Ok… so I think I found the start of the problem…
with the input pot set to “comfortable” listening levels, which is barely a smidge above minimum… the grid of first stage triode is seeing the rolloff.
If i disconnect the pot from the grid, and measure off just the wiper , the rolloff is gone.
(PS not sure why it’s showing a weird LF boost, the picoscope measures this straight from its own output… not sure if it’s some sort of calibration issue or not.)
with the input pot set to “comfortable” listening levels, which is barely a smidge above minimum… the grid of first stage triode is seeing the rolloff.
If i disconnect the pot from the grid, and measure off just the wiper , the rolloff is gone.
(PS not sure why it’s showing a weird LF boost, the picoscope measures this straight from its own output… not sure if it’s some sort of calibration issue or not.)
I would grab that Bode plot program and use that for a better look at the frequency response.
https://www.picotech.com/library/picoapp/frequency-response-analyzer-with-bode-plots
You want to test at about 1 watt output, which you can figure out by generating a 1K sine wave and the RMS measurement tool in the scope program to determine the input voltage that gives you 1 watt output. Then close Picoscope and open the Frequency Response Analyser. Channel A goes to the input and Channel B to the ouput. Set the input voltage for the 1 watt output, set the FR you want to measure, and enter 50 in the Steps box. This will give you a very fine-grained look. You can also look for phase shifts, which isn't critical in a no-feedback amp but would be interesting to see.
Square wave testing will tell you a lot more than a simple FR plot, though. You can also test for power at frequency by sending a sine wave through the amp at various frequencies until the wave clips and use the RMS measurement to calculate the power output at higher frequencies.
I'm not sure why the LF response looks boosted, it's a function of the Picoscope but I can't remember the explanation.
I assume you have a dummy load, like this:
https://www.amazon.com/Parts-Express-Non-Inductive-Dummy-Resistor/dp/B0002KRDGC/ref=sr_1_10?crid=2ANQC8TT8AU80&keywords=dummy+load+100w&qid=1685205282&sprefix=dummy+load,aps,226&sr=8-10
https://www.picotech.com/library/picoapp/frequency-response-analyzer-with-bode-plots
You want to test at about 1 watt output, which you can figure out by generating a 1K sine wave and the RMS measurement tool in the scope program to determine the input voltage that gives you 1 watt output. Then close Picoscope and open the Frequency Response Analyser. Channel A goes to the input and Channel B to the ouput. Set the input voltage for the 1 watt output, set the FR you want to measure, and enter 50 in the Steps box. This will give you a very fine-grained look. You can also look for phase shifts, which isn't critical in a no-feedback amp but would be interesting to see.
Square wave testing will tell you a lot more than a simple FR plot, though. You can also test for power at frequency by sending a sine wave through the amp at various frequencies until the wave clips and use the RMS measurement to calculate the power output at higher frequencies.
I'm not sure why the LF response looks boosted, it's a function of the Picoscope but I can't remember the explanation.
I assume you have a dummy load, like this:
https://www.amazon.com/Parts-Express-Non-Inductive-Dummy-Resistor/dp/B0002KRDGC/ref=sr_1_10?crid=2ANQC8TT8AU80&keywords=dummy+load+100w&qid=1685205282&sprefix=dummy+load,aps,226&sr=8-10
Last edited:
Thanks for the tips and patience with my ignorance 🙂 hopefully this program will work on my computer.
yes, i’ve been using a dummy load but it’s probably fairly inductive… a 50w chassis mount resistor… i’ll look into getting one of these non inductive units.
I was able to improve the response last night by
1) removing a 15k grid stopper resistor on the input of stage 1 that I didn’t realize I had put in… that was the biggest issue…
2) removed the 1.8k grid stopper on the 300b, this didn’t make too much of a difference but i’ll put it back in if i get oscillation issues.
3) decreased the value of the stage 2 grid leak resistor from 500k to 20k - this one has a decent improvement.
I have not yet tried reconfiguring the G2 to run in pentode mode yet, that is next… currently going town the rabbit hole of the best way to get a nice stiff 150v to feed it. Maybe a VVR?
Speaker output before:
Speaker output after:
yes, i’ve been using a dummy load but it’s probably fairly inductive… a 50w chassis mount resistor… i’ll look into getting one of these non inductive units.
I was able to improve the response last night by
1) removing a 15k grid stopper resistor on the input of stage 1 that I didn’t realize I had put in… that was the biggest issue…
2) removed the 1.8k grid stopper on the 300b, this didn’t make too much of a difference but i’ll put it back in if i get oscillation issues.
3) decreased the value of the stage 2 grid leak resistor from 500k to 20k - this one has a decent improvement.
I have not yet tried reconfiguring the G2 to run in pentode mode yet, that is next… currently going town the rabbit hole of the best way to get a nice stiff 150v to feed it. Maybe a VVR?
Speaker output before:
Speaker output after:
Decreasing the grid resistance on the second stage may give you a bit of HIF boost, but it's drastically reducing the load on the first tube. I wouldn't think this is a good solution. But yes, you're making progress! I woulod seriously consider a touch of global feedback. You're unlikely to run into stability issues, and it would boost the performance of the OPT.
wow, i played around with some NFB and it does make quite a bit of difference! I haven’t listened to it yet but seems to measure a lot better.
