6L6GC AB2 Amp

[SGregory]

chrish,
A Very nice build so far.
Your zeners should work nicely. Although I went the route of self protected, I have not had one fail since the change.

In answer to your question regarding the CCS under the FET source, I will say the jury is still out. I am just now getting geared up to do some serious measuring, my EAR in limited tests says the CCS sounded more detailed. There was a brief discussion on this (i couldn't find the thread) that had similar arguments as using a LED for bias. Keep in mind my EAR opinion is only one channel and I only had a resistor swapped in under the FET for a short time. Whether it stays or not will be based on tests and measurements and then finally whether I can listen to it. If I can't listen to it I will through the measurements out and set it so that I can ;-)

Keep Up the good work.

 

[chrish]

Thanks for the comments!

I went through this thread and found some comments from Tubelab regarding the same issue. I think his opinion was that it did not produce much of a difference, so went for the simple approach. This thing is already complex enough, so I stuck with the original design. When it is up and working I may consider tweeks, but more components will probably mean more mounting hardware. I want to keep a clean top plate, and drilling blind holes will be difficult without a complete disassembly.

Hope to get some signal wiring done this weekend. Looking forward to seeing some results finally!

Thanks again,

Chris

 

[chrish]

Just a quick update...

Managed a little time this afternoon to do some wiring. Have wired up the input stage to the MOSFETs. All appears to be OK, with nice clean output from both diff amp stages and the MOSFET source followers. B+ regulating at 395V - close enough to the 400V I was aiming for. Too much hassle to change the set resistor. Hopefully I will finish wiring the output stage some time in the next day or two.

Cheers,

Chris

 

[chrish]

Finished the output tube wiring this morning. I wired it triode mode, strapped with a 470R resistor.

Brief problem with oscillation. Was oscillating at some combinations of volume and frequency. No problem with output tubes removed, so took a guess that it required some grid stoppers. First thing I tried was some 680R carbon comp right on the grid pins of the output tubes. It appears to have solved the problem.

Something else I noticed, when I drove the amp in to clipping, the 75v voltage regulator tube began to dim. I am guessing that the MOSFET source followers are drawing more current here than the VR circuit can handle. My initial impression is to leave it as-is, and use it as a visual clipping indicator. Any problems with this?

I tried playing music through it. Just iPod through a bookshelf speaker, but it worked.

I will tidy up the wiring this afternoon and post some shots.

Cheers,

Chris

 

[whitelabrat]

but it worked.

Good work!

 

[rknize]

Something else I noticed, when I drove the amp in to clipping, the 75v voltage regulator tube began to dim. I am guessing that the MOSFET source followers are drawing more current here than the VR circuit can handle. My initial impression is to leave it as-is, and use it as a visual clipping indicator. Any problems with this?

I would say it's just right. Maybe put a scope probe on the rail to make sure it doesn't spike up and out of regulation during recovery, but it shouldn't as long as tube doesn't completely go out and need to re-strike. You aren't going to listen to sine waves much, I suspect.

 

[chrish]

Thanks!

I hooked it up to the main system, and it plays. Have a few minor issues to sort out, but isn't that half the fun? I have two versions of the 105v VR tube. Some RCA that have a nice pale purple glow and some Russian, with a neon orange. The Russian ones glow, with the glow flickering with the music. The RCAs, one will strike up, but end up going out after a minute, and one will not strike at all. Maybe have to tweak the current limit resistor. Had calculated 500R required. I suspect the power supply is sagging with the load and the limit resistor is dropping too many volts to allow the tube to strike. Fortunately I have a high wattage 1K variable in this position (the dark green one in the photo).

Worst problem is the amp has a high pitched "whistle" coming from the speaker. This whistle comes on and off according to the load on my Mac Mini computer that is doing duty as a home theatre PC. If I turn the Mac Mini off, no whistle. I suspect the problem is from its laptop-style external power supply. Switch mode? Anybody had any problem like this?

Also need to drill some vent holes in the bottom plate. There is quite a lot of dissipation going on, and I only have a slot above the top of the heat sink at the rear for ventilation.

Some photos. Tubes are Russian "test tubes", have Sylvania 6SN7 and Philips JAN 6L6WGB to replace. Yes, with the amp working that is a hint of 'purple glow' on the top of the 6P3S

Red and black probe jacks behind tubes for bias sense. The two holes to the right of these allow a small screwdriver to access multi-turn trim pots for the bias adjust.

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

 

[SGregory]

Again, very nice work. Will you be testing harmonics and such? Very curious to compare results.

I also hear a "whistle" from my HTPC periodically. Using a MAudio 192 card for that, so I came to a similar conclusion that it is related to the computer PS. Probably will build an external DAC one day to solve the problem.

 

[rknize]

The source is probably the power supply, but how it is getting into your amp may be more of a ground loop issue. PCs are generally grounded as well. Is the PC and amp plugged into the same outlet/power strip/etc?

 

[chrish]

Yes, there is a lack of power outlets so all tv and hifi are connected to the same outlet. Agree it is probably a ground loop issue. Have the signal ground connected to the chassis at the input end and have used star grounding. Apart from this sound, it is quiet. I have another home made tube amp plugged in to the same outlet (baby Huey EL84) that is dead quiet. I may try a signal ground lift circuit to see if that helps.

As for distortion analysis, I don't have any fancy test gear, just a second hand scope, a cheap audio oscillator that makes pretty good sine and square waves and a couple of reasonable DMMs. Square wave response looked pretty good at 1kHz, with just a hint of ringing at 10kHz. Top of the square slopes down ever so slightly, indicating a little HF rolloff? Will post picks of response later. BTW, not using any feedback at present.

Will keep posted!

Chris

 

[SpreadSpectrum]

As for distortion analysis, I don't have any fancy test gear...

I scavenged a big heatsink out of an industrial motor inverter and just mounted some TO-220 thick film resistors to it. Then I made a 1K stepped attenuator and put it in parallel with the load along with some diodes to clip the output if it gets too high. Then I ran that to a soundcard. Cheap and simple.

The inverter had fans on it, too which I mated to an old laptop power supply. It's probably good for 100+W I figure.

As for the VR tube, you want to be running it on the hot side because your load will never decrease from the quiescent value(and it looks cooler), but it will increase with grid current. I would expect it to dim when you draw grid current(since more current is going into the power supply's load), but I would make sure it is not going out. That might cause problems.

I had to solder the mosfet source and source resistor directly to the socket in order to keep oscillations away on my AB2 KT88 amp. I probably should have mentioned that to you earlier. It pays to minimize lead inductances if you are going to forgo a grid stopper (which is necessary for low distortion AB2).

 

[chrish]

Here is a snap of the square wave before fixing the oscillation. I think tnis one was at around 1kHz.

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

Here is a sine wave with the oscillation (now fixed):

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

Here is the output from each phase of the input stage:

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

Here is the same inputs (each phase of the input stage) with the 'sum' function of the scope, with the same scale. Note a little distortion:

An externally hosted image should be here but it was not working when we last tested it.

Uploaded with ImageShack.us

If there are any tests I can do with the limited equipment available, I am happy post the results for you.

Cheers,

Chris

 

[SGregory]

Lets see some squarewave pics at 1kHz, 2kHz, and 10kHz at 1Watt with the osc corrected. Have you calculated / measured open loop gain? You can also use ferrite beads per George's recomendation to stomp out oscillations when grid stopper are troublesome, as in AB2.

I am just now getting audiotester to work well enough. Using Pete Millett's sound interface REALLY helps, but you can get some close measurements without it as long as you are carefull of the soundcard. Or are you using a MAC?

In my OPUS I also saw some tilt on the squarewave, initially. It is a low frequency roll off issue. Part of my problem was the input cap was to small for the impedance. I also lowered the impedance of the RC decoupling on my bias circuit. I don't think you have the later. Even still, I have a hint as seen in the attached photo.

 

[chrish]

OK, so for the oscillation, try to shorten the leads as much as possible, and use as small a resistor that works? Re-locating the MOSFETs would require a lot of re-work.

As for soundcard, I am using a Macbook Pro on the desktop and the Mac Mini as the HTPC. Nothing I can put a soundcard on...

Mods may have to wait a few days. Have a licence renewal exam Monday then away with work till Friday.

Thanks for the help and input.

Chris

 

[SpreadSpectrum]

Using Pete Millett's sound interface REALLY helps...

I really like that thing. However, I built a 1M input impedance version, so I get better noise performance coming straight off the stepped attenuator on my dummy load for amp output testing. I love being able to use 100X scope probes for measuring distortion on CCS loaded stages swinging a lot of volts. It is awesome to be able to probe that without loading it down and causing a bunch of distortion.

 

[SGregory]

Chrish,
Maybe you may have an unknowing freind who would be able to loan you a x86 machine to play with. ;-)

Spread, I am thinking about building a second interface, perhaps I will try your mod.

 

[chrish]

A quick update...

Well, as for the oscillation, I tried just about everything. I tried placing the MOSFEt right next to the output tube with a very short lead, I tried 100R plate resistors, then with the air core inductor wound around the resistor. I tried grid stoppers and ferrites, all to no avail. In the end I noticed that the oscillation changed when I touched my signal generator. I have a BNC T junction on my battery powered signal generator. One connection to the scope, the other to the input RCA of the amp. I removed the T connector and tried just connecting the signal generator directly to the amp and using a scope probe at the grid of the first stage to detect input signal. Oscillation gone. I removed all the stoppers and left the MOSFET close to the output tube. Appears to working OK now.

Now, some questions.

I have re-read the thread a few times and always pick up some good advice that I did not see or was too ignorant to understand. I used Tubelabs recommended procedure for setting up: Set current on first stage to get 160 volts at the plates, set current on second stage to get best signal (highest amplitude clean waveform) on the scope at about 320 volts. Drive at the output of the MOSFETs is around 150 volts P-P. All appears to be OK. The current for each of the tails in the diff amps is 8mA. That is 4mA per triode section. When I look at the anode resistors value, that looks to be about right for the design. My question is that I have read (MJ in Valve Amplifiers) that the 6SN7 is most linear in the 7-8mA range. Clearly, to get that current through the tubes, I am going to have to halve the value of the anode resistors. Is this required with the 6SN7 diff amp? The current values for the anode resistors are those recommended by Tubelab. Past experience is that setups recommended by Tubelab are usually for a reason. I am sure there is one here that I have missed or too dumb to understand Unfortunately my test equipment does extend to being able to do sophisticated distortion measurements.

Tubelab also mentioned adjusting the grid of the second triode in the first stage to help correct some DC offset. Not quite sure what this is describing. I suspect it might be similar (or the same) effect I noticed of having a small difference in the plate voltages of the first stage turn in to a large difference in the second stage. I trimmed the plate resistors within a couple of tenths of a percent of each other - still got the same effect. I have a whole bunch of Russian 6SN6 equivalents, so subbed in tubes until I got one two that gave identical plate voltages. Unfortunately I don't have well matched pairs of my "good" 6SN6s, so interested in implementing Tubelab's suggested circuit (or similar).

As for how does it sound, not sure yet. My main system is bi-amped with an active crossover. I have a Baby Huey doing duty for the mid range and an SE 6P15P for the tweeters. The new amp has a greater gain than the Baby Huey, so a direct replacement is not possible without a whole lot of re-tweeking the setup of the active crossover. Don't want to do that until I have sorted the big issues of the amp. Also will need it in the workshop as the model for the second mono-block.

Thanks Guys!

Chris

 

[Tubelab_com]

I have read (MJ in Valve Amplifiers) that the 6SN7 is most linear in the 7-8mA range.....The current values for the anode resistors are those recommended by Tubelab. Past experience is that setups recommended by Tubelab are usually for a reason. I am sure there is one here that I have missed or too dumb to understand

There is another reason that does not involve missing something or being dumb. The values shown in my diagram were the actual resistor values that were in place in my board when I drew up the diagram. I tend to start with values that are a bit on the conservative side especially when I have intentions of running the plate voltage up quite a bit higher than you are using. My amp is still in its infancy state and I still haven't decided on which output tubes to use, or what voltage to run them at.

If I remember correctly I was running 7403's at about 500 volts last time I had it running. Higher B+ voltage will produce more current with a given resistor value. It is also possible that I have changed the resistors since the schematic was drawn. Since about 1200 miles seperate me from the amp and my notes I really don't know right now.

From my experiments the 6SN7 has a fairly wide sweet spot (current and voltage where it works good) but I do tend to favor highish currents. In my circuits where the tube is CCS loaded I tend to run the current rather high, high enough where you have to consider the total tube dissipation. In this case the tube is resistively loaded so reducing the plate load resistor will reduce the gain a bit. There is a trade off between gain, drive voltage ability, and distortion. I was playing with tubes that required considerably more drive voltage than the 6L6GC, so my "optimum values" may not be the same as yours. It is more likely than not that I would change the resistor values as needed once I settled on the final output tube choices, power output level and B+ voltage. After the amp is running I use an FFT analyzer, and listening tests to tweak parts values for optimum sound.

Tubelab also mentioned adjusting the grid of the second triode in the first stage to help correct some DC offset

I have a large box of used 6SN7's. This is one of those tubes that I tend to collect when I find them cheap. I have found that matched sections are rare and don't mean much about the quality of the tube in other respects. I have a few of those RCA red based tubes that the audiophools drool over. They do sound nice, but the sections are far from matched. I wired a pot to the grounded end of R17 (no longer grounded). A cap is connected from this junction to ground. I used a .1 uF. The ends of the pot are connected through resistors to a positive and a negative regulated supply. The resistor values are adjusted so that the pot will vary the grid voltage from about +1 to -1 volt. This gives you a balance knob. Initially set it so that the voltages are equal at the plates of the second tube. It can be tuned for lowest distortion after the amp is running.

Good luck with your amp. I hope to get back to mine in the not too distant future. As with all projects I have come up with a dozen new ideas to try with this amp, but they involve Pentode wired sweep tubes and local negative feedback for big power output (ideas learned from playing with the red board). So far I would leave it just like it is for 6L6GC's and reasonable power levels, with some parts value tweaking as needed.

Unfortunately my test equipment does extend to being able to do sophisticated distortion measurements.

Maybe its time to look for a used X86 PC with a sound card. Even a Pentium 4 class machine with a 16 bit 96KHz sound card is a very useful tool. THD measurements are good, but being able to see the harmonic spectrum in real time while tweaking is cool. If I had only one tweaking tool this would be it. A fancy interface isn't needed. My first one was an 8 ohm load resistor and a 500 ohm pot. I am still using it until I get Petes fancy interface built, and I haven't even started on it yet.

 

[chrish]

Thanks again Tubelab, really appreciate the advice and assistance. I have learned quite a bit on this long process thanks to you and others! Regarding your balance control, I am guessing that using the regulated + and - supplies for the MOSFET followers will work? You say +/- 1 volt, so I am thinking of using a couple of zeners on the previously mentioned +/- supplies to give this, with appropriate current limiting resistors to give a milliamp or so of current? I will draw up a small schemo this afternoon.

As for computer for sound distortion measurements, I do have a 10" netbook computer that runs XP (Asus eee PC 1000H). Perhaps I can find some software that will be useful with this as it is conveniently small for my workbench.

Thanks again!

Chris

 

[Tubelab_com]

You say +/- 1 volt, so I am thinking of using a couple of zeners on the previously mentioned +/- supplies to give this, with appropriate current limiting resistors to give a milliamp or so of current?

I used some blue LED's that I had lying around. They give about 2.4 volts which makes setting the tiny 3/4 turn PC board trim pot a bit touchy, but not unreasonably so. They are wired off of the +/- mosfet supplies on my last PC board.

As for computer for sound distortion measurements, I do have a 10" netbook computer that runs XP (Asus eee PC 1000H). Perhaps I can find some software that will be useful with this as it is conveniently small for my workbench.

Does the netbook have connections for using the internal sound card? If so do you have a line level input? If so you need a circuit to reduce the speaker level output of the amp being tested to something the computers input can eat. I used a 500 ohm pot wired across my test load. This works, but accidentally turning the pot up too much will blow up your computers input.

I use Win Audio MLS software which is not cheap. Many forum members use Audiotester or RightMark. I am sure that there are others out there. Once you find software, run this test. Before trusting any measurements connect the PC's output right back to its input. Run distortion and frequency response tests. This is your measurement threshold. You will be able to accurately judge equipment that is at least 10 times worse than these measurements. This sounds bad, but my old computer measured something like 0.02% distortion and the frequency response was within a tenth of a db or so from 10 Hz to 40 KHz. Find me a tube amp that can do this! Unfortunately that computer died and I haven't built a new one yet.