It means:
If your house has big rooms....... big enough to have room for LP players, and 1000s of LPs, you probably have space for loudspeakers, and don't need headphones.
The sound system that initiated me in the audio world in 1974 was an integrated amplifier Copland with MM valves, loudspeakers 3 way passive Dynaudio and TT Linn Sondek LP12 with Linn MM K9 cartridge. After a couple of years I discovered Stax but never had enough money to buy because I wanted the combo: headphone amp + headphones.
We have to consider ultrasonic filtering. I know many designs do not trouble to do it.
A single pole can be added by parallel cap across the output resistor.
Then again, the whole analogue subsytem, from DAC out to headphone socket - is open-loop.
Feedback can present problems if the forward path of an amplifier is not able to keep pace with the spectrum of quantisation and other noise. With open-loop design, and discrete components, there is far less reason to suspect trouble - maybe!
What DACs are we using? ESS9018? PCM170x ?
Volume control is another consideration - it's a shame to break up this single-current-path stage to accommodate a pot. Are the DAC-internal digital volume controls good enough yet?
A single pole can be added by parallel cap across the output resistor.
Then again, the whole analogue subsytem, from DAC out to headphone socket - is open-loop.
Feedback can present problems if the forward path of an amplifier is not able to keep pace with the spectrum of quantisation and other noise. With open-loop design, and discrete components, there is far less reason to suspect trouble - maybe!
What DACs are we using? ESS9018? PCM170x ?
Volume control is another consideration - it's a shame to break up this single-current-path stage to accommodate a pot. Are the DAC-internal digital volume controls good enough yet?
erno borbely told me the PCM1704 is the best DAC chip ever made, so I took his word, been working with it for years never found anything better. It doesn't allow digital volume control so a potentiometer or autoformer is unfortunately a requirement but its filtering requirements are much less since it doesn't operate in the hf's that sigma delta chips do, generally a CLC filter is enough.
PCM1794A or PCM1792A -- the best DACs from TI with 132dB SNR (in one particular circuit as per its datasheet). Also it has one little function that helps solve the next problem:
Another
unusual proposition is Reed relay switched resistor ladder. The DACs above have zero crossing signal, so if the volume has to be changed, relays wait for the zero crossing to switch, which eliminates a click due to switching. The best relay I could find switch in 0.25ms. The contacts are sealed in vacuum with an option of mercury coating to enable switching of very weak signals without loss or arcing. Combined with zero crossing it should be sufficient for silent volume regulation. Of course that requires a microcontroller to drive the relays. There is a much easier option for that in a single chip (such as TI's LM1972), but that means a FET switch in signal path at all times.I agree with what you are saying completly but there is always the chance that .75W isn't enough, what if in a year the new HE6 takes 2W. After building the "perfect" SET for Grados I look for this DHT build to provide exponetially more power so the latest and greatest will "work". I like playing with designs and research, but actually sort of desipse casework, so I am after something that will last
The more I think about it we aren't going to come to consesus on a DAC, and its a bit of scope creep.
This 4p1l shunt cascode to 2a3/6c4c would fit all the critea of the original project definitions in the first page. It sims incredibly well and really I don't think the 2A3 is an expensive tube, the real secret is that the 6C4C with coleman filaments is a bit of an equalizer in that it is reported as good as the 2c4c. So we have a parts cost of well under $150 including all active devices not inculding chassis and iron.
Its powerful enough that one could use a huge range of off the shelf transformers with the 16 ohm tap and a parallel resistor to match impedance. I'll probably go for custom iron. Best bet is two 8 ohm secondaries. Run in parallel = speaker amp, in series center tap grounded gives a 32 ohm balanced headphone out.
All we need is a PCB
Last edited:
It seems like your design goal should be for a case and layout that easily allows component changes and additions rather than for an omnipotent electrical circuit. Maybe that sounds a little smart-alecky but it is truly intended as something constructive and pragmatic about how you are approaching this mentally.
My PCB already exists for shunt cascode.... but this application would require some updates, and add the shunt regulation for the bias and cascode voltages.
The shunt regulator design is already laid out as a separate board. It's an ultra-low-noise design with discrete-transistor error amp.
A new PCB (or PCB set) is possible, but will need some development & 'productionising' work to make it suitable for widespread use.
One consideration is that surface mount parts are really needed for best performance - including chip ferrites and SOT23 transistors. This would also allow much smaller PCBs - eg the shunt regulator is only 78 x 32mm, as it stands, and looks rather like my filament supply boards!
Maybe I will need to assemble them.
