Member
Joined 2009
Paid Member
Thinking of DHT driving DHT, two stage only which means I need some gain from the driver. The 3A5 has a mu of 15 (data sheet attached), not exactly high. Is there a DHT with more mu than this one ?
I know the Fi 2A3 has a 'weak' driver (< 1mA plate current supposedly) but I wonder - does the 3A5 have what it takes to drive a 2A3 and
after all is said and done, is the 3A5 very linear ?
I found only one set of curves of any use... and they are attributed to a website that no longer exists. Where I found them is here: Phono Preamplifier with 3a5 Directly Heated Triode
edit: here's how I figure out the gain of the amp. mu of 3A5 x mu of 2A3 = 15 x 4 = 60
with 5k OPT into 8R the voltage ratio is 25, so the gain of the amp is 60/25 = 2.4
that's not really going to do me much use is it....
I know the Fi 2A3 has a 'weak' driver (< 1mA plate current supposedly) but I wonder - does the 3A5 have what it takes to drive a 2A3 and
after all is said and done, is the 3A5 very linear ?
I found only one set of curves of any use... and they are attributed to a website that no longer exists. Where I found them is here: Phono Preamplifier with 3a5 Directly Heated Triode
edit: here's how I figure out the gain of the amp. mu of 3A5 x mu of 2A3 = 15 x 4 = 60
with 5k OPT into 8R the voltage ratio is 25, so the gain of the amp is 60/25 = 2.4
that's not really going to do me much use is it....
Attachments
Last edited:
Maybe not, but maybe. Leaving aside the fact that you have likely overestimated the gain you will realize, if you feed your amp with a 2Vrms signal and have a gain of 2.4, you end up with 4.8Vrms on the output. W = V^2/R = (4.8)(4.8)/8 = ~2.8W. So, you are not quite pushing the 2A3 all the way, but you are actually pretty close. If you have a preamp, or a source with a hot signal, then you are OK.
The loadline you have drawn is pretty useless as it doesn´t allow a descent swing and also is in a very unlinear area. With a correct working point the tube is very linear. If you use http://www.revintage.se/triodecalc.xls you can check both max outputvoltage and 2nd order distortion.
If you parallell the two halves(guess you planned that already) and go for something like 4mA/110V, using CCS load you could maybe make it work. Then do a loadline with 2A3s gridleak resistor as load, with parallelled halves you off course double the the load(ie if grid leak is 100k you use 200k).
If you parallell the two halves(guess you planned that already) and go for something like 4mA/110V, using CCS load you could maybe make it work. Then do a loadline with 2A3s gridleak resistor as load, with parallelled halves you off course double the the load(ie if grid leak is 100k you use 200k).
I've been building with 300b SET and with 2a3 PP. Need the watts for my Alpair 10s. With the 300b SET I use a 46 as a driver into the Hammond 126C interstage. I could also use a 4P1L or a 31 or a 71A or a 2a3 into that interstage. I don't use a cathode bypass (they suck) and the 46 into the 126C works nicely without one. I thought it was the best of the bunch, but haven't really tried the 4P1L enough.
With DHTs you're pretty much looking at 3 stages - I use a 26 preamp with 5 ohm (40W) filament bias, starved filaments of 1.2v, 95v anode and two Hammond 156C anode chokes in series and with a FT-2 output cap. This is so good that three stages work fine for me.
The 3a5 is a fine tube - no question. I haven't used it with 2 halves joined but I've used it in PP in line stages and DAC output. A similar double triode is the 1J6 - take you pick. 3a5 is leaner, 1J6 more lush. I'd suggest using either with two Hammond 156c plate chokes in series - bottom to bottom and one phase reversed so they are hum bucking. They need to be kept away from magnetic fields since they can pick up hum, but they sound glorious with the higher Ra tubes like 3a5 or 1J6 or 26 - you have inductance of 300H so you'd probably not need a cathode bypass. Filament bias would be even better if the voltages stack up - it's really only convenient for low bias voltages hence running the 26 in my preamp at 4.6v bias on the cathode. I guess you could use the 26 like that as a driver. On the agenda is direct coupling the first two stages - maybe even 26 to 26.
There are no higher mu DHTs worth using. Ra goes through the roof with tubes like 40. With my 2a3 PP amps I've been using a differential 6SN7 driver with the 90K half of a Hammond 124b used as an anode choke. This works like a dream. I've also tried a pair of 4P1L here which works OK. If it has to be indirectly heated I'd look at a 2C22. Possibly a ECC40 which is as good as a E80CC. But I'd stick to DHT solutions and 3 stages
andy
With DHTs you're pretty much looking at 3 stages - I use a 26 preamp with 5 ohm (40W) filament bias, starved filaments of 1.2v, 95v anode and two Hammond 156C anode chokes in series and with a FT-2 output cap. This is so good that three stages work fine for me.
The 3a5 is a fine tube - no question. I haven't used it with 2 halves joined but I've used it in PP in line stages and DAC output. A similar double triode is the 1J6 - take you pick. 3a5 is leaner, 1J6 more lush. I'd suggest using either with two Hammond 156c plate chokes in series - bottom to bottom and one phase reversed so they are hum bucking. They need to be kept away from magnetic fields since they can pick up hum, but they sound glorious with the higher Ra tubes like 3a5 or 1J6 or 26 - you have inductance of 300H so you'd probably not need a cathode bypass. Filament bias would be even better if the voltages stack up - it's really only convenient for low bias voltages hence running the 26 in my preamp at 4.6v bias on the cathode. I guess you could use the 26 like that as a driver. On the agenda is direct coupling the first two stages - maybe even 26 to 26.
