KT120 Amp build

[kward]

The amp with 5751 input tube, and other refinements

Wow, its hard to believe it's been a year since I finished this amp. I've been slowly refining the design over the last six months or so, and I'm finally ready to publish some updated results.

There were a couple of things I didn't like with the original design as published on post #499:
1. The amp had a little too much gain.
2. The amp had too much feedback (15.5 dB) for my tastes.
3. With that much feedback, the sensitivity, at 1.15V RMS wasn't quite high enough.
4. I thought I could make it sound better.

The biggest changes were to swap out the 12AX7 front end tube for a 5751 and reduce feedback to 12 dB. This gave an input sensitivity of 0.75V RMS. Other changes were to reset the output stage idle current for a more optimal distortion reading, fine tune the phase inverter AC balance, and finally, re-tune the HF stability of the amp for all of these new parameters and conditions.

First of all, the new schematic is attached. Only the first page (of 3) changed, so I won't repeat the other two pages as given on post #499.

The 5751 tube was almost a direct plug in replacement. The only thing I changed was the cathode resistor (R4) to 2.2K (was 2K previously). I did this to provide a little buffer when the amp is driven to full power. Input sensitivity is 0.75V RMS or about 1.1V peak (as measured on my scope). The cathode bias now sits at 1.2V, which provides just a little buffer so that the grid-cathode voltage doesn't quite swing all the way to zero volts at max input signal swing. In practice, it turns out I could have left the cathode resistor at 2K, since it didn't change output distortion one bit at max power.

Additionally, I upped the quiescent current in the output stage to 72 mA per tube. This puts each tube dissipating about 35 watts (at 475V plate), which is roughly 60% of max. Lowest distortion (0.175% THD) at max power at 1KHz was achieved at a whopping 100 mA per tube! Too hot for my tastes, so I compromised at 60% of max dissipation (72 mA).

Here are the measurements I took:

Max power on the 8 ohm tap using an 8 ohm dummy load: 56 Watts, or 60 Watts with output stage cathode resistors bypassed.

Input sensitivity: 0.75V RMS

Frequency Response at 1 watt (relative to 1KHz) on an 8 ohm dummy load:
8Hz: -0dB
20Hz: -0dB
1KHz: -0dB
46KHz: -1dB
57KHz: -2 dB
65KHz: -3 dB

Power bandwidth (relative to 56 watts at 1KHz on 8 ohm dummy load): 18Hz (onset of notch distortion) to 30KHz (-1dB)

Stability: Stable under any loading conditions, including no load.

Output resistance on the 8 ohm tap: 1.5 ohms

THD measurements single channel driven at 1dB down from max power (or about 44.5 watts):
20Hz: 0.009%
1KHz: 0.24%
5KHz: 0.47%
10KHz: 0.64%
20KHz: 0.80%

How does removing the 10ohm resistors on the output stage cathodes affect THD? Using the same conditions as above:
20Hz: 0.009%
1KHz: 0.27%
5KHz: 0.49%
10KHz: 0.68%
20KHz: 0.85%

So there is some lowering of distortion using a small amount of resistance in the cathode circuits of the output stage.

First stage gain (with 12 dB feedback): 7.14

Second stage gain: 6.4 (from each output to ground)

Here's what an 8KHz square wave looks like (same shape as 10KHz but fits better on the scope screen). Very nice:



And finally, here's what the amp looks like today. I'm using a 1960's GE 5751 grey plate, and the Shuguang Treasure CV-181T in the phase inverter sections:



In summary, I continue to be impressed with the Triode Electronics A431S clone output transformers. These are simply excellent performers both on the low end and on the high end.

Also I need to mention that I built my sister a close copy of this amp last year but using the Hammond 1650R output transformers. The Hammond 1650R is a fine output transformer, BUT not to the level of performance of the A431S. The Hammonds don't go as low or as high, and the high frequency square waves don't look nearly as nice--they have fairly serious undulations across the tops of the square waves which I just could not eliminate, except at the expense of frequency response. These undulations are indicative of winding resonance, I believe. Of all the Hammond OPTs I've tried, I'm most favorable to the 1650H. That seems to be the sweet spot in the Hammond line.

Back to this amp, with the changes made as described, I am really liking what I'm hearing. The amp is a notch or two better sounding and performing than it was a year ago. In the last several weeks while finalizing the changes, I've been listening mostly to the Mannheim Steamroller Fresh Aire III and IV. I'm very familiar with these albums and I've never heard them sound so good.

Also, perhaps of lesser interest, but still noteworthy in my book, if you recall, this amp has a tube voltage regulator for the frontend stages. I tried several different tubes for the voltage regulator pass tube: Winged C 6L6, and Shuguang KT66. In both the Winged C and Shuguang cases, the front end just didn't sound as good (sounded a little edgier or harsher on the high notes). I settled on the Tung Sol new production 6V6 as the sweet spot tube for this amp for this position. I'm not sure why I can hear a difference...after all DC is DC, but I can definitely hear a difference when swapping out voltage regulator pass tubes. Also, I settled on my old favorite the Sovtek 12AX7 LPS in the error amplifier position, out of several I tried: Mullard 12AX7 new production and 1980's stock GE 5751.

 

[zekk]

Hi Kevin..... Thanks for the up-dates! Followed your's exactly despite using 7F7 and 7N7 at the front. Deep bass without any pitchyness!

However, I did something differently and that's connecting up the cathodes directly to ground, after setting the bias. Got this idea from my H.H.Scott 250 monoblocks. :thmbsp:

Later on, I'll change the GE 6550 for the GEC TT22 for a taste!

Enjoy The Music!

Regards, Zekk.

 

[dcgillespie]

Kevin -- I must have missed your last post with your updates. Very, very nice!

As you get even more into the experimentation, you might revisit the cathode resistors again, as it is the IM distortion that is reduced so significantly by having them in place, with the cathodes strapped together.

Great, great job!!

Dave

 

[kward]

Hi Kevin..... Thanks for the up-dates! Followed your's exactly despite using 7F7 and 7N7 at the front. Deep bass without any pitchyness!

However, I did something differently and that's connecting up the cathodes directly to ground, after setting the bias. Got this idea from my H.H.Scott 250 monoblocks. :thmbsp:

Later on, I'll change the GE 6550 for the GEC TT22 for a taste!

Enjoy The Music!

Regards, Zekk.

Nice. Looks like we ended up settling on the same front end tubes in their respective bases.

Kevin -- I must have missed your last post with your updates. Very, very nice!

As you get even more into the experimentation, you might revisit the cathode resistors again, as it is the IM distortion that is reduced so significantly by having them in place, with the cathodes strapped together.

Great, great job!!

Dave

Thanks again Dave. Your help was invaluable.

FYI to anyone reading, The 20 Hz THD measurement quoted in post 521 should be taken as suspect until I can figure out what is wrong with my distortion analyzer.

 

[Johnferro]

Kevin - what a great thread! This ties together much information in one place. On a different thread you provided a suggestion (from your KT120 amp circuit) on how to balance the long tailed pair that I implemented for a pair of 6L6 mono blocks. This has worked well and the inputs to the Power tubes are well balanced now. My question on your build is this - what does the step network on the plate of your pre-amp tube do. I have read this thread but must have missed the explanation.

 

[kward]

John, For starters, see post #181 where Dave provided this explanation. Recall C6 is the cap that straddles the feedback resistor, and C5 is part of the step network.

Kevin -- C6 does in fact control the transient response when the global loop is closed. C5 is a shorter loop that controls the open loop response relative to the global loop. It functions to prevent the global loop from trying to establish moon shot HF response levels. Both are very output transformer specific and best determined after the basic build is completed.

In my testing I've found that explanation to be right on. How does just one extra resistor and capacitor do this? The step network reduces the gain of the stage at higher frequencies--up to a point (the "knee" of the step response curve) where gain levels off. At that "knee point" the resistor in the step becomes more dominant than the reactance of the capacitor at that frequency. The summation of the resistance and reactance in the step network form an impedance that is in parallel with the plate resistor--in effect lowering the plate resistance at higher frequencies. Lower plate resistance means less gain for the stage. It's kinda like saying, anytime the tube sees a high frequency, I'm going to turn down the gain at the speed of the input signal so those high frequencies don't cause the amp to go unstable, but on lower frequencies, I'm going to leave the gain of the tube at its original setting--all done automatically with just one resistor and one capacitor.

