MT Audio Design 'Sep Yep'

Some Comments on the Design

This design is an attempt to solve the problems with single ended triode amplifier design. The main problems are a high output impedance (due to no feedback rather than SE design), high distortion and output transformer design problems.

The trick used is a distributed load with 25% at the cathode and 75% at the anode. This lowers the output impedance with nearly 50%, and the cathode load also introduces some local feedback, which lowers the distortion with 5-6 dB without introducing any other forms of distortion, because this feedback acts directly on the input and does not alter the load (except at very low frequencies where the load is not a perfect coil). This design does not alter the harmonic spectra as push-pull designs do, and it shall be nearly as linear as a PP design all the way up to full power. The used tube is one of the most linear there is (2x300B in parallel). Well, that's the pro's. Now the con's. The driver stage is very difficult to design because of the high swing needed from the driver (nearly 200V), and it has to deliver some current too. The other disadvantage to PP design is that the power supply is a lot more difficult to design, as the PSRR is very low.

I will build it as a three stage design with transformer coupling all the way and with split load in the output stage and driver stage. While testing some alternatives I checked out the possibilities to make a resistance loaded 300B driver stage (+/-90-100V swing) with B+ less than 500V. The result was surprisingly bad, it seems hard to make such a stage with ordinary tubes (6SN7, 5687) with distortion levels below 2-3%! Most 300B designs I have seen use such a driver stage and the 300B's are loaded at around 2.5-3kOhm, which gives about 5-6% 2nd harmonic at full power, then there is cancellation between driver and output stage. The result is probably THD around 3%, but the 3rd harmonic will be close and cancellation distortion will occur. With 6BX7 or 6BL7 it is possible to get below 2% THD with low current (3-4mA). If you must use resistance load, use the tubes at low current and low voltage. This shall give much lower distortion in the driver stage, but the problem will come when the 300B starts asking for more current on the grid. SRPP and mu-followers are interesting also, or why not use a cathode follower as driver, this must be a lot better than the ordinary common cathode driver.
I have had some trouble designing the driver stage, but when I found the curves for the 6BX7 on the Glass Ware (TubeCad) site the problems vanished. This is the tube to use as a driver! I have ordered four JJ(Tesla) 300B from Robert Losconi (good prices) and I will use them as output tubes (PSE) and each half of the 6BX7 shall drive one 300B using its own interstage transformer (1:1.125) with 25% split load. Using this constitution the transformer action is very good and the output impedance should be low enough to drive the 300B's into Class A2 without serious problems. The driver stage is able to swing well over +/- 200V with 0.2% 2nd harmonic distortion! The down-side of this driver stage is that the amplification factor is low, leaving +/- 60V swing to the input stage. I have decided to use two 6BL7 in parallel with interstage transformer (1:2.25). This input stage delivers the needed swing with 2nd harmonic distortion < 0.5%. The input sensitivity is around 1.5V RMS for full Class A1 power (16-17W).
It is possible to get good performance with simple circuits, but transformers must be used for normal B+ (400-450V). The transformers must also be of good quality.

I hope I can finish the project soon, as I think this looks promising. The present status of the project is that I have bought the tubes I need. Unfortunately the iron costs a lot of money, so this will have to wait for some time. Maybe I will be able to finish this project in the second part of 1999.

This design is only in the planning stage, and I can not guarantee that it actually works. Feel free to try it, and please inform me about the result.

Power Amplifier Stage PSE 300B

Power amplifier schematics

Power Supply

'Sep Yep' Power Supply

I found an interesting active ripple reduction method on the 'Griffon Valve Pages', and I use this as active regulators for the input and driver stages. I think this is an interesting method. Parallel regulators have a nice feel about them, these shall not cause as much distortion as serial regulators (remember that the PSU is in series with the output valves). In this design the driver and input stages are isolated from the output stage, as the second coil causes an almost constant current flow at the regulator tap.

The ripple reduction depends on the Gm. The regulator output impedance Zo = rp/mu = 1/Gm = 1/2.66(mA/V) = 376 Ohm. Zo is about ten times lower than the load resistor. This means that the ripple found in the load resistance will be reduced approximately 10 times. With this load the 6BX7 has 1.6% 2nd harmonic and 0.1% 3rd at 1V on the grid (SE Amp CAD result). The distortion from the ripple reduction ought to be minimal.

To calculate the B+ ripple caused by the alternating current drawn by the two 300B we use the following assumptions:
The 10H coil delivers a constant DC current of 160mA.
Energy is drawn from the capacitor bank at positive alternating current and delivered to the capacitors at negative current.
The formula used for ripple calculation:
Vr = (T x I) / C
Vr = ripple voltage
T = period time = 1 / f
f = frequency
I = current
C = capacitor value
With f = 20Hz, T is 50 ms.
I = load current (AC component) = +/- 120mA
If we can accept Vr to be 1% of the B+ at full 20Hz power (this never happens), C is calculated as follows:
C = (50ms x 120mA) / 4V = 1500uF
Normally such low frequencies does not exist in music and certainly not at this high level. With normal music the fundamental note will be at least 6dB down, which means a reduction of the ripple to 2V with a 1500uF capacitor bank. This is still an extreme case, so 1500uF seems to be a decent value, the ripple voltage will be well below 1V at normal music signals. The distortion caused by ripple voltage is in the worst case below 0.5%, and with normal music signals it should be negligible, but still there is room for improvements here.
It is essential that the capacitors are of good quality, and I plan to use Elna Cerafine types with polypropylene bypass capacitors. This is expensive, and to begin with I will use 2 x 470uF Elna Cerafine per channel with 10uF bypass capacitors.
The ripple voltage is a mean looking signal, which clearly will make life difficult for the output tubes, so the power supply is an extremely important part of single ended designs where the PSRR is low (1dB according to SE Amp CAD, but a little higher in reality because of the split load). The purpose of the split load is however not to compensate for ripple voltage, it is meant to operate only on input signals.
If the ripple reduction method used for the input and driver stages turns out well, I may implement this on the output stage too, but I need an extra filament supply for this and the used tube(s) must also be able to deliver quite a lot of power. I could of course use the extra 5V filament with a 300B as regulator tube, but it is a bit expensive (but so are Elna Cerafines) and the Gm is a bit low on the 300B. With a higher Gm tube (around 10) an almost perfect regulation could be made with capacitor value as small as 330uF. I have found an operating point at 390V that is actually a bit better than the one I have designed with here, a little less power of course but lower distortion (especially 3rd harmonic). I need a tube shunt regulator with Zo = 30 and capable of delivering at least 10mA with low distortion, this tube I have not found yet. A series regulated power supply could also be interesting to try sometime in the future. If you are interested in regulated PSU's, take a look at the
'J.C. Verdier site'.

In the July 1999 issue of
'Tube CAD Journal' the tube shunt regulator is discussed, take a look.

300B Bias Supply

I am not finished with the design of the Bias supply.

Load-line 6BX7 driver stage

6BX7 driver SE Amp CAD results

6BX7 driver stage loadline

The driver stage is made with two 6BX7 driving one 300B each using a 25% split load interstage transformer (Lundahl LL1660, 1:1.125). The distortion at full swing is VERY low.
This is a magnificent tube! It can actually swing well over +/-300V with less than 1% 2nd harmonic into 80 kOhm. With the required swing of +/- 200V (actually +/-185V)  it is capable of this into loads as low as 50 kOhm (even lower with more current). At -200V the output impedance is 1.9 kOhm (reduced to 690 Ohm by the split load), and the driver stage shall be driven IN PHASE with the output stage to use this low output impedance at 0V on the 300B. I have been told that the grid current at 0V is as high as 1mA for the 300B, and this requires a low output impedance in the driver stage. 1mA at -96V on the grid means that the input impedance is 67 kOhm with a 220 kOhm grid resistor. The split load output stage increases this into 130 kOhm per 300B. One 6BX7 could make this, but to be on the safe side I have chosen to use two, each with its own interstage transformer. The use of two transformers is mainly due to the fact that the transformer I have decided to use can only stand 10mA, but it also makes biasing simple. It makes it very easy to transform(!) the circuit into a push-pull stage, and I will try this to evaluate the advantages of SE and PP. The amplification factor in the driver stage is 3.1 (including the step-up transformer). This leaves the input stage with 60V for full Class A1 swing.

Load-line 6BL7 input stage

6BL7 input stage SE Amp CAD results

6BL7 input stage loadline

To make it a three stage amplifier I have chosen to use two 6BL7 tubes in parallel with an interstage transformer (Lundahl LL1660. 1:2.25). Two tubes are needed to get good low frequency response, the transformer is 33H and with the 5 kOhm output impedance of the 6BL7 the result is -3dB at 12Hz. The required swing is around +/- 30V and the 6BL7 makes this into quite low impedances with low distortion. The load impedance is (each 6BX7 has approximately 150k input impedance): (150 kOhm x 2.8) / 5 = 84 kOhm. The factor 2.8 is the increase in input impedance from the split load 6BX7 and the factor 5 is the reduction in the 1:2.25 step-up transformer. This stage is connected to be IN PHASE with the driver stage, I do not want any distortion cancellation between the stages as I think that is not a good design. As you can see from the SE Amp CAD results above the distortion is VERY low even with swing well over the required.
The amplification factor is 13.8 x 2.25 = 31, which means that the input sensitivity is around 1.5V RMS for full Class A1 output power.

Load-line PSE 300B

300B PSE output stage SE Amp CAD results
300B PSE output stage SE Amp CAD results at 1W

300B PSE output stage loadline

300B PSE output stage transfer function

The results from the SE Amp CAD above are all before cathode feedback.
The cathode load is 1/4 of the total load, which will cause a change of the amplification factor as follows:
A0=3.4, beta=1/4 -> A=3.4/(1+(0.25x3.4))=1.84 (5.3dB feedback)
This means that the driver stage must be able to swing around +/- 175 Volts for full class A1 power (17.4W) with THD at 2.3% without global feedback. Together with the cathode feedback (5.3dB) the estimated distortion at full power is around 1.2% 2nd order and well below 0.1% 3rd order without global feedback. The output impedance should be below 1 Ohm. Not bad for a single ended triode design. With distortion cancellation from a normal driver stage with 1% 2nd harmonic the distortion figures could be much better, but I think it is better to keep the harmonic spectra as it is, and cancellation between stages will also add some disharmonic distortion. Probably a better solution would be to make the output stage push-pull which is easy (alter some transformer connections and make the output stage 50% split load). I have tried to make this design with simple circuits, good linearity and good driving capability of each stage. Maybe the output stage could gain a little by higher load impedance and an even lower output impedance in the driver stage. The input and driver stages could perform even better with more current, but this is not possible with the interstage transformers I have chosen.

SE Amp CAD scenarios for WE300B

I have tried some load conditions for the WE300B using Glass Ware SE AMP CAD.

Below are eight scenarios with a single 300B at 425V, 80mA into different load impedances. The red on yellow is the one I intend to use.

Load imp.

5120 Ohm

5000 Ohm

4000 Ohm

3000 Ohm

2800 Ohm

2600 Ohm

2400 Ohm

2200 Ohm

Power output

8.31 W

8.44 W

9.63 W

11.0 W

11.3 W

11.6 W

11.8 W

12.0 W

Z out (8 Ohm)

1.63 Ohm

1.66 Ohm

1.87 Ohm

2.20 Ohm

2.28 Ohm

2.38 Ohm

2.49 Ohm

2.61 Ohm

2nd (dB)

-31.3 dB

-31.1 dB

-28.8 dB

-25.7 dB

-25.0 dB

-24.1 dB

-23.3 dB

-22.4 dB

3rd (dB)

-54.3 dB

-53.6 dB

-46.6 dB

-39.1 dB

-37.4 dB

-35.7 dB

-33.8 dB

-32.0 dB

2nd (%)

2.7 %

2.8 %

3.6 %

5.2 %

5.6 %

6.2 %

6.9 %

7.6 dB

3rd (%)

0.2 %

0.2 %

0.5 %

1.1 %

1.4 %

1.6 %

2.0 %

2.5 %

Note the rapid increase in 3rd harmonic at lower load impedance. The 5.12k load gives 15dB lower 3rd harmonic than the more commonly used 3k load! The power loss is only 1.2 dB. With 25% split load the distortion figures are reduced an extra 50%, and the output impedance is reduced with 50% too.

Below are four scenarios with 300B at 425V and different bias currents into a 5.12k load impedance. The red on yellow is the one I intend to use.


80 mA

70 mA

60 mA

50 mA

Power output

8.31 W

8.37 W

8.48 W

8.34 W

Z out (8 Ohm)

1.63 Ohm

1.70 Ohm

1.78 Ohm

1.89 Ohm

2nd (dB)

-31.3 dB

-29.9 dB

-27.9 dB

-25.3 dB

3rd (dB)

-54.3 dB

-48.9 dB

-43.2 dB

-36.5 dB

2nd (%)

2.7 %

3.2 %

4.0 %

5.4 %

3rd (%)

0.2 %

0.4 %

0.7 %

1.5 %

MT Audio Design Logo

This site uses frames, but if you have found this page via a search engine and do not see any frames, you can use the button below to get to the main frame of the MT Audio Design site.