Welcome! Need support, you got it. Or share your ideas and experiences.
Return to MagneQuest/Peerless Forum
24.6.199.104
Finally here is the schematic for the 9 watt version of the CCS based transconductance amp. Its very similar to the EL84 version With the following differences:Different tube (of course!) I originally designed this for the EL509 but the 6550 or KT88 will also work, you might have to change one resistor value to handle different tubes, there will be a separate section on this below.
Input impedance has been raised to 100K instead of 30K, closer to a more normal tube input impedance, easier to drive from passive pres, BUT more sensitive to picking up noise, layout is a bit more critical.
Q3 and Q4 are now connected to their own voltage divider, they are only rated for 500V which was just a little too close for comfort to the 450V B+. R25 and R26 should be 1 watt metal film types.
All caps are motor run or other high voltage film types except for C5, its a 2.2uf electrolytic 50-100V. C1 is probably the most sensitive to "quality", I've had very good results with motor run caps for this. The exact value is not that important, I've run it from 5uf to 50uf, the amp works anywhere in that range, you will get different tradeoffs in "bass impact" "airines" with different values, feel free to experiment. Different tubes will also need different values. Start with 25uf and go from there.
I chose the Robin Hood PF OPT for this design, its a perfect match and will sound exceptional without costing a fortune. You can also use an EXO-50 if you have them on hand, Yes it will probably sound better than the RH, but not by a huge amount, only you can determine if its worth the price. A TFA2004 will also work but you won't get the full power output. I designed this for an EXO-03 for its low price, you could go with a BAC-80 and get a little more power output, but the price goes up quite a bit, the EXO-03 and the RH-PF are a real "sweet spot" for this design.
If you build it as in the schematic the sensitivity for full output power is about 3V RMS. If you are using an active preamp with some gain this should be fine. (It ought to work great with an FPIII) If you are using a "passive pre" or just want a little more gain, make R1 47 ohm and R10 68K, this gives an input sensitivity of 2V RMS for full power at a VERY slight decrease in sonic purity. With 2V you should be able to get full power from a CD player or DAC output with a passive pre. I don't recommend going to any higher gain.
Tube choice, as mentioned above you can use EL509, 6550 or KT88 tubes. I don't recommend El34 or 6L6. The NOS EL509 takes a special large 9 pin socket and plate cap, but I really like these tubes. There is supposedly an octal based version out now but few have seen one of these. The more traditional 6550 and KT88 will also work. The gain of the amp is NOT tube dependant, it is determined soley on the value of R1. The CCS automatically adjusts the bias for the proper current so tube rolling is really easy and you don't have to worry about "runaway" or any such weird effects. You might have to tweak R3 to get the screen voltage in the range for proper operation. All of these tubes will work just fine in this circuit with a fairly broad range of screen voltages so its not very critical. Even if its set wrong it will NOT harm anything, you just won't get full power output.
There is one adjustment you will have to make, adjusting the pot R11. This sets the idle current, the manufacturing tolerances on the MOSFETS are fairly loose so you need to adjust this. You want to adjust R11 so there is 65mA or so flowing through R1. For the 75ohm R1 thats about 4.8 volts across R1. For the 47ohm R1 thats 3 volts.
You will need to experiment to find the best output tap to use, its going to depend on the speaker and cabinet used. The transonductance amp is specifically designed to handle single driver speakers that have a rising impedance in the bass. I usually set mine for 16 ohms so the bass region which rises to 16 ohms gets full power. The midrange has lower impedance so it doesn't get full power but it doesn't need it. If your cabinet does not exhibit very much bass impedance rise you might be better off going with the 8 ohm tap. Again experiment and find out what sound best to you.
As a little side note I've been testing out the El84 version with a spectrum analyzer, its got some pretty good specs, the noise floor is 120db below 1 watt and I can't see ANY harmonics at all so I don't know what the distortion number is, but its pretty darn low if they are all less than 120db down!! There IS some 120Hz PS bleed through, but its pretty far down, and no visible IM distortion from the PS on audio signals (I expected to see some 120Hz side bands, but they are completely absent).