The other consideration is that I am looking for a DAC myself, and don't like any I have seen so far. The I/V converters don't seem completely satisfactory either....maybe I like that DAC current-out design enough that I will make a PCB with this - and an ESS9018 added, if only for my own system. I design computer boards for a day job, so the layout problems are well understood.
Anyway, I'm working on the basic boards, and if they look worthwhile for manufacture, I can always get some fabbed up.
The shunt regulator design is already laid out as a separate board. It's an ultra-low-noise design with discrete-transistor error amp.
A new PCB (or PCB set) is possible, but will need some development & 'productionising' work to make it suitable for widespread use.
One consideration is that surface mount parts are really needed for best performance - including chip ferrites and SOT23 transistors. This would also allow much smaller PCBs - eg the shunt regulator is only 78 x 32mm, as it stands, and looks rather like my filament supply boards!
Maybe I will need to assemble them.
The other consideration is that I am looking for a DAC myself, and don't like any I have seen so far. The I/V converters don't seem completely satisfactory either....maybe I like that DAC current-out design enough that I will make a PCB with this - and an ESS9018 added, if only for my own system. I design computer boards for a day job, so the layout problems are well understood.
Anyway, I'm working on the basic boards, and if they look worthwhile for manufacture, I can always get some fabbed up.
Attachments
I agree, I am still kicking myself for having electraprint-print gap my 5k:32 ohm transformers for 20 mA
should have at least done 35 ma. The nice thing about parafeed is config/tube changes can be accomadated a bit easier.
The tough part is a headamp for sensitive phones HAVE to be a two chassis build with the power transformers/chokes in a separate chassi 3 feet away.
I put together a nice looking case that looks classic tube amp even with a big heatsink for the shunt reg, but its too small to add coleman regs or make it a two stage amp. But I am not big on asthetics, I just want it too sound good and be safe, but being an engineer (ChE) I have to live safety and can't stomach open builds even for proto typing. Next build will definately be the bigger case's ( I like the hammond aluminum case that come with wood sides, they are cheap and look good painted with texture paint.)
PCM1794A or PCM1792A -- the best DACs from TI with 132dB SNR (in one particular circuit as per its datasheet). Also it has one little function that helps solve the next problem:
Another
This might work well - with shunt-cascode, we only have to switch IN one extra shunt resistor to decrease gain. It doesn't need a pot.
I control volume like this now, only using a rotary switch (not click-free) so I am seeking something better.
Sounds great, Rod!
As for I/V DACs, one thing that I've tried VERY successfully (with PCM1794 and others) is to use a 10-ohm resistor for I/V. The output signal needs significant gain, which merely means an extra tube stage in the amp. It is a more transparent-sounding solution than active I/V or transformer I/V.
As for I/V DACs, one thing that I've tried VERY successfully (with PCM1794 and others) is to use a 10-ohm resistor for I/V. The output signal needs significant gain, which merely means an extra tube stage in the amp. It is a more transparent-sounding solution than active I/V or transformer I/V.
I started a thread on which DAC board to feed directly into the grids of a 4P1L, 26 or whatever. I presently have a RAKK DAC mk 1 which does this (uses a 75R I/V resistor), and looking for a ESS type board. I don't know what output devices the boards have (ODAC etc) and if they can be bypassed.
http://www.diyaudio.com/forums/digi...ac-board-directly-drive-grids-tube-stage.html
http://www.diyaudio.com/forums/digi...ac-board-directly-drive-grids-tube-stage.html
I've done exactly that (see the other thread) and with a 1:1 transformer there was not enough gain. Other than that, no problem. Most voltage DACs have an output impedance around 100 ohms and don't mind a low impedance load, so they should not have much trouble driving the grid via a step up transformer. I have a few that I could try.
Thanks, that agrees with what I've been seeing. But if I need the shunt cascode for the >50 gain, isn't it the shunt doing the amplifying? Trying to keep the gain as much in the DHT as possible, you understand.
The PNP transistor separates the signal portion of the anode-current from the anode itself.
With the anode - and its low impedance - out of the way, the signal current can develop a higher voltage on the load. It is the low impedance of the anode that squashes the gain.
But the signal current is 100% generated by the triode, so you're not getting any transistor fakery.
the signal current you get is:
Iout (mA) = Vin (the grid voltage) x gm (the triode's mutual conductance in mA/V)
So none of the gain is from the transistor... this is present only as a support function, that allows the triode to maximise the use of its generated signal.
With the anode - and its low impedance - out of the way, the signal current can develop a higher voltage on the load. It is the low impedance of the anode that squashes the gain.
But the signal current is 100% generated by the triode, so you're not getting any transistor fakery.
the signal current you get is:
Iout (mA) = Vin (the grid voltage) x gm (the triode's mