Burning Amp BA-3

I think that was a typo. 72 watts dissipation is sizzynezzes even on a teflon frying pan. He goes on to say on it will dissipate approx 300 watts two chans. Indicating he really means 500mA per device.

They can take 800mA easily, given enough sink. Few probably have that for 6 deep. Diff between 400 and 600 mA in my old 3 deep build were significant.

But really. For comfort temps all year round for 20 years w/o babysitters, 6U500 is in order for 6 deep @ 500mA per device. Gianluca will build it, matter of cost.
 
Last edited:
  • Like
Reactions: 1 user
Member
Joined 2022
Paid Member
Not a typo. From the article:
“The 250 mA setting results in dissipation of about 6 watts per transistor for the output stage shown. If you have lots of heat sinking, you could consider twice as much bias, for 3 amps of bias per channel. This would be about 300 watts per chassis”.

Note: could.
 
Last edited:
  • Like
Reactions: 1 user
Member
Joined 2001
Paid Member
I think the "dissipate 300W two channels" in the BA-2 article is the typo here. You can bias the output stage at much lower current in the push-pull stage and still get the same output power. It makes no sense to me to run the push-pull stage at 3 or 3.5 amps of bias current. Most people build the push-pull versions (BA-2, F4, etc) for a good reason. That being said, the SE version sounds amazing if you know what you're getting yourself into and have a plan to deal with it.
 
  • Like
Reactions: 1 user
Hello BA-3 - builders,

the name of the BA-3 (also BA-1) is program: burning amplifier... :flame:
:hot:
More aluminum and good heattransfer helps.
Cheers
Dirk ;)
 

Attachments

  • BA-1 aluminum frame_heatsink_transforme2r.jpg
    BA-1 aluminum frame_heatsink_transforme2r.jpg
    91.8 KB · Views: 87
  • BA-1 moboblocks next to each other.jpg
    BA-1 moboblocks next to each other.jpg
    85 KB · Views: 87
I think the "dissipate 300W two channels" in the BA-2 article is the typo here. You can bias the output stage at much lower current in the push-pull stage and still get the same output power. It makes no sense to me to run the push-pull stage at 3 or 3.5 amps of bias current. Most people build the push-pull versions (BA-2, F4, etc) for a good reason. That being said, the SE version sounds amazing if you know what you're getting yourself into and have a plan to deal with it.
I am not sure of the actual intent. He recommends matching the output devices @ about 500mA, then recommends at least a 2:1 margin on the tranny, recommending 600vA. So I would say this points towards 500mA stock per device.

Plus, accepting that more Iq is not merrier, means I need to rethink the world :rofl:

That said, whatever original intent, it might still sound as good with less, I have no reason or knowledge to challenge that. I just always crank it to max.
 
Last edited:
Member
Joined 2001
Paid Member
The right amount of kill is..... over. Then you just do what you want. Reason I went with big monoblock chassis right from the beginning with 400VA transformers in each, in order to not limit myself with all these different varations and projects. My hardware is way overkill for this puny SissySIT R.3 in there now, being another P-P circuit at 1.7A. I had the BA-1 OS in these for a while with 3.2 amps of bias. The heatsinks were actually getting to do some real work at that point.
 
  • Like
Reactions: 1 user
Member
Joined 2019
Paid Member
FWIW - No typos that I see in the BA-2 article...

With 24V rails and 3A bias per side per channel => 72W dissipation per channel.
(Note - He's talking about 250mA across 1R with 12 devices => 6 pairs). 6W per device

Call that a "non-burning, burning amp"... and all of that is consistent. I started off there with my build.

Then he goes on to say...

"If you have lots of heat sinking, you could consider twice as much bias, for 3 amps of bias per channel. This would be about 300 watts per chassis"

That's also consistent. 2x72x2 => 288

The max I ran on mine with a 4U/500 was ~125W per side (250W total in the chassis)

At the time, I was running a 24V supply (ish) and 0R22 source resistors => 95-96mV across the source resistors. It was a burning amp.

A more tame, but wonderful setting (which when I measured it this morning it was set to) was 76mV across the source resistors... => Looks like I was shooting for 100W per side of dissipation. This was much more manageable.

Reason for all that...