There are no higher mu DHTs worth using. Ra goes through the roof with tubes like 40. With my 2a3 PP amps I've been using a differential 6SN7 driver with the 90K half of a Hammond 124b used as an anode choke. This works like a dream. I've also tried a pair of 4P1L here which works OK. If it has to be indirectly heated I'd look at a 2C22. Possibly a ECC40 which is as good as a E80CC. But I'd stick to DHT solutions and 3 stages
andy
A 2A3 is normally built like a 3A5, dual-plate with the elements parallelled. After having seen Andys recommendation of 3A5 we have used it in a 1545 DAC interface with parallelled halves. Sounds great. http://www.diyaudio.com/forums/tubes-valves/169073-dht-buffered-nos-dac.html
On paper tube data looks good with S=3,6 mA/V , Ri=4,2k and mu=15.
It will not work with resistive load as the loadline will be to steep. So why not go for two 156C in series(at below 8mA) as Andy suggests. With a CCS heater through the cathode resistor no bypass will be needed. Sensitivity will be something like 2,3Vrms.
Think I must sim this
. Any figures on 156C capacitance?
On paper tube data looks good with S=3,6 mA/V , Ri=4,2k and mu=15.
It will not work with resistive load as the loadline will be to steep. So why not go for two 156C in series(at below 8mA) as Andy suggests. With a CCS heater through the cathode resistor no bypass will be needed. Sensitivity will be something like 2,3Vrms.
Think I must sim this
. Any figures on 156C capacitance?
Last edited:
The 3A5 can easily answer the call!
With shunt cascode (the best sounding 300B driver I have made) you can drive 2A3 to full output with only 0,6V rms.
The 3A5 is very easy to use with filament bias, so I have drawn that, and drive it with my Filament Regulators (shameless, I know).
Shunt-cascode + Filament bias means NO capacitors in the signal path, except the output coupler.
With shunt cascode (the best sounding 300B driver I have made) you can drive 2A3 to full output with only 0,6V rms.
The 3A5 is very easy to use with filament bias, so I have drawn that, and drive it with my Filament Regulators (shameless, I know).
Shunt-cascode + Filament bias means NO capacitors in the signal path, except the output coupler.
Attachments
Yes, Lars they will hate it!
But they do not appreciate that the BJT is simply using its mighty gm to hold the anode voltage steady, while passing a variable current.
You can compare a BJT vs triode in this capacity, very easily. Just make a standard series-cascode stage with any triode on the bottom, any triode on the top. If you swap the UPSTAIRS triode around, it is quickly evident that high-gm triodes sound best. Last of all, compare the best triode with an NPN BJT - which easily surpasses all.
But they do not appreciate that the BJT is simply using its mighty gm to hold the anode voltage steady, while passing a variable current.
You can compare a BJT vs triode in this capacity, very easily. Just make a standard series-cascode stage with any triode on the bottom, any triode on the top. If you swap the UPSTAIRS triode around, it is quickly evident that high-gm triodes sound best. Last of all, compare the best triode with an NPN BJT - which easily surpasses all.
Hi Andy,
2N5401 and MPSA92 behave similarly & either type will work fine - but the MPSA92 is easy to buy.
V1 is a voltage source of 101V. It can be made from a resistor-divider + emitter-follower (for low impedance drive) or it can be a shunt regulator set to about 101V.
I use the divider circuit in my amp, and the sound is outstanding - but shunt regulation is potentially an improvement. That's where I am working now.
2N5401 and MPSA92 behave similarly & either type will work fine - but the MPSA92 is easy to buy.
V1 is a voltage source of 101V. It can be made from a resistor-divider + emitter-follower (for low impedance drive) or it can be a shunt regulator set to about 101V.
I use the divider circuit in my amp, and the sound is outstanding - but shunt regulation is potentially an improvement. That's where I am working now.
But maybe the tube-talibans will puke when seeing the BJT
Screw them! The music is what matters. Check Morgan-Jones β follower out or even my penchant for DC coupled ZVN0545A source followers.
JMO, carefully constructed hybrids of tubes and "sand" do a better job than either technology alone.
Member
Joined 2009
Paid Member
This is very encouraging, I like the idea of a DHT driving a DHT, but I always felt that two stages is better than three.
I'm not a stranger to sand (most of my DIY experience is sand) but I'd much prefer it was used as a load, not directly in the signal path - Rod, in post #8 there's this "Q1" ? - was the intention that it provides gain or just a means to couple two different dc-bias points ?
Lars, can you post your simulation (the forum now allows attachments directly from LTspice) ?
Last edited:
With a shunt-cascode, the amplification is entirely determined by the triode.
The transistor Q1 simply forces the triode to operate with fixed anode voltage, and vary the current according to the grid input. So the triode works on a vertical load line (the line is at 100V & varies from 4mA idle and 2,25mA to 5,75mA at full signal - numbers per each half of the 3A5). This allows many triodes to give their best possible performance, even with gains of many times mu).
The fact that the B+ capacitor is eliminated from the signal path makes far more difference than adding a "grounded grid" NPN. And notice also, that the input AND output are referred to ground, so the power supply rejection is greatly improved compared to other architectures. EMI is also highly rejected for this reason.
The transistor Q1 simply forces the triode to operate with fixed anode voltage, and vary the current according to the grid input. So the triode works on a vertical load line (the line is at 100V & varies from 4mA idle and 2,25mA to 5,75mA at full signal - numbers per each half of the 3A5). This allows many triodes to give their best possible performance, even with gains of many times mu).
The fact that the B+ capacitor is eliminated from the signal path makes far more difference than adding a "grounded grid" NPN. And notice also, that the input AND output are referred to ground, so the power supply rejection is greatly improved compared to other architectures. EMI is also highly rejected for this reason.