Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › A lo-cost HT-supply true transformer-coupled tube headphone amplifier
New Posts  All Forums:Forum Nav:

A lo-cost HT-supply true transformer-coupled tube headphone amplifier - Page 6

post #76 of 108
Thread Starter 
Originally Posted by elliottstudio View Post

Are you going to build it out on that PCB you designed? That would be a treat.


Looking at your first stage tube circuit, I've tried using a diode (and an LED too) for the cathode bias and measured more THD than with a plain resistor. Does the CCS load on the plate help alleviate some of the distortion?


I'm still looking at the positioning of the PT, OPTs and Choke for the rebuild. The PCB requires planning, or at least thinking about the whole layout before you start. That's one of the reasons I'm attracted to it, because of the discipline that it enforces, but it's more inflexible than point-to-point. I might go for a combination approach, split the PCB into a number of subcircuit assemblies. I liked the layout I had in the first build, it was just too cramped. I'd like to follow the general scheme for the PCB, but there are problems. It might be necessary to have holes in the PCB to allow wires from the top of the chassis to pass through to the component side of the PCB. I would like to stick with a single-sided PCB because I can make it myself (no through hole plating), a doublesided PCB that size would be expensive to have manufactured. I can't get PCB mount dual-gang 4-way switches, so I'd have some PCB mount parts passing through the front panel, but the 4-way switches would have to go to headers. I can't use a 10 * 6 PCB which I would like to because although the chassis I'm looking at is 10 * 6, it has a lip on the bottom, so actually inserting the PCB is problematic.


While the toroidal transformer suits me, it was cheap and simple, in some ways I'd prefer to revert to the back-to-back arrangement I drew in the other thread. I'd like to have 12V and run the heaters in series because the DC would be a lot less fussy with a peak voltage ~17V, there'd be more overhead for the regulator, and you'd only need one, because the current would be halved. You'd be able to use a 110-0, 110-0 primary on the first transformer which would make the build directly accessible to builders in the US, but sourcing transformers which are available in both Europe and the US is a pain. I'd really prefer to have both power transformers PCB mounting which adds to the difficulty if anything. 


It's all a bit of a puzzle. The way I usually deal with these conundrums is to let the problem stew around in my head until my subconscious solves it. Sometimes a new piece of information comes to light which pushes the decision in a particular direction. Maybe in this case it's not possible to get an absolutely optimal solution. I'm just not going to go at it like a bull at a gate like I did the first time round.


The theory says that the CCS plate load results in a horizontal load line. This should result is a reduction in distortion, the spacing between where lines on the characteristic intersect the load line becomes more constant. Whether this pans out in practice, I don't know.


I'll look at replacing the diodes with resistors in the sim. and see what effect it has on the THD. I'm just following the recommendations from Morgan Jones' book and the suggestions I've had here for the moment. Not that I feel that the sim. is to be taken as gospel anyway...


Transistors for the CCSs arrived.



post #77 of 108

It will be interesting to see the difference between diode and cathode resistor THD on your sims with a CCS load. My circuit uses a plate resistor load so not as linear to begin with, but the diode bias increased the measured distortion by an order of magnitude if I remember correctly.


I wonder if the measured THD and the simulated THD would be the same in your design for each case (resistor vs diode)?

post #78 of 108
Thread Starter 

Overall distortion (distortion at the transformer output) rose from ~1.425% to ~1.48% when I replaced the diode with a resistor.


This is with the amplifier driven very close to grid current in the second stage, which is the limiting condition governing the power which can be obtained.


Distortion at the output of the first stage is 0.775% with the diode, rising to 0.79% with the resistor.


