6SL7 SRPP / KT77 Class-A Push-Pull Tube Amplifier
6SL7 SRPP / KT77 SIPP Tube Amplifier - Ultra-Linear Class-A Push-Pull
Photograph 1: 6SL7 SRPP / KT77 Push-Pull DIY Tube Amp with External Power Supply
This project (Oddwatt 225) is a follow on to a previous tube amp project of mine. The Oddwatt, a ECC802S SRPP / EL84 (6BQ5) Push-Pull Tube Amp that worked so well that I thought I would pursue the concept a bit further. That's the great thing about diy projects, they never end. Something else that never ends is the caution you must exercise when building tube equipment. This amplifier uses dangerous voltages that have the potential to deliver a fatal shock. Please be careful. If you aren't experienced with such voltages, then perhaps this is not the project for you.
I had a few goals in mind when I started. Mostly, I wanted a bit more power. My main speakers (Dayton MTMs) are rather inefficient (87 dB/W). Neither my modified S-5 Electronics K-12M Tube Amplifier Kit with about 5 watts per channel nor the Oddwatt with around 10 per channel would deliver the levels of sound I wanted. There were also some modifications I wanted to try with the new amp. A few limits were put on the project as well. I already had Edcor output transformers from a project that didn't get built. Since transformers are a major cost, I didn't want to buy more. I decided that since the new amp was a follow on to the earlier one I would start a naming system. The new name is based on an old time tradition of listing the number of channels (2) and the power output per channel (25). As you will see later other versions up to a 264 and perhaps higher are possible (at least the theory says). As I sort of alluded to, I encountered a rather unexpected but really pleasant discovery in building the amplifier. The design is easily scalable from about 5 watts per channel to over 50 (and perhaps as much as 80). Changing one resistor in each channel will allow the use of any of the power tubes that share the base layout of a 6L6GC. This includes 6V6, 6CA7, EL34, KT66, KT77 and KT88. There is a good selection of these tubes at Tube Depot, both NOS and new production. With a switch to swap the resistors, you can go from one tube type to another without fuss. I'll explain how this works later. The primary limitations to the swaps are the dissipation of the tubes and the exact value of the B+ voltage available. Anything from about 250 to 500 (depending on the output tubes) can be used. For the highest power versions using 6550s and KT88 there are a few other simple changes if you go to B+ voltages over about 400.
Now with the transformers a limiting factor, it made the initial selection of tubes a lot easier. I really wanted to use all octal tubes. Those big bottles really look cool glowing in the dark. Based on price, the choice was originally narrowed down to either 6V6s or 6L6GCs. Either would work well with the output transformers. The 6V6s would not be more powerful than the EL84s in Oddwatt so the choice was for the JJ 6L6GC tubes. These are inexpensive and readily available. Shortly thereafter I discovered that the newly manufactured JJ KT77 tubes were available so I switched to them for the final project. They have distortion levels little a bit lower than the 6L6GCs. The price was only a dollar more per tube.
SRPP / SIPP Tube Amplifier Design with CCS
Figure 1: 6SL7 SRPP / KT77 Push-Pull Tube Amp Schematic
The basic circuit is the same as in the earlier project. I use a SRPP driver stage feeding a self-inverting push-pull (SIPP) output stage which is based on the Compact Hi-Fi Power Amplifier by Melvin Leibowitz. The SIPP has the advantages of a conventional push - pull output stage, but requires only a single ended driver. If you look at the stage carefully, you will see the output stage looks a lot like the phase inverter used as a driver in more conventional amps. There are a few drawbacks to the SIPP stage as well. There's no such thing as a free lunch in electronics. The stage operates in Class A mode only, the cathode circuit can not be bypassed and you can't apply plate to grid NFB on the outputs. None of these are serious problems, just considerations. The driver stage is an SRPP and is unremarkable in design with the possible exception of the application of a small amount of global NFB. The NFB can be omitted, but with it in place, the gain is better defined and scope traces of high frequency square waves have a better shape. Harmonic distortion is about half a percent lower at all power levels as well. The constant current source (CCS) is composed of a National Semiconductor LM317 voltage regulator IC and a fixed resistor. If you connect it as shown it will be a CCS. The output current is 1.25 divided by the resistor value in ohms (or you can use the LM317 current regulator calculator). In this case 1.25 / 10 = 125 milliamperes (mA). Make sure that the regulator's adjust terminal is grounded and the resistor is from the output to ground. The regulator needs a heat sink as it will dissipate about 3 watts in the KT77 version. In KT88 and 6550 versions a resistor of about 150 ohms at 10 watts needs to be in series with the CCS. This is because they operate at cathode voltages above the rating of the LM317. Remember also that the regulator's tab is connected to the power and will need an insulator if you fasten it to the chassis. I also use heat sink grease to help keep it cool. The 25 ohm variable resistor located in the output tube cathode circuit is for balancing the current through the tubes. The one ohm resistors allow for measurement of the current passing through each tube. The number of DC millivolts (mV) across each resistor equals the number of mA passing through that tube. I was able to balance mine to 0.1 mA. Each tube should draw between 60-65 mA in this circuit. The B+ supply would seem a bit low for a 6L6GC / KT77 based amp. This would certainly be the case for anything other than Class A operation. The B+ supply of 350 volts will provide a maximum output in the 36 watt range. If you like math, do the numbers and you'll see. This voltage and current is comfortably within the limits of a 6L6GC or KT77. Additional voltage will not increase the output (in Class A mode). The lower voltage allows for the use of less expensive filter capacitors and power transformers. Tube life should be increased as well. I do not believe this circuit will work in any other class of operation. If someone knows differently, I would like to know about it. For those of you who are observant, you probably noticed that I used 8K output transformers. The load required by the tubes is closer to 4K. The way this works is that the transformers are rated at an 8 ohm load on the 8 ohm tap. If you use the 16 ohm tap for 8 ohm loads, the reflected impedance to the output tubes is cut in half. If you don't believe it check the math on how to calculate the output transformer load resistance. Likewise the 8 ohm tap can be used for 4 ohm loads. The power transformer is an Edcor XPWR101. This transformer is a beast. It can deliver 2000 mA at 240 volts AC. With a full wave bridge into a capacitor filter it delivers 350 volts DC.
Figure 2: Oddwatt 225 Power Supply Schematic
The 12 volt heater winding is also rectified with a full wave bridge. The heaters of the power tubes are connected in series (pairs) to use 12 volts DC. This was done to ease the current levels required. The resistor R* in the heater circuit reduces the DC voltage by about 2.4 volts to the 12.6 needed by the tubes. The value will have to be determined experimentally based on the output of the heater supply and the load current. This resistor is dependent on the tubes used. A one ohm resistor will drop 3 volts at 3 amps. Be aware that this resistor will get hot and need to be at least a 20 watt size (I squared times R = 9 watts continuous dissipation). As an alternative, you can use power rectifier diodes to drop the voltage. Each diode will reduce the voltage by 0.7 volts. This is what I actually do in my amp. The rectifiers must have a current rating greater than the heater draw. They need only to be low voltage inexpensive types. Just string enough of them in series to get the value you need. Remember that each of the diodes will dissipate over 2 watts and will require ventilation. The 200k and 100k ohm resistors connected from the B+ and ground to the heater circuit provide a voltage reference for the heaters. They are at about 1/3 the 300 volt B+ above ground. This is necessary to protect the SRPP driver tube cathodes. During start up, the upper cathode can see most of the B+ supply voltage. This can easily exceed the heater cathode break down rating and result in tube failure. This did occur in the Oddwatt amp. Do not ground any part of the filament circuit. I had another fortunate occurrence in this part of the circuit. Using the values of resistors in the power supply (PS) filters and the B+ voltage, it worked out that the time to charge up the filter capacitors on the SRPP plates was the same as the time to the heaters needed to warm up. The voltage never exceeded the value required by the circuit. The heaters can be AC powered and this will simplify the circuit. They will still need to be above ground potential to protect the driver tubes. I used DC on the heaters to eliminate all sources of potential AC hum in the amp chassis (the interconnect from the remote power supply carries only DC). Is it worth the trouble? The jury is out on that, but it is a very quiet amp.
Oddwatt 225 Tube Amplifier Construction
The build is straightforward and not complicated. I used some terminal boards with pre-mounted gold tube sockets for the "modules". These came from Antique Electronic Supply and are optional, but they worked great and I highly recommend them. The amp was constructed in four parts. Two output sections, a driver section and the power supply. The modular arrangement allows for flexibility in the final assembly. For a chassis I used a wooden shadow box and a piece of Plexiglas. The power supply is external because of space limitations. It is on a small metal chassis. I used a DIN 6 pin cable and mating plugs to connect the PS to the amp. RS232 or other cables would work as well. Just be sure the voltage rating of the cables is sufficient to handle the B+. As a result of the separation of the amp and PS, I added additional B+ and filament filters inside the amp case. The tube cover is a copper plated storage bin with one side cut out. It came from a home improvement store for $6.
|JJ 6L6GC / JJ KT77
Ruby and Solen
Russian PIO (thanks Rodney)
Antique Electronic Supply
Table 1: Major Parts Used
Photograph 2: 6SL7 SRPP / KT77 Push-Pull DIY Tube Amplifier - Inside
For additional tube amp design and construction tips, see my design and construction tips and suggestions for vacuum tube amplifiers. Also, I have previously posted some suggestions for a tube amplifier wiring color code that you can use during construction.
Oddwatt 225 Project Costs
The largest expenses are the transformers and tubes. They were around $250US. Everything else was around $150. So if you have access to the parts, the amp could be duplicated for around $400US. Is it worth it? I think so. I would expect to pay a lot more for this level of output and refinement. There are a number of kits out there that cost less, but they usually have a lot less output. Commercial tube amps in this size are at least twice the cost and often much more. Besides there is the intangible benefits inherent in diy projects.
Oddwatt 225 Performance Measurements
Measurements were made with the following equipment: HP 331A distortion analyzer, Tenma Low Distortion Signal Generator (0.05% residual THD)and a Velleman PCSU1000 oscilloscope to computer interface. For those of you unfamiliar with the PCSU1000 it is a dual trace 60 MHz front end that connects via USB to a PC. It has all the functions of a conventional dual trace scope, plus will do spectrum analysis and all sorts of measurements. The software provided is easy to use and intuitive. My B&K dual trace scope now resides in the shed.
It was a surprise from the start. My initial goal was 20 watts at less than 2% distortion. The end product with KT77s was 25 watts at 1.7%. At 12 watts it is 0.5% and at 6 watts it is below 0.2%. Bandwidth was 8 Hz to 21 kHz +0/-1 dB. I deliberately tapered the high end response as there is a transformer resonance at 60 kHz. Without the adjustment, the -1 dB frequency was 40 kHz. The maximum output at 3.6% distortion was 33 watts. Being conservative, I decided that 25 watts was a good rating. With 6L6GCs the power is the same, but the distortion is about 0.3% higher across the full range. The 6L6GCs seemed to have more of a gutsy bottom end, but measurements did not confirm any significant differences. Hum and noise level are at about the 1 mV level at the outputs. Most of this is wideband noise. None of which is audible at the speakers.
|Frequency (Hz)||Power (W)||Distortion (%)|
Table 2: Distortion Performance of Oddwatt 225
The Noise Floor was -68 dB (un-weighted) at 5 W. Second Order Harmonics were typically -57 dB (un-weighted) across 20 Hz to 20 kHz. Higher Order Harmonics were not significantly above the Noise Floor across 20 Hz to 20 kHz.
Square waves are well behaved with typical tube type shapes. For folks not familiar with how amps behave with square wave inputs, I'll give you the 5 cent tour. Square waves do not pass through inductors well. An old and still apparently valid rule of thumb is that to pass a square wave a circuit must respond at both 10 times below and 10 times above the frequency of the signal. This is a nearly impossible task for tube power amps with reasonably priced transformers. So if you see a "clean" square wave at say 100 Hz then the circuit has good response at both 10 Hz and 1000 Hz. There are often little squiggles at the leading edge of the reproduced wave. These indicate the tendency of the amp to oscillate and become unstable as well as phase shifts. Again, if measured at the output terminals, these are rather common. Fortunately, music is composed of mostly sine waves and thus the problems are usually more technical than audible. The second harmonics were -57 dB throughout the response range. All other harmonics were buried in the noise floor.
