TBP-Zero Class A: Research and build

Hi folks,

Not sure if this is the first BUILD thread, but I am sure there's very limited discussion about this power amp.

TBP-zero is a class A amp design without emitter resistors, fully-balanced architecture, and looks very complex to build. I've never build class A without emitter resistors, this is a new challenge.

Thanks to my friends, D.L. and Luke Chu, they help with simplifying the design, making it easier to build and more DIY-friendly. The fully-balanced symmetrical design is modified into single-ended, that means half of materials and cost!

The project will be open-sourced, and Eagle sch and .brd files would be published later for further discussion.

Please be noted that some details on the magazines will conflicted.

Any suggestions are welcomed. ;)
 

Attachments

  • TBP-Zero-clone-S_DC-srv_LIM.asc
    22.1 KB · Views: 181
  • Single.jpg
    Single.jpg
    96.3 KB · Views: 1,313
  • post-69184-0-34216000-1372316455.jpg
    post-69184-0-34216000-1372316455.jpg
    98.1 KB · Views: 1,285
  • FFT.jpg
    FFT.jpg
    74.2 KB · Views: 1,080
  • 22154273_1972378102778403_7300731344173834352_n.jpg
    22154273_1972378102778403_7300731344173834352_n.jpg
    70.7 KB · Views: 1,071
  • 22135729_1972375122778701_1497850650095998450_o.jpg
    22135729_1972375122778701_1497850650095998450_o.jpg
    79.9 KB · Views: 766
  • BRD1.jpg
    BRD1.jpg
    1.1 MB · Views: 649
  • tb_zero_int_01.jpg
    tb_zero_int_01.jpg
    23.8 KB · Views: 558
  • LT1.png
    LT1.png
    221.3 KB · Views: 1,203
Last edited:
Eagle boards and schematics are zipped.
Also with the MJ Radio magazine copies.
 

Attachments

  • 無線と実験 2013-09 P38.jpg
    無線と実験 2013-09 P38.jpg
    379.2 KB · Views: 444
  • 無線と実験 2013-09 P36.jpg
    無線と実験 2013-09 P36.jpg
    308.6 KB · Views: 457
  • 無線と実験 2013-09 P35.jpg
    無線と実験 2013-09 P35.jpg
    341.4 KB · Views: 487
  • 無線と実験 2013-08 P39.jpg
    無線と実験 2013-08 P39.jpg
    216.4 KB · Views: 484
  • 無線と実験 2013-08 P37.jpg
    無線と実験 2013-08 P37.jpg
    355.1 KB · Views: 436
  • 無線と実験 2013-08 P36.jpg
    無線と実験 2013-08 P36.jpg
    291.5 KB · Views: 461
  • 無線と実験 2013-08 P35.jpg
    無線と実験 2013-08 P35.jpg
    306.5 KB · Views: 551
  • 無線と実験 2013-09 P39.jpg
    無線と実験 2013-09 P39.jpg
    363.4 KB · Views: 420
  • TBP.zip
    115.8 KB · Views: 185
A kind of diamond buffer... with 1 ohm base stoppers on the outputs. It might actually be stable - base ballasting works if you use enough of it. Might need a few more ohms, might not. Matching better be good on the parallel sets regardless. And that includes thermal gradients on the heat sink. But hey - if you have runaway or current hogging, increase the base stoppers.
 
A kind of diamond buffer... with 1 ohm base stoppers on the outputs. It might actually be stable - base ballasting works if you use enough of it. Might need a few more ohms, might not. Matching better be good on the parallel sets regardless. And that includes thermal gradients on the heat sink. But hey - if you have runaway or current hogging, increase the base stoppers.

Good suggestions, thanks!
About the grounding, is this topology fine for the design?
 
The one and only
Joined 2001
Paid Member
A kind of diamond buffer... with 1 ohm base stoppers on the outputs. It might actually be stable - base ballasting works if you use enough of it. Might need a few more ohms, might not. Matching better be good on the parallel sets regardless. And that includes thermal gradients on the heat sink. But hey - if you have runaway or current hogging, increase the base stoppers.

Unfortunately at the point where base resistors help, they have an effect similar to Emitter resistors.
 
Hi

Built the JLH years ago with parallel output pairs for higher power. My friend matched them thermally using his fingers :)

In an EF losing the REs reduces z-out but then you have to make sure the devices are matched and share current dynamically. In a CE output stage, RE is tied to the rail so z-out is really low and you still have the benefit of RE to enforce sharing and for easy test-point access.

Some designers argue that in either case, RE should float on top of the load, providing a minimum z-out so z-out will remain positive regardless of signal current phase, back EMF, etc.
 
Well, that's what I called challenge!
This is a rare design to DIY hobbyists, I think. :D
Hi erikovsky, All,

I agree. It is a very interesting and a promising topology for DIY.
To help simulate this I have got an electrothermal version of the single ended version above running in LTspice. See attached file.
diyAudio_TBP-Zero-SE-cct.png

I found if the drivers Q12 and Q13 are mounted on a thermal spreader plate with the output transistors then most of the thermal runaway can be stopped.
The remainder of the thermal control is by mounting the spreader Q21 on the heatsink. The changes in idle current are 304mA at turn on, 331mA warmed up no signal and 294mA at 28Vpk (50W). So it looks like it can be safely if the thermal linages are done right.
diyAudio_TBP-Zero-SE-ElTh9-30s-plot.png

My version looks like only one pair but don't be fooled, there are 3 in parallel, using area scaling factor "m" for the BJT model.
The electrothermal version has 9 widgets which makes it very slow and hard to get it to start. The above plot took 43 minutes and 1.3GB of plot memory.
Q5 and Q6 are rearranged as current sources with 12V DC. The servo is reconfigured to drives the input stage source node resistor directly; I found that was necessary to get the electrothermal simulation to run (that is not needed in a real build). For the similar reason, the servo is powered by voltage sources.