These are the figures automatically calculated by the program. I trust the figures calculated by the program for SS amps where the FFT doesn't show a downward slope to the right, but I don't trust the tube sims because I don't know why they show this slope which I suspect throws the calculation off, so I treat them as valid only in comparative terms. I had been led to understand that the sloping noise floor was associated with a DC offset, but even taking the output across a large resistor coupled with a large capacitor does not eliminate it.



post #79 of 108
I would be very interested in one of these when the design is perfected, if parts are easily sourced, great work.
post #80 of 108
Thread Starter 

You should have no trouble obtaining the parts, I see you live in Portugal, I am in the UK. I sourced the majority of them from a German supplier, tube-town.net, they deal in Euros, the rest can be obtained from ebay.



post #81 of 108

Cool, I'm in england next week too, but i can get parts easily i think, now to wait for the board to be ready, how much will be the aproximate cost?

post #82 of 108
Thread Starter 

Here's a BOM with costing, major component prices are accurate, small component prices may vary from those shown. You can probably save a few Euros by being careful. I have simply marked generic components 'ebay supplier'



Category Quantity Reference Value Price (Euro) Total Source
Resistors 2 R1,R2 5k1 0.1 0.2 ebay supplier
Resistors 2 R3,R7 1k2 0.1 0.2 ebay supplier
Resistors 3 R4,R27,R32 500k 0.1 0.3 ebay supplier
Resistors 5 R5,R13,R17,R20,R33 1M 0.1 0.5 ebay supplier
Resistors 6 R6,R9,R25,R29-R31 10k 2W 0.17 1.02 tube-town.net
Resistors 2 R8,R12 33R 0.1 0.2 ebay supplier
Resistors 2 R10,R34 2M2 0.1 0.2 ebay supplier
Resistors 2 R11,R35 3M6 0.1 0.2 ebay supplier
Resistors 4 R14,R18,R21,R26 4k7 0.1 0.4 ebay supplier
Resistors 4 R15,R19,R24,R28 56R 0.1 0.4 ebay supplier
Resistors 2 R16,R22 1k 0.1 0.2 ebay supplier
Resistors 1 R23 100k 0.1 0.1 ebay supplier
Capacitors 4 C1-C3,C8 220uF 350V 3.95 15.8 tube-town.net
Capacitors 2 C4,C6 100n 0.03 0.06 ebay supplier
Capacitors 1 C5 22000uF 16V 0.5 0.5 ebay supplier
Capacitors 2 C7,C9 100uF 350V 3.45 6.9 tube-town.net
Capacitors 2 C10,C12 3pF 0.03 0.06 ebay supplier
Integrated Circuits 2 U1,U2 LM317L 0.5 1 ebay supplier
Integrated Circuits 2 U3,U4 6N6P 2.7 5.4 ebay supplier
Transistors 2 Q3,Q4 MJE340 1 2 http://www.ebay.co.uk/itm/200590106893?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649
Transistors 2 Q5,Q6 MJE350 2 4 http://www.ebay.co.uk/itm/180643655858?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649
Diodes 8 D3,D4,D6,D7,D18-D21 1N4148 0.1 0.8 ebay supplier
Diodes 2 D5,D8 Vzen=5V 0.2 0.4 ebay supplier
Diodes 4 D13,D15-D17 Shottky 0.25 1 ebay supplier
Diodes 1 D14 RED LED 0.1 0.1 ebay supplier
Miscellaneous 1 BR1 400V 1.2 1.2 ebay supplier
Miscellaneous 1 FU1 160mA 2 2 ebay supplier
Miscellaneous 2 J1,J2 1_4_IN_TRS_JACK 1.5 3 ebay supplier
Miscellaneous 1 L1 10H Hammond 158M 16.55 16.55 tube-town.net
Miscellaneous 1 RV1 100k log 1.5 1.5 ebay supplier
Miscellaneous 1 SW1 SW-DPST-W 1.5 1.5 ebay supplier
Miscellaneous 2 SW2,SW3 SW-DP4T 2.17 4.34 http://www.ebay.co.uk/itm/Rotary-Switch-3-Pole-4-Position-way-Shorting-SW195-/150614736838?pt=UK_BOI_Electrical_Components_Supplies_ET&hash=item2311567fc6
Miscellaneous 1 TR1 tube-town.net 30VA toroidal 22.5 22.5 tube-town.net
Miscellaneous 2 TR3,TR4 Hammond 125BSE 36.4 72.8 tube-town.net
Socket 2 SCK1,SCK2 Noval Ceramic 1.59 3.18 tube-town.net
Chassis 1 CHAS1 AL 1444-16 11.6 11.6 tube-town.net
Power connector 1 POW1   1.5 1.5 ebay supplier
  Euro       183.61  
  UKP       148.24  
  USD       238  


I have had a new idea about how to arrange the transformers, 4-way rotary switches and other components.


If I get 10 sets of PCBs made of the size necessary they will be maybe 15 Euros to you including shipping. A doublesided PCB, properly manufactured, gives me a few more options when laying it out. I can have a vertical daughterboard to carry the 4-way switches, have ALL the components PCB mount with header plugs for the transformers.


It's a lot of money for me to lay out though, so the best thing, I think, is maybe to build it point-to-point, get some RMAA results, work up a 3D CAD model and then see if I can get enough takers to make it worth making 10 sets of PCBs.


Give me some time to try drawing the layout in the software. Maybe I can still make the first PCB myself, that would reduce the risk.



post #83 of 108
Thread Starter 

Here's the new PCB with routing completed, showing grounding highlighted.






This shows the chassis in outline with transformer positioning and 4-way switch mounting arrangement. Some of the track spacings are a bit tight, and the whole thing needs a bit of fine tuning, but it shows that the routing is possible. Regulators and CCS transistors will be mounted off-board on the chassis for heatsinking.




Oh, I'll be able to fabricate this at home, no problem.



Edited by wakibaki - 12/10/12 at 9:53am
post #84 of 108
Thread Starter 

I haven't got round to the PCB rebuild, but I did find the problem with the original build. As I thought, the problem was with a 220u cap in the PSU which I dented while wrestling with the heater supply. I was reluctant to go for a repair, I was thinking of a complete rebuild, but after leaving the amp to sit for a few weeks, I figured a way to do the repair that wasn't too ugly. I replaced the cap, fortunately I had a spare on hand, and the tick, tick, ticking sound went away.


I'm not 100% happy with the overall result, there's not much gain, and the sound isn't as clean as I'm used to. I'll give it a bit more listening though, and maybe fiddle with the gain. Now that its built and going, I'm not so keen on cannibalizing it.


I've found some even cheaper OP transformers, 3.5W Fender reverb drivers, they're only about $10 each. Again they only go down to 100Hz at full output, but at about half-a-watt they should be OK. They've got a very high primary impedance. I'm thinking about buying a few and doing some more experiments with them. That woud obviate the necessity to dismantle this one. 'Course I'll need some power TX's, a couple of chassis, some HV caps and maybe some tubes too...



post #85 of 108
Great, can't wait till you get it up and running to your taste!
post #86 of 108
Thread Starter 

A few more sims:


All these are fed from a low impedance source, the amplifier worst-case input impedance is ~25k. It drives 16 ohms with gain, not much gain, but it's a reasonable impedance transformation, and it's sensitivity is in the correct range for typical sources, 1- 2V for full output, depending on load impedance, and it'll drive 600 ohms to 9V peak. It's a bit better than a buffer, about what you might expect for a headphone amp to meet a present-day requirenent. These kind of numbers are what you might reasonably be expected to cater for, but not with a lot to spare, whereas it would be easy to batter the requirement into submission with gain and voltage and current and feedback using transistors. 


You have to remember that the transformer is idealized, there are current sources in the circuit, and these simulations are particularly revealing of the tubes performances, but indicative of the best performance of which each tube is capable. This means that the tube's performance may in practice be masked by nonlinearities in the transformer not modelled in the simulation. There are in addition nonidealities in the current sink implementations not accounted for and other factors which mean that no real-world circuit will achieve the idealized performance shown here.


Keeping the power output to the minimum truly required makes it easier to overspecify the O/P transformer in practice, which at least brings the theoretical performance closer.


This one is DC coupled into 16 ohms, fully driven. 



It turns out that capacitor coupled is best. Here's 1mW into 16 ohms.



This is Vmaxout into 600 ohms...



Take it from me, they are better all round than the DC coupled version.


It's obvious, as well, from a component cost point of view, capacitor coupled makes the most sense, it means that the bypass on the stage 2 current sink can be a much lower voltage part. 


So there'll be a revised circuit along in a minute, with cascoded FET current sources. This makes the circuit versatile, because you can plug in just about any tube you want, any SE transformer you want, and there's a lot of leeway with B+. Although it's drawn for identical tubes, it's hardly difficult to take 2 channels through 2 different tubes, if you're wiring point-to-point.


If anybody thinks that the THD should be better than it looks, I can only say, you wanted that tube sound, didn't you?


Here's what it looks like with 6SN7:



Better, but not a helluva lot better.



post #87 of 108
Thread Starter 

Just fiddling around, listening to this amplifier as I have built it, I started tapping on one of the tubes, suddenly the volume level jumped up quite a lot, accompanied with a loss of fidelity. Gave me quite a fright.


Debugging these things is so much more difficult than solid state, because you can't just poke around in there with quite the same degree of freedom. I hate dangerous voltages. Tubes get hot.


It seems there's some kind of wiring problem in there that I wasn't aware of until now. I'd probably have found it before now if I wasn't constrained to be so cautious. This is quite good news though, because it seems there's quite a lot more gain available than I had previously thought. Because it was an experiment my expectations weren't very disappointed, I just figured that the circuit didn't perform quite as per the sim. No surprise there.


The sound up until that point was quite good, just a bit quiet, I'm optimistic now that it will turn out to be quite successful. 


I've got some more (3.5W reverb) transformers coming + a new chassis and some other components for a chokeless PSU and HV CCS plate load, etc. etc. I'm going to try another assembly that will be a lot less cramped than the first, maybe I can get it to work a bit better this time. I've learned a bit about how to go about the assembly, i.e. wire the bases before screwing them to the chassis. It should actually be a bit cheaper without the choke. Bits should come Friday or Saturday.


Got to get some sleep.



post #88 of 108
Chopsticks are your friend when poking around in a tube amp.
Edited by looser101 - 2/10/13 at 9:03am
post #89 of 108
Thread Starter 


post #90 of 108
Thread Starter 

I did some more sims. I improved the transformer model, I've put in some parasitic resistances, tweaked the coupling (leakage). I need a transformer spare in my hand to measure.


Then I put in some 6SN7's and optimized. 6SN7's have a couple of things going in their favour.


1) They're better at lower currents than the russian tubes. The tubes are best >= 8mA I read, but the CCS's help the linearity, the sim works best at the values shown, and I like the fact that the standing current in the transformers is low.


2) They're in current manufacture.


3) They don't cost a fortune. They're not as cheap as the russian tubes, but they're nowhere prohibitive. 


4) They've got better distortion characteristics.


This is 1V in for 2V out at 1/2 volume pot: All voltages quoted are PEAK.



This is the AC response at 1/2 volume:



This is the FFT at 1mW  into 16 ohms:



This is Vmaxout into 600 ohms:



This is 1V in for 4V out into 38 ohms:



This THD of 1.3V in for 5.5V into 38 ohms. No point in trying to gild the lily. This is where you can really see the output suffering



6% THD. Not too shabby. It's not anything it pretends to be. It's buffer from 10k to a few ohms, there's a little bit of gain.


Now there's a question of implementation.


The Supertex LR8 regulators will easily supply these currents, 2 as plate loads, 2 as VR's for the txfrmr stage, then we'll have LM317 CCS's for the cathodes.


Low current means the ripple on the supply will be less, and this time we'll have a CRCRC filter.




Slightly higher voltage available than that shown.



And a couple of PCBs





New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › A lo-cost HT-supply true transformer-coupled tube headphone amplifier