Yanked out the front end, and I'm trying to figure out how to make it more modular to try with new front-ends and/or just run it as a big fat crippled F4 and mate it to my external PSU. The external PSU is now at 32V rails, and I wanted to bring the bias down to "null" before hooking it up to a 32V supply and rebiasing. W/O the FE, I don't have to worry about precious Toshiba JFETs. I don't need the voltage swing, but I don't want to swap the donuts in the PSU.

So... looks like initial settings will be ~4A Iq per channel with 32V rails and split the difference @ 125W per side of dissipation. I'll shoot for 70mV across the source resistors. I may start off a bit lower to see how it measures and then bring it up to see if it performs any better with higher rails / higher bias.

If I'm lucky, and it doesn't burst into flames, I may try some tunes with it by tomorrow.
 
  • Like
Reactions: 1 user
Member
Joined 2012
Paid Member
Higher bias current gives lower THD throughout the range of power output. So that is the advantage of higher bias current. The disadvantage of course is higher energy dissipation.

Higher rail voltage allows for higher power output. Output voltage cannot exceed the rail voltage.
 

Attachments

  • art_leave_classa.pdf
    1.9 MB · Views: 43
  • art_leaving_class_a_2019_redux.pdf
    1.2 MB · Views: 39
  • Like
Reactions: 1 users
Member
Joined 2023
Paid Member
Swapped out the 24V transformers with a pair of 300VA 18V Anteks. Sinks settled in around 55C. The rails are running 20V at idle, which seems a little low, but it's drawing a lot of power, so I suppose that's not unexpected. The sound is very nice - airy, natural, crisp. The bells in Henry Mancini's Baby Elephant Walk are remarkably bright and lifelike. The whole Hatari album sounds brand new. I think I'm going to like this amp a lot. It may be the best sounding out of the F5, F6, and Aleph J I built.
 
  • Like
Reactions: 3 users
Member
Joined 2019
Paid Member
Higher bias current gives lower THD throughout the range of power output. So that is the advantage of higher bias current. The disadvantage of course is higher energy dissipation.

Higher rail voltage allows for higher power output. Output voltage cannot exceed the rail voltage.
:nod:

Here is a quick and dirty run-through I did with the BA-2 output stage sans input stage confirming what we know. Still fun though. :)

This was with 24V rails (ish).
The notations for each plot are load and mV across 0R22 source resistors. So, if my math is correct (always suspect)...

20mV => ~1A1 Iq
100mV => ~5A4 Iq

The signal generator I was using only puts out 7V8rms max, so I didn't push it too hard.

Edited to add - What I also 'like' is that at the higher Iq, the distortion at 4R becomes closer to and then becomes lower than at 8R. For me, this is a good thing with my speakers.
 

Attachments

  • PWRTHD.jpg
    PWRTHD.jpg
    283 KB · Views: 46
Last edited:
  • Like
Reactions: 3 users
Member
Joined 2019
Paid Member
A few interesting (to me) observations. Note - Below is only wild thought / conjecture. I haven't done anything back-to-back with any kind of control.

The distortion of the BA-2 OS on its own is higher than I had expected. I haven't compared it to the F4 etc. I've only (not apples to apples) compared it to previous checks with the BA-3 IPS that I did when tweaking P3. My previous measurements are suspect at best, but they may be directional.

tl;dr - I was wondering why when removing an entire amplification stage (with no feedback) why my distortion (possibly) increased. I had assumed it would decrease. What do I know?

I don't think I've solved the riddle, but I'm going to fool with a few things when I get more time.

What I'm wondering is...

The distortion (that I get) from the BA-2 OS is 'positive 2nd' or close enough (see example attached). When I tweaked the BA-3 IPS, was I creating a negative 2nd, thus cancelling some of the positive 2nd from the OS? Nelson mentions in the article that he tweaked for the lowest distortion, and his is an order of magnitude better than my "bare" OS. My BA-3 IPS + BA-2 OS was more in-line with expectations. I've seen things like this (distortion 'cancelling') to be true between a pre-amp and amp, but I had never experienced it first hand with an adjustable IPS / OS. Now... I need to set up some things to see if it's all just nonsense, or if there's something to it. I also want to go back and try the 3-pair OS with higher bias. I'd imagine that the Iq per device is more of a factor for distortion than total Iq across all devices. I also want to learn how to measure the output impedance properly. I was 'chasing' a low(ish) output impedance with the build thinking it would work well with a few pairs of speakers. Who knows if I'm robbing Peter to pay Paul.

