bootstrapsCCS+T-TMC

[AndrewT]

would bsp122 be a suitable alternative?

 

[dadod]

Oh it is, think of the high input impedance of a mosfet and how it will affect loading on the LTP, on this forum there is only one other member I know of that has recently discovered it too, thinking he was the first, it was patented in late 1980s. Ive been using it for 15 years.

Soon D Self will have it in his book, I bet.

I use emitter follower before VAS(beta enhancer) and in this case I've got high impedance.
Even small MOSFET are with high imput capacitance(50-200pF) and this is not so good, loads LTP as well.
JLH used small MOSFET in his 80W mosfet amp(1984).

 

[homemodder]

Dadod, the mosfet has much higher impedance than a BJT. I hope you understand me correct, Im saying the beta enhancer should be a mosfet. Just put one in your sim and be surprised, no need for any other circuit modifications except when you tune for sound as a higher current through mosfet sounds a bit better. The much higher input impedance far outway the input capacitance.
I can give more info but it will have to wait till Monday.

Andrew it should work fine.

 

[jcx]

it doesn't matter since input Z of vas isn't limiting gain

the mirror output conductance is a limit, can be bootstrapped away

 

[dadod]

it doesn't matter since input Z of vas isn't limiting gain

the mirror output conductance is a limit, can be bootstrapped away

Could you elaborate it, preferably with circuit diagram?
dado

 

[homemodder]

it doesn't matter since input Z of vas isn't limiting gain

the mirror output conductance is a limit, can be bootstrapped away

Im not worried about the gain, the idea is to lower the vas output impedance, one can nearly halve the distortion by using a mosfet as beta enhancer if you choose the mosfet carefully.

JCX have you looked into lowering THD due to vas loading effects ?

Dadod try it out on your sim, then come tell if it drops THD or not.

 

[Bigun]

the beta enhancer adds gain within the VAS/cdom loop to make it more linear but you are still relying on cdom to close this local loop, for good or bad. I think you can enjoy the same distortion reduction by cascoding the VAS device.

 

[jcx]

the vas open loop gain depends on both source and load circuit Z too

there are some complicating issues, degen of the mirror raises the output Z

combining mosfet vas "beta enhancer" with degen in the mirror does show a region of increased vas open loop gain - basically single pole over audio

but bjt in the same position can give quite good (sim, ~140 dB) gain at audio and has a higher high frequency gain

at such high levels of low frequency gain I expect more parasitic limitations to show up


in sum - mosfet vs bjt "beta enhancer" - it depends on surrounding circuit properties but mosfet can show increased low frequency gain, bjt can be faster due to the better gm,C tradeoff

if you use Bob's 20 KHz THD as a metric I expect the bjt to give better numbers, mosfet could be better @ 1 KHz THD - again depending on other circuit properties - such as output stage current gain - just 2 Q "Darlington" output with bjt's low frequency input Z will completly swamp the potential added vas open loop low frequency gain of the mosfet "beta enhancer"

 

[dadod]

there are some complicating issues, degen of the mirror raises the output Z

combining mosfet vas "beta enhancer" with degen in the mirror does show a region of increased vas open loop gain - basically single pole over audio

but bjt in the same position can give quite good (sim, ~140 dB) gain at audio and has a higher high frequency gain

at such high levels of low frequency gain I expect more parasitic limitations to show up


in sum - mosfet vs bjt "beta enhancer" - it depends on surrounding circuit properties but mosfet can show increased low frequency gain, bjt can be faster due to the better gm,C tradeoff

if you use Bob's 20 KHz THD as a metric I expect the bjt to give better numbers, mosfet could be better @ 1 KHz THD - again depending on other circuit properties - such as output stage current gain - just 2 Q "Darlington" output with bjt's low frequency input Z will completly swamp the potential added vas open loop low frequency gain of the mosfet "beta enhancer"

I did quick test with mosfet "beta enhancer" and, really, mosfet was better at 1kHz(cca half distortion) but bjt was a bit better at 20kHz.
I wanted to eavaluate it more but I have problem with LTspice now. I bought a new PC with Winows 7 home edition 64bits and LTspice blocks PC after a few steps.
Does someone had similar problem and what is a solution??
dado

 

[dadod]

would bsp122 be a suitable alternative?

I am not sure, quite high imput capacitance. Could you send spice model, I can't find it?
dado

 

[AndrewT]

I don't have a Spice Model.
I just looked up Farnell and searched for jFETs, then downloaded the datasheets of candidates.

