Another Bass Amp: LNFB, Shunt, UL20%, PCFB

[roberto]

Hello everyone,

I would like to share with you a design (not yet finished) for a bass amp I'm planning to build in the next months.
The idea behind is to use as much as possible the local feedback and avoid global negative feedback, blending old technologies with new ones.

In attachment you can find the preliminary schematic and the loadlines with different percentages of feedback for UL operation.
The eq is not definitive, I would implement a Dumble-like eq with the switch to have James and TMB on the same amp.

On the second stage there's the classic LNFB found on some Dumbles. ...so it will be a Dumble? Nope.

The PI has a CCS on the tail instead of the classic resistor, and is connected to a negative rail.
To bring GU50s' grid positive and go into AB2, there's a Powerdrive system supplied on the negative side at around 3 times the bias voltage, and on the positive side at around maximum swing plus 50 V circa.
The current is chosen in order to have the right slew-rate at the entyre bandwidth.

Power amp is a sextet of GU50 supplied at 900V on the plates and 300V on the screens. g3 is not shown but it's positive at around +30V to square-up the curves and increase the ratio Ia/Ig2. Raa is 2k7.
I've chosen 20% Ultralinear connection because, from literature, this value ensures a big reduction in output impedance (around -80% depending on tubes) while still having "pentodeish" characteristics.
Another reason why I've chosen 20% is that with such a difference of voltage between plates and screens: with plates going down from 900 to 50V, screens go down from 300 to 130V.

Then there's a 10% shunt feedback from plates to grids on the output tubes, again from literature this will drop output impedance down to triode like values (between 10 and 20 times, EG EL84 in pentode is 38-40 kOhm and in triode is 2-3 kOhm).

On the output it is shown positive current feedback, that lowers the output impedance even more, but it will be a positive-to-negative current feedback, to adapt the output impedance to the cab and... daily preferences.

From simulations it will exceed 500 Wrms at 10% THD, that is mainly 2nd, 3rd and partially 5th harmonic.

I'm thinking if implement or not a serial loop before the PI, just to add further equalisation and a rack compressor.

Any suggestion to improve the design is very welcome.

Thanks
Roberto

[JazzGuitarGimp]

Nice! I am definitely following this one. What simulation app are you running, Roberto? Does it come with tubes in its library? And does it do a decent job of simulating tubes in real world applications?

Ciao,
Lou

[roberto]

Thanks Lou,

I will collect all the starting point I used to come to this design, in order to have a better idea of the project.

[roberto]

The design was originally based on:
- the Baby Huey by gingertube;
- the shunt negative feedback;
- "A new look at positive current feedback" ( https://www.americanradiohistory.com/Ar ... 957-11.pdf page 56);
- "A new approach to negative feedback design" ( https://www.audiofaidate.org/it/articol ... edback.pdf );
- this thread ( GU-50 Push-Pull AB OT Impedance/Load Line Questions );
- this thread ( Bass Power Amp with GU50 );
- this link for g3 ( http://www.audioxpress.com/assets/uploa ... et2903.pdf );
- and this link ( http://www.pmillett.com/tubebooks/tubed ... _tubes.pdf );

...I will post further data this evening.

[roberto]

What simulation app are you running, Roberto?

It's LT SPICE

Does it come with tubes in its library?

This is the GU50 model (without g3):

