450Ω in the primary is certianly high (my Sowter SA08s measure 80Ω cold). This Resistance adds a little more than 1Ω to the speaker circuit, in addition to the ra of the parallel triodes.
2Ω total is a lot for an 8Ω load, and might affect the bass on some speakers that were not designed for such high-Z drive - compare the impedance curve of the speaker.
At the beginning of the thread I described my experiments with different output transformers. Amplifiers without feedback highly depend on output transformers.
Do you like the mid and high regions with this amp? Why not explore a more radical route to good bass?
You could adopt bi-amping, and use a transistor amp to drive the low bass, and the 4P1L to take over from around 100Hz.
This approach improves the mids and highs, as well as the bass, and offers a lot of fun in the making of it.
You can use an old stylee analogue crossover at line level, but if you use digital sources, much better to use your PC processor as a DSP engine.
You can get an XMOS board to do USB to I2S, and connect it to a 8-channel ES9028PRO DAC, for example.
DIYINHK sells these DIY boards at a decent price. They need a little finishing, but it's a low cost way to a very good multi channel source, and opens the door to DSP. These are the boards I have bought and will get to work on shortly.
Forget DSP boxes that convert A>D and back again - and so destroy the sound; they can't compete - or even do very good DSP (limited FIR taps...)
The crossover can be built with CamillaDSP installed on any recent OS. It's not too tough to configure, and allows wide adjustments to taste, without the high cost of big air core inductors or caps.
Anyway, just a suggestion.
Hello Rod. Well, I've thought about bi-amping in the past but always resisted. First, too complex and I don't build with solid state. Second, I don't really need deep bass - I'm in an apartment with people up and downstairs. I have 3 SE amps, and the other two have decent bass performance, so it's not a necessity to use the 4P1L amp. I may need more gain anyway depending on how my AD1865 DAC project goes. Figuring out what valves to use in the I/V output. So although it's a good idea in general, bi-amping isn't for me personally. Certainly others would appreciate it.
Seems like a sound solution. I am at verge to go biamping with DSP, Class-D for subwoofers to 60Hz and Altec 414 midbass up to 350Hz. Above that passive filtering for a 3- way horn combination with a DHT SE 4P1L-6B4G. Will anyway help me get all levels and crossover frequencies right. Probably enough with Class-D for the subs, though.
I agree that biamping with active crossovers can be a very, very good solution. Close to a necessity for those of us who enjoy low power SETs and still want a proper LF response.
Bass is always problematic for SE output transformers (the airgap kills the inductance) and the ones with decent bass often have a less than stellar HF response.
Bass is always problematic for SE output transformers (the airgap kills the inductance) and the ones with decent bass often have a less than stellar HF response.
Hi Zintolo
Two processors are involved: One is your PC (Laptop or Raspberry Pi for example). The PC processor provides the kind of compute power (and large memories) needed to perform the DSP for crossover and EQ functions. A modern OS is needed, so you can install CamilaDSP
This marvellous software tool from HenrikEnquist (here on diyAudio) opens the door to easily-configured DSP crossovers, room EQ and plenty more - without the sound-degrading effect of external DSP-boxes that start with analogue inputs (so you go D→A and then A→D, then DSP, then D→A).
Now we have the opportunity to DIY with multichannel digital outputs without the tradeoff of corrupted sound quality.
Next we need to get the audio data streams out of the PC (USB is the usual route) and present them to the multiple DAC channels we need, so that loudspeakers can be built wiithout coils - and compromises.
The next processor does this work - the XMOS X-core from a British firm that develops custom multicore processors. They are ideal for real-time high-defintion and multichannel audio processing. Their development software include optimised drivers for USB and I2S, so they find their way into many audio applications, including DIY boards for low cost experimentation. I like the DIY in HK boards fo this work.
They have completed the most time-consuming work - though you have to add a few parts to finish, but the boards are very good value.
Add one of their 8-channel ES9028PRO DAC boards, and you are ready to start work!
Two processors are involved: One is your PC (Laptop or Raspberry Pi for example). The PC processor provides the kind of compute power (and large memories) needed to perform the DSP for crossover and EQ functions. A modern OS is needed, so you can install CamilaDSP
This marvellous software tool from HenrikEnquist (here on diyAudio) opens the door to easily-configured DSP crossovers, room EQ and plenty more - without the sound-degrading effect of external DSP-boxes that start with analogue inputs (so you go D→A and then A→D, then DSP, then D→A).
