Building a valve amplifier

[horias2000]

I managed to perform a quick measurement with the feedback closed. Below you can see the frequency response (log frequency scale).

I'm also inserting the openloop response with log frequency scale, just for reference.

And I'm attaching the square wave response for 1kHz, 500Hz and 100Hz. The frequency response does look better than open loop but I still get a HF rolloff way too early. I'll look into that when I have more time.

 

[horias2000]

I removed C7 and the frequency response got a lot better (as expected).

So I now have -0.7dB at 20kHz and -3dB above 40kHz. I will add a smaller capacitor so that I will get -3dB at around 20kHz. I'm not sure what is the best practice here. Is it ok to have -3dB at 20kHz or I should go a bit higher than 20kHz?
As expected, the square wave response improved significantly (as seen below).

1kHz

and 10kHz

 

[horias2000]

I did a quick performance evaluation of the amplifier and the results are shown below. I'm happy with this amount of distortion (0.035%) but the 50Hz noise (or oscillation) needs to be dealt with. So I will focus on that in the coming days/weeks. If I can manage to get rid of that (or at least reduce it significantly) I will be very happy with the performance. The measurement was done with 60mA quiescent current through the KT88 output tubes. This can be increased a bit (65 - 70 mA). So the THD will most probably get better. The measured frequency response looks really good as well (apart from the 50Hz issues). I plan to run the other module in ultra linear and compare the performance between the two modes (pentode vs ultra linear).

 

[horias2000]

Just a quick update. I thought that the 50Hz could be coming from the ac filament supply/wires. I tried routing the wires as far away as possible from any sensitive parts and I also twisted the two wires together. Nothing, the 50Hz was still there. I tried using another 6.3V winding but with no improvement. I know that sometimes the 50Hz filament ac voltage can leak inside the small signal tubes. Therefore I used a regulated lab power supply for the two input tubes and BINGO! The 50Hz noise went down almost completely. I then looked at the spectrum and the frequency response again and this time it looks much better. I will most probably use DC for the low signal filaments or I will look online for tubes that do not exhibit this issue so much.

I must say that the performance of the amplifier looks to be very good in my opinion. Frequency response is really flat in the audio band and THD and THD+N look respectable in my opinion. With the 360Vdc supply I can get around 20W before clipping. I know this is not a lot but for my needs it's more than enough. More power can be achieved with a higher supply voltage.

This is for 1W output on 8OHM load

This is for 5W output

 

[Remk]

That looks good. On a pcb you almost have the need to feed the heaters with DC.
Make shure they are not flooting; use 2 resistors or a pot betweed heaters and ground if not already.

 

[horias2000]

As I said earlier, the second power amplifier PCB will be setup to work in UL mode so that I can assess witch mode has better performance (UL or pentode). I managed to do this today and in order for it to work in UL mode I had to perform a few modifications.

1. the grid voltage of the KT88 valves had to be lowered from around -18Vdc to around -40Vdc.
2. the R-C compensation network that I placed on the anodes of V1 was modified to 1k + 330pF. This gave a -3dB frequency of around 40kHz
3. the 100nF capacitor placed on the feedback input of V1 was increased to 200nF. This improved the LF response.

The closed loop frequency response is shown below. It looks good but in pentode mode it was a bit better. Not a big difference though.

The THD was the big surprise. It's clear that the 2nd harmonic is a lot bigger than it was in pentode mode. In pentode mode the THD was 0.017% (1W) and in ultra linear it is 0.096% at 1W. This is a big difference in my opinion. It gets even bigger at 5W, 0.038% vs 0.21%. Output power was about the same, 20W in UL mode and around 18W in pentode mode. Is this THD difference something normal? I was really expecting the THD in UL to be very similar to the THD in pentode mode but this is clearly not the case.

The conclusion is that I will modify the second module to work in pentode mode and I will use pentode mode for the final build. I will keep this thread updated with construction photos once I start.

1W output power

5W output power

 

[horias2000]

I know it is a long time since I updated this thread but this doesn't mean that I wasn't working on the project. After all the measurements done on the boards, the next step was to order an enclosure and to start fitting everything inside it. I chose an enclosure from ModuShop, the Galaxy. I tried to keep the costs as low as possible while still looking good. I already spent a big budget on the project so far so I didn't want to go overboard with the enclosure. I had some minimal customizations done on the front and back panel. I also wanted to have the top cover customized but this can only be done on aluminium covers and the cost was to high. I chose a top cover from steel and I drilled it myself. I will try to give it a paint cover before installing it.

