Do you think a common mode choke before the regulator could help? I have a few I could try. Anything else that could be put in between the power supply and regulator? Bigger caps don't seem to affect the hiss much.
Hmmm - I could try a 5 ohm resistor and put the current up. That could work - good thinking. A 12v DC supply would be very handy.
Andy
Hmmm - I could try a 5 ohm resistor and put the current up. That could work - good thinking. A 12v DC supply would be very handy.
Andy
Thanks Rob!
Yes you are right, the origin of the noise is the LT1085. This chip, and almost every other regulator IC, uses a bandgap reference to control its output level. These works fine for normal industrial applications, but they are just too noisy for high quality audio. Here is the calculation for the approximate noise level for the LT1085, connected as a current regulator:
The data sheet specifies the noise as:
0.003% of Vout, Typical, 25 deg C, RMS value, 10Hz-10kHz.
When connected as a CCS, the regulator applies the reference voltage: 1,2V across a 1,2-ohm
resistor, so the output noise is:
Vn = 1,2V x 0.003% = 36uV rms;
and this noise is translated into current by the 1,2-ohm sense-resistor:
36uV/1,2R = 30uA.
So the CCS injects RMS noise of 30uA into your filament current.
Now, the filament shares its terminals, internal wiring and all with the CATHODE current. So the filament current (which just heats the tube) is DIRECTLY mixed with the cathode-current - ie our Music Signal!
For 26 preamp with the cathode current set at 6mA this 30uA of rms noise is 0,5% of the full current range - and even higher % than this, if we only consider the working range of current of the 26 (the current swing required to play music).
Naturally 0,5% of noise, caused by the filament current is too much. Not only you will hear Hisssss, but the music signal will have sidebands added in the frequency domain (eg signal 1kHz, noise at 10Hz will give sidebands at 1010 and 990Hz - and this
repeats for all the other frequencies in the noise). The sidebands will give clouding of the sound, and spoil the stereo image.
The noise at 0.1 to 10Hz is extra large in bandgaps (and other noise sources), and though you do not hear the sound of it as hum, it is always there, generating sidebands, degrading the sound.
With switching regulators, there is not usually any specification for the **reference** noise, which you can be sure means it is very bad! But the high frequency ripple with switchers will also create sum & difference sidebands, and degrade the sound.
With my regulators, the self-noise is the minimum it can possibly be: just a pair of transistor junctions, whose self-noise is at least 100x lower than a bandgap diode.
This is one reason why regulators made with ICs sound bad. There are plenty of others, such as the way that the Music signal gets mixed up in the feedback loop of the IC regulator, which should only be controlling filament current.
Andy, the old (2004) circuit required higher voltage, but the new PCBs only need 4V (guaranteed), and usually works quite well with 3,6V above the filament voltage.
Yes you are right, the origin of the noise is the LT1085. This chip, and almost every other regulator IC, uses a bandgap reference to control its output level. These works fine for normal industrial applications, but they are just too noisy for high quality audio. Here is the calculation for the approximate noise level for the LT1085, connected as a current regulator:
The data sheet specifies the noise as:
0.003% of Vout, Typical, 25 deg C, RMS value, 10Hz-10kHz.
When connected as a CCS, the regulator applies the reference voltage: 1,2V across a 1,2-ohm
resistor, so the output noise is:
Vn = 1,2V x 0.003% = 36uV rms;
and this noise is translated into current by the 1,2-ohm sense-resistor:
36uV/1,2R = 30uA.
So the CCS injects RMS noise of 30uA into your filament current.
Now, the filament shares its terminals, internal wiring and all with the CATHODE current. So the filament current (which just heats the tube) is DIRECTLY mixed with the cathode-current - ie our Music Signal!
For 26 preamp with the cathode current set at 6mA this 30uA of rms noise is 0,5% of the full current range - and even higher % than this, if we only consider the working range of current of the 26 (the current swing required to play music).
Naturally 0,5% of noise, caused by the filament current is too much. Not only you will hear Hisssss, but the music signal will have sidebands added in the frequency domain (eg signal 1kHz, noise at 10Hz will give sidebands at 1010 and 990Hz - and this
repeats for all the other frequencies in the noise). The sidebands will give clouding of the sound, and spoil the stereo image.
The noise at 0.1 to 10Hz is extra large in bandgaps (and other noise sources), and though you do not hear the sound of it as hum, it is always there, generating sidebands, degrading the sound.
With switching regulators, there is not usually any specification for the **reference** noise, which you can be sure means it is very bad! But the high frequency ripple with switchers will also create sum & difference sidebands, and degrade the sound.
With my regulators, the self-noise is the minimum it can possibly be: just a pair of transistor junctions, whose self-noise is at least 100x lower than a bandgap diode.
This is one reason why regulators made with ICs sound bad. There are plenty of others, such as the way that the Music signal gets mixed up in the feedback loop of the IC regulator, which should only be controlling filament current.