Code: Select all

*
*
* Generic pentode model: GU50
* Copyright 2003--2008 by Ayumi Nakabayashi, All rights reserved.
* Version 3.10, Generated on Thu Dec 23 17:30:30 2014
*               Plate
*               | Screen Grid
*               | |  Control Grid
*               | |  |  Cathode
*               | |  |  |
.SUBCKT GU50 A G2 G1 K
BGG   GG   0 V=V(G1,K)+1
BM1   M1   0 V=(0.10436355*(URAMP(V(G2,K))+1e-10))**-0.94152046
BM2   M2   0 V=(0.61437126*(URAMP(V(GG)+URAMP(V(G2,K))/3.6950521)))**2.4415205
BP    P    0 V=0.0013346082*(URAMP(V(GG)+URAMP(V(G2,K))/6.0143635))**1.5
BIK   IK   0 V=U(V(GG))*V(P)+(1-U(V(GG)))*0.00086083603*V(M1)*V(M2)
BIG   IG   0 V=0.000667**408*URAMP(V(G1,K))**1.5*(URAMP(V(G1,K))/(URAMP(V(A,K))+URAMP(V(G1,K)))*1.2+0.4)
BIK2  IK2  0 V=V(IK,IG)*(1-0.4*(EXP(-URAMP(V(A,K))/URAMP(V(G2,K))*15)-EXP(-15)))
BIG2T IG2T 0 V=V(IK2)*(0.952380952*(1-URAMP(V(A,K))/(URAMP(V(A,K))+10))**1.5+0.047619048)
BIK3  IK3  0 V=V(IK2)*(URAMP(V(A,K))+11100)/(URAMP(V(G2,K))+11100)
BIK4  IK4  0 V=V(IK3)-URAMP(V(IK3)-(0.0010136338*(URAMP(V(A,K))+URAMP(URAMP(V(G2,K))-URAMP(V(A,K))))**1.5))
BIP   IP   0 V=URAMP(V(IK4,IG2T)-URAMP(V(IK4,IG2T)-(0.0010136338*URAMP(V(A,K))**1.5)))
BIAK  A    K I=V(IP)+1e-10*V(A,K)
BIG2  G2   K I=URAMP(V(IK4,IP))
BIGK  G1   K I=V(IG)
* CAPS
CGA   G1  A  0.1p
CGK   G1  K  8.4p
C12   G1  G2 5.6p
CAK   A   K  9.2p
.ENDS

And this is how I modded 12ax7 model to fit 6N2P-EV values:

Code: Select all

*
* Roberto's note: 6N2P-EV model based on:
* Generic triode model: 12AX7
* Copyright 2003--2008 by Ayumi Nakabayashi, All rights reserved.
* Version 3.10, Generated on Sat Mar  8 22:41:09 2008
* Roberto's note: modified following http://www.radiotechnika.hu/images/6N2P-EV.pdf
* Roberto's note: input capacitance CGK = 2.35 pF
* Roberto's note: output capacitance CAK = 2.5 pF
* Roberto's note: transfer capacitance CGA = 0.55 pF
*               Plate
*               | Grid
*               | | Cathode
*               | | |
.SUBCKT 6N2P-EV A G K
BGG   GG   0 V=V(G,K)+0.59836683
BM1   M1   0 V=(0.0017172334*(URAMP(V(A,K))+1e-10))**-0.2685074
BM2   M2   0 V=(0.84817287*(URAMP(V(GG)+URAMP(V(A,K))/88.413802)+1e-10))**1.7685074
BP    P    0 V=0.001130216*(URAMP(V(GG)+URAMP(V(A,K))/104.24031)+1e-10)**1.5
BIK   IK   0 V=U(V(GG))*V(P)+(1-U(V(GG)))*0.00071211506*V(M1)*V(M2)
BIG   IG   0 V=0.000565108*URAMP(V(G,K))**1.5*(URAMP(V(G,K))/(URAMP(V(A,K))+URAMP(V(G,K)))*1.2+0.4)
BIAK  A    K I=URAMP(V(IK,IG)-URAMP(V(IK,IG)-(0.00058141055*URAMP(V(A,K))**1.5)))+1e-10*V(A,K)
BIGK  G    K I=V(IG)
* CAPS
CGA   G    A 0.55p
CGK   G    K 2.35p
CAK   A    K 2.5p
.ENDS

And does it do a decent job of simulating tubes in real world applications?

Do not expect them to fit when simulating a lead channel, but for a clean channel they are quite good (except GU50 for the variation of its curves based on g3 voltage).

[roberto]

Some more of the data used to come to the design (please note that SS part is not what I would use, just what I've found on LTSPICE):

Multiloop feedback:
https://jrossmacdonald.com/jrm/wp-conte ... edback.pdf
https://jrossmacdonald.com/jrm/wp-conte ... lifier.pdf

Combined feedback:
https://books.google.it/books?id=BGMnhl ... 4&lpg=PA54

Other integrations of new and old technologies:
http://www.bartola.co.uk/valves/2018/04 ... rd-part-i/
http://www.bartola.co.uk/valves/2013/03 ... de-driver/

The choice of the 10% shunt feedback:
http://www.mcmlv.org/Archive/TubeTheory ... 0Tubes.pdf
"The high power sensitivity of the beam power tube requires the value of only 10 per cent feedback to effect a loud-speaker damping equal to that obtained with class A1-operated low-impedance triodes."

Another kind of shunt feedback:
page 14 of this pdf: http://www.tubebooks.org/file_downloads/RCA_HiFi.pdf

Here the suggestion to raise g3 voltage:
https://www.diyaudio.com/forums/tubes-v ... ntode.html

Some articles of mid-1955 about UL by Langford-Smith, where he suggests caps between plates and screens, not usually implemented in UL design:
https://frank.pocnet.net/other/AWV_Radi ... _20_05.pdf
https://frank.pocnet.net/other/AWV_Radi ... _20_06.pdf
https://frank.pocnet.net/other/AWV_Radi ... _20_07.pdf

Three articles about CCS:
https://audioxpress.com/assets/upload/f ... 101_P1.pdf
https://audioxpress.com/assets/upload/f ... 101_P2.pdf
https://audioxpress.com/assets/upload/f ... s_0907.pdf
Last link contains what seems one of the best solutions, the single Vbe at page 59.

Hints about running in AB2:
https://music-electronics-forum.com/sho ... post278792
"1) Choose the Drain Voltage such that on the maximum positive signal swing on the gate and hence also on the source you have at least 25 to 30V left across the mosfet drain to source. This effectively minimises any modulation of the device capacitance with the audio signal. Probably not critical in a MI Amp but quite noticeable in the smoothness and detail of the top end in a HiFi Amp. No point in going higher on the drain voltage, all that does is increase the power dissipation in the mosfet.
2) Don't use huge power mosfets, I use ZVN0545A (600mW) when I can or a 2 to 5 Watt rated device when the 600mW is'nt enough.
3) Direct couple the mosfet source to the output tube grid with just the grid stop in the path
4) Apply the output tube bias to the mosfet gate.
5) Dont forget the protection zener (12 or 15V) between gate (zener cathode) and source (zener anode). Right on the mosfet gate and source pins.
6) DO use a gate stop on the Mosfet, 220 Ohms to 1K will do, Mosfets have lots more gm than tubes and accordingly are even more prone to parasitic oscillations because of it.
7) The source load resistor should return to a negative rail of at least 3 times the output tube bias voltage (rule of thumb).
8 ) If you want to go a little overboard (or if you are building a HIFi Amp) replace the source follower load resistors with current sources, they don't need to be particularly great current sources, the "Ring of Two" bipolar transistors works very well, a high beta, garden variety small signal BC548C for example on the bottom with an MJE340 above it for 300V withstand is what I routinely use. Th Current Source Loads do sound better than simple resistive loads but again that is my HiFi Amp experince talking. You may not notice the difference with a MI Amp.

OH! and do remember to fit the output tube screen resistors with the resistor body right up against the tube socket pin, just like you do for the grid stop. Screen resistors have a grid stop function as well as their screen dissipation limiting function.

This arrangement (source follower drive of the Output tube grid) give one huge benefit which is very often overlooked. Output tubes suffer from a fair amount of grid current noise (random grid currents). With the low impedance drive of a source follower, this noise current is shunted to signal ground and the amp is MUCH quieter. HiFi Amps I've built using this scheme have the "blackest" background of just about any amp I've ever heard, tube or SS."

