• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Looking to build a 25-30 WPC Push Pull amp - Need Help

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I need suggestions for a stereo push-pull amp producing about 25-30 WPC using something like a 6L6 or el34 (or something similar, I am flexible)

This will be my first amp build and I was looking for something relatively straightforward, not overly complex, and not a bank breaker.

I was looking at the Simple P-P, and would still love to build that, but I want something for this Fall and the Simple P-P looks like it's design may be delayed for a while.

Any help would be appreciated.

ps. One small feature I would like to add are VU meters for each channel, but not sure how that's accomplished.
 
Do you already have a preamp or is an "integrated" amp the thing you have in mind? If you have a preamp, does it have gain?

You said simple. IMO, that means 7591 "finals", as they are as easy to drive as "12" W. types (EL84s and 6V6s).

BTW, 7591s work into the same 6.6 KOhm primary trafos that 6L6s work into.
 
The only equipment I currently have is a Lafayette 224a integrated amp. Don't have a pre-amp.
Can you use the 224a as a preamp or is that not a good idea?

I suppose if I had to build both a preamp and an amp, then maybe building and integrated amp is the way to go. Makes it more complex, I imagine.

I am willing to hear advice. Nothing is set in stone.
 
hey-Hey!!!,
Try an E-Linear circuit. No loop NFB used. It is a long-tail-par input stage of pentodes with the plate loads connected to the screen taps of a U-L output TX instead of a separate, decoupled supply. 6AU6 work well, as do many other small-signal pentodes. I attached a pdf drawing of the amplifier stage.
cheers,
Douglas
 

Attachments

  • 6au6_st70.pdf
    60.3 KB · Views: 2,349
83stang said:
That's a sweet easy circuit how does that preform? Does it use the same power supply at the ST70? or is it different? I'm in the same boat as you SigloOne glad you started this thread!

hey-Hey!!!,
It can( and should ) use the rest of the St.70 power supply and chassis. As to 'how it sounds?', the answer is, GREAT!. I built a Mk.III pair up with this circuit using the 6AC7 and most recently my pair of W6m using nearly the same parts. The 6AU6 is also excellent, but the 6AC7 delivers a more sensitive amp.

On the power supply, put a small cap right after the rectifier, then the choke, then parallel everything else after the choke as all the power and signal current goes to a single node( the OPT's center tap ).

The dropping R to the screen grids is about 100k, and you can reduce the g2 bypass cap to about 0.1 uF( it's PP so one increases; the other decreases, it is the difference that needs bypassing ).
cheers,
Douglas
 
Doug's "exolinear" connection is a clever implementation of "plate to plate" short loop NFB.

Budget O/P "iron", such as Edcor and Hammond, badly needs (IMO) the linearizing help a few dB. of global NFB provide. Expensive, "boutique", trafos, with superior bandwidth and linearity, are in order, for circuitry that omits a global NFB loop. Keep in mind the fact that David Hafler expected the now well regarded O/P trafos he designed at Dynaco to be used in combination with global NFB. Dyna "iron", both OEM and clone, is better than Hammond/Edcor.

For net HD spectrum reasons, I like the 12AT7 in PP amps. Doug's circuit can be implemented using cascodes, with 'T7 sections "downstairs" and DN2540N3 MOSFETs "upstairs", instead of pentodes, in the LTP. Set the tail CCS for 6 mA.

Adding global NFB is easy. Lift the non-inverting grid from ground and apply the global feedback to that grid. You end up with an "exolinear" "El Cheapo". ;) Say Doug, would peaking inductance as part of the LTP loads upset operation of the "plate to plate" NFB?

BTW, it's a very good idea for LTP B- to rise after B+ rises. If B- is applied before B+ the grid will be positive with respect to the cathode and tube damage can be expected.
 
Eli Duttman said:
Doug's "exolinear" connection is a clever implementation of "plate to plate" short loop NFB.

Budget O/P "iron", such as Edcor and Hammond, badly needs (IMO) the linearizing help a few dB. of global NFB provide. Expensive, "boutique", trafos, with superior bandwidth and linearity, are in order, for circuitry that omits a global NFB loop. Keep in mind the fact that David Hafler expected the now well regarded O/P trafos he designed at Dynaco to be used in combination with global NFB. Dyna "iron", both OEM and clone, is better than Hammond/Edcor.

For net HD spectrum reasons, I like the 12AT7 in PP amps. Doug's circuit can be implemented using cascodes, with 'T7 sections "downstairs" and DN2540N3 MOSFETs "upstairs", instead of pentodes, in the LTP. Set the tail CCS for 6 mA.

Adding global NFB is easy. Lift the non-inverting grid from ground and apply the global feedback to that grid. You end up with an "exolinear" "El Cheapo". ;) Say Doug, would peaking inductance as part of the LTP loads upset operation of the "plate to plate" NFB?

BTW, it's a very good idea for LTP B- to rise after B+ rises. If B- is applied before B+ the grid will be positive with respect to the cathode and tube damage can be expected.

couple of things...
12AT7 is too high plate Z to go cascode; it will eat up too much headroom. Type 5687 is very good, as is type 6H6Pi.

It is not a good idea to have B+ before or after the negative rail. If it comes up first, it pulls current from the B+ caps( OK if they're film, not a good idea if they're 'lytics ). Same issue if it comes up after, B+ charges the negative rail reverse polarity. BAD for 'lytics. Put a 18V, 500 mW zener from ground to cathode( banded end to cathode). Then the cathode never goes much below 0.7V negative and poses little risk to other lytics.

E-Linear, E for voltage, and Emotive Audio where it was first developed and named....and it is plate-to-grid around the output stage.
cheers,
Douglas
 
Bandersnatch said:
E-Linear, E for voltage, and Emotive Audio where it was first developed and named....and it is plate-to-grid around the output stage.
I know I'm being anal and evil here, but the Pièce de résistance of the "E-Linear" circuit, being anode to grid feedback in the output stage, is described in detail on pp. 332-334 of Langford-Smith's Radiotron Designer's Handbook, 4th edition. Incidentally, that was published in 1952. It is further described in more approachable language in John Broskie's TubeCad Journal in March 2001 - including walk-throughs of how to calculate values and providing practical examples of its use. It seems the innovation is that feedback to the grid is taken as a from a tap on the output transformer as a ratio of the anode voltage rather than from the anode directly.

It's not a personal attack on anyone, just saying that there's very little new under the sun.

As an aside, I would be concerned about the circuit posted in post no. 4 being able to operate in anything but strict and complete class A. The differential pair phase splitter is dependent on identical loading of both valves. However, as the output stage enters class B, one of the output valves is driven into cutoff, where its gain is approximately zero (further negative grid voltage doesn't generate additional positive anode voltage, and in UL, the screen is at a constant ratio of anode voltage).

As I see it, in this state, the loading of the half of the differential pair driving the output valve in cutoff would see the full 'open loop' load impedance looking into the output stage grid, while the other half driving the output into saturation will be seeing the intended load impedance, being reduced by the feedback. In this situation the loading of the differential pair will differ by a factor of approximately the gain of the output stage, which is significant, and would lead to severely asymmetrical drive and premature clipping.
 
Bandersnatch I'm a little confused with the bias one, two and negative bias tap. are they connected together to the bias tap of the power supply? and the t1 and t2 taps are they connected to the transformer? or to somewhere else? what kind of power can you derive from this circuit? sorry about the questions, just intrigued that's all!
 
The grids of the 6AU6s are referenced to ground, so the CCS is returned to the C- (bias) supply to give the CCS a bit more voltage in which to breathe. Otherwise, it would only have the Vgk of the 6AU6 to work with, which is a couple of volts at most.

T1 and T2 look like test points to sense the current through the output valves. By Ohm's Law, the current through them creates a voltage over the 10 ohm cathode resistors, so we can use that voltage to indirectly measure that current.
 
audiousername said:

I know I'm being anal and evil here, but the Pièce de résistance of the "E-Linear" circuit, being anode to grid feedback in the output stage, is described in detail on pp. 332-334 of Langford-Smith's Radiotron Designer's Handbook, 4th edition. Incidentally, that was published in 1952. It is further described in more approachable language in John Broskie's TubeCad Journal in March 2001 - including walk-throughs of how to calculate values and providing practical examples of its use. It seems the innovation is that feedback to the grid is taken as a from a tap on the output transformer as a ratio of the anode voltage rather than from the anode directly.

It's not a personal attack on anyone, just saying that there's very little new under the sun.

As an aside, I would be concerned about the circuit posted in post no. 4 being able to operate in anything but strict and complete class A. The differential pair phase splitter is dependent on identical loading of both valves. However, as the output stage enters class B, one of the output valves is driven into cutoff, where its gain is approximately zero (further negative grid voltage doesn't generate additional positive anode voltage, and in UL, the screen is at a constant ratio of anode voltage).

As I see it, in this state, the loading of the half of the differential pair driving the output valve in cutoff would see the full 'open loop' load impedance looking into the output stage grid, while the other half driving the output into saturation will be seeing the intended load impedance, being reduced by the feedback. In this situation the loading of the differential pair will differ by a factor of approximately the gain of the output stage, which is significant, and would lead to severely asymmetrical drive and premature clipping.


You are quite correct about very little being new under the sun.

A "stiff" CCS in the LTP's tail forces symmetry between the 2 sides. Increasing current on either side is accompanied by a matching decrease on the other side, regardless of the loads being driven. Does that fact alter your calculus? The gain argument makes more sense to me, when the LTP tail is resistively loaded. Would switching to full pentode mode and applying a regulated 300 VDC to the EL34 screen grids improve the situation?
 
83stang said:
Bandersnatch I'm a little confused with the bias one, two and negative bias tap. are they connected together to the bias tap of the power supply? and the t1 and t2 taps are they connected to the transformer? or to somewhere else? what kind of power can you derive from this circuit? sorry about the questions, just intrigued that's all!

Each power tube gets its own bias adjustment. The grid resistors( 91k or so IIRC ) get attached to the wiper of the pot that is connected across the negative supply and ground. We'll use this same negative rail to give the CCS some headroom. The T1&2 are indeed bias current measurement points as described by Audiousername.

As to power, FB does not in any way alter the output power. In this case, the delivered grid drive can be altered( and reduced ) past the point of power reduction if too high a tap percentage is used for E-Linear connection.

I have preferred to use 20% taps for the front end, and 30-40% for the power final's g2( depends on what power tube ). That needs custom output TX winding and coil geometry to support such choices. Installing a tap at the end of a layer is easiest. For example the Peerless S-265-Q with its 20 layer primary( 10 on each side of the 10k a-a load CT )can have taps installed in 10% increments. The S-271-S has only 16 layers of primary so 12.5% is its minimum increment. The Dynaco outputs are slightly different. Consult your winder...mine's Heyboer.
cheers,
Douglas
 
audiousername said:

I know I'm being anal and evil here, but the Pièce de résistance of the "E-Linear" circuit, being anode to grid feedback in the output stage, is described in detail on pp. 332-334 of Langford-Smith's Radiotron Designer's Handbook, 4th edition. Incidentally, that was published in 1952. It is further described in more approachable language in John Broskie's TubeCad Journal in March 2001 - including walk-throughs of how to calculate values and providing practical examples of its use. It seems the innovation is that feedback to the grid is taken as a from a tap on the output transformer as a ratio of the anode voltage rather than from the anode directly.

It's not a personal attack on anyone, just saying that there's very little new under the sun.

As an aside, I would be concerned about the circuit posted in post no. 4 being able to operate in anything but strict and complete class A. The differential pair phase splitter is dependent on identical loading of both valves. However, as the output stage enters class B, one of the output valves is driven into cutoff, where its gain is approximately zero (further negative grid voltage doesn't generate additional positive anode voltage, and in UL, the screen is at a constant ratio of anode voltage).

As I see it, in this state, the loading of the half of the differential pair driving the output valve in cutoff would see the full 'open loop' load impedance looking into the output stage grid, while the other half driving the output into saturation will be seeing the intended load impedance, being reduced by the feedback. In this situation the loading of the differential pair will differ by a factor of approximately the gain of the output stage, which is significant, and would lead to severely asymmetrical drive and premature clipping.

Just a few points, the output stage cannot 'enter' class B. It is either biased there in the beginning or it isn't. Just in the same fashion that it cannot 'leave' class A'; it is either there in the idle/zero signal, or it is not.

Any circuit is going to behave more linearly in Class A bias. I took this as an experiment to employ it on an AB1 amp. That happened to be a Dynaco Mk.III and it worked quite well. Previously I had only tried it on Class A biased amps. The LTP needs high output Z, and in that case the loading does not vary much at all. If the LTP output Z goes low, as in that of triodes, things can and have done less well than expected. Cascodes and pentodes seem to be high enough.

Langford Smith speaks of the typical Plate-to-grid implementations and goes through several permutations of that in order to deliver some fraction of the final's plate signal back to the preceeding stage. The use of the U-L screen tap is the primary inovation. Small as it may be, it is new ground.
cheers,
Douglas
 
and it is plate-to-grid around the output stage

Darn tootin that "plate to plate" NFB is really plate to grid. Please notice that I use quotation marks.

what kind of power can you derive from this circuit?

This discussion is centered around small signal circuitry and NFB considerations. The "routine" expectation of 30+ WPC from a PP pair of EL34s continues to be valid.

12AT7 is too high plate Z to go cascode; it will eat up too much headroom. Type 5687 is very good, as is type 6H6Pi.

The 5687/ECC99/6Н30П {6n30p} are (sic) too linear. The reason the 'T7 is so very good in PP amps is its HD pattern, which is skewed towards 2nd order. When combined with the inherent cancellation of even order HD products of PP "finals", the net HD spectrum is an ear pleasing "waterfall" of 2nd > 3rd > 4th > 5th. :)

If "check valve" B+ decoupling is used, upwards of 450 V. will be available to a 'T7/FET cascode setup. If 27 KOhm load resistors are used in combination with an IB of 3 mA., approx. 370 V. are available to operate the cascode. That should be PLENTY.
 
Wieslaw Lipowsk said:
Ping "Bandersnatch" please >Try an E-Linear circuit

Would you please explaining to me the Bias circuit for the 6AU6 duo? Or, where I cac I find a suitable schematic with the regard of it?

Greetings

Wieslaw

hi Wieslaw,
The pair is biased with the current regulator. Grids are set to ground and the cathode goes to what ever voltage is required. The DN2450 MOSFET current regulator is very similar to those set out by Gary Pimm( google will get you to his page ).

On the 12AT7, use of that tube( on the bottom with a MOSFET on top) will need 150V across it, add 10V for the FET and we've gone through more than a third of the available B+. The pentode is decidedly preferable; it can swing below the g2 voltage. With the cascode, that upper gate voltage is about the limit...beyond that it's 'here be Monsters'.

For this one, we want linear, and low plate Z along with adequate gm to provide the sensitivity. The low plate Z reduces the required gate node voltage. for the 5687, or 6H6Pi this would be ~65V above the cathodes( for a pair of FQP2N60's ).

A pair of C3g with g2 at 75V ( still running the 33k plate loads and ~8 mA of plate and g2 current at idle ) is IMO far better. The *VERY* inexpensive 6AC7 works very well here.
cheers,
Douglas

attached is a schematic with the CCS used as a triode's plate load. ignore the output at the lower source, connect negative to the other end of the set resistor( now shown to the triode's plate) and the cathodes to the upper drain( now shown as B+)
 

Attachments

  • aklsamplifier.pdf
    47.1 KB · Views: 641
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.