 

[homemodder]

the beta enhancer adds gain within the VAS/cdom loop to make it more linear but you are still relying on cdom to close this local loop, for good or bad. I think you can enjoy the same distortion reduction by cascoding the VAS device.

The cascode will only make a very limited THD improvement after the use of beta enhancer at the cost of another transistor, more complexity and not to mention the extra care youll have to take to keep the circuit stable. The mosfet can halve distortion especially below 5 khz, is not more complex and it disrupts the currents less in the LTP helping current balance.

 

[jaycee]

I've been playing around with this compensation scheme myself in LTSpice - looks like I was beaten to it! Very cool that our schematics are almost identical.

One thing I added was a CFP driver/predriver stage - a la Roender SymAsym. This seems to bring the benefits of a CFP output stage without the stability hassle.

 

[dadod]

I simulated BJT "beta enhancer" with BJT verse NJF(LSK170A). I can't find spice models for aby MOSFET suitable for "beta VAS enhacer".
So here are my result. Output votage wase set at 44V PP, 0.3V of the imput signal.
NJF current was set to 4.6Ma, lowest distortion.
Here is BJT:

Fourier components of V(vout)
DC component:-0.00101986

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.192e+01 1.000e+00 -0.13° 0.00°
2 2.000e+03 8.609e-06 3.928e-07 84.97° 85.10°
3 3.000e+03 8.780e-06 4.006e-07 16.62° 16.74°
4 4.000e+03 5.988e-07 2.732e-08 97.10° 97.23°
5 5.000e+03 8.081e-07 3.687e-08 14.21° 14.33°
6 6.000e+03 3.141e-07 1.433e-08 79.41° 79.54°
7 7.000e+03 9.456e-07 4.314e-08 21.98° 22.11°
8 8.000e+03 2.508e-07 1.144e-08 27.12° 27.25°
9 9.000e+03 7.327e-07 3.343e-08 22.47° 22.59°
Total Harmonic Distortion: 0.000057%

Fourier components of V(vout)
DC component:-0.00102965

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.190e+01 1.000e+00 -2.55° 0.00°
2 4.000e+04 3.509e-05 1.602e-06 -36.75° -34.20°
3 6.000e+04 5.646e-05 2.577e-06 129.50° 132.05°
4 8.000e+04 1.777e-05 8.115e-07 -49.03° -46.48°
5 1.000e+05 2.613e-05 1.193e-06 166.38° 168.93°
6 1.200e+05 1.910e-05 8.720e-07 -50.70° -48.15°
7 1.400e+05 4.381e-05 2.000e-06 165.29° 167.84°
8 1.600e+05 1.454e-05 6.639e-07 -54.00° -51.45°
9 1.800e+05 4.526e-05 2.066e-06 167.94° 170.48°
Total Harmonic Distortion: 0.000456%

Now NJF:
Fourier components of V(vout)
DC component:-0.0169394

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.192e+01 1.000e+00 -0.13° 0.00°
2 2.000e+03 8.969e-06 4.092e-07 91.49° 91.62°
3 3.000e+03 8.480e-06 3.869e-07 21.49° 21.62°
4 4.000e+03 4.864e-07 2.219e-08 99.09° 99.22°
5 5.000e+03 8.040e-07 3.668e-08 13.90° 14.03°
6 6.000e+03 2.983e-07 1.361e-08 83.31° 83.43°
7 7.000e+03 9.438e-07 4.306e-08 23.38° 23.51°
8 8.000e+03 2.410e-07 1.100e-08 26.28° 26.41°
9 9.000e+03 7.303e-07 3.332e-08 23.29° 23.42°
Total Harmonic Distortion: 0.000057%


Fourier components of V(vout)
DC component:-0.016956

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.190e+01 1.000e+00 -2.55° 0.00°
2 4.000e+04 2.614e-05 1.193e-06 -69.89° -67.35°
3 6.000e+04 7.047e-05 3.217e-06 126.34° 128.89°
4 8.000e+04 1.826e-05 8.336e-07 -48.57° -46.02°
5 1.000e+05 2.669e-05 1.219e-06 166.83° 169.38°
6 1.200e+05 1.898e-05 8.666e-07 -49.84° -47.29°
7 1.400e+05 4.495e-05 2.052e-06 165.45° 168.00°
8 1.600e+05 1.444e-05 6.593e-07 -52.05° -49.50°
9 1.800e+05 4.633e-05 2.115e-06 168.67° 171.22°
Total Harmonic Distortion: 0.000488%

There is no significant difference, so BJT is good enough.
For MOSFET if somone can find spice model I will do simulation.
dado

 

[dadod]

I've been playing around with this compensation scheme myself in LTSpice - looks like I was beaten to it! Very cool that our schematics are almost identical.

