Dead Simple 1920's Triode Audio Amp
Sigurthr, Thu Dec 05 2013, 01:26AMHey everyone!
I finally got around to checking out my in-law's Victrola Radiola, and while it turned out to be a total loss for the radiola system, I was able to salvage the power supply and several triode tubes. I initially wanted to use only valves for the amp but I didn't have suitable impedance matching transformers for the inter-stage couplings. I tried with a second UX-226 and another 120:6.3 transformer (like what I'm using to match the high impedance output to a modern speaker) but it wasn't enough to yield proper grid drive for the output tube. I'd have needed three or four stages at the level of amplification I was getting just due to losses in impedance mismatches to get close to the performance I got using a silicon preamp to drive the single triode's grid. I had to put a 100uF cap in series with the output of the preamp to limit the bass response as my little dinky speaker couldn't handle the strong low frequencies. Ideally if I were building this out of more than just what I had laying around I'd have used a proper loudspeaker, proper impedance matching transformers, and used sockets for the tube(s) instead of just direct soldering to the base pins. I probably wouldn't have mounted the speaker using a hard drive magnet either, haha.
Overall topology is this: Mono audio is supplied from MP3 player, goes into LM386 preamp for voltage amplification to +-2.5V (LM386 is powered by 5Vdc from rectified and filtered 5VAC filament line), which is then fed through a high pass filter and impedance matching and isolation transformer (4 to 50ohms, step up). The signal then drives the grid which is biased at -20V referenced to cathode. Output is pulled from a winding on the B+ transformer and then fed into an impedance matching and isolation transformer (120:6.3, step down) to drive an 8Ohm modern midrange speaker.
Filmed on an iPhone, so forgive the taps and clicks as I manually focused the camera.
Re: Dead Simple 1920's Triode Audio Amp
Sigurthr, Wed Dec 11 2013, 03:21AM
12/9
Realized there was an interstage transformer included in the original Radiola 18. Ripped it out and remade my original double triode circuit:
https://www.youtube.com/watch?v=7RQTCzoibtA
Will have to see about getting some four pin sockets, rectifying and filtering the filament lines (to remove hum), and a proper speaker.
It feels good to say that after years of wanting to; I have finally built a fully tube amp.
12/10
I did some experimenting with negative feedback on my dual triode amp today. While it worked well to suppress noise, the loss in gain was quite noticeable. I guess that's why they didn't really employ it for triodes, but rather tetrodes and pentodes. I also experimented with positive feedback for additional gain, but it was pretty marginal.
Data for others to benefit from: Negative Feedback is easiest to implement via a series resistor between the output valve's audio-pickoff transformer's secondary and the first stage valve's cathode. One leg of the transformer secondary will yield negative feedback, while the other will yield positive feedback. Adding L and or C will affect the frequency response of the NFB loop to shape the output sound.
Negative Feedback + less series resistance = less gain, stronger effect of sound shaping.
Negative Feedback + more series resistance = more gain retained, weaker effect of sound shaping.
Adding a series capacitor forms a high-pass filter in the negative loop, which has a net Low-Pass effect on the sound.
Adding a series inductor forms a low-pass filter in the negative loop, which has a net High-Pass effect on the sound.
Positive feedback loops are more intuitive; more resistance = less additional gain. Series L and C values do not have "inverse of normal" effects on the sound. Depending upon the bias points used and the series resistances chosen positive feedback can introduce a lot of distortion as the tube goes into full conduction and begins to flat-top the signal.
I tried many, many values and ranges of R and C for the NFB loop but couldn't find a sweet point of gain vs low-pass effect that made the output sound better than without the NFB loop. I'm using a higher frequency range small speaker without an enclosure so there is a large inherent high-pass effect. Despite achieving significant results in low-pass filtering I still couldn't balance the sound profile. It would go from "muddy AND tinny" to "thin" with no sweet spot in between. I tried 0-500kOhms linear taper and 1pF to 1uF in x10 increments. The best altered sound I achieved was with a combination of 2.2kOhms + 80nF/47nF (one was better at one song, another for another) and a Positive feedback loop with only a series junkbox mystery inductor somewhere on the order of 100-100uH. The PFB enhanced the bass response while the NFB reduced the treble. Unfortunately nothing seemed to reduce the 60Hz hum (I suspect I'd need a notch filter) or the real high frequencies, above about 7kHz, which was very strange. The low pass effect of the NFB loop should have rolled off everything over the 3dB frequencies but really it just had the effect of a negative Gaussian curve. The result was upper-midrange scooping with linear reduction in bass/treble just due to loss of overall gain. Not sure why that happened.
In the end I decided to stick with the 1920's topology of no feedback and just deal with the frequency response by other means (mostly LC filtering the output of the impedance matching transformer to the low impedance speaker).
I tried briefly to rectify and filter the heater lines to nix the hum but had no luck. Oddly enough the heater wouldn't glow under DC, even though they do from my bench supply. This power supply (for the amp) uses an odd means of grid bias by having the heaters (and consequently the cathodes) at about 23V relative to true ground and then grounding the grids. This makes it difficult to alter without screwing things up. I suspect that due to the large negative bias and zero isolation something ran afoul (hard to believe with 1 diode and 1 cap!) in my wiring. I had rectified and filtered an unused filament line before but it was the 5V line for the 227 tube that I'm not using, and that one's heater isn't held so far negative to ground. Plus, there was a layer of isolation because of the impedance matching transformer used between the silicon preamp and the grid of the 226.
I'd certainly consider implementing NFB again if I can get a proper sharp slope 60Hz notch filter figured out.
Sigurthr, Wed Dec 11 2013, 03:21AM
12/9
Realized there was an interstage transformer included in the original Radiola 18. Ripped it out and remade my original double triode circuit:
https://www.youtube.com/watch?v=7RQTCzoibtA
Will have to see about getting some four pin sockets, rectifying and filtering the filament lines (to remove hum), and a proper speaker.
It feels good to say that after years of wanting to; I have finally built a fully tube amp.
12/10
I did some experimenting with negative feedback on my dual triode amp today. While it worked well to suppress noise, the loss in gain was quite noticeable. I guess that's why they didn't really employ it for triodes, but rather tetrodes and pentodes. I also experimented with positive feedback for additional gain, but it was pretty marginal.
Data for others to benefit from: Negative Feedback is easiest to implement via a series resistor between the output valve's audio-pickoff transformer's secondary and the first stage valve's cathode. One leg of the transformer secondary will yield negative feedback, while the other will yield positive feedback. Adding L and or C will affect the frequency response of the NFB loop to shape the output sound.
Negative Feedback + less series resistance = less gain, stronger effect of sound shaping.
Negative Feedback + more series resistance = more gain retained, weaker effect of sound shaping.
Adding a series capacitor forms a high-pass filter in the negative loop, which has a net Low-Pass effect on the sound.
Adding a series inductor forms a low-pass filter in the negative loop, which has a net High-Pass effect on the sound.
Positive feedback loops are more intuitive; more resistance = less additional gain. Series L and C values do not have "inverse of normal" effects on the sound. Depending upon the bias points used and the series resistances chosen positive feedback can introduce a lot of distortion as the tube goes into full conduction and begins to flat-top the signal.
I tried many, many values and ranges of R and C for the NFB loop but couldn't find a sweet point of gain vs low-pass effect that made the output sound better than without the NFB loop. I'm using a higher frequency range small speaker without an enclosure so there is a large inherent high-pass effect. Despite achieving significant results in low-pass filtering I still couldn't balance the sound profile. It would go from "muddy AND tinny" to "thin" with no sweet spot in between. I tried 0-500kOhms linear taper and 1pF to 1uF in x10 increments. The best altered sound I achieved was with a combination of 2.2kOhms + 80nF/47nF (one was better at one song, another for another) and a Positive feedback loop with only a series junkbox mystery inductor somewhere on the order of 100-100uH. The PFB enhanced the bass response while the NFB reduced the treble. Unfortunately nothing seemed to reduce the 60Hz hum (I suspect I'd need a notch filter) or the real high frequencies, above about 7kHz, which was very strange. The low pass effect of the NFB loop should have rolled off everything over the 3dB frequencies but really it just had the effect of a negative Gaussian curve. The result was upper-midrange scooping with linear reduction in bass/treble just due to loss of overall gain. Not sure why that happened.
In the end I decided to stick with the 1920's topology of no feedback and just deal with the frequency response by other means (mostly LC filtering the output of the impedance matching transformer to the low impedance speaker).
I tried briefly to rectify and filter the heater lines to nix the hum but had no luck. Oddly enough the heater wouldn't glow under DC, even though they do from my bench supply. This power supply (for the amp) uses an odd means of grid bias by having the heaters (and consequently the cathodes) at about 23V relative to true ground and then grounding the grids. This makes it difficult to alter without screwing things up. I suspect that due to the large negative bias and zero isolation something ran afoul (hard to believe with 1 diode and 1 cap!) in my wiring. I had rectified and filtered an unused filament line before but it was the 5V line for the 227 tube that I'm not using, and that one's heater isn't held so far negative to ground. Plus, there was a layer of isolation because of the impedance matching transformer used between the silicon preamp and the grid of the 226.
I'd certainly consider implementing NFB again if I can get a proper sharp slope 60Hz notch filter figured out.
Re: Dead Simple 1920's Triode Audio Amp
Ash Small, Wed Dec 11 2013, 09:58AM
If you're using a heater secondary for the rectified DC you should be able to compensate for the grid bias, because it's an isolated supply. Are the diodes dropping the voltage too far?
I'd try a rectified 12V transformer, and resistor voltage divider, or even a battery, to see if DC heater current eliminates the mains hum. Once you've established it's effectivness, you can devise something a bit more refined.
Ash Small, Wed Dec 11 2013, 09:58AM
Sigurthr wrote ...
I tried briefly to rectify and filter the heater lines to nix the hum but had no luck. Oddly enough the heater wouldn't glow under DC, even though they do from my bench supply. This power supply (for the amp) uses an odd means of grid bias by having the heaters (and consequently the cathodes) at about 23V relative to true ground and then grounding the grids. This makes it difficult to alter without screwing things up. I suspect that due to the large negative bias and zero isolation something ran afoul (hard to believe with 1 diode and 1 cap!) in my wiring. I had rectified and filtered an unused filament line before but it was the 5V line for the 227 tube that I'm not using, and that one's heater isn't held so far negative to ground. Plus, there was a layer of isolation because of the impedance matching transformer used between the silicon preamp and the grid of the 226.
I tried briefly to rectify and filter the heater lines to nix the hum but had no luck. Oddly enough the heater wouldn't glow under DC, even though they do from my bench supply. This power supply (for the amp) uses an odd means of grid bias by having the heaters (and consequently the cathodes) at about 23V relative to true ground and then grounding the grids. This makes it difficult to alter without screwing things up. I suspect that due to the large negative bias and zero isolation something ran afoul (hard to believe with 1 diode and 1 cap!) in my wiring. I had rectified and filtered an unused filament line before but it was the 5V line for the 227 tube that I'm not using, and that one's heater isn't held so far negative to ground. Plus, there was a layer of isolation because of the impedance matching transformer used between the silicon preamp and the grid of the 226.
If you're using a heater secondary for the rectified DC you should be able to compensate for the grid bias, because it's an isolated supply. Are the diodes dropping the voltage too far?
I'd try a rectified 12V transformer, and resistor voltage divider, or even a battery, to see if DC heater current eliminates the mains hum. Once you've established it's effectivness, you can devise something a bit more refined.
Re: Dead Simple 1920's Triode Audio Amp
Proud Mary, Wed Dec 11 2013, 10:07AM
Part of the hum may be from ambient 50/60 Hz field. Are all points at high impedance - especially grid circuits - shielded? Are all earth points on the triodes taken to a single point on the chassis to prevent hum loops?
Proud Mary, Wed Dec 11 2013, 10:07AM
Part of the hum may be from ambient 50/60 Hz field. Are all points at high impedance - especially grid circuits - shielded? Are all earth points on the triodes taken to a single point on the chassis to prevent hum loops?
Re: Dead Simple 1920's Triode Audio Amp
Sigurthr, Thu Dec 12 2013, 01:09AM
Here's the circuit of the amp as shown in my second post:
I tried directly posting the image but it kept being way too large, even though I resized it smaller and smaller. For some reason when I post it here using the IMG tags it would blow it up to original size. I never did figure out the internal image hosting aspect of 4HV. Imageshack has always worked well for me, but for some reason it isn't working well here.
When I tried rectifying the filament line for the 226 I simply placed two series diodes (1N4003) and one capacitor (50V 470uF) in parallel with the heater line to the tube. I measured the voltage across the cap as 1.48V DC. Measuring the AC voltage across the terminal strip with no 226 hooked up showed 1.44Vac.
I figured that if I were to put any isolation between the terminal strip and the 226's filament then the cathode would no longer be positive ~20V in relation to ground, and thus the grid bias would be removed. When I was using the silicon preamp I rectified and filtered the 5Vac heater line which is unused, and there was an isolating transformer between the output of the LM386 and the grid of the UX-226, so as to not interfere with the grid bias.
Unfortunately nothing is shielded and the power supply itself is not chassis grounded. I will have to find a three conductor cord to ground the power supply. Likewise the grid circuits are not physically far removed from the heater circuits. Once I get the sockets in the mail I'll wire everything up a lot more professionally.
I can live with the hum if I have to, but the poor bass response is what bothers me most. Adding negative feedback via a 85nF film cap between the 165V B+ and the 1.5V Heater terminals cuts the gain by a little under 3dB or so, and helps to attenuate some of the higher frequencies, but still there is a pretty sharp bass roll off below about 300Hz. I pulled up a schematic of the power supply and output unit and it uses a 0.5uF cap to couple the audio from the 165V B+ line, so I figured perhaps this is causing the weak bass. But when I put 470uF in parallel with the 0.5uF there was zero noticeable increase in bass response. None of the feedback I experimented with really added much attenuation to the 5kHz and up part of the spectrum either, which I thought was very odd. No matter how large I made the feedback capacitor the high end of the treble still blasted through.
Sigurthr, Thu Dec 12 2013, 01:09AM
Here's the circuit of the amp as shown in my second post:
I tried directly posting the image but it kept being way too large, even though I resized it smaller and smaller. For some reason when I post it here using the IMG tags it would blow it up to original size. I never did figure out the internal image hosting aspect of 4HV. Imageshack has always worked well for me, but for some reason it isn't working well here.
When I tried rectifying the filament line for the 226 I simply placed two series diodes (1N4003) and one capacitor (50V 470uF) in parallel with the heater line to the tube. I measured the voltage across the cap as 1.48V DC. Measuring the AC voltage across the terminal strip with no 226 hooked up showed 1.44Vac.
I figured that if I were to put any isolation between the terminal strip and the 226's filament then the cathode would no longer be positive ~20V in relation to ground, and thus the grid bias would be removed. When I was using the silicon preamp I rectified and filtered the 5Vac heater line which is unused, and there was an isolating transformer between the output of the LM386 and the grid of the UX-226, so as to not interfere with the grid bias.
Unfortunately nothing is shielded and the power supply itself is not chassis grounded. I will have to find a three conductor cord to ground the power supply. Likewise the grid circuits are not physically far removed from the heater circuits. Once I get the sockets in the mail I'll wire everything up a lot more professionally.
I can live with the hum if I have to, but the poor bass response is what bothers me most. Adding negative feedback via a 85nF film cap between the 165V B+ and the 1.5V Heater terminals cuts the gain by a little under 3dB or so, and helps to attenuate some of the higher frequencies, but still there is a pretty sharp bass roll off below about 300Hz. I pulled up a schematic of the power supply and output unit and it uses a 0.5uF cap to couple the audio from the 165V B+ line, so I figured perhaps this is causing the weak bass. But when I put 470uF in parallel with the 0.5uF there was zero noticeable increase in bass response. None of the feedback I experimented with really added much attenuation to the 5kHz and up part of the spectrum either, which I thought was very odd. No matter how large I made the feedback capacitor the high end of the treble still blasted through.
Re: Dead Simple 1920's Triode Audio Amp
Proud Mary, Thu Dec 12 2013, 10:11AM
If you load your image into the Attachments thread, you can link to it from there.
A directly heated small signal valve designed for battery applications will amplify any residual ripple on your filament supply just as surely as if you stuck a signal on the grid. Use batteries if you can't get it smooth enough.
Providing grid-bias to directly heated valves always involves extra circuitry, such as chokes in the filaments lines, which is one of the reasons that spurred on the development of the indirectly heated cathode. Battery powered radios designed before very roughly 1950 had a separate grid bias battery to overcome this problem. Any hum or ripple on a grid bias supply will modulate the current through the valve.
I think you will find that if you use a suitable output transformer providing a reasonable impedance match at adequate bandwidth, your audio will sound much better.
P.S. "Filament" is for directly heated valves, and "heater" for indirectly heated cathode types.
Proud Mary, Thu Dec 12 2013, 10:11AM
Sigurthr wrote ...
Here's the circuit of the amp as shown in my second post:
I tried directly posting the image but it kept being way too large, even though I resized it smaller and smaller. For some reason when I post it here using the IMG tags it would blow it up to original size. I never did figure out the internal image hosting aspect of 4HV. Imageshack has always worked well for me, but for some reason it isn't working well here.
Here's the circuit of the amp as shown in my second post:
I tried directly posting the image but it kept being way too large, even though I resized it smaller and smaller. For some reason when I post it here using the IMG tags it would blow it up to original size. I never did figure out the internal image hosting aspect of 4HV. Imageshack has always worked well for me, but for some reason it isn't working well here.
If you load your image into the Attachments thread, you can link to it from there.
Sigurthr wrote ...
When I tried rectifying the filament line for the 226 I simply placed two series diodes (1N4003) and one capacitor (50V 470uF) in parallel with the heater line to the tube. I measured the voltage across the cap as 1.48V DC. Measuring the AC voltage across the terminal strip with no 226 hooked up showed 1.44Vac.
When I tried rectifying the filament line for the 226 I simply placed two series diodes (1N4003) and one capacitor (50V 470uF) in parallel with the heater line to the tube. I measured the voltage across the cap as 1.48V DC. Measuring the AC voltage across the terminal strip with no 226 hooked up showed 1.44Vac.
A directly heated small signal valve designed for battery applications will amplify any residual ripple on your filament supply just as surely as if you stuck a signal on the grid. Use batteries if you can't get it smooth enough.
Sigurthr wrote ...
I figured that if I were to put any isolation between the terminal strip and the 226's filament then the cathode would no longer be positive ~20V in relation to ground, and thus the grid bias would be removed. When I was using the silicon preamp I rectified and filtered the 5Vac heater line which is unused, and there was an isolating transformer between the output of the LM386 and the grid of the UX-226, so as to not interfere with the grid bias.
I figured that if I were to put any isolation between the terminal strip and the 226's filament then the cathode would no longer be positive ~20V in relation to ground, and thus the grid bias would be removed. When I was using the silicon preamp I rectified and filtered the 5Vac heater line which is unused, and there was an isolating transformer between the output of the LM386 and the grid of the UX-226, so as to not interfere with the grid bias.
Providing grid-bias to directly heated valves always involves extra circuitry, such as chokes in the filaments lines, which is one of the reasons that spurred on the development of the indirectly heated cathode. Battery powered radios designed before very roughly 1950 had a separate grid bias battery to overcome this problem. Any hum or ripple on a grid bias supply will modulate the current through the valve.
I think you will find that if you use a suitable output transformer providing a reasonable impedance match at adequate bandwidth, your audio will sound much better.
P.S. "Filament" is for directly heated valves, and "heater" for indirectly heated cathode types.
Re: Dead Simple 1920's Triode Audio Amp
Sigurthr, Tue Dec 17 2013, 02:00AM
Worked out a suitable negative feedback scheme and fixed the bass response via controlled backpressure of the loudspeaker. Cleaned her up all nice and fancy too.
Sigurthr, Tue Dec 17 2013, 02:00AM
Worked out a suitable negative feedback scheme and fixed the bass response via controlled backpressure of the loudspeaker. Cleaned her up all nice and fancy too.
Re: Dead Simple 1920's Triode Audio Amp
Proud Mary, Tue Dec 17 2013, 10:26AM
Good to see you've got it running!
Proud Mary, Tue Dec 17 2013, 10:26AM
Good to see you've got it running!
Print this page