As I said in previous posts, I had to use DC for the filaments of the input and driver tubes as the 50Hz component was way to high. For this I will use a separate toroidal transformer and two separate LM317 regulators (one for each channel). Another important thing is to not leave the 6.3Vdc for the input/driver tubes floating. So the ground of the 6.3Vdc connects to the main ground of the amplifier.

I managed to install and wire the left channel and I also performed some quick measurements to make sure everything works just fine. I am again baffled by these measurements and how good these are considering that I use tubes . Of course you get much better results from solid state amps but I have to say that these are not bad results in my opinion (0.02% THD at 1W).

I also chose to include the PHNO pre-amp inside the amp. That is the PCB in the upper right corner. Input switching will be done by relays and a rotary switch.

As per the advice I received, I also included a fan that is powered from the 6.3Vdc of the right channel. I'm not sure if this will suffice or not but I can always use a higher voltage to power it if this is needed. I really do not want this fan to be audible.

I hope to have the time to finalize all the wiring and actually listen to this things as I didn't have the chance to hear it yet

 

[SIY]

You're making me itch to get my next tube amp built.

If it's not too late... you might want to look at powering the heaters in a way to minimize common mode noise. There's a nice discussion of this in Morgan Jones's book. Here's how I did it in the raw supply of my little EL84 amp.

Note that no DC heaters are used, but the amp is quite hum-free.

 

[horias2000]

I know what you mean . I really want to finalize this build and listen to the amp. But when I will listen to it I'll be think of the next project . It's a continuous itch.

Interesting schematic. By the looks of it they use 12.6Vac for the filaments and the center tap of the transformer goes to the center tap of the EL84 filaments. The interesting part is the connection between the center taps to the resistive divider of B+ 2 and the 10uF capacitor.

As for my design, as far as I can see, the JJ KT-88 are not that problematic when it comes to hum. Having the 50Hz component at -100dB does not constitute a problem in my view . But the ECC82 and ECC83 are definitely more problematic. I'm not sure which one is the the one causing the issue or if both are affected in the same way. Due to the construction of the case I chose (aluminium sides) I can easily cool down the two LM318s. I still have to perform a longer term temperature measurements but I do not think there will be any major issues.

 

[SIY]

I know what you mean . I really want to finalize this build and listen to the amp. But when I will listen to it I'll be think of the next project . It's a continuous itch.

Interesting schematic. By the looks of it they use 12.6Vac for the filaments and the center tap of the transformer goes to the center tap of the EL84 filaments. The interesting part is the connection between the center taps to the resistive divider of B+ 2 and the 10uF capacitor.

As for my design, as far as I can see, the JJ KT-88 are not that problematic when it comes to hum. Having the 50Hz component at -100dB does not constitute a problem in my view . But the ECC82 and ECC83 are definitely more problematic. I'm not sure which one is the the one causing the issue or if both are affected in the same way. Due to the construction of the case I chose (aluminium sides) I can easily cool down the two LM318s. I still have to perform a longer term temperature measurements but I do not think there will be any major issues.

There's a few things going on there. If you want to remove both differential mode and common mode noise, both sides of the heater need to be driven and bypassed balanced. I didn't have a center tap on the 12V heater winding, so I artificially created one by series connecting the output tube 6V heaters to make a 12V heater. This allowed the 12V input stage heaters to be driven and bypassed balanced- generally, the hum from output tube heaters is pretty negligible, it's the input stage you need to worry about.

Second, the heater-to-cathode leakage can be greatly reduced by elevating the heater potential to be positive with respect to the cathode. In this circuit, the "virtual" CT is at AC ground (the 10u/100V cap ensures it) but DC is elevated by about 20% of the B+ voltage (the voltage divider formed by the 51k and 200k resistors).

If you want to be ultra fancy and use DC on the heaters, it will take two regulators, one for each side. Here's what I did for my MC phono preamp:

"H common" goes to the same kind of voltage divider that the CT goes to in the power amp schematic. The values I show here (100k and 33k) for the voltage divider reflect the difference in the B+ and won't be optimum for a power amp- you can just use the same 200k/51k as in the previous schematic.