Andy, the old (2004) circuit required higher voltage, but the new PCBs only need 4V (guaranteed), and usually works quite well with 3,6V above the filament voltage.
Wow! Considerable progress here - thanks guys! I now have the 26 preamp running at approx dhtrob's operating points - it's extremely good. 95v on anode, 3.4mA, 5 ohm cathode resistor. It's 4.6v bias running slightly starved filament at 1.2v. So my cheap 12v DC supplies are working just fine!! No hiss.
I'm using two Hammond 156C chokes in series and FT-2 teflon output caps. This is possibly the best 26 preamp I've made yet. Incredible - it was made as a cheap compromise and blow me, it outperforms my other one with 10 ohms cathode resistors, 135v anode and LL1660/5mA. Considerable saving of cost. It's very detailed indeed. This has been a very worthwhile exercise, and it's a blueprint for a very cheap build.
I suspect dhtrob didn't come to this operating point by accident!
Try it one day!! No chokes in the filament supply even. Wow.
Andy
I'm using two Hammond 156C chokes in series and FT-2 teflon output caps. This is possibly the best 26 preamp I've made yet. Incredible - it was made as a cheap compromise and blow me, it outperforms my other one with 10 ohms cathode resistors, 135v anode and LL1660/5mA. Considerable saving of cost. It's very detailed indeed. This has been a very worthwhile exercise, and it's a blueprint for a very cheap build.
I suspect dhtrob didn't come to this operating point by accident!
Try it one day!! No chokes in the filament supply even. Wow.
Andy
Great stuff Andy! I do have my 26 running at 6mA with the LL1660 and 10 ohm cathode resistor. However, I started playing with starved filaments and operating points with the 4P1L, 71a and now the 01a, which is great for me! Love the thoriated filament sound btw.
I will build a 26 with gyrator load and try 95V and 3.5mA to compare results. It would be a great candidate to add a cathode/follower stage to reduce Output impedance as per my previous posts.
Oh no, another preamp on the way!!
Cheers,
Ale
The 26 in this iteration blew the VT-67, 31 and 01A out of the water. It was really no contest. The 26 is smoother and all round more musical. This 95v operating point is a bit leaner and more detailed with the 1.2v starved filaments. Treble is crisper and it takes out that slightly warm and tubby sound. Well, I don't mean tubby really, the 26 can't ever be that bad. It also restores the balance in my 300b SET - that needed to be a little leaner and more detailed (it has 46 drivers).
I have had great results with my PP amp - currently 01A diff pair direct coupled to 31 interstage coupled (LL1660/PP) to 2a3s. If it's good with that, then it's going to be better still with 26 pairs direct coupled to ??? and LL1660/PP to the 2a3s. So input tube is 26, sorted. Driver should be a pair of 10Y really - I have 46 and 4P1L as options. If the 4P1L works then the gain would be good. This is all looking very promising. I'm going to try out some 6C4C. I'm preferring the old Chinese biplate 2a3s to the Sovtek monoplates, so the old Svetlana biplate 6C4C may be good too. My sound system is really sounding good these days. Pretty happy with my Alpair 10s in 21 litre MLTL. Actually they're infinite baffle right now with plastic bags stuffed in the port.
Andy
I have had great results with my PP amp - currently 01A diff pair direct coupled to 31 interstage coupled (LL1660/PP) to 2a3s. If it's good with that, then it's going to be better still with 26 pairs direct coupled to ??? and LL1660/PP to the 2a3s. So input tube is 26, sorted. Driver should be a pair of 10Y really - I have 46 and 4P1L as options. If the 4P1L works then the gain would be good. This is all looking very promising. I'm going to try out some 6C4C. I'm preferring the old Chinese biplate 2a3s to the Sovtek monoplates, so the old Svetlana biplate 6C4C may be good too. My sound system is really sounding good these days. Pretty happy with my Alpair 10s in 21 litre MLTL. Actually they're infinite baffle right now with plastic bags stuffed in the port.
Andy
Actually I did experiment with the '26, but was more or less forced into the 95V B+ and -4V Vg1 'cause of limitations on the heater supply..
@ Aje: Rubber stand-offs are arriving this week...
Rob
@ Rob, many thanks you are a star!
Andy/Rob,
Just wanted to double check and did a quick plot on my tracer of a NOS RCA 26 ST valve. Filaments starved at Vf=1.2V and 800mA as per Andy's operating point.
Ra increases to 11K (3K more than a 01a) but if Andy reports good bass response I take is worth trying it?
Initially I thought the operating point might be a bit on the low side, but At Vg=-4V and Va=100V is a good linear region. Perhaps a bit marginal at -4V bias to respond to input signal transients and end up driving a bit of grid current?
