Thanks for letting us watch (those who consider themselves students and spectators) Nice job Kevin!
KT120 Amp build
- Thread starter kward
- Start date
s-petersen
Scott
One question.Can c12 and c13 be elecrtolytics if there is no Dc to keep them formed?
Kevin -- You should find that there is indeed DC across these caps -- along with the AC component they are passing, so there should be no worries.
One comment however concerns data on the power amplifier page. You indicate that the screen tap percent for the A431 clone is 40%. Do you know this to be a fact? The reason I ask is that the screen percent of a bonafied A431 is in fact 33%. If the clone is not 33%, then that represents a substantial change for the clone, making it not a clone at all. On the other hand, it would be very easy to simply assume the taps were at 40%, since that seems to be the percent that all of today's transformer wizards think represents "UL". In reality of course, nothing could be further from the truth.
Dave
One comment however concerns data on the power amplifier page. You indicate that the screen tap percent for the A431 clone is 40%. Do you know this to be a fact? The reason I ask is that the screen percent of a bonafied A431 is in fact 33%. If the clone is not 33%, then that represents a substantial change for the clone, making it not a clone at all. On the other hand, it would be very easy to simply assume the taps were at 40%, since that seems to be the percent that all of today's transformer wizards think represents "UL". In reality of course, nothing could be further from the truth.
Dave
I measure 69 VDC across each of C12 and C13. Interestingly, I measure 52 VAC from either side of each cap to ground. Strange behavior this voltage doubler.
I emailed Triode Electronics today asking this question, because I just ASSumed everyone did 40% taps these days. Wrong again. Their tech said: "A clone is a clone in every respect or they can't call it a clone, except for the Teflon insulated leads. The taps are 33%."
So I've fixed the "final" schematic at post 499.
Also, Brutus (the name I gave the amp) has developed his first problem tonight. The heaters to the KT120s take about 90 seconds to warm up. Putting a volt meter probe on the positive leg of the bridge rectifier for the filament circuit of the KT120s shows its not conducting until about 85 seconds. If you recall, I changed out the original 12A large bridge rectifier for this smaller one to try to get rid of some switching noise that was seeping into the circuit ground and thus causing hum. The 8A bridge I put in its place was an old one--maybe 5 or 6 years--that I pulled from one of my other salvaged builds. This is the second time I've had a bridge rectifier do this. I'll pop over to RadioShack tonight and get a new one and also a big heat sink to set it inside of. That should cure it.
Kevin -- remember too that the bridge rectifier is not only passing the operating current for the KT120 tubes (about 4.5 amps in your case), but also the charging current for the filter caps as well. This will push the peak current very close to -- if not at -- the 8 amp repetitive peak current rating of your existing device. If it is not very well heat sunk, it is likely running hotter than a fire cracker as well.
For this application then, the current rating of this device needs to be significantly upgraded from the existing 8 A rating, and for long life operation, provided with adequate heat sinking as well.
Dave
For this application then, the current rating of this device needs to be significantly upgraded from the existing 8 A rating, and for long life operation, provided with adequate heat sinking as well.
Dave
Kevin -- for perspective on your heater rectifier, on my "ultimate" build, the heater supply provides a regulated 6.300 vdc at a continuous 8.5 amps in an amplifier similar to yours, and employs a 16 amp transformer, a 100,000 uF filter cap, and two 25 amp bridge diodes connected in parallel and mounted under the heat sink shown here. You can see the blue 100K cap in the bottom right of the pic, and the filament power transformer to the right of the heat sink. It has two 8 amp windings which are paralleled to easily handle to total current needs of the supply based on continuous operation. The transformer and heatsink become mildly warm in the summertime, but because of the conservative ratings of all these components, they have operated flawlessly for over 25 years of daily use, and will likely last my lifetime.
Dave
Dave
Attachments
Maybe my first relevant question. On Dave's recommendation I'm reading Morgan Jones' valve amplifies. I wish I was absorbing as much as the stuff going over my head. I happened to have just finished chapter on power supplies.If I understood correctly if the larger rectifier were noisy (because its larger?). You might be better off using Schottky rectifiers to build the bridge. Did I misunderstand?
Rust -- You have entirely too much idle time on your hands! 
Dave
Dave
I am in the middle of a feedback/stability evaluation on another project, which brought to mind the discussion surrounding circuit response and waveform displays with the project of this thread -- and in particular, the discussion of overshoot on a 10kHz square wave.
In an earlier thread, I mentioned that the overshoot is due to the inductance of the output transformer, and is ultimately unavoidable in typical quality output transformer designs. I also mentioned that finished amplifiers that do not display the overshoot typically have a reduced frequency response as well.
Since the discussion surrounded the tuning of the feedback and stability circuits of the active amplifier, I thought it might be helpful, if not instructive to show what happens to a 10 kHz square wave when passed through nothing more than an Acrosound TO-330 output transformer. When I mean nothing more, I mean just that. The output of the square wave generator is connected directly to the secondary of the transformer, and the scope is connected directly to the primary.
The waveform in the pic clearly shows the leading edge overshoot created by the inductance of the transformer. This particular series of transformers has an inherent response within +/- 1 db to 100 kHz, which is very good for the size an power level capability it represents.
It follows then that if a 10 kHz square wave display from the finished amplifier is notably different than that produced by the transformer alone, then the amplifier circuits must either be tapering off or emphasizing response either below or beyond the capabilities of the basic transformer, either of which will produce performance problems.
In the amplifier in this thread, with the waveform and flat response produced to 50 kHz, in a design using a transformer that is rated to 60 kHz, it indicates that the response of the active circuits is well matched to that of the transformer, and that good stability will be achieved.
Response can be actively increased or decreased relative to that of the transformer's, but stability will be compromised if going above it, and sonic presentation will generally be compromised either way you go -- either because of the effects of instability from a poor feedback amplifier (too high a response), or because the stability networks are affecting response down into the audible range (too low a response) -- all IMHO of course.
If nothing else, it also shows why casually swapping out OPTs in an established circuit can often lead to big problems!
Dave
In an earlier thread, I mentioned that the overshoot is due to the inductance of the output transformer, and is ultimately unavoidable in typical quality output transformer designs. I also mentioned that finished amplifiers that do not display the overshoot typically have a reduced frequency response as well.
Since the discussion surrounded the tuning of the feedback and stability circuits of the active amplifier, I thought it might be helpful, if not instructive to show what happens to a 10 kHz square wave when passed through nothing more than an Acrosound TO-330 output transformer. When I mean nothing more, I mean just that. The output of the square wave generator is connected directly to the secondary of the transformer, and the scope is connected directly to the primary.
The waveform in the pic clearly shows the leading edge overshoot created by the inductance of the transformer. This particular series of transformers has an inherent response within +/- 1 db to 100 kHz, which is very good for the size an power level capability it represents.
It follows then that if a 10 kHz square wave display from the finished amplifier is notably different than that produced by the transformer alone, then the amplifier circuits must either be tapering off or emphasizing response either below or beyond the capabilities of the basic transformer, either of which will produce performance problems.
In the amplifier in this thread, with the waveform and flat response produced to 50 kHz, in a design using a transformer that is rated to 60 kHz, it indicates that the response of the active circuits is well matched to that of the transformer, and that good stability will be achieved.
Response can be actively increased or decreased relative to that of the transformer's, but stability will be compromised if going above it, and sonic presentation will generally be compromised either way you go -- either because of the effects of instability from a poor feedback amplifier (too high a response), or because the stability networks are affecting response down into the audible range (too low a response) -- all IMHO of course.
If nothing else, it also shows why casually swapping out OPTs in an established circuit can often lead to big problems!
Dave
Attachments
Hi Shelly -- Absolutely! I do it routinely when I'm working with feedback networks for the exact reason you cite.
Dave
Dave
Hello. Kevin. I have fixed all voltages as you suggested and gave it a serious listen today. It sounds great to me especially in lower frequencies. I have a feeling that I hear all music preserved on record and the amplifier could go lower if necessary but there was simply nothing to reproduce. It is hard to speak about the sound but the feeling was there. I use Infinite Slope speakers with 10 and 12 inch woofers. Your observations may differ with smaller speakers. Despite all the talk about feedback and its positive impact I use none. It is stable as is. Thanks again. Robert.
I got a TO-220 heat sink from RadioShack and it cooled down the bridge rectifier significantly. I had my wife touch her finger (carefully) on the bridge (she was curious what I'm doing down in my dungeon) and she said "ow, that's hot." (I did it first to make sure it wouldn't burn).
But it's much cooler than it was. I've ordered a big ole heat sink that should take care of the rest of the heat. I'm also upgrading the bridge itself to a 10A device that will pass 200A surge. Thanks for the tip. By the way, my wife gave me the best compliment of all. She said, "Boy that amp you built sure sounds good." That's the best kind of compliment anyone can get.
And in fact, when underloading the feedback cap at 470 pF (too low of a value), I get a significant rise in frequency response that peaks at about 74kHz. But when putting on the 680pF, it gives the results I quoted for frequency resposnse flat to 50kHz, and ever so slightly dipping past that...no more bump at 74kHz. I assume this is following a 6 dB/octave slope but I didn't measure that, however I did sweep the frequency generator out to 800 kHz or so, and there were no additional anomalies (had to keep turning up the sensitivity of my scope to see the wave form, clear down to 2 mV/division at 800 kHz).
This brings up an interesting issue of another "famous brand" commercially produced amp that I have in my listening room. I reviewed its measurements that Stereophile published in the early 90's for it, and it shows the same kind of frequency spike at 60kHz or so. Hmmm... :headscrat
Schottkeys or fast/ultra-fast recovery diodes, both of which should reduce the switching noise. There's something about those big bridge rectifiers 12A and above--they are just noisy. At least every time I try them I always get switching noise which sounds like buzz into the circuit. I knew I had switching noise when I connected the ground of the RCA input jack to the negative post of the one of the 22000 uF filter caps right after that big rectifier. But that location is electrically the same ground at that location as the circuit ground bus. When connecting the ground of the RCA input jack to the ground bus directly, the noise went down significantly, but still there...that was the clue.
Moving down in size to an 8A bridge allowed the circuit to go totally silent. I assume a 10A bridge which has the same form factor will be equally silent.
Living with the amp for a week
I've got about 60 hours on the amp now. Really didn't open er up full throttle till tonight. Thought I might try running my SACD player wide open directly into the amp. I put on Leonard Coen "By the river's dark" from the album 10 new songs. Bass is tight like a horse kick to the chest and the top end is silky smooth.
If there's any drawback so far it might be this amp is very revealing of mediocre source material, more so than I expected. Little things are noticeable like changing the error amplifier 12AX7 tube from a Mullard to a Sovtek.
Now I'm starting to think passive preamp.
I've got about 60 hours on the amp now. Really didn't open er up full throttle till tonight. Thought I might try running my SACD player wide open directly into the amp. I put on Leonard Coen "By the river's dark" from the album 10 new songs. Bass is tight like a horse kick to the chest and the top end is silky smooth.
If there's any drawback so far it might be this amp is very revealing of mediocre source material, more so than I expected. Little things are noticeable like changing the error amplifier 12AX7 tube from a Mullard to a Sovtek.
Now I'm starting to think passive preamp.
Living with the amp for two more weeks
Finally I've been able to nail down how this thing sounds. When all the dust settled I found the amp had just a bit of leanness in the midrange, a hint of brightness on the top end, and the depth of the sound stage as well as harmonic content wasn't quite as deep or rich as I wanted. So I made a few tweaks, and I think this is where I'm going to leave it.
The amp sounds really very nice to me now.
If anyone's interested, here are the feedback and step values I used:
Feedback: R = 3.9 kOhms, C = 560 pF, tapped from 16 ohm secondary
Step: R = 20 kOhms, C = 33 pF
Finally I've been able to nail down how this thing sounds. When all the dust settled I found the amp had just a bit of leanness in the midrange, a hint of brightness on the top end, and the depth of the sound stage as well as harmonic content wasn't quite as deep or rich as I wanted. So I made a few tweaks, and I think this is where I'm going to leave it.
- I dialed down the feedback (was 18 dB) to 15 dB. This also gives a sensitivity of about 0.9 VRMS, which will work very nicely when driven from a passive preamp.
- Dialing down the feedback meant I needed to retune the step network and feedback cap. A 10 kHz square wave looks very similar to the last one I posted but the initial overshoot is about 3/4 the height of what it was before. All of this then gives a frequency roll starting at about 43khZ, instead of 52kHz as before,
The amp sounds really very nice to me now.
If anyone's interested, here are the feedback and step values I used:
Feedback: R = 3.9 kOhms, C = 560 pF, tapped from 16 ohm secondary
Step: R = 20 kOhms, C = 33 pF
Congrats Kevin, I appreciate you taking us along the way. Hopefully I learned something. I suspect based on the topology and components it sounds a lot like by Bob Latino VTA ST 120 ( of course I could be entirely wrong) . I really like the family tree too.
balle clorin
New Member
LTspice model KT120 as Triode
I used Curve Captor by Andrei Frolov to make this model of the Sofia KT120 triode curves.
* KT120So2k8 LTSpice model
* Modified Koren model (8 parameters): mean fit error 1.664mA
* Traced by erik 25/12/2013 using Curve Captor v0.9.1
* from sofia tracer
* kt120so2 LTSpice model
.subckt kt120so2k8 P G K
Bp P K I=(0.09616726061m)*uramp(V(P,K)*ln(1.0+(-0.1983471212)+exp((0.5067247971)+(0.5067247971)*((28.86336585)+(2.153515075m)*V(G,K))*V(G,K)/sqrt((-0.1596450898)**2+(V(P,K)-(-9.567928489))**2)))/(0.5067247971))**(1.391876027)
Cgk G K 10.2p ; 0.2p added
Cpk P K 9.3p ; 0.5p added
Cgp G P 29.5p ; 0.5p added
Rpk P K 1G ; to avoid floating nodes
d3 G K dx1
.model dx1 d(is=1n rs=2k cjo=1pf N=1.5 tt=1n)
.ends
guidelined to make models here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&cad=rja&ved=0CDYQFjAC&url=http%3A%2F%2Foleate.free.fr%2F___%2Fpong%2FDoc%2F2%2520Doc%2520Pr%25E9pa%2F9.%2520Tipe%2FDoc%2FSpice%2FHOW%2520TO%2520BUILD%2520YOUR%2520OWN%2520TRIODE%2520SPICE%2520MODEL.doc&ei=7967UsqyCaidyQOjroGgCQ&usg=AFQjCNEB88SCJMInXQkYxtcShILRbDXKaQ&bvm=bv.58187178,d.bGQ
I used Curve Captor by Andrei Frolov to make this model of the Sofia KT120 triode curves.
* KT120So2k8 LTSpice model
* Modified Koren model (8 parameters): mean fit error 1.664mA
* Traced by erik 25/12/2013 using Curve Captor v0.9.1
* from sofia tracer
* kt120so2 LTSpice model
.subckt kt120so2k8 P G K
Bp P K I=(0.09616726061m)*uramp(V(P,K)*ln(1.0+(-0.1983471212)+exp((0.5067247971)+(0.5067247971)*((28.86336585)+(2.153515075m)*V(G,K))*V(G,K)/sqrt((-0.1596450898)**2+(V(P,K)-(-9.567928489))**2)))/(0.5067247971))**(1.391876027)
Cgk G K 10.2p ; 0.2p added
Cpk P K 9.3p ; 0.5p added
Cgp G P 29.5p ; 0.5p added
Rpk P K 1G ; to avoid floating nodes
d3 G K dx1
.model dx1 d(is=1n rs=2k cjo=1pf N=1.5 tt=1n)
.ends
guidelined to make models here:
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&cad=rja&ved=0CDYQFjAC&url=http%3A%2F%2Foleate.free.fr%2F___%2Fpong%2FDoc%2F2%2520Doc%2520Pr%25E9pa%2F9.%2520Tipe%2FDoc%2FSpice%2FHOW%2520TO%2520BUILD%2520YOUR%2520OWN%2520TRIODE%2520SPICE%2520MODEL.doc&ei=7967UsqyCaidyQOjroGgCQ&usg=AFQjCNEB88SCJMInXQkYxtcShILRbDXKaQ&bvm=bv.58187178,d.bGQ