Now we have the opportunity to DIY with multichannel digital outputs without the tradeoff of corrupted sound quality.
Next we need to get the audio data streams out of the PC (USB is the usual route) and present them to the multiple DAC channels we need, so that loudspeakers can be built wiithout coils - and compromises.
The next processor does this work - the XMOS X-core from a British firm that develops custom multicore processors. They are ideal for real-time high-defintion and multichannel audio processing. Their development software include optimised drivers for USB and I2S, so they find their way into many audio applications, including DIY boards for low cost experimentation. I like the DIY in HK boards fo this work.
They have completed the most time-consuming work - though you have to add a few parts to finish, but the boards are very good value.
Add one of their 8-channel ES9028PRO DAC boards, and you are ready to start work!
I'd like to ask a question about the tonality of PSE 4P1L and how to drive them. I've always found the tonality rather thin. Bright and clear but lacking in body and fullness. A pair of amorphous OPTs was not a good match - not enough fulness in the sound. I also tried a 10Y as driver and the same thing - bright and clear but lacked body. I'm currently using a pair of 27 mesh and they seem to have put some fulness back into the sound. I imagine 26 would be the same.
What are your experiences?
What are your experiences?
Anatoliy [Wavebourn] used a cathode follower to drive the 4P1L grids. You could try it, or better still a power-FET follower.
RC coupling works badly with many output valves - and worse still with low power amps, since the grid will draw current unilaterally near the positive Vgk point, and give temporary offset to the grid (in the cold direction). This makes the sound thin and weak IME. Don't think that the grid will not reach pos Vgk - it usually will, even with fairly quiet playing.
RC coupling works badly with many output valves - and worse still with low power amps, since the grid will draw current unilaterally near the positive Vgk point, and give temporary offset to the grid (in the cold direction). This makes the sound thin and weak IME. Don't think that the grid will not reach pos Vgk - it usually will, even with fairly quiet playing.
For proper loading the 4P1L PSE requiring at least 20-24H primary inductance 2k5-3k5 transformer.
If it lower, the LF behaviour is worsening.
3k-3k5 is the best compromise.
Sample stepped load for 4P1L PSE:
If it lower, the LF behaviour is worsening.
3k-3k5 is the best compromise.
Sample stepped load for 4P1L PSE:
Could you go into a little more detail of how RC coupling, which I use because I like the tonality, can be optimised? Is it a question of the current through the driver tube, or the grid voltage - larger=better? Or the Ri of the tube? Or anything else?
No need to change the driver itself.
The problem is asymmetrical grid current: the coupling cap charges as positive grid region is approached. No opposite effect occurs at negative end of the voltage swing.
The result is a charged coupling cap - meaning unwanted extra negative bias voltage, and reduced anode current.
A follower buffers the driver against grid current, and supplies it without loading the driver.
DC couple the follower to the grid for best results; RC couple to the followers input
The problem is asymmetrical grid current: the coupling cap charges as positive grid region is approached. No opposite effect occurs at negative end of the voltage swing.
The result is a charged coupling cap - meaning unwanted extra negative bias voltage, and reduced anode current.
A follower buffers the driver against grid current, and supplies it without loading the driver.
DC couple the follower to the grid for best results; RC couple to the followers input
Thanks Rod. How can we optimise the driver in a RC topology without a follower? What factors could be optimised as above - grid voltage of output valve, current and Ri of driver, type and value of coupling cap and anything else? Just looking for the "least bad" configuration of a simple RC driver stage, if you like.
Large value of anode resistor in the driver.
The old Marconi Osram circuits had 10 or 15k grid resistors for power tubes PX25A; they look like stoppers, but are set high values to address this problem.
The old Marconi Osram circuits had 10 or 15k grid resistors for power tubes PX25A; they look like stoppers, but are set high values to address this problem.
Large value of anode resistor in the driver...... I try and use 5x the Ri of the tube. So for my 27 driver, Ri=9K, I use 47K anode resistor. Same for type 26 with Ri = 7.5K and 36K anode resistor.
Does this help? Is 5x Ri enough?
Does this help? Is 5x Ri enough?
The ra has little influence.
47k driver load is pretty good.
If you want to see whether this is a problem for you, just add 33k - 47k in series with the grid (like a stopper)
If the sound is unchanged, this is not the trouble in this stage.
47k driver load is pretty good.
If you want to see whether this is a problem for you, just add 33k - 47k in series with the grid (like a stopper)
If the sound is unchanged, this is not the trouble in this stage.