 

[horias2000]

I managed to put everything together and to have a listen to it. I have to say that I really like the way it sounds. I can't say that it sounds better than a transistor amp but it's crispy clear all over the frequency band and low frequencies sound great. I guess the output transformers do a really good job.
I still have to do some more tweaking:

• I want to implement a delay for the anode voltage so that I do not stress the tubes at start-up. The heaters will start and after 30s or 1min the anode voltage will be applied.
• I still have to do something about the heaters for the KT88's. Each of these require 1,6A for the heater, so in total I need 6.4A and the transformer windings provide only 2x 3A. I still have a spare 6.3Vac winding that can output 2A. So I have two options here, I either connect all three 6.3Vac windings in parallel (3A + 3A + 2A) or I wire the 2A winding to one of the KT88s. Is there a preference here? I now have the two 3A windings in parallel going to all four KT88s.
• I want to improve the internal wiring. The earth wire needs to be well fastened to the chassis. I have to wire the internal led of the power switch.
• I want to perform some temperature measurements on some of the internal components. I will be using a PicoLog TC-08 temperature logger and measure the temperature for some internal components like the smaller toroidal transformer used for auxiliary voltages and the heaters of the 12AX7 and 12AU7 tubes, the big electrolytic capacitors, the MOSFETS and the LM317/LM337 regulators. I will also look at the temperature from the PCB surface near the output tubes. Just to make sure it doesn't get too hot.

If you guys think there are other things I can do in order to prolong the lifetime of the output tubes, then please let me know.

 

[SIY]

I managed to put everything together and to have a listen to it. I have to say that I really like the way it sounds. I can't say that it sounds better than a transistor amp but it's crispy clear all over the frequency band and low frequencies sound great. I guess the output transformers do a really good job.
I still have to do some more tweaking:

• I want to implement a delay for the anode voltage so that I do not stress the tubes at start-up. The heaters will start and after 30s or 1min the anode voltage will be applied.
• I still have to do something about the heaters for the KT88's. Each of these require 1,6A for the heater, so in total I need 6.4A and the transformer windings provide only 2x 3A. I still have a spare 6.3Vac winding that can output 2A. So I have two options here, I either connect all three 6.3Vac windings in parallel (3A + 3A + 2A) or I wire the 2A winding to one of the KT88s. Is there a preference here? I now have the two 3A windings in parallel going to all four KT88s.
• I want to improve the internal wiring. The earth wire needs to be well fastened to the chassis. I have to wire the internal led of the power switch.
• I want to perform some temperature measurements on some of the internal components. I will be using a PicoLog TC-08 temperature logger and measure the temperature for some internal components like the smaller toroidal transformer used for auxiliary voltages and the heaters of the 12AX7 and 12AU7 tubes, the big electrolytic capacitors, the MOSFETS and the LM317/LM337 regulators. I will also look at the temperature from the PCB surface near the output tubes. Just to make sure it doesn't get too hot.

If you guys think there are other things I can do in order to prolong the lifetime of the output tubes, then please let me know.

Much bird has been whipped about delayed anode B+, with dark mutterings about cathode stripping. Bottom line: for "normal" voltages and indirect heated cathodes, it's likely a non-issue. I give you amps like the Marantz 8B and 9, Dyna SCA-35, and H-K Citation II which had absolutely fine output tube life.

Also, 317/337 run very well hot, actually improving in performance. So if you see, say, 75-90 degrees C on them, leave well enough alone.

 

[horias2000]

Much bird has been whipped about delayed anode B+, with dark mutterings about cathode stripping. Bottom line: for "normal" voltages and indirect heated cathodes, it's likely a non-issue. I give you amps like the Marantz 8B and 9, Dyna SCA-35, and H-K Citation II which had absolutely fine output tube life.

Also, 317/337 run very well hot, actually improving in performance. So if you see, say, 75-90 degrees C on them, leave well enough alone.

That's interesting. I was reading that delayed anode voltage application can prologue the life of the output tubes significantly. I even saw recommendations to have half voltage on the heaters all the time when the amp is turned off in order to reduce thermal cycling. I will use the amp as it if for now and I will see how long the tubes last.

What about the heater windings issue? What do you think is the best approach? Have three output tubes connected to the two 3A windings in parallel and one tube to the 2A winding or put all three windings in parallel and go to all four tubes? I prefer to put them all three in parallel as I do not have to re-wire that part of the amp.

Below you can see the specs of the mains transformer.

 

[SIY]

I'm always wary of paralleling windings, mostly out of fretting about potential mismatches of exact voltage and the likelihood that I'll screw something up and accidentally put them in antiphase.

 

[horias2000]

I'm always wary of paralleling windings, mostly out of fretting about potential mismatches of exact voltage and the likelihood that I'll screw something up and accidentally put them in antiphase.

I agree. I don't like paralleling windings myself but the reasoning was that the two 3A windings are most probably wound together and therefore there should be no voltage difference between the two. The "hot" wires were also marked and this made it easier to not connect them in antiphase. This mains transformer is made for el34 tubes, hence the 3A filament windings. I will most probably leave it as it is or I will use the 3A windings separately for each channel. I'm drawing 3.2A instead of 3A. It should not affect the transformer to much. Again, many thanks for the input. Highly appreciated.

 

[egellings]

If you are in doubt about paralleling, say, two 6.3V windings, you can always try paralleling them with an AC ammeter in series with the two wires on one end and look for any circulating current due to unequal voltages across the windings when the transformer is powered up. If you see only a couple of milliamps or so, then it would be safe to parallel the windings.

 

[Eulipian]

One of my next projects is a valve amplifier. I built many solid state amplifiers and I'm very happy with these. I always considered valve amplifiers as being vastly inferior to solid state one but nevertheless I would like to build one to measure it and to listen to it. From all the available schematics and projects I saw, the one made by Bob Cordell (C7 KT88) stuck with me and I'm planning to build it. The only notable difference from the schematic published by BC is that I plan to use an ultra linear output transformer. I have a question regarding output transformers in general. Is there a reason to avoid using troridal ones? There is a company (Toroidy) that builds very good transformers in general. I bought several custom transformers from them for the power supply of solid state amplifiers and I can say these are high quality products. As I do not have that much experience with tube amplifiers (I'm still in the process of reading Morgan Jones's book) I do not know if toroidal transformers have any downsides when used as output transformers in tube amplifiers. I can still buy the "classical" Hammond transformers as these have very good reviews allover the internet. Any help is highly appreciated.

Take this FWIW from a non-engineer latent hobbyist: one source of old Hammond iron of high quality (though perhaps not the voltage taps you need) are Hammond organ and reverb chassis. Referred to in the hobby as "organ donors" they are used at the other end of the audio realm to build boutique guitar tube amps which distort at volumes and frequencies certain people like. I do not know if which of these have correct taps for the kinkless tetrodes. Again, ignore this if it's not helpful.

 

[egellings]

Take this FWIW from a non-engineer latent hobbyist: one source of old Hammond iron of high quality (though perhaps not the voltage taps you need) are Hammond organ and reverb chassis. Referred to in the hobby as "organ donors" they are used at the other end of the audio realm to build boutique guitar tube amps which distort at volumes and frequencies certain people like. I do not know if which of these have correct taps for the kinkless tetrodes. Again, ignore this if it's not helpful.

Toroidal OPTs are excellent for tube amps. The biggest problem with them is that the core can be easily saturated with magnetism if there is any substantial DC current imbalance in the primary that would cause core magnetization. I do not know if toroidal OPTs have gapped cores or not. A gap in the core would reduce susceptibility to becoming magnetized.

 

[horias2000]

Just a short update. I measured the currents between the paralleled windings and these are around 1.5uA. So basically zero . This is the same for the smaller 2A winding. So in my opinion there is no problem to have all three windings in parallel (3A + 3A + 2A). The mains transformer still gets hot but now at least I'm not stressing it outside its specs. I measured a bit below 60C on top of it (it's potted inside a metal enclosure) after 3h of continuous music playing. The transformer does not have anything specified about it's thermal performance in the datasheet so I contacted Toroidy and they quickly replied and told me that 75-80C is considered normal for this transformer and that it should never get above 115C. They also mentioned that it does not have a thermal fuse inside the windings. Considering that I measured a bit below 60C, I would say that inside the transformer it was 75-80C. So it should be fine.

In the coming days I will post the latest version of the schematic and BOM so that you guys can build it as well if you want to do so.

 

[horias2000]

I was curious if the measurement setup mistake affected the THD measurement for the tube amp. So I measured the THD again but this time with the actual speaker connected, not a resistor load. The speakers I have are 4OHM so I measured on the 4OHM output. It seems that the THD performance is even better than what I measured previously.

I will repeat the measurement on the 8OHM output with a resistive load so that I can compare like for like but this is looking really good. To have a THD of -81dB using tubes is really something .

On another subject, I noticed something when turning the amp off. There is a thump in the speakers (nothing really loud, just that you hear it) and I see a small light coming from the tubes. I was not able to pinpoint the actual source of the light. I think that it might have something to do with the fact that the +10V I use for the driver tube is also used by the PHONO preamp and due to this fact, the 10V will decrease very fast when I turn the amp off. Is this something that will have a negative impact on tube life? To verify this, I will perform a test by disconnecting the PHONO preamp and see is this happens again or not.