Anyway, enjoy the plot and draw the load lines
Attachments
I've been doing some extended listening to the new 26 preamp with my 300b and 2a3 amps, and I'm now coming back to preferring the 300b with 46 driver and Lundahl LL1620/80mA. With the new operating point the 26 preamp is sounding more detailed and the overall level of detail into the 300b SET is stunning. All of which means I'll have to go back and revise my previous 26 preamp. More experiments! But making some serious progress. As they say - necessity is the mother of invention. If I hadn't had the hair-brained idea of using cheap 12v DC power supplies none of this would have happened.
andy
andy
The power supply for the new 26 preamp is very simple. It has a small mains transformer secondary 120v, rated 50mA, and an EZ80. Then a 2uf input cap to a 15H choke rated 30mA. After that a 50uf cap, 2.4K dropper resistor and another 50uF cap. All capacitors are polypropylene - I never use electrolytics.
My usual HT supply has an AZ1 mesh and is a bit more elaborate.
I'm building another 26 preamp exactly the same! Need to clone this. I'm out of Hammond 156c chokes so next one will be Lundahl LL1667/10mA. Otherwise the same. Will be interesting to compare plate chokes.
Andy
My usual HT supply has an AZ1 mesh and is a bit more elaborate.
I'm building another 26 preamp exactly the same! Need to clone this. I'm out of Hammond 156c chokes so next one will be Lundahl LL1667/10mA. Otherwise the same. Will be interesting to compare plate chokes.
Andy
Hello,
Can you provide a link to a schematic of your 26 preamp? If it's been mentioned before, please forgive me, but this is an extremely long thread. Also, what phono preamp are you using?
Regards.
David
Well, I tried out a clone of the 26 preamp that has me transfixed. You're not going to believe this, guys, but the original two Hammond 156C plate chokes in series sound better than the Lundahl LL1667/10mA. May be inductance - Hammonds are 300H while the LL1667/15mA is rated at 270H. So the 10ma version should be similar to the Hammonds. The static resistance is different - 7K for the Hammonds and 2.3K for the Lundahls. Maybe the capacitance is a factor - two chokes in series may mean the Hammonds are lower? Haven't tried to measure. I think I've pretty much cloned everything else - only small differences. Now this surprises me, but it's good news economically!
I will have to do some extended listening, but with the Lundahls some of the magic has gone - slightly thicker sound. The Hammonds are very transparent. I'll have to check the filament regs to see if the current is exactly the same in starved filament - that may be a factor.
Andy
I will have to do some extended listening, but with the Lundahls some of the magic has gone - slightly thicker sound. The Hammonds are very transparent. I'll have to check the filament regs to see if the current is exactly the same in starved filament - that may be a factor.
Andy
Last edited:
Progress report: I received my autoformer volume controls, front panel display board, rear input board and remote from Bent Audio today. Excellent quality, very substantial parts. Six remote-selectable relay activated inputs, two outputs plus a tape output, 1dB steps on the VC, with a range of -51dB to +7dB. I've rearranged my preamp-on-a-board to fit them in and shorten the signal path, and measured to determine the box size I'll need. I have a snapshot below with the preamp, the display and input boards, and the PS parts in back (still waiting on the power transformer).
Front Panel Express can make the boxes I'll need, I just have to set up the front and rear panels on their CAD software.
Front Panel Express can make the boxes I'll need, I just have to set up the front and rear panels on their CAD software.
Attachments
Last edited:
I use a Tribute 1:2 interstage in reverse. It has 160He on the (now) primary, very good results.
Imho the static resistance is VERY important here. If you want to try out something interesting: Just put a 100˜150ohm resistor between anode and choke in series. You will hear a noticeable difference..
Another thing: When playing at approx 97Volt on the plate, even a 0.4Volt difference on the cathode will sound different (and using other brands '26 as well? ..)
Rob
Hi Rob,
I tried out the LL1667 with an added 4.7K anode resistor to make it equivalent to the 2x Hammond 156Cs in static resistance. The sound was better - cleaner. But still not as good as the Hammonds.
I'm intrigued by all this. The next step is using the Hammonds for the driver stage in my 300b amp. Currently that's 46s into Hammond 126Cs. Its good as it is but I'm wondering if I can now sneak in a 10Y with 2x 156Cs as a driver at lower voltage and current with starved filaments....... Not sure if it would like the 8mA limit on the Hammonds but worth a whirl. Could be direct coupled from the 26 also.
Andy
I tried out the LL1667 with an added 4.7K anode resistor to make it equivalent to the 2x Hammond 156Cs in static resistance. The sound was better - cleaner. But still not as good as the Hammonds.
I'm intrigued by all this. The next step is using the Hammonds for the driver stage in my 300b amp. Currently that's 46s into Hammond 126Cs. Its good as it is but I'm wondering if I can now sneak in a 10Y with 2x 156Cs as a driver at lower voltage and current with starved filaments....... Not sure if it would like the 8mA limit on the Hammonds but worth a whirl. Could be direct coupled from the 26 also.
Andy