A word on the power supply, the venerable MQ PGP8.1 will work to drive one channel of this amp which is why I put it on the schematic (This IS the MQ forum after all), but given all the posts flying around on low DCR PS design it may not be the optimal solution. There are so many different transformers and chokes available that will work for this design that its hard to come up with the one "best PS", so all I'm going to say right now is you need 450V at 70mA B+ and up to 2A at 6.3V for heaters, how you get there is up to you! I'll certainly be glad to help design a PS for this amp given some diretion from the user as to how they want to go.
Oh yeah, what is this thing good for? This is specifically designed for single driver speaker systems, Lowthers, Fostex, AER etc. With this amp the current through the speaker is proportional to the input voltage, this works great for speakers without crossovers, but does not work very well for most multiway speakers with crossovers. The inspiration for this was Nelson Pass's work over at firstwatt.com.
A note on construction, you can either build the CCS part by hand on perfboard or use a PC board. Gary Pimm sells little boards that can be used very well with this design. (Just make sure you use these values, not the values he has in his instructions) Try and keep the input and CCS as far away from the PS as possible. If you can't physically separate them a metal shield (aluminum plate, piece of unetched PC board etc) between CCS board and PS helps a lot. The big MOSFETS might get pretty hot (it depends on the tube and screen voltage) and so need heat sinking. I would just do it to begin with and not worry about it. You can either mount a small heat sink to the transistor or mount the transistor on the underside of the board and screw it directly to the top plate (or other mounting plate). A small standoff raises the board the right distance from the plate. Gary has some nice pictures of this on his site. You can get the the IRF820s with either insullated or non-insullated tabs, the inullated is more expensive but a lot easier to mount to a heatsink, you don't need any special electrically insullating hardware. (the insullateded version is IRFI820 (note the extra 'I' in the part number))
I guess thats it for now, have fun!
John S.
![]()
Follow Ups:
John,Just in time, when I received my pair of EL509, I saw your new post on this amp I' m going to try.
Just a qick question: what is the voltage range for the screen grid (g2)?
John, Great design and execution, especially targeting the single driver speaker load. 9W basically from one stage, too. Amazing. Has this been done in differential, with a PP output? --keto
![]()
Gary Pimm has done a SS only differential PP version which he thinks sounds better than anything else he has ever done. I still like tubes, they handle the higher voltages a lot easier and you don't need big masssive heatsinks like you do for the SS only version.I've done a differential line level version (my preamp is this design without the tube) which works very well. Going differential causes all kinds of problems though, when you put a CCS on the tail or as a load for a follower the whole thing starts oscillating. It does not like the high impedance load. It turns out that with the very high impedance load any random RF in the environment (even the recovery spike from PS diodes!) sets the things off. Even a fairly high resistive load keeps that away from the gates and its fine. Gary came up with a couple ways to make the circuit less sensitive to this and they seem to work quite nicely. First up the voltage between Q1 and Q2 and second add a small resistor between them. With both these in place a long tail pair works great, no separate phase splitter necessary, it should make a very nice PP amp, I just have not tried it.
Nice looking circuit. I have a bunch of NOS EL509's sitting around
so I might have to give this a try.
Now, lets see, where should I put this on my tube, to-do, list? :)
Oh well, I'll figure that out later.
It's excitng to see your circuit since I know what it's like with Pass' transconductance amp.
I wonder why 9 watts for Lowther? ;-) I wanted to see your EL84 circuit but could not find it. Looking at your circuit I thought about small one like ELL80 transconductance amp.Are you feeding this amp with low DCR supply?
Here is the link for the original El84 version. It was a little under powered (even for Lowthers) with large full scale music so I decided to make something a little larger and went overboard, the result is an 813 based SE parafeed version that goes up to about 30 watts! Its still a work in progress but sound fantastic even in its current breadboarded version.The El509 version was going to be what I made until I got hooked on doing a 813 version.
There are a lot of people with fostex full range speakers that want to try this type of amp but the El84 was just too wimpy so I have gotten a lot of requests for the larger but still inexpensive version.
Thanks for the interest!
John S.
![]()
Since I have already (breadboard) built the EL84 version (and happily listening to it for the past 6 months) driving my FE167 Fostexes, I can confirm John's finding that, even for a full-ranger, EL84's 2 watts can prove inadequate for louder and dynamic passages. Hence my desire to build a higher power version of this great implementation.
John,
Great design again,many thanks for publishing. I tried a version of your EL84 design but with 12GN7s (3 times the Gm) after tweaking it in it worked nicely with my AER open baffle speakers.One thing I do with pentodes in pentode move that you might like to try, I use a VR tube between screen grid and cathode bypassed with a zobel network as per Mark Kelly's posts on the subject. I, and others who have tried it, find it opens out the sound and greatly improves the bass response - not so much in extension but in tone and texture. It removes a certain wooliness that cap decoupling has... Sometimes this means a hunt for the right VR tube and it does increase the current draw but the sound quality improvement has been more than worth it...
Again many thanks for publishing your designs and thoughts.
caio
James,I have already built the EL84 version of the CCS Gm amplifier and I'm going to try the higher power EL509 version.
Since I' m not aware of Mark Kelly's suggestions you are referring to, could you please give me some more data on this and a link to the relevant posts?
As far as I can understand from your description, this is feeding the screen grid from a regulated voltage supply. But, what about this zobel network?
Hello Evangelos,It is shunt regulating the screen grid with reference to the cathode (this is important!). The VR tube looks like an inductance to the screen grid and it also produces some hf and rf hash (noise) normally one would use a cap to shunt the noise away from the screen grid but with the VR tubes inductance this can form a nice oscillator which even if it isn't oscillating per se can cause the noise to become more pronounced. The solution Mark Kelly came up with was to neutralise the VR tubes indiuctance with a zobel network across the tube and this also quietens the tube and bleeds the noise away. For minimu noise a typical 40mA VR tube needs 14mA or more current flowing through it and gets quite noisy for less than 5mA flowing through it. Mark measured the insuctance of the tubes and found a zobel network of 100nF with 390R worked with a typical VR150 tube.
Search Tube DIY for the posts.
hope this helps
Sorry - got the zobel values the wrong way around! Cap is 300nF and resistor is 100R...
Thanks James, I'm going to be trying this out on my 813 version of the amp. It takes 450V on the screen so I need 3 tubes in series, I presume thats going to change the values needed so I'll have to experiment a little bit.
Hello John,Yes, it will look like the three inductors in series but the zobel works best if applied to each VR tube individually so you can use the same values for each VR150.
I'm trying to find Mark Ks post where he gives example inductances but no luck yet... Maybe its on hs website...
Hello James,Thanks for the information. I did a search in Tube DIY and found some posts regarding this issue. I will study them.
Hmm, lets see, if I drove the VR tube with another CCS that would be the ideal configuration I think. I'll have to try that someday. I had thought of trying that for awhile but never had time. Thanks for bringing it up.
Hi James,If I remember correct, you are the gentlman who showed the beautiful plexiglass 'Quasar' on BD design forum. I also read recently that you speaker is getting better. How nice.
Once a Oris user, your OB is my next target. I've been keeping AER3 and 12" Altec woofers for 3 years now. Bert's network design for Qausar seems to be a "serial" network which is for transconductance amp like on Pass site. Am I correct? If so, this transconductance amp is right on, this time with tubes thanks to John Swenson.
Just mumbling, 'cause for now in my house Quasar OB is the best I can afford, and good to see you here.
Cheers
C.Y
Hello C.Y.Yes that's me :) Thank you for the kind remarks about my speakers and they are getting better all the time - some break-in period. 3 years and counting. They do use a serial network for the speakers and work nicely with transconductance amplifer just like Nelson Pass explains on his site. Of course the Quasars will work nicely with all amps as they are high effeciency and an easy load to drive being 10 ohms and looking mainly resistive but transconductance amps do work very well with them.
I find the bass to lower mid-range integration of the Quasars to be excellent - something that isn't really true of the Oris setup.
ciao
This post is made possible by the generous support of people like you and our sponsors: