I recently finished my all DHT low power 845 SE amp and have been tweaking and listening to it for the last six weeks. This is my first attempt at an own 'design'. I was trying to read up on all aspects of an 845 build as best I can. One of the topics was AC or DC on the filaments of course. Apart from a few, notably Jim Doyle who has done a lot of experimentation with low power 845, a lot seem to suggest that DC is the only viable solution if excessive noise was to be avoided.
I want to share my own experience for the 'googlers' that follow with their future projects. I make no comments about SQ, I've not gotten around to comparing in any case, only the amount of hum with AC on the 845 filaments.
My amp is all DHT using a 102d(modern)-RC-45/2a3-IT-845. Only the 102d filaments are on DC using a Delon doubler from a 2.5vac CT Tx with dropping resistors before rectification then CRC. The last two stages are on AC fils. All fils are on 100R 2W pots.
I measured the AC on the 8R speaker post with the speakers as load. With the 2a3 in the driver position, I can only null down to about 5-7mv. With the 45 in the same position I can null down to 3-5mv. (These are about the ranges of my first 2a3 SE also on AC fils.). On 89db and 94db speakers I'm using, these levels are only faintly audible from the listening position 3m away in a small room. (In fact the Cambridge Audio 640H music server is noisier!) I don't have more sophisticated measuring tools to tell me more.
The hard part is nulling to these levels on the pots. Both driver and output valves have to be, IME, nulled in tandem. There is a sweet spot when both stage's pots are in positions that stop them from 'exciting' the hum out of each other. The last mv is shaved from the input's pot once its at its lowest from adjusting the other two pots. (Product of my 'dirty dc' implementation I suppose.). The hard part is getting the knack to find the sweet spot which changes if one set of valves are replaced. The longest has been 30mins finding it! Without the careful fiddling, the level is around 15-17mv on a quick adjustment after fresh turn on. I did try to null in stages with preceding valves pulled, didn't work out in my case.
So to my conclusion, it can be done on AC. Whether its an optimal choice with the 845, I'm yet to find out for myself. But first, I need to learn how to do simulations so I can look deeper into my little circuit. Anybody want to run simulations for me?
I want to share my own experience for the 'googlers' that follow with their future projects. I make no comments about SQ, I've not gotten around to comparing in any case, only the amount of hum with AC on the 845 filaments.
My amp is all DHT using a 102d(modern)-RC-45/2a3-IT-845. Only the 102d filaments are on DC using a Delon doubler from a 2.5vac CT Tx with dropping resistors before rectification then CRC. The last two stages are on AC fils. All fils are on 100R 2W pots.
I measured the AC on the 8R speaker post with the speakers as load. With the 2a3 in the driver position, I can only null down to about 5-7mv. With the 45 in the same position I can null down to 3-5mv. (These are about the ranges of my first 2a3 SE also on AC fils.). On 89db and 94db speakers I'm using, these levels are only faintly audible from the listening position 3m away in a small room. (In fact the Cambridge Audio 640H music server is noisier!) I don't have more sophisticated measuring tools to tell me more.
The hard part is nulling to these levels on the pots. Both driver and output valves have to be, IME, nulled in tandem. There is a sweet spot when both stage's pots are in positions that stop them from 'exciting' the hum out of each other. The last mv is shaved from the input's pot once its at its lowest from adjusting the other two pots. (Product of my 'dirty dc' implementation I suppose.). The hard part is getting the knack to find the sweet spot which changes if one set of valves are replaced. The longest has been 30mins finding it! Without the careful fiddling, the level is around 15-17mv on a quick adjustment after fresh turn on. I did try to null in stages with preceding valves pulled, didn't work out in my case.
So to my conclusion, it can be done on AC. Whether its an optimal choice with the 845, I'm yet to find out for myself. But first, I need to learn how to do simulations so I can look deeper into my little circuit. Anybody want to run simulations for me?
What are you trying to simulate? Have you used ltspice yet? If not, I found it very intuitive and easy to learn. A couple of threads here showed me how to model tubes and transformers. Then you just draw your circuit and go....
I did find a few complete spice files here, to help me figure out how to run frequency sweeps or distortion analysis, by copying and learning how they set it up.
Anyway, if you haven't used spice, it's definitely worth spending a few hours learning. (and then countless hours simulating )
I did find a few complete spice files here, to help me figure out how to run frequency sweeps or distortion analysis, by copying and learning how they set it up.
Anyway, if you haven't used spice, it's definitely worth spending a few hours learning. (and then countless hours simulating
)I wanted to simulate really just to see whether the pen and paper calculations I made were accurate or whether there are other issues in the circuit I have not accounted for that a simulation would shed light on. I've not tackled learning the simulation side before as I wanted to get a 'pen and paper design' build and subjective tweaking under my belt first. The way I figured it, it'll help me make a better judgement of my own preferences for future builds. Now I'm satisfied (for now ) of the adjustments I've made (component values) and happy with the sound of this project, I want to scrutinise it further.
I'm also looking to learn more about scopes and measurements....
) of the adjustments I've made (component values) and happy with the sound of this project, I want to scrutinise it further. I'm also looking to learn more about scopes and measurements....
I built an amp that runs 845's or 211's. It is not low power. I run 1050 volts and get 40 WPC. Tubes are 5842 -> 45 -> 845. All use DC heating. The residual hum across my speakers is under 1 mV. The meter reads the same with the amp off and unplugged, so that reading is the test setup residual hum. The power supply is on a separate chassis and there is no AC at all in the power amp chassis.
I tried AC heating and found that the hum could be reduced to the 3 or 4 mV level on the 845. The 211 was worse at about 6 mV. The issue is not the hum, it is IMD. The 60 Hz (50 Hz) will intermodulate with the audio signal creating mixing products. Run a 1 KHz tone through the amp, and you will get the 1 K tone out....WITH some 1060 Hz and some 940 Hz (1050 and 950 in the UK). You will also get some hum at 60 Hz AND its harmonics, 120 Hz, 180 Hz...
A sensitive analyzer will also pick up the IMD products of the hum AND its harmonics. So a 1 KHz tone will generate output at 1000 Hz, 60 Hz, 120 Hz, 180 Hz, 820 Hz, 880 Hz, 940 Hz, 1060 Hz, 1120 Hz, 1180 Hz, and others at a lower level. Granted, these IMD products are at a low level, but remember that music is not a single tone, there are multiple signals at multiple levels.
These generate multiple IMD products with each other AND the 60 Hz hum. While plainly visible on the FFT analyzer, the IMD was not itself audible in the amp. The amp just had a vague fogginess that contributed to listener fatigue. After a few hours I would grow tired of listening, or more correctly, uninterested in listening. DC heating fixed all that, and every DHT amp I have built since uses DC heating.
Note that there is a real issue with DC heating and high Mu tubes. Consider the 811A with a Mu of 160. With a 6.3 volt heater the bias voltage is 6.3 volts higher at one end of the heater compared to the other. This is a significant difference on a high Mu tube, so one end of the filament will be responsible for most of the emission, and the other end will be near cutoff. This can lead to some distortion, and possibly tube wear. This is not an issue with 45's or 845's.
The best choice for DHT filament power is probably a high frequency sine wave, something in the 200+ KHz region so that any residual "hum" and its IMD will be way above the audio range, and never pass through the OPT. Anything other than a clean sine wave can result in noise or "hash" in the output. Mosfets and toroids are the hot ticket here, but my early experiments just made smoke, so more work is needed.
I tried AC heating and found that the hum could be reduced to the 3 or 4 mV level on the 845. The 211 was worse at about 6 mV. The issue is not the hum, it is IMD. The 60 Hz (50 Hz) will intermodulate with the audio signal creating mixing products. Run a 1 KHz tone through the amp, and you will get the 1 K tone out....WITH some 1060 Hz and some 940 Hz (1050 and 950 in the UK). You will also get some hum at 60 Hz AND its harmonics, 120 Hz, 180 Hz...
A sensitive analyzer will also pick up the IMD products of the hum AND its harmonics. So a 1 KHz tone will generate output at 1000 Hz, 60 Hz, 120 Hz, 180 Hz, 820 Hz, 880 Hz, 940 Hz, 1060 Hz, 1120 Hz, 1180 Hz, and others at a lower level. Granted, these IMD products are at a low level, but remember that music is not a single tone, there are multiple signals at multiple levels.
These generate multiple IMD products with each other AND the 60 Hz hum. While plainly visible on the FFT analyzer, the IMD was not itself audible in the amp. The amp just had a vague fogginess that contributed to listener fatigue. After a few hours I would grow tired of listening, or more correctly, uninterested in listening. DC heating fixed all that, and every DHT amp I have built since uses DC heating.
Note that there is a real issue with DC heating and high Mu tubes. Consider the 811A with a Mu of 160. With a 6.3 volt heater the bias voltage is 6.3 volts higher at one end of the heater compared to the other. This is a significant difference on a high Mu tube, so one end of the filament will be responsible for most of the emission, and the other end will be near cutoff. This can lead to some distortion, and possibly tube wear. This is not an issue with 45's or 845's.
The best choice for DHT filament power is probably a high frequency sine wave, something in the 200+ KHz region so that any residual "hum" and its IMD will be way above the audio range, and never pass through the OPT. Anything other than a clean sine wave can result in noise or "hash" in the output. Mosfets and toroids are the hot ticket here, but my early experiments just made smoke, so more work is needed.