BTW you can run the non-thermal files in the attached zip and they only take a few seconds to run for the THD etc.
BTW2 I think the "zero" in TPB-zero refers to no emitter resistors and not zero global fb.

For help on using the electrothermal 'widgets' see this post https://www.diyaudio.com/community/threads/better-power-mosfet-models-in-ltspice.266655/post-6885269 and used in the standard topology in this post https://www.diyaudio.com/community/threads/bob-cordells-power-amplifier-book.171159/post-6886335.
 

Attachments

  • diyAudio_TBP-Zero-org.zip
    39.4 KB · Views: 114
An autobias version of the TBP-zero-SE

After designing a wideband autobias diamond amplifier here https://www.diyaudio.com/community/...ching-auto-bias-power-amp.375141/post-6857523 I thought why not add autobias to the TBP-zero-SE (above post). Here it is:
diyAudio_TBP-Zero-MOS-diode-Autobias_cct.png

The MOSFETs are added to the emitters of the 3 pair of power transistors in parallel and these MOSFETs gates are wired back to drains after subtracting their threshold voltages which makes them act like Schottky diodes, except with a positive temperature coefficient (diodes normally have a negative temperature coefficient).

What's gained by this? Well, we now have a bias feedback loop so the bias spreader transistors do not need to be mounted on the heatsink and bias current is very stable when the power transistors try to thermally run away and so the thermal connection of drivers Q12 and Q13 is not so critical. We effectively have two levels of thermal control. The MOS-diodes have a low dynamic resistance (about 20mR) so effectively we still have no added series resistors in the emitters, so we can still call it a 'zero emitter resistance' amplifier. Here's the thermal response:
diyAudio_TBP-Zero-MOS-diode-Autobias-ElTh8-30s-plot.png

It starts cold at 376mA, reaches 376mA at idle, then running hot at 50W falls slightly to 344mA. And that's without the spreader Q17,Q18 on the heatsink! M1 and M3 are thermally linked with a thermal washer and freestanding on the PCB without any extra heatsinking (ditto M2 and M4). The thermal lag between these MOSFET pairs does not appear to be significant in terms of causing a short term rise in idle current but some bench tests are still needed to prove this. But it looks very promising from these sims.
 

Attachments

  • diyAudio_TBP-Zero-MOS-diode-Autobias.zip
    37.8 KB · Views: 99
Using cob cancellation circuit to reduce the distortion rate sharply
 

Attachments

  • A680.jpg
    A680.jpg
    382.2 KB · Views: 236
  • TBP A.jpg
    TBP A.jpg
    488.8 KB · Views: 215
  • TBP B.jpg
    TBP B.jpg
    127.1 KB · Views: 199
  • TBP C.jpg
    TBP C.jpg
    193.9 KB · Views: 189
  • TBP D.jpg
    TBP D.jpg
    238.3 KB · Views: 199
  • TBP E.jpg
    TBP E.jpg
    155.1 KB · Views: 213
  • TBP F.jpg
    TBP F.jpg
    161.2 KB · Views: 200
  • TBP G.jpg
    TBP G.jpg
    160.3 KB · Views: 183
  • TBP H.jpg
    TBP H.jpg
    97 KB · Views: 162
  • TBP I.jpg
    TBP I.jpg
    23.6 KB · Views: 167
  • TBP J.png
    TBP J.png
    8.4 KB · Views: 210
This is what you are trying to clone [TB website] ?
Hi Patrick,

Nice to hear from you again. I hope you are well and all the best for the new year.

My part is not cloning but applying electrothermal widgets using the LTspice file supplied in Post 1 (first attachment).
Post 336 here: One area where they are really helpful is output stages that do not use any emitter resistors or source resistors to check the thermal stability and bias loops in standard amplifiers. They are also useful for checking for current hogging of parallel devices.
My sims needed mods to the supplied file to get the widgets to run -- seems to be due to the number of nodes added by the subcircuit making the SPICE matrix a lot less sparse and the algorithms at some point can't find the dc operating point to get the trans running. So I had to make changes to the servo supply and any cross-rail bias resistors R51 and CCS J5. Also reducing the number of widgets from 13 to 9 by using area factor 'm' for the parallel power transistors. And sometimes also relaxing Reltol to get it run.

Despite these difficulties the circuit does run allowing safe experimenting with no emitter resistor output stages and trying different mounting arrangements of the power transistors drivers and spreader.

I hope it opens a new era for power amp design -- a bit like going from 2D to 3D with our power amp sims.

Also the effect of no emitter resistors versus using emitter resistors on the harmonics can be seen and possibly heard (if you play some music through your simulation and listen to the recorded file). You can do this test without the electrothermal widgets, fortunately!

Re: Is this Class-A?
I can't read the Japanese article so I don't know. The website link you supplied does not say what Class it is. The thread starter, erikovsky, says it is Class-A. But the sim file supplied for the singe ended "clone" runs with around 300mA idle current (total for 3 in parallel) which is Class-AB (with Class-A up to nearly 1W into 8 ohms).

I don't see any reason why it would have to be run in Class-A up to full power. Class-AB costs less and assuming sound quality is acceptable in Class-AB, then to me it is the best choice for this topology.

As an aside, converting a BTL to single ended I think the load should be halved to 4 ohms. The original version is 360W into 8 ohms and the SE version should be 180W into 4 ohms. The SE circuit supplied above is 8 ohms giving 90W.

@levinson mark. Thanks. I hope I have not offended anyone with my changes like the cob cancellation. Maybe it needs to be added back?

@analog_sa. You appreciation is much appreciated.