Figure 3: 20Hz Square Wave at 5W
Figure 4: 500Hz Square Wave at 5W
Figure 5: 1000Hz Square Wave at 5W
Figure 6: 5000Hz Square Wave at 5W
Scalability with Other Tubes
As I mentioned in the beginning, this amplifier is scalable. You can build a lower or more powerful version. You can also adjust it to use different output tubes. The primary considerations are maximum voltage rating, maximum cathode current and maximum dissipation. I generally suggest that with the existing B+ voltage you select the cathode current such that the dissipation is limited to around 75% of the rating. It should be noted that even taking into consideration that the cathodes are above ground potential, the B+ supply is right at the max for most 6V6 tubes. I suggest using a lower voltage power supply for 6V6s. All the other tubes can easily handle voltage. To change the cathode current all that is necessary is to change the resistor in the CCS. Using the formula of 1.25/R (or a current regulator calculator) you can calculate the value needed. It will probably require a pair of resistors in parallel to get the correct value. The resistor values are critical. The deviation of one ohm will make substantial changes in the current flow and possibly damage the tubes. One way to make the swapping of tubes easy is to use a rotary switch. By selecting the values of resistors, you can use a 10 ohm one permanently wired in place which allows a current flow of 125 mA and connect other selected ones in parallel with it to get higher current flow. For example a 48 ohm in parallel with the 10 would equal 8.3 ohms and result in 150 mA. A 22 ohm in parallel with the 10 ohm would result in 180 mA. I suggest that you not use a variable resistor as it would be very easy to get a value too low and fry the tubes. In theory a pair of KT88s could deliver as much as 80 watts in this circuit. I would not run them that hot, but 50-60 should be an easy goal.
Other Possible Changes
You can substitute 12AT7s for the 6SL7s without circuit modifications except to the heater circuit. Using other tubes is possible, but the values in the SRPP circuit will have to be adjusted. I found that 12AU7s and 6SN7s did not provide sufficient gain for the circuit to operate properly. 12AX7s could be used if the circuit values are adjusted. The input volume control is optional and a fixed 100k resistor can be used in its place if you use the amp with a preamp (I did make this change in mine). There is also no reason that you can't put an input selector switch and use it like an integrated amp with the limitation that you will need to feed it signal sources that have about 1 V or mores. I have found the B+ relay to be unnecessary in the many hours of use the amp has undergone during testing, but it is a good way to protect the power tubes from excess voltages when the filaments are warming up. You can use a standby switch or an optional circuit that has an adjustable delay timer and turn on the B+ after about 15 to 45 seconds. The most sensitive are the drivers and as I mentioned, careful selection of power supply filters protects the ones in the circuit as I built it.
Photograph 3: 6SL7 SRPP / KT77 SIPP Tube Amp with Cage
How does it sound? This is always a sticky question. What I like and hear could be and probably is different from what someone else would hear. On projects like this there is no good way to make across the board comparisons. That said I like the sound of this amp. I prefer the warmer sound of tube amps. I do have several amps to compare it to. I have two other tube amps, several solid state amps (Marantz, Kenwood, Pioneer, Audio Source and Sony), Super T amps and a DIY LM3875 non-inverting gainclone kit. This amp has a typical warm tube sound. It has good inner detail, good sound stage, and plenty of output. The bottom end is much better than my other tube amps. The words that come to mind on the sound of this amp are robust and refined. It doesn't have the power of the gainclone, but it has better inner detail. The 6L6GC version seems to be more robust in sound with apparently a little less detail. For rock music the 6L6GC version might be more likeable to many listeners. For other music types the KT77 version is the one of choice. As with any tube amp, they work best with fairly efficient speakers. I really like the sound through the pair of Altec Lansing 1010s I have (93 dB/W). With my huge subs coming in at 50 Hz (driven by the gainclone through a 24 dB/octave electronic crossover) the sound is stunning.
What I Might Do Differently?
Even though I am very happy with the amp, in the best spirit of diy, there are some things that could be changed. For one, I might attempt to build it on a single chassis. It makes the build easier, but would also be very heavy (~40 pounds). An alternative would to be to build it as two mono-blocks (with two power supplies). I would custom order the power transformers to match the load. I would order the output transformers to match as well. I obtain my transformers from Edcor. They are very helpful and will custom wind ones if you want. I got lucky this time with stock transformers. In addition to the various options for power output, I have a variation in mind using 6SN7s driving 6SN7s as a low power version for headphone use or low level speaker use. It could deliver about one watt. There is no reason that the amp can't run triode or pentode modes, I just think the best compromise for power output and sound quality is the ultra linear mode.
If anyone out there builds an amp like or similar to this one, I would like to hear from them on their thoughts, changes or recommendations. DIY Audio is an ongoing thing with me and I always welcome new ideas.
UPDATE 29 October 2008: Bruce has completed a mono block version of this amplifier using the Group B output tubes (6550, KT88, KT90). For more information about his latest tube amplifier project, see the OddBlocks - Class-A Push-Pull KT88 Tube Amplifier (12SL7 Driver) project.
UPDATE February 2009: I have made a few refinements since I first built the amplifier. First, the standard LM317 voltage regulator IC is only able to withstand a maximum of 37 volts. I have not had or heard of any failing, but I suggest replacing it with the LM317HV that is good for a maximum of 57 volts. They are available from several sources. Mine came from Jameco, part number 837927 at $1.59 each. This is cheap insurance in my book. It is probably essential in the long term reliability of any amps built with KT88s because of their higher bias voltage. Second, I retract the recommendation to use 12AT7s or variants in place of the 6SL7s. They just don't sound as good. The type 5751 is a better (and direct pin for pin drop in) match. Alternatively, I would use a 12SL7 that are available from AES (tubesandmore.com) for under $5. It will make the wiring a bit easier as well. For more information about the LM317 voltage regulator IC, see my notes onThe Care and Feeding of LM317 and LR8 Integrated Circuit Regulators, Particularly in Valve Circuits.
A pending change that is in the works is the use of a power FET as the CCS. I am experimenting with IRF730s although others with suitable voltage and power ratings ought to work as well. There are a number of advantages to using a FET over the LM317s. First the LM317 requires a minimum voltage of about 4 volts to operate properly. Operation below that level is undefined and could result in dramatic distortion levels. The FET can go essentially to zero. The dynamic impedance of the FET is much smaller than the LM317 as well. Both characteristics should improve the linearity of the output stage at high power levels and are likely to translate into a slight increase in power output. The only difficulty I have run into so far is that the FETs are temperature sensitive and require generous heat sinks. The LM317s are by contrast rather stable thermally.
I remain pleased with the sound of the amplifier and have experienced no failures (always a possibility with a diy project) and no unusual behavior. It is in daily operation and has accumulated several hundred hours of "on" time. There did not seem to be any real "break in" period. The sound was refined at the start and remains such.
Good Listening, Bruce.
For the latest information about the OddWatt Audio SIPP tube amplifier projects, see discussion thread on the Oddwatt Push-Pull Tube Amplifiers in the Forum.
UPDATE - October 2009
The Oddwatt Audio OddBlock KT77 tube amplifier kits are available for purchase from OddWatt Audio. The KT77 kit is offered as a monoblock amplifier (so you will need two for stereo) and comes with all the required parts including a drilled and painted chassis. The kits come with NOS 5751 JAN Philips and a matched pair of JJ KT77 tubes.
UPDATE - December 2012
Regarding the audio output transformer load. In this project the B+ is lower (350V) and the KT77 tubes seemed to work better into an effective 4k output load by using the taps as noted in the schematic. However with the higher B+ this has turned out not the case. and the Oddwatt 225 amplifiers are not as good as the Oddblock tube amplifiers.