is simply set by the emitter resistors whilst being governed by Ohm's law.

Another thing to check is whether you sourced genuine Sanken's? Other people that built this design have seen heat related issues that were due to using fake Sanken's. Did you buy these from an official Sanken distributor?

 

[pinnocchio]

HI Sander,

Yes, all parts are as per original BoM and sourced at Digikey and Mouser, no eBay for me.

I'm using a very thin layer of thermal paste in between the plate and heatsink and the transfer is actually very good (I've got tremendous pressure between the plate and heatsink due to all the screws). My heatsinks are just not big enough to handle the 100W per channel so by using 3x 1 ohm resistors, I'm running at 59 watts which seems to be ok (heatsink @ 60 Celcius) but a little hotter than I wanted.

I've just bought some new 1 ohm 5 watts 50ppm non-inductive resistors since the 3 watts are getting hot and making the bias drift a little.

Testing all this shortly and will report back

Thanks
Do

 

[sonnya]

Wrong! Bias is 2.5A per transistor, which yields about 70W per Sanken device, which is about as high as we figured we could push it whilst keeping a close eye on reliability. As per the original article efficiency when driven at 100W/8R is around 40%.

I must admit i was wrong. I was looking at the current limiter.

T29 and T30 sets the current. Still the bias current is subject to the temperature of T29 and T30 as it the change with a rate of 2.3mV/K per transistor

I have my amps running at 1 - 1.5 Amp per mosfet (Rated 250Watt) with +/-35Vdc rails which gives 35 - 53Watt per mosfet. Higher than that, solderjoints starts to deteriorate over time. (within a year)

BR

Sonny

 

[SSassen]

@Do,

As for the scaling, the following applies:

Obviously not everybody is willing to put up with an amplifier dissipating about 250W to enable 100W output in an 8-ohm load. In many cases about 50W would be sufficient, in which case the bias currents for the output stage and the supply voltages can be scaled accordingly. Suppose we would like to have 50W into a 6-ohm load the peak current will be 4.1A, which results in a bias current of 2.05A. As mentioned the bias current is set by the emitter resistors of the output transistors. The voltage drop across these four parallel resistors is 0.65V, hence they would need to be 0.32-ohm in value, or rather 1.2-ohm per resistor. The output voltage would need to be 25V, to be able to supply that voltage we need a little more than half that per supply rail, hence 14V would be sufficient. Both output stages will now dissipate 2x2.05Ax2x14V=115W, about half that of the 100W configuration.

If, in your case, you only reduced the bias current and not the supply voltage, you are still dissipating a lot of heat.

 

[SSassen]

I have my amps running at 1 - 1.5 Amp per mosfet (Rated 250Watt) with +/-35Vdc rails which gives 35 - 53Watt per mosfet. Higher than that, solderjoints starts to deteriorate over time. (within a year)

Transistor rating has little to do with it, it is all about die area and the package size. In this case the ring-emitter die of the Sanken transistors is huge by comparison to a 'normal' power transistor and the MT-200 package allows for a lot more heat being dissipated from a single transistor than a typical TO3P package for example.

 

[sonnya]

Transistor rating has little to do with it, it is all about die area and the package size. In this case the ring-emitter die of the Sanken transistors is huge by comparison to a 'normal' power transistor and the MT-200 package allows for a lot more heat being dissipated from a single transistor than a typical TO3P package for example.

Agree, surface area is very important.


Sent from my iPhone using Tapatalk

 

[pinnocchio]

Sander,

The mod you suggest to do by changing C1/C2 from 33pF to 68pF and C4/C9 from 1nF to 220pF. I had done those changes in the build but reverted when I had issues which I was not too sure about. Now that I have confirmed 100% that the emitter resistor heat was according to what Jacco was saying (heat generated from the emitter of the Sanken that travels through the copper traces) then I can continue the project but would like to know how well these changes were tested? Through simulation only or some users made the changes on their build? If only simulation, how confident are you that this change won't actually cause issues?

Thanks
Do