For 20% ultralinear, from the original patent:
https://patents.google.com/patent/US2710312A/en
See parameters at 20% UL (for that specific tube):
https://patentimages.storage.googleapis ... page-1.png


To ensure the right slew rate for the Powerdrive:
I need 100 Vpp more reasonably at 50 kHz slewrate will be:
2 x pi x f x V = 2 x 3,14 x 50 kHz x 100 = 31,4 V/us

Three GU50s in parallel will have around 45 pF input capacitance.

C = i x dt / V
dV/dt = 31,4 V/us

then
45 pF = i / slewrate

so
i = 45 pF x 31,4 V/us = 1,41 mA

So it can be enough to have 470 Ohm on the CCS.
I will keep 390 Ohm and 1,7 mA.

[didit]

Hello -

Obviously big ideas project.

I've a few small thoughts:

• Bass amps of our traditional designs and my experience have larger coupling & cathode bypass caps widening the bottom end. Have you considered this?

• The GU50 is somewhat different beast than 6L6. Some quick reading suggests relatively fragile screens. Have you some modelled engineering or prototype testing to confirm 20% operating point will work?

Otherwise seems simultaneously quite ambitious and impressive, and I cannot see anything else to question.

Best .. Ian

[roberto]

• Bass amps of our traditional designs and my experience have larger coupling & cathode bypass caps widening the bottom end. Have you considered this?

Thanks Ian! I've checked the bandwidth and seems good, but for sure I will try bigger bypass caps on cathodes, while coupling should be enough but I will investigate as well. The reason why they are not so high, is that bass instrument usually has 2nd and 3rd harmonic that predominates even on fundamental.

• The GU50 is somewhat different beast than 6L6. Some quick reading suggests relatively fragile screens. Have you some modelled engineering or prototype testing to confirm 20% operating point will work?

I will post the plots of anode vs screen voltages for secondary emission. On top of that, increasing g3 seems to reduce the screen current vs anode current, so it should improve even more the situation.

Otherwise seems simultaneously quite ambitious and impressive, and I cannot see anything else to question.

Thanks, it is absolutely ambitious for me, and I hope final results will be impressive as well.

Thanks again Ian!

[dorrisant]

Subscribed!

[roberto]

This is another version that will simplify the supply of the phase inverter.
Plate load is 2k4 and UL is always 20%, but now shunt is just 2%.

This is what I get from the simulation: it's more than 600 Wrms at 10% THD.

Code: Select all

Harmonic	Frequency	 Fourier 	Normalized	 Phase  	Normalized
 Number 	  [Hz]   	Component	 Component	[degree]	Phase [deg]
    1   	1.000e+03	9.804e+01	1.000e+00	    5.75°	    0.00°
    2   	2.000e+03	8.452e-01	8.621e-03	  -78.36°	  -84.11°
    3   	3.000e+03	6.901e+00	7.039e-02	   18.54°	   12.79°
    4   	4.000e+03	5.270e-02	5.376e-04	  -78.72°	  -84.47°
    5   	5.000e+03	5.878e+00	5.996e-02	 -148.24°	 -154.00°
    6   	6.000e+03	2.655e-01	2.709e-03	  133.24°	  127.49°
    7   	7.000e+03	2.019e+00	2.060e-02	 -131.16°	 -136.91°
    8   	8.000e+03	1.115e-01	1.137e-03	  154.20°	  148.45°
    9   	9.000e+03	7.825e-01	7.981e-03	   49.46°	   43.71°
Total Harmonic Distortion: 9.550485%(9.559542%)

[roberto]

Thanks to Tubelab I'm simulating another approach for shunt and local feedback:
local feedback is applied to g1 through a voltage divider from plates to ground, g2 is at 300V, g3 at +15V, anodes at 900V, cathode driven through a source follower.
One mosfet per tube, with independent bias and automatic through comparators: this comes handy with this kind of tubes that cannot be found matched.

This simplifies the power supply and the output transformer as well.