One thing I added was a CFP driver/predriver stage - a la Roender SymAsym. This seems to bring the benefits of a CFP output stage without the stability hassle.

Did you make real amp, and if yes, how it sounds.
You are using triple OPS as I undersood.
I allready started with triple and TT OPS and did same preliminary listening.
A simulation shows 5 to 10 times less distortion compare with the amp from this thread.
I finishing this project with enclosure, and I will show the result.
The sound I like so much.
dado

 

[Bigun]

The cascode will only make a very limited THD improvement after the use of beta enhancer at the cost of another transistor, more complexity and ...

I was suggesting to cascode the VAS and NOT use any beta enhancer. The cascode works just as well and may have other benefits as it nicely controls the operating conditions for the VAS device. Doesn't mean it will sound better, as it pretty much eliminates 2nd harmonic.

One thing I added was a CFP driver/predriver stage - a la Roender...

I can say it works well, I used this on my TGM3 amplifier. It isolates the VAS from the lumps and bumps of the output stage and yet the CFP stays operating in Class A (it doesn't work well in AB or B), and presents low drive impedance to the EF output pairs. I like it a lot.
http://www.diyaudio.com/forums/solid-state/167369-designing-tgm3-output-triples.html

 

[homemodder]

I was suggesting to cascode the VAS and NOT use any beta enhancer. The cascode works just as well and may have other benefits as it nicely controls the operating conditions for the VAS device. Doesn't mean it will sound better, as it pretty much eliminates 2nd harmonic.

But if you looking for measured performance the cascoded vas alone performs much worse than a beta enhancer. Remember here that the vas is loaded, and in this situation a cascoded vas basically doesnt do anything for linearity except if you use triple output stages, I think you have read D Self. The cascode causes the vas output impedance to increase and this is what we dont want, we want it to decrease to remove vas loading contribution. This is where the mosfet beta enhancer comes in, it doesnt have anything to do with gain, but the output impedance decrease.

 

[homemodder]

Dadod for models try, the zetex mosfets, all their models are available on their website. I used ostrippers ltspice file of his amp and using a mosfet I could easily halve even the THD of his super low TMC amp at 1Khz. At 20 Khz the the drop is smaller but this is very related to the part that one uses. Even using a large part like irf240 the distortion drops at 1Khz.

 

[Bigun]

...a cascoded vas basically doesnt do anything for linearity except if you use triple output stages...

Now that you mention it, I was only simulating the cascoded VAS with my TGM3 design, which uses a CFP+ERF (triple) output. In the end I didn't try it out as simulations indicated it killed the low order harmonics all too well...

 

[ostripper]

By Bigun- In the end I didn't try it out as simulations indicated it killed the low order harmonics all too well...

Is that a bad thing ? Is it something that can be heard ?

(below 1 ) is the "base" thread topic amp. Not bad - a PPM amp. Besides what a bootstrap does in feeding back EMF in a local loop back to the voltage stage , a bootstrapped amp will only slightly deviate from it's current sourced "cousin" - both will simulate nearly the same and just sound slightly different.

(below 2 ) is the "harmonic killer" I was surprised at it , even. With just half the loop gain , it blows the lin amp away. So I built one (I actually just needed a subwoofer amp). Hooking it to a full range speaker convinced me it had some real possibilities. Cascodes and the reduction of the early effect DO make for better sound , the simulations (first watt - class A) show the VAS itself is nearly distortionless , even compared to the LIN circuit. In both simulations ,the higher power FFT's just show the contributed Xover distortion differences between a EF2 and 3. The VAS itself, in the cascoded version, still shows much better linearity.

Both simulations were done with Cordell defined models (no OEM supplied ones) , 4R , 3 pair-mjl 1302/4281 op @ 85ma bias.

I did the "no - no" homemodder mentioned , and ran a cascoded VAS on a EF2 (for 6 month's). It's sound was good at low levels but quite poor at higher levels. All this fooling around paid off - have you ever tried to modulate the impedance of the cascodes with a bootstrap ? The downside of this route is the ever increasing complexity (more parts)... quite the compromise.

OS