 

[Johnferro]

Got it, that explanation makes sense to me as well. I will experiment with the KT88 mono-blocks. It looks like as your design progressed, you moved the knee lower in frequency on your step network. As I mentioned in the other post, the current circuit has 6.6K OPTs and the PT is 350-0-350 which yields about 440 volts on the plates at idle. I am hoping to boost the power a bit and am considering the CXPP100-MS-4.2K OPTs from Edcor as well as the XPWR239 PT which has a 420 V secondary which should yield 540 VDC on the plates of the KT88s at idle. My preliminary look at the load line with that OPT and running the tubes at 540 volts on the plates shows the amp running above the max dissipation of the tubes for a lot of the time. However if I use KT120s with 60 watt dissipation, things look more reasonable. I would welcome your thoughts on this.

 

[kward]

In fact, I have another custom wound toroid that is 420-0-420 on the HV wind that I was going to use, but in the end I decided not to since 60 watts seems plenty for my needs. I don't have any extensive experience running KT120s harder than what turned out on my build, but I do believe with ~550+ volts on the plates, the tube could support probably 90 watts output (push pull, ultralinear, fixed biased) without putting undue stress on the tubes.

What I did do for a short time experiment was bias up my output stage to about 100 mA per tube (but the plate voltage dropped to about 450V). I ran it that way for a few weeks just to see what would happen. Worked fine, but I'm not sure how many years the tubes would last under those conditions.

The other thing I would do differently if doing this build again today is use a full wave bridge instead of a full wave center tap--would have made for a less expensive transformer.

 

[oldgoatsrock]

Thread Index

1. Post #1: General discussion of design goals, circuit topology, and improvements
2. Post #238: Build schematic
3. Post #278: Bill of Materials
4. Post #343: Chassis layout
5. Post #351: Construction & Testing
6. Post #454: Measurements and Tuning
7. Post #499: Final schematic with 12AX7 input
8. Post #521: Final schematic with 5751 input

On this thread I intend to document the build for a 60 watt per channel KT120-based amplifier. I want to try the KT120 tube because I've never used it before, and everything I read about it indicates its a really great tube.

This is a continuation of http://www.audiokarma.org/forums/showthread.php?t=492619.

Attached are the schematics. I’d like some feedback on the circuit design. Is there some dumb thing I’ve done that if changed will improve the design? Is there anything missing that would greatly improve the sound or performance? Fire away with your comments.

After I get feedback on the circuit design, I plan to document the build on this same thread with a parts list and pictures as I progress through it. It’ll take a while (2 or 3 months) but I think it will be a good thing to share in the public forum. I hope tubes never die, and I’m doing my part to ensure that they don’t.

I’d love someone else to try this build also. It’s not exactly a beginner build, but you’ll have a great sounding high performance amp when you’re finished.

===========

Here’s the circuit description and design notes for those who are curious: (warning, the rest of this is kinda long)

One thing you should know is that I’m only showing one channel for the audio frequency amplifier (schematic page 1), so you’ll need to build page 1 of the schematic twice for stereo. The rest of the circuitry (schematic pages 2 and 3) are sized and shown as-is for two channels.

My design principles for this amp were to have no preconceptions of what works or doesn’t work and to make it as absolutely simple as possible while still maintaining very high quality sound reproduction. Input sensitivity is 5V RMS to drive the output stage to full power. (Edit--this has now been addressed. Input sensitivity is now about 1.15V RMS). This amplifier will produce about 60 watts per channel.

Schematic Page 1: (Audio amp)

The first stage is a common cathode voltage amplifier. I’ve tried other configurations for the first stage, such as a mu follower, SRPP, constant current sources, even a pentode, and they don’t make it sound better and just add complexity. I’ve also tried other tubes for this stage such as a 6DJ8/6922, and I don’t like them as much in this position. The good ole’ 12AX7 is a perfect tube for this stage and there are lots of choices for tube rolling, including many good NOS choices. It might be interesting to try a pentode strapped as a triode. If I did that, I’d consider the EF86.

Capacitor C23 is used to reduce HF instability. This is one part of the circuit I haven’t tried before so I’m guessing at the capacitor size. Somewhere between 330 pF and 470 pF seems about right and actually anything in that range will probably work just fine.

Circuit ground (labeled G in the schematics) is expected to be a ground bus. The circuit ground should be connected to the chassis at one point only, near the input tube. When I build it, the ground bus will be a 12 gauge bare copper wire pulled from a length of Romex.

The second stage is a cathode coupled differential pair phase splitter. I’ve tried the cathodyne configuration also, and while it’s a very simple phase splitter, I’d need a lot of headroom to run the output tubes and I didn’t have that kind of voltage available. I’ve also tried other tubes for this stage, including the 6H30p and 12AU7. The 6SN7 is the best tube I know of for this stage. It has inherently low distortion, allows very large signal swings, and has an octal base so it’s easy to work on when wiring it up.

The first stage is direct coupled to the second stage. I’ve tried capacitor (AC) coupling these stages and it doesn’t improve performance and adds complexity when instrumenting the feedback circuit.

The output tubes run at ~485 volts on the plates and are biased at about 60 mA per side. This is a medium idle for these KT120’s (!!) so these tubes should last a good long time. There is a balance adjustment circuit that is intended to be accessible from the top side using a screwdriver and a volt meter to keep each side of the push-pull pair at equal idle current. Since I didn’t make the bias of each output tube separately adjustable (seemed like added complexity that was unnecessary), you’ll need to use a matched quad for best left/right channel balance. You can direct substitute 6550, KT88 or KT90 and the amp should work just as well.

If you don’t change the fixed bias setting, you can even directly substitute a KT66 (but not a 6L6!!) and it will work pretty well, although it will idle the KT66’s at about 60% of capacity and output power will drop to about 45 watts per channel. Maybe try this if you’re really bored and want to tube roll, but ideally it’s not designed directly for a KT66. Just don’t try a 6L6. It’ll likely red plate them or fry the screen grids, or both.

The two output transformers that I know will work well with this amplifier are the Hammond 1650R and the Triode Electronics A431 Dynaco clone. Both should give really good performance, but the Hammond will reduce the output power to about 70 watts per channel since it's primary winding is a little higher impedance.

Schematic page 2 (Power supply)

Here’s where it gets a bit more interesting. The power transformer is a toroid. I’ve never had good luck with stacked EI core power transformers — they are more prone to mechanical buzz and can create hum problems in the audio circuit. Toroids are designed to keep most of the electromagnetic field inside the confines of the core. The toroid I’m using is from toroid.com. I chose it because it’s an “off the shelf” part and is a multi-wind, so I can fit everything on one chassis. R26 is a thermistor that limits current surge so the fuse doesn’t blow when you flip on the power switch.

Capacitors C9 and C10 are special. They are relatively new from ClarityCap, called the TC series. These are film caps and have extremely low ESR so they deliver energy almost instantaneously. Also these caps are rated at 700 volts so there is no need to stack them to get the required voltage tolerance. Downside is they are HUGE! If you don’t want to use the Claritys, you can easily substitute two stacked 220 uF electrolytics rated at 350V in place of each Clarity, and the amp will still sound really good.

I added a high voltage (HV) regulator so that at large signal swings (high output power), the voltage supplied to the input tube and phase splitter tubes will remain more or less constant. This greatly reduces distortion for both the first and second stages. It also improves bass response at high volumes.

Here’s where I may get some questions about why I did it this way. All filaments, including the power tubes, are DC supplied. I did this out of consideration for total heat generation and hum reduction, plus the availability of the windings on the toroid—they were there. I’m running the output tube filaments in series per pair of tubes, so that only half of the current is needed to power all four filaments (about 3.85 amps). A straight up all parallel connection using AC would require the transformer to supply almost 8 amps. This DC configuration will generate significantly less transformer heat and won’t saturate the core as much. The toroid is going to get a little warm, but I intend to mount it top side, so it will dissipate heat easily. The amplifier itself should run fairly cool.

Schematic page 3: (HV regulator and power supply)

The HV regulator is a series-pass design. I’ve tried many tubes for the series-pass element, such as a 6C19p, 6AS7, EL34, and a 6L6. They all work well, but when I tried the 6V6 (a new production Tung-Sol), I about peed my pants! It sounded that good. One thing I would like to try is a cold cathode constant voltage source instead of the 150V Zener. I think that might require a two chassis build though, as it’s probably going to be tight to get everything here on a single chassis. That might be one area I will experiment with in the future. I hear cold cathode voltage sources in voltage regulators can sound really awesome though.

The filament supply for the HV regulator is DC supplied because I needed to keep the voltage between the cathode and filaments within specification, so I’ve raised the filament bias to 266 volts. The 12AX7 error amplifier receives the same raised voltage on its filament.

Well, that’s the design. This should be a fun build and I think it’ll be a cool running gorgeous sounding little amp.

Kevin

Kevin, I'm new to AK and came across your threads in regards to your KT120 build. This peeked my interest because I'm interested in finding an amp to use with a Pilot 216-A preamp. Like you I enjoy taking on projects from start to finish. I don't have any formal training in electronics however with the information that you as well as others on AK have provided id like to give it a go with your Kt120 mkIII build. My question is do you still have and are willing to share your BOM and layout information mentioned in this thread?
Hope all is well and look forward to any advice.
R/

 

[kward]

Hello. Wow this thread is a blast from the past!

You definitely want to build the MkII or MkIII. As I recall the MkIII utilizes EFB while the MkII does not.

Unfortunately I do not have a BOM for either the II or III. But what I can offer is some guidance as you build it if you create a new thread about your build and document your progress.

If you want to contact me privately through a private message on AK, I can answer questions that way also.

-kevin

 

[oldgoatsrock]

Thanks for your future help, as you can tell I currently lack basic social media skills and will attempt contacting you via messaging.

Spencer

 

[crispycircuit]

Maybe repost the Mk 3 schematic for us?

 

[nerdorama]

In case Kevin is tied up, I save a copy.

 

[oldgoatsrock]

In case Kevin is tied up, I save a copy.

Thanks for the post. One of the things that impressed me about Kevin's work as well as other AK members is the attention to detail. I've been using this schematic to come up with a BOM and almost have it. Keeping in mind that this discussion took place several years back I'm currently researching costs. Trying not to reinvent the wheel I'll most likely stay with the same brand components unless my research comes up with something different.

R/

 

[kward]

I had a lot of help on the first one (the one discussed on this thread) . The Mk II and Mk III were refinements to it as I learned more along the way.

Concerning brand of components, most of the passive components are not critical of brand/manufacturer, however I tend to like to use JJ can caps in the power supply, Panasonic or Nichicon electrolytic caps in other places, Vishay 2W metal film resistors, Belton sockets, and Illinois coupling caps.

Unless the landscape has changed since 2018 when I built the Mk III (which is probably has a bit), I think you will be hard-pressed to find a less expensive power transformer or choke than what's specified on the schematic. Edcor power transformers are reasonably good but oftentimes buzz when operating normally (bobbin rattle typically), so special care needs to be taken when preparing them for use. I pot mine with thermally conductive epoxy and mount them on neoprene washers. Others have not had as good of experiences when potting their Edcor power transformers. Nonetheless, I have had decent results with a simple cold pour thermally conductive epoxy inside the end bells. It does takes a bit of prep to mount the power transformer on neoprene (rubber) washers so that the iron core is grounded--you need to expose the top side of the mount flanges so that the mount bolts conduct from the end bells to the chassis plate.

You really need to stick with the brand of output transformer that is specified on the schematic (Triode Electronics A431S), OR otherwise use an authentic copy (Dynakit Parts A431), OR otherwise use original (old stock) A431 Dynaco output transformers. If you use other budget brands/types of output transformers, I really think the amp's performance will suffer. The output transformers will probably be the single most expensive components of the amp (other than the chassis itself, depending on how extravagant you go). The amp was tuned specifically for Dynaco A431 output transformers, or the authentic copies from Triode or Dynakit parts, so deviating from that will reduce performance.

The EFB circuit is the frosting on the cake for this amp. It brings the amp's performance up a notch into what I would call "near top tier" performance (notice the THD numbers quoted on the schematic). Build the EFB circuit exactly as shown.

One more thing--avoid the temptation to want to "upgrade" the amount of capacitance in the power supply. Use exactly what's specified on the schematic for sizes. The power supply has been tuned to the best of my ability when built in 2018 for best operating performance. One upgrade you could do in the power supply is use Schottky diodes. They are expensive critters though for the size you'd need.