Before I return to the NFB implementation, I do want to explore the pentode mode of the second stage.
My experiments with Pentode connection so far (using just resistive dividers totaling around 100k and capacitors on the screen) i’m getting asymmetrical distortion, i’m guessing my screen supply just doesn’t have low enough impedance…
Before I return to the NFB implementation, I do want to explore the pentode mode of the second stage.
My experiments with Pentode connection so far (using just resistive dividers totaling around 100k and capacitors on the screen) i’m getting asymmetrical distortion, i’m guessing my screen supply just doesn’t have low enough impedance…
Screen resistors with large resistance values, that come from B+ need to be bypassed.
Either a capacitor from the screen to Ground; Or a capacitor from the screen to the Cathode.
Just Saying.
Either a capacitor from the screen to Ground; Or a capacitor from the screen to the Cathode.
Just Saying.
Yes, I've tried both a capacitor from screen to ground, and screen to cathode, even tried an external power supply. Ultimately decided to leave it triode-strapped.
I decided to listen to the current setup for a few hours with a remote-controlled relay to engage/disengage the NFB.
I much much prefer the sound without NFB - there doesn't seem to be a significant difference in HF content, perhaps a little brittle even at lower volumes.
The biggest difference I find is that the sustain of notes/percussion hits and overall dimensionality... as soon as I turn NFB on it goes kind of flat.
Ultimately I think I'll abandon the 6AW6A.. another issue I've run into is that I can't seem to get the amount of swing into the grid. I get maybe 3w before it starts clipping
I think I have what is the “final” design for this project.
- I’m sticking with the 6AW8A as the driver. Original design only had enough swing to output about 3.5W @ 6ohms before clipping and running out of steam. Revised design feeds the pentode plate approx 245v instead of 150v and, increased the plate resistor from 6.8k to 10k for a bit more slack, which results in a the driver output going from 70Vpp to 160Vpp, end result is that gives me the whole 8w out of the OPT with very little distortion.
- Pentode remains triode strapped - had trouble getting a sine wave to be symmetrical… always got flat tops and pointy bottoms… even used a stiff external G2 supply and varied the voltage… nothing was as symmetrical as having it triode-strapped.
- No NFB: While I recognize the merits of NFB, after A/B listening for hours with a remote controlled relay to engage/disengage it, I ultimately prefer the sound without it… seems to kill the decay of sounds. Granted, maybe my NFB was a crappy implementation.
- Reinstated “HF boost” hack at the middle stage - Probably a bit unorthodox but I like it better than NFB.
To use a tube like this, I was probably “supposed” to use the more powerful 6BM8 like a normal person but I gave all those away to a HAM guy. And for some reason, I just can’t shake their association with the nasty Rogers/Philips consoles that are so ubiquitous in my area lol. I have a big bag of 6AW8As so might as well use them.
- I’m sticking with the 6AW8A as the driver. Original design only had enough swing to output about 3.5W @ 6ohms before clipping and running out of steam. Revised design feeds the pentode plate approx 245v instead of 150v and, increased the plate resistor from 6.8k to 10k for a bit more slack, which results in a the driver output going from 70Vpp to 160Vpp, end result is that gives me the whole 8w out of the OPT with very little distortion.
- Pentode remains triode strapped - had trouble getting a sine wave to be symmetrical… always got flat tops and pointy bottoms… even used a stiff external G2 supply and varied the voltage… nothing was as symmetrical as having it triode-strapped.
- No NFB: While I recognize the merits of NFB, after A/B listening for hours with a remote controlled relay to engage/disengage it, I ultimately prefer the sound without it… seems to kill the decay of sounds. Granted, maybe my NFB was a crappy implementation.
- Reinstated “HF boost” hack at the middle stage - Probably a bit unorthodox but I like it better than NFB.
To use a tube like this, I was probably “supposed” to use the more powerful 6BM8 like a normal person but I gave all those away to a HAM guy. And for some reason, I just can’t shake their association with the nasty Rogers/Philips consoles that are so ubiquitous in my area lol. I have a big bag of 6AW8As so might as well use them.
Thanks! I’m very happy, the midrange just gives me goosebumps, especially when sitting it the sweet spot of my speakers.
I like it so much that i’m going to take half of it apart again to do some thermal improvements so I can leave it on for hours on end without worrying about the filament supply melting down and or cooking the capacitors.
I like it so much that i’m going to take half of it apart again to do some thermal improvements so I can leave it on for hours on end without worrying about the filament supply melting down and or cooking the capacitors.
Well, initially the driver circuit’s resistor values were copied off a random schematic for a spud amp I found off the internet and “adapted” to run the grid on a 300B instead of an OPT + speaker. The original circuit was set up more for lower voltage, higher current.
I wasn’t exactly “smart enough” to look at the datasheets either, but I kind of forced myself to. Given my ADHD, I only learn enough to do what I need to do, and clearly my prior knowledge was insufficient.
Looking at the “typical application” of the pentode side, it was saying 150v @ 15ma (2.25w PD) This tube’s maximum rating is 330V at the plate. I wanted more swing so I figured 250v @ 9ma would keep me at a similar dissipation while giving me more headroom to work with, also the added advantage of being closer to the B+ voltage so there’s less i need to dissipate/waste.
Another thing I needed to make sure was that the grid voltage stayed within acceptable range to achieve these operating conditions.
In the end I ended up building a “driver tube” jig with a bunch of meters that let me monitor plate volts, grid volts, plate current, etc… along with scope monitoring the output… picking the right cathode, plate, supply resistors is a bunch of trial and error because the voltages behave like they’re all points on an elastic band. I’m sure this could have been calculated mathematically but that’s no fun to me.
At this juncture you should retest with R9 = 0 ohms (jumpered or deleted) and recheck your HF response. You may also find that you have greater swing at the plate of U2B (second stage). What is the current bias on the second stage? To drive your 300B to full output you need to be able to swing about 160Vpp minimum
That was the situation I had before implementing R9 and C9 , I definitely had a 1.3db drop at 20khz. I added R9 and adjusted it conservatively to drop the rest of range about 0.8db, ie. 0.5db drop @ 20khz which is in line with the rest of my amps. Even with this adjustment, I am still able to achieve 160Vpp
Current bias on the second stage is about 9ma at 245v
You just got a very elaborate tone control.
Still I cannot figure how you get 420V from Hammond 300BX?
Still I cannot figure how you get 420V from Hammond 300BX?
Actually hf response is a typical weak point of Lundahl output and interstage transformers. The small input transformers are pretty good in that regard but the bigger ones I have used in the past all had quite high interwinding C and none of them had a -3dB point beyond 20kHz.
Checking an output or interstage transformer? There are lots of tests, including this one . . .
Use a square wave generator with 'X' output impedance, and connect a series resistor, so that the generator plus series resistor equals the driving impedance of the tube or preamp you will use with the transformer.
Then terminate the output or interstage transformer with the impedance it will drive in your actual circuit.
(Not always easy to do, but worth the effort).
Connect an oscilloscope to the output of the transformer. Set the generator to 1kHz, then to 10kHz as below:
Look as the downward slope of the 1kHz generator's square wave.
Reset the square wave generator to 10kHz, and look for 'ringing', note the ring frequency.
Then measure the rise time. 0.35/rise time = -3dB bandwith.
That is just a very few of the many tests I run on my transformers.
\
Use a square wave generator with 'X' output impedance, and connect a series resistor, so that the generator plus series resistor equals the driving impedance of the tube or preamp you will use with the transformer.
Then terminate the output or interstage transformer with the impedance it will drive in your actual circuit.
(Not always easy to do, but worth the effort).
Connect an oscilloscope to the output of the transformer. Set the generator to 1kHz, then to 10kHz as below:
Look as the downward slope of the 1kHz generator's square wave.
Reset the square wave generator to 10kHz, and look for 'ringing', note the ring frequency.
Then measure the rise time. 0.35/rise time = -3dB bandwith.
That is just a very few of the many tests I run on my transformers.
\
Last edited:
5u4gb with capacitor input followed by no resistors. I do remember I spent a lot of time with the arrangement of components to get this.
22uF->2H 40 Ohms->220uF-(split from here per channel)-> 10H 276 Ohms ->56uF
I simulated this in PSUD beforehand and it’s very close to the inrush limits of the rectifier tube
Just an update on this project to use the 6AW8A as a driver tube for a 300b - it works and sounds, I've found the right combination of parts and even eliminated the separate filament transformer by using Schottky diodes to rectify the 5VAC to 5VDC RMS unregulated CRC but the hum pot is able to tune it out. Very little waste dissipation now.
However! I borrowed an Audionote Kit One 300B amp from a local shop I do repairs for, and rigged up a remote controlled relay system for A/B comparing my 300b amp to the ANK1 which is an amazing sounding amp... I don't notice a huge difference in frequency response, but the ANK1 sounds more "3D".... So.. I think i'll be trying a whole new driver topology at this point.
I'm now toying with the idea 6AU6 on the input (I was attracted by the low grid-plate capacitance), followed by half a 6BX7 which both channels will share. For some reason I'm more drawn to the sound of amps where there's a shared tube between channels, technically it's worse for separation but it somehow doesn't seem to matter at listening time.
However! I borrowed an Audionote Kit One 300B amp from a local shop I do repairs for, and rigged up a remote controlled relay system for A/B comparing my 300b amp to the ANK1 which is an amazing sounding amp... I don't notice a huge difference in frequency response, but the ANK1 sounds more "3D".... So.. I think i'll be trying a whole new driver topology at this point.
I'm now toying with the idea 6AU6 on the input (I was attracted by the low grid-plate capacitance), followed by half a 6BX7 which both channels will share. For some reason I'm more drawn to the sound of amps where there's a shared tube between channels, technically it's worse for separation but it somehow doesn't seem to matter at listening time.