I would much appreciate if you elaborate: what did you do, why, what was the result, what brought you to your conclusions. Please walk me through that experience, with all details.
One problem I see with parafeed is that it seems to be little known, little used and thus poorly understood. People love DHT SET to bits and argue only about minutiae, but I see very few references on parafeed, and no consensus at all. Another problem is transformers. Lundahl has a long line of transformers for all kinds of tube amps, but zero parafeed output ones. In such a realm where people's impressions means Lundahl's business, I find disturbingly little interest from the likes of Lundahl to parafeed. Also there is an irremovable capacitor there, which also forms an irremovable resonant circuit with transformer primary, which is also disturbing in a tube headamp. On the contrary, DHT 4W Class A amps are a well tested range of designs, with headphones, small speakers, and a range of DHTs well tested, described, criticised and liked -- enough designs and usage experiences to analyse.
Please do.
All Spice Sim 2 stage CCS-cap coupled 2 stage 5k:38 with 38 ohm load
1V 1W
6n6p-2A3 0.11% 1%
6n6p-4P1l 0.12% 1.3%
3A5-4P1L 1.40% 1.5%
4P1l-4P1l 1.40% 1.4%
26-4P1l 1.40% 1.5%
5842-2A3 0.18% 1.70%
What is the Data telling us?
#1. A DHT input tube probably isn't worth the trouble, the loss in gain and sensitivity, potential for microphonics, etc does not gain any reduction in THD
#2. A 4P1l is a decent alternative to the 2A3 for the budget conscious.
#3. All the tube combos are relatively similar in THD, the main factor in sound will be the quality of the output transformer.
Next I will work on sims with Parafeed. If you read the ECP documents it explains why Parafeed allows a better quality transformer for SET. Also read the DHT P-P headphone amp where Tribute recommends a P-P topology entirely so one can use a better quality transformer, parafed SET is the another alternative.
Yes the 1V out is meant to represent low level so its ((1V/1.44)^2)/38= 12.7 mW into 38 ohms.
The 1 Watt is 1 Watt RMS into 38 ohms. I'm using a fairly generic transformer model from modified from Slegals website 5k:38. All biasing are at "optimums" based mainly on Ale's website for the DHT's. For the input tube just a separate B+ & the generic spice CCS. Cathode resistor+cap.
Obviously real world results will be higher THD, that is why I say the transformer is the dominating factor , it is just obvious to me. Also avoid 3H leaning tubes like the 5842, D3A, 6922, etc for this 2 stage amp.
It is why Dsavisk came to the conclusion that for headphone SET parafeed is the way to go, you work the tube's less because of the efficiency/permeability of the core, you have less secondary dcr/inductance, less loss of low level detail.
Its the same reason they use hi nickey alloy (permalloy) transformers when dealing with microphones, line inputs, phono preamps, why they used the nickle cinemag transformer at the superbowl studio this year, etc. Good cores are used with low level milliwatt detail content we need for headphones.
What throws a wrench in all this is of course the modern orthos.
So I should say 99% of all headphones. I have completely ruled out the Hifi-Man HE-6, talking to folks who have tried $10k commercial DHT SET's there wasn't enough power.
But I recently bought a newer model the HE-500 which is much more efficient and "tube Set compatible", sounds pretty good at about at roughly 0.25W. The other orthos the LCD-2/3 are efficient enough that they should work from the 38 ohm secondary.
For HD800 and the Beyers we will want probably a 150 ohm tap plus parallel resistor.
Last edited:
Probably the transformer model ( I am actually using a 40 ohm load):
.SUBCKT OutXFMR_5.0k:40 1 2 3 4
L1 1 5 46H ; primary inductance
R1 5 2 103 ; primary resistance
L2 3 6 0.366H ; secondary inductance
R2 6 4 2.4 ; secondary resistance
K1 L1 L2 0.9996 ; coupling coefficient
C1 1 5 5p ; primary side capacitance
C2 3 6 30p ; secondary side capacitance
C3 5 6 200p ; primary to secondary capacitance
.ENDS
Could be the bias points, I will put this in a zip file and upload tonight, I would like to compare notes also.
.SUBCKT OutXFMR_5.0k:40 1 2 3 4
L1 1 5 46H ; primary inductance
R1 5 2 103 ; primary resistance
L2 3 6 0.366H ; secondary inductance
R2 6 4 2.4 ; secondary resistance
K1 L1 L2 0.9996 ; coupling coefficient
C1 1 5 5p ; primary side capacitance
C2 3 6 30p ; secondary side capacitance
C3 5 6 200p ; primary to secondary capacitance
.ENDS
Could be the bias points, I will put this in a zip file and upload tonight, I would like to compare notes also.
I use rar not zip will need to figure that out but here is a screenshot, note that the transformer model is extremely optimistic, its actually a parafeed transformer model. Real world results with an SE gapped Iron core would I expect be much worse.
Attachments
Last edited:
Honestly, I would not put the effort into building a headphone amp that promises to give me the listening pleasure of 1.4% distortion at 1W. Nothing wrong with your simulation. I'm just saying that for me this would be extremely discouraging. If it turns out that my combination of tubes ends up with similar levels of distortion I don't think I'll go ahead with the project.
A real output transformer will add even more distortion, especially SE gapped. The only way to do significantly better is a 1000+V GM70, heck I've simulated a million combinations. $5k to $10k commercial DHT headphone amps have the same relative levels of THD, this is SET and the norm. Remember its the first milliwat that counts and a black background.
You are comparing to a preamp OTL distortion numbers which swing much lower plate voltage those numbers are meaningless to a headphone amp. The 4p1l is straining to push a watt into 40 ohms.
Sounds like you want a Hybrid, the Tubmos output stage (from the Beau's hybrid) hooked up to a 4p1L will be best for your tastes.
Last edited:
Why don't you try to use a simple anode resistor and see if you can cancel a good amount of H2? In theory it shouldn't be difficult with a 4P1L driving a 4P1L.
You could start with 3.3K and 5.6K. These, in parallel with 100K next stage grid resistor, will make 3.2K and 5.3K respectively. With 3.3K and 750R cathode resistor it should run at about 205V/20 mA; with 5.6K and 620R cathode resistor it should run at about 170 V/19 mA.
A good output transformer will not add significant distortion at least between 100Hz and 10KHz. For very low distortion at low levels amorphous core should be the choice. I don't like permalloy for OPT's. It is a little bit of no-sense while it can work great at signal level.
You could start with 3.3K and 5.6K. These, in parallel with 100K next stage grid resistor, will make 3.2K and 5.3K respectively. With 3.3K and 750R cathode resistor it should run at about 205V/20 mA; with 5.6K and 620R cathode resistor it should run at about 170 V/19 mA.
A good output transformer will not add significant distortion at least between 100Hz and 10KHz. For very low distortion at low levels amorphous core should be the choice. I don't like permalloy for OPT's. It is a little bit of no-sense while it can work great at signal level.
The 4P1L is straining to push 1 watt into any load, even the OPT primary. Well, 1W is not too bad, but beyond that it's working hard. And there are plenty of headphones that will be happy with 1W. It works for me.
This is mainly signal level, all the studios have been using nickle for decades. I think it will be very difficult to find a transformer winder who will make a custom 1-3W amorphous core, probably can't even order the laminations in the size we need. I doubt a big 35W amorphous core would be a good match.
Nickle works well because it maintains a low THD through the bass at low signal levels, this is where iron has issues (low signal levels.) For most headphones all we really care about is the first couple milliwatts. The issue is nickle alloy must be a parafed.
Because you have to buy in $2k lots from the foundary, to have a transformer winder shell out that kind of dough for one guy is pie in the skie stuff. Then you have to find a competent winder. Trust me I've been thru all of this. Ran into the same issue with 49% nickel Lams.
Here is a picture of what ElectraPrint sold me from their left overs from a large order but it is a true embarassment. See the huge gap between the copper and the lams? Not good at all. Its not their fault really they just tried to couble together what they had left from a big job.
Attachments
It could do if you go for para-feed.
If you want 1W output with 40 mA DC current is no more signal level. In such case I think permalloy is a little bit of no-sense for a normal OPT because having to deal with such DC current you need put quite some gap and so you loose one of its main advantages: high permeability. If you can't get enough inductance then you will get distortion anyway at low frequency.
The amorphous core has a low coercive field, high resistivity (i.e. low loss, up to 80% less than iron) and practical permeability (i.e. using a gap for DC operation) comparable to the best iron core. It is not as good as the nickel for low THD at low levels but it is a good compromise. You can go for a Lundahl if you don't have other options.
Last edited:
I agree, parafeed - nickle. Singlefeed - amorphous. Both smaller than the usual speaker amps. Thats the issue.
I would jump on a nice c-core amorphous 50 mA gap 1W at 30 hz with a custom secondary.
The other option is the electra-print silverplated windings, they claim excellent low signal detention. If two 10 ohm secondaries can give a balanced 40 ohm match as I've been told that would be a great option, but I can't find the math to prove how that works.
All Spice Sim 2 stage CCS-cap coupled 2 stage
5k:40 opt, 40 ohm load
1V 1W
6n6p-2A3 0.11% 1%
6n6p-4P1l 12.00% 1%
3A5-4P1L 0.14% 1.50%
4p1l-4p1l 0.14% 1.40%
26-4p1l 0.15% 1.40%
5842-2A3 0.18% 1.70%
6n6p-300B 0.05% 0.47%
6n6p-gm70 0.01% 0.05%
Added the 300B and GM70 for the THD OCD SS guys (sorry Iko
)
Yes the 6n6p can drive the 300B or the GM70 a single watt.
5k:40 opt, 40 ohm load
1V 1W
6n6p-2A3 0.11% 1%
6n6p-4P1l 12.00% 1%
3A5-4P1L 0.14% 1.50%
4p1l-4p1l 0.14% 1.40%
26-4p1l 0.15% 1.40%
5842-2A3 0.18% 1.70%
6n6p-300B 0.05% 0.47%
6n6p-gm70 0.01% 0.05%
Added the 300B and GM70 for the THD OCD SS guys (sorry Iko
)Yes the 6n6p can drive the 300B or the GM70 a single watt.
No, they sell single pieces too. I got a few cores from them a little while ago. I'm not in any way associated with that company. To give you an idea what kind of prices they ask for amorphous cores, look what they quoted me some time ago:
>> Mass (g) PN 1-9pc 10-24pc
>> 99 BMCC 4 $5.38 $4.89
>> 154 BMCC 6.3 $7.48 $6.79
>> 172 BMCC 8 $8.35 $7.59
>> 198 BMCC 10 $9.42 $8.56
>> 248 BMCC 16A $11.56 $10.50
>> 281 BMCC 16B $13.09 $11.89
>> 337 BMCC 20 $15.70 $14.25
>> 379 BMCC 25 $16.92 $15.36
>> 454 BMCC 32 $19.39 $17.60
>> 530 BMCC 40 $21.62 $19.63
>> 586 BMCC 50 $23.89 $21.70
>> 703 BMCC 63 $27.99 $25.42
>> 938 BMCC 80 $36.44 $33.09
>> 1,055 BMCC 100 $38.93 $35.35