Insights and others' experience would be welcome. In the interim, I'm just fooling around and having fun.

Either way... thanks to Papa!!!!
 

Attachments

  • 4R 14W Resisdual.jpg
    4R 14W Resisdual.jpg
    226.2 KB · Views: 32
Member
Joined 2019
Paid Member
Well, I never said reading comprehension was my strong suit... it must rank near my short term memory, but I forgot.

What's exciting, is that although I thought I was achieving "odd" results with my BA-2 output stage when I first started screwing around with it along with the BA-3 FE, I think I'm actually getting nicely similar results to Papa.

Originally, I was reading the BA-3 distortion plots as if they were the distortion plots for the "BA-3 amplifier" i.e. the BA-3 FE + the BA-2 output stage.

After noodling a bit more with the BA-2 OS and wondering WTF was going so wrong and why my distortion was so high for the OS alone... I was wondering what the FE brought to the party in terms of distortion as it runs through the voltage range needed to achieve a certain power output. So, I ran it and reran it "like an amplifier" and told the software I was running an amp with an 8R load. Then, I reread the article AGAIN, and noticed that Nelson said...

"The data following are for a circuit with the parts and bias as shown and with P3 set for minimum distortion. Here's the distortion curve of the BA-3 front end vs output watts (as if it is driving a perfect follower output stage)"

Emphasis added is mine.

What's hilarious and perhaps kinda cool, is that I thought to do a similar graph before actually understanding what Nelson was putting in the article. Yes, I'm a dodo. It's well known by now.

Here's Nelson's version and mine below.

My guess re: the major difference in 'simulated' output power is that I set my 'bench' PSU to 24V0, and I was running Fairchild MOSFETs. In this case, the bench PSU was a standard linear PSU of common components seen around the threads. I was clipping right around 20Vp on the negative going waves. My guess is that Nelson was running Toshiba MOSFETs with lower Vgsth. My guess is that's also why his distortion was a bit better than mine. I don't remember whether I used LS JFETs or Toshiba for this FE, but I'll check. I'm going to compare the Fairchild MOSFETs to the Toshiba once I switch out the FEs and re-bias etc. etc. Nelson's plot also show a little lower distortion with 32V rails, so since that's where my 'real' PSU is currently sitting, that's encouraging.

1698609197781.png

And here's mine...
1698609243714.jpeg

Also for fun, I examined the residual at various P3 positions to see what's happening.

I was correct that on one end of the pot travel the distortion is negative 2nd, and on the opposite end of the pot travel, the distortion is positive 2nd. That will be more interesting to play with mated back to the 'positive 2nd' BA-2 OPS.

Here are a few FFTs and their associated distortion residuals at 1Vrms (ish) output. Note, no matter what I did, the distortion started to become 3rd dominant at higher outputs ... I didn't study exactly where it happened or run a bunch of different settings and plots, but when checking out what things looked like at higher (call it > 4Vrms or so) output everything was 3rd dominant. Further piddling is needed to check that out. Apologies that the graphs aren't labeled. File names should be indicative. Note, I did not tweak bias when P3 was adjusted.

If THD numbers are your thing, you can notice distortion can be altered by orders of magnitude with the turn of a knob.

With the pot fully one way...
Fully CCW FFT Fully CCW FFT

Roughly middle
Middle FFT Middle Residual

Fully the other direction
Full CW FFT Full CW Residual

Lowest distortion I could find in a few mins
Low Distortion FFT Low Distortion Residual

It's also important to note that all of this testing took place with a really poor grounding setup, everything strewn across the floor, and clip leads everywhere. Also, the residual DEFINITELY followed the mains ripple similar to what I had inquired about in a thread I started asking about "wandering" distortion. So, everything is likely directional, but things could have been set up a little better. I'll be curious to see how it performs from a measurement standpoint once I get it 'properly wired'.

I have another FE equipped with Toshiba MOSFETs (I'll check the JFETs), a better PSU, and some additional time. Now that I think I'm beginning to grasp what's happening, I'm ready to have more fun.

As always... old is new again.

Thank you @Nelson Pass!! This is far too fun.
 
  • Like
Reactions: 1 users