Yes, the residual hum's IMD effect across the audible frequency is the main consideration for me whether it is apparent now or not. I have observed that there are things made apparent in listening tests only by their absence, or presence for that matter. I went DC for the 102d as the mu of 30 for the DHT input tube made me weary of trying AC at all.
In any case, I am looking at my options for trying out DC in the future. It will require me splitting the current circuit in to monoblocks as there will be no room to implement a sophisticated DC supply on the current set up. (A bridge followed by CRC excepting, I can shoehorn those in.). Since I've never delved in to DC fils supply options, I have been bewildered a bit. I thought that bridge-CRC method would be enough for my A/B comparison but looking more in to the subject, it seems, that would not be a fair crack for a DC comparison. I've even read opinions that the bad reputation of DC fils, in some quarters, comes down mostly to inadequate supply configurations. Fair enough, I can see how that can be.
I'm currently looking at Rod Coleman boards as they seem to come highly recommended by people who use them.
In any case, I am looking at my options for trying out DC in the future. It will require me splitting the current circuit in to monoblocks as there will be no room to implement a sophisticated DC supply on the current set up. (A bridge followed by CRC excepting, I can shoehorn those in.). Since I've never delved in to DC fils supply options, I have been bewildered a bit. I thought that bridge-CRC method would be enough for my A/B comparison but looking more in to the subject, it seems, that would not be a fair crack for a DC comparison. I've even read opinions that the bad reputation of DC fils, in some quarters, comes down mostly to inadequate supply configurations. Fair enough, I can see how that can be.
I'm currently looking at Rod Coleman boards as they seem to come highly recommended by people who use them.
In the 1990s, ac-heat for DHTs was often regarded as the only game in town, as everyone who tried dc found bad sound one way or another.
The cause is touched on in George's (Tubelab) post. The filament voltage skews the biasing of the DHT along the length of the filament - the positive end is closer to the HT ground, and so biased to more current.
The difference in bias also skews gm (mutual conductance) of the DHT, so that when a signal is applied to the grid, the diverging gm produces a differential (music) signal across the ends of the filament.
If you connect a voltage regulator to the filament, the voltage reg's feedback path will sense this music signal, and use its slow, clumsy Ampere-level pass-transistor to try to null it out.
Not to mention the electrolytic capacitor connected across this music signal - always needed with voltage-feedback loops.
With raw rectified dc, you connect the filament to large capacitors that also have to handle the forward (recharge) pulses from the rectifier, and any reverse-recovery pulses. No wonder that this is the worst-sounding option.
My kits are designed to provide current-only control, so that no output capacitors need make contact with the filament, and the feedback sensing is remote from the filament voltages. The current is stable, using temperature compensation. And the noise is reduced by 40dB or more compared to what can be achieved with LT108x or LM317 parts - & you've probably seen what many other DIYers think of the resulting sound!
The cause is touched on in George's (Tubelab) post. The filament voltage skews the biasing of the DHT along the length of the filament - the positive end is closer to the HT ground, and so biased to more current.
The difference in bias also skews gm (mutual conductance) of the DHT, so that when a signal is applied to the grid, the diverging gm produces a differential (music) signal across the ends of the filament.
If you connect a voltage regulator to the filament, the voltage reg's feedback path will sense this music signal, and use its slow, clumsy Ampere-level pass-transistor to try to null it out.
Not to mention the electrolytic capacitor connected across this music signal - always needed with voltage-feedback loops.
With raw rectified dc, you connect the filament to large capacitors that also have to handle the forward (recharge) pulses from the rectifier, and any reverse-recovery pulses. No wonder that this is the worst-sounding option.
My kits are designed to provide current-only control, so that no output capacitors need make contact with the filament, and the feedback sensing is remote from the filament voltages. The current is stable, using temperature compensation. And the noise is reduced by 40dB or more compared to what can be achieved with LT108x or LM317 parts - & you've probably seen what many other DIYers think of the resulting sound!
As is, I'm quite happy with the AC sound of my creation. But, I also do place a lot of weight on the opinions of guys that have made much more amps than I. This is why I want to try to implement a good DC supply and hear for myself. What Rod mentioned rings true for sure as my first tube gurus got started in the hobby in the 90s are firmly entrenched in the AC camp.
Because of my lack of experience, future complications and costs in these things are easily written of as educational funding. Like the experiments in the science lab that the professors see year after year knowing the results but pushing students to think and learn...
Because of my lack of experience, future complications and costs in these things are easily written of as educational funding
. Like the experiments in the science lab that the professors see year after year knowing the results but pushing students to think and learn...- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.