Hey folks.
I've finally built my amp after months of learning and feedback thanks to you folks.
Here's the final design(x2, since it's stereo).
I think it sounds great, however, I have a couple of questions.
First, I had a problem with bass, especially repetitive like with four to the floor music. My speakers would "pop" very often. I'm not sure of how to describe it better. They would just jolt audibly and ruin the sound. Both channels would do it at the same time. I thought because I was using my prototype tubes(which I thought I surely damaged being a new learner), I'd swap them out. Still happened. I used a ECC81 in place of the ECC83S, same problem. I thought maybe it was a PSU issue, but even after switching off the AC it runs without distortion for at least 2-3 seconds, so I figured the tubes are sufficiently supplied.
I put the ECC83S back in and the problem never occurred again. As a software engineer by trade, I hate unreproducible issues. The only lead I have is that I have basically reseated the ECC83S and perhaps made a better connection.
I bought these sockets, as I heard porcelain was the worst for tolerances:
The Tube Store - 9 pin socket
They seem just as bad but I went with it. The pin alignment is wobbly and only fixed because I used solid core wire to solder to it. Do folks of this forum have a preferred socket that is solid and has no physical variance?
Does anyone think there is something wrong with my circuit that could otherwise explain this popping? (although I can't reproduce it)
Second part of my question. I have a partial bypass cap on the KT88 at the end. I used a run-of-the-mill electrolytic. In the books and online I've seen so many people say you should use an "audio grade" capacitor for bypassing, but they never say much more about it.
Can someone tell me exactly what qualities make a cap audio grade? Low ESR? Something else?
I've finally built my amp after months of learning and feedback thanks to you folks.
Here's the final design(x2, since it's stereo).
I think it sounds great, however, I have a couple of questions.
First, I had a problem with bass, especially repetitive like with four to the floor music. My speakers would "pop" very often. I'm not sure of how to describe it better. They would just jolt audibly and ruin the sound. Both channels would do it at the same time. I thought because I was using my prototype tubes(which I thought I surely damaged being a new learner), I'd swap them out. Still happened. I used a ECC81 in place of the ECC83S, same problem. I thought maybe it was a PSU issue, but even after switching off the AC it runs without distortion for at least 2-3 seconds, so I figured the tubes are sufficiently supplied.
I put the ECC83S back in and the problem never occurred again. As a software engineer by trade, I hate unreproducible issues. The only lead I have is that I have basically reseated the ECC83S and perhaps made a better connection.
I bought these sockets, as I heard porcelain was the worst for tolerances:
The Tube Store - 9 pin socket
They seem just as bad but I went with it. The pin alignment is wobbly and only fixed because I used solid core wire to solder to it. Do folks of this forum have a preferred socket that is solid and has no physical variance?
Does anyone think there is something wrong with my circuit that could otherwise explain this popping? (although I can't reproduce it)
Second part of my question. I have a partial bypass cap on the KT88 at the end. I used a run-of-the-mill electrolytic. In the books and online I've seen so many people say you should use an "audio grade" capacitor for bypassing, but they never say much more about it.
Can someone tell me exactly what qualities make a cap audio grade? Low ESR? Something else?
I'm surprised by the value of the cathode resistor. A total of 1000 Ohm looks way too high to me. What is your B+ voltage and what is the current running through a KT88? But if this detail is already discussed in an other thread, than just forget about my remark.
About wobbly "pin accepters" (what are they really called in English?) on tube sockets. My experience is that once you soldered them in, they will be OK. With some it's hard to get the tube in for the first time (sometimes I use a very thin piece of metal to wiggle an unwilling "pin accepter" a bit into shape). With difficult sockets I use an old/duff tube to make shure I don't damage a good tube but I untill now no tube (not even an old/duff one) ever got damaged in the process (well, as far as I know anyway...). For noval sockets, I have a plug with 9 pins, of which the tips of the pins are pointy. They tend to slide in easier than normal pins. Once the plug fits, the tubes will also fit.
I can't help you with the popping problem. Strange (but good at the same time) that now it's not reproducible anymore.
In my audio amplifiers I try go for low ESR capacitors for decoupling at the cathode. But to be honest, I find it hard to note any substantial audible difference, so for me it's more a matter of 'the idea' (why not make it better right away).
About wobbly "pin accepters" (what are they really called in English?) on tube sockets. My experience is that once you soldered them in, they will be OK. With some it's hard to get the tube in for the first time (sometimes I use a very thin piece of metal to wiggle an unwilling "pin accepter" a bit into shape). With difficult sockets I use an old/duff tube to make shure I don't damage a good tube but I untill now no tube (not even an old/duff one) ever got damaged in the process (well, as far as I know anyway...). For noval sockets, I have a plug with 9 pins, of which the tips of the pins are pointy. They tend to slide in easier than normal pins. Once the plug fits, the tubes will also fit.
I can't help you with the popping problem. Strange (but good at the same time) that now it's not reproducible anymore.
In my audio amplifiers I try go for low ESR capacitors for decoupling at the cathode. But to be honest, I find it hard to note any substantial audible difference, so for me it's more a matter of 'the idea' (why not make it better right away).
Last edited:
It is high and untraditional, but mucking around with values, this set well with my ears. Plate to cathode is 418V, plate to GND is 468, cathode to grid is -50.8. Current-wise, I read across one of the 500ohms as a voltage drop of 25.5, so around 51mA.
Yeah I hate that I can't make it happen again with a solid reason. In software, the worst thing to do is pretend it never happened(blame it on sunspots or planetary alignment at the time) only for it to come up again with a vengeance. I'm just hoping my theory is correct, but wanted to make sure there's nothing obvious in the circuit that could do it.
Yeah, I get that for sure. I'm using what I already have "in stock", and this is just for my own amusement anyway. Is low ESR what "makes" an audio cap? I'm just trying to understand the common definition.
Thanks Robert for all of your responses. You've really untangled a ton of my understanding through this process.
Like I said, it's higher than traditional, but I'm curious about your thoughts on it. I'm going through this primarily to learn.
OK, just my thoughts.
The KT88 in triode mode can take 40 Watts of plate+screengrid dissipation. You are now running them at 21.4 Watt (0.051 x 418), so rather 'cold' (less power output).
A (partial) non-decoupled cathode resistor creates local negative feedback but also highers the output impedance of the KT88. There's not much wrong with that (and the global NFB compensates part of this if I understand this correctly myself).
But if it sounds best to you this way and if it is loud/powerfull enough for you this way, than it must be the right way.
The KT88 in triode mode can take 40 Watts of plate+screengrid dissipation. You are now running them at 21.4 Watt (0.051 x 418), so rather 'cold' (less power output).
A (partial) non-decoupled cathode resistor creates local negative feedback but also highers the output impedance of the KT88. There's not much wrong with that (and the global NFB compensates part of this if I understand this correctly myself).
But if it sounds best to you this way and if it is loud/powerfull enough for you this way, than it must be the right way.
This isn't the first time I've heard this(although in a different form), and as I'm learning, still trying to wrap my head around it. Could you point out specifically the part that is concerning, and maybe a recommendation to resolve it? I've been asking not to be handheld, but this is one thing that keeps coming up that I just can't seem to get my head around.
Maybe I have an absolutely incorrect understanding of this, and I get that the load line wants to get as close to the maximum dissipation. I tried designing this a bunch of times and it ended up being a failure until I took a different approach. I started backward. I targeted a certain voltage swing(loaded OT against an 8 ohm resistive dummy load), worked back through the OT to see what the voltage swing would need to be out of the KT88, worked out the needed grid for that B+ and impedance, etc.. When biasing in this case, it seemed to me that running it hot or "colder" would not make that much of a difference. How do I put it... the slope of the load line is determined by my OT, which does not change. My cathode bias will determine how "hot" I'm running my tubes at idle, but the effective swing at either side of that should not change too much, right? Let's put it this way, if I put a jumper between the unbypassed resistor and ground while the music is playing, I hear a small tick(shifting bias of course) but no seriously discernible difference in volume or audio quality. Maybe I'm pooping all over the theory here, but doing a simple empirical tests(and even originally using way lower values like 100/150, complete bypass), I thought it sounded good and didn't have to use the typical lower values. I'm not sure what else to say 🙂 I originally up-biased it because I was already running my power transformer too hot(110mA out of 200mA max) and I didn't build the 2nd channel yet!
I'm getting a flat response(on a resistive loaded scoped output anyway) and getting upper-90dBs SPL in my listening room driving two MTM towers. I'm not sure I want it much louder to be honest, and if you're saying I'm running my tubes cold, maybe I won't have to buy new ones for a while 🙂
LTspice isn't friendly with PSUs, but here's roughly what it is.
Loaded, V1 comes out to about 480V I believe. B+, B++ and B+++ are directly coupled on both channels(e.g. B+ is tied directly to the ends of both OT primaries). The random 180k inserted there is just a bleeder.
For V1, it's a simple full wave UF4007 rectifier setup. Primary has a fuse and switch on it.
Here's what she ended up looking like, in case anyone was curious.
There's definitely a list of things I wish I had wired better, had I done this before, but overall worked out.
There's definitely a list of things I wish I had wired better, had I done this before, but overall worked out.
This topic is discussed in almost every publication about feedback. In short, and simplified:
Capacitors (mainly the coupling capacitors between stages), inductors and output transformers cause phase-shifts. That would not be a problem if the phase-shifts would be the same for every frequency but this is not the case. Negative feedback 'works' because the feedback signal has (enough) opposite phase to the signal where it's being 'reintroduced' in the amplifier. Instability occurs when parts (certain frequencies, or parts of the frequency band) of the feedback signal are (allmost) in phase with the original signal, while the total amplifier still does amplify these parts. Instead of negative feedback for these part(s) of the signal you wind up with positive feedback, so with oscillations.
Since there's not enough info on your output transformers, I don't think somebody could propose the right cure 'just like that'. I advise you to start with looking at the output signal with an oscilloscope while feeding the amplifier square wave signals of different frequencies. On the internet you can find publications in which pictures are shown of what to look for (and what you don't want to see).
Addition: I just saw the pictures of your build. Looks good (including the wiring)!
Last edited:
Very smart amplifier! Those power resistors that are bolted to the top plate really kill two bird with one stone with the convenient wiring anchor points - I will look into those for my next project.
I am interested in your grounding solution - it does not look like the typical star grounding solution, with a ground bus going to a central point near the power supply.
I am interested in your grounding solution - it does not look like the typical star grounding solution, with a ground bus going to a central point near the power supply.
I get it now. The concern that is being brought up is basically how this would handle a transient response, and this would show bandwidth, ringing, all sorts of problems. I'm extremely familiar with what it takes to make an accurate square wave(more from Fourier's than electronics), so yes. I thought everybody was talking about something else for some reason. I did already plan on shooting some square waves into it soon now that I have the proper equipment. My iPhone didn't have the bandwidth required to even output a square wave no matter the frequency. I will share my findings! Thank you for explaining it!
You're exactly correct. I'll admit I was a little lazy on the grounding and wanted to see what I could get away with. The top plate is aluminum and I did not insulate the bolts. Further, the grounds are connected via the red terminals you see need the preamp/driver tubes, those are then fed from a single wire coming from the 45º(left) terminal near the power supply. I figured if it was that bad I could always fix it.
If you couldn't tell, I'm working a little more empirically than by theory(isn't that supposed to be the fun part of building tube amps? 🙂 ), and I really have to shove my ear into the woofer to hear any hum, my beard scratching the cone, so I'm not sure if I'm going to centralize that better, yet.
For better low-frequency stability, stagger the time constants of the coupling networks more - make C4 10 nF. Input cap too - why pass frequencies that will saturate the output transformer? Also increase R7 to 10K, R17 to 33K for better decoupling. Input stages don't need that much voltage anyway.
A small cap across the feedback resistor is likely needed for best high-frequency stability. Best selected with a square wave test - enough to limit overshoot. Usually in the 47 to few hundred pF range.
A small cap across the feedback resistor is likely needed for best high-frequency stability. Best selected with a square wave test - enough to limit overshoot. Usually in the 47 to few hundred pF range.
Last edited:
Finally got around to testing this. At full power(before it started distorting a sine into an 8ohm resistive load@1kHz). So it raises a couple of other questions. Is there an objective measurement for when an overshoot becomes defined as a ring?
Here's the edge zoomed it(this is actually the negative edge, since that looked worse for some reason).
The ring comes out to about 80kHz(which surprised me, thought it would be lower), and dampened fairly quickly.
So that leads into my next question. This overshoot is a problem because it doesn't scale with frequency, so when I get into the upper values of the audible range, it really starts to "eat into" the line, if that makes sense. This leads to my next question...
At what frequency is typically the reasonable maximum for an amplifier to stop outputting a clean square wave? A 1kHz one should be easy as the harmonics are 3kHz, 5kHz, 7kHz, etc.. easily 9 harmonics in the audible range. When you get up to 10kHz, we're talking harmonics of 30kHz, 50kHz(!!), none of them in the audible range.
I know the real thing I'm looking for is what is considered a good slew rate, but just in terms of impulse testing, what do you folks typically consider reasonable when square wave testing?
Here's the edge zoomed it(this is actually the negative edge, since that looked worse for some reason).
The ring comes out to about 80kHz(which surprised me, thought it would be lower), and dampened fairly quickly.
So that leads into my next question. This overshoot is a problem because it doesn't scale with frequency, so when I get into the upper values of the audible range, it really starts to "eat into" the line, if that makes sense. This leads to my next question...
At what frequency is typically the reasonable maximum for an amplifier to stop outputting a clean square wave? A 1kHz one should be easy as the harmonics are 3kHz, 5kHz, 7kHz, etc.. easily 9 harmonics in the audible range. When you get up to 10kHz, we're talking harmonics of 30kHz, 50kHz(!!), none of them in the audible range.
I know the real thing I'm looking for is what is considered a good slew rate, but just in terms of impulse testing, what do you folks typically consider reasonable when square wave testing?
Last edited:
I just did some quick tests on the scope. The upper 180º is at around 90.9kHz, however it was very rolled off by that point(about -11dB compared to the same output at 1kHz), and the lower-end 180º is at about 4.3Hz(I could barely measure this with my crappy scope!) coming in at around -14dB compared to 1kHz as a baseline. My lower -3dB seems to be about 5Hz, and the upper -3dB seems to be around 60kHz.
I get my tube sockets from Cricklewood electronics UK, they are porcelain and gold plated for about £3GBP.
As said previously your output valve cathode resistors are way too high.
Check the negative feedback is the right phase or you will get instability problems.
As said previously your output valve cathode resistors are way too high.
Check the negative feedback is the right phase or you will get instability problems.
I have tried lower values on my prototype(between 100 and what I have now), but all it does is runs my tubes (and PSU) far hotter with marginal gain increase(less than 1dB more). I'm not sure what I'm missing here.
I finally went through the laborious process(I have an old fashion analog scope, and Excel) of getting the frequency/phase response, from 4Hz to 100kHz. I couldn't go much lower as the slope was too low for my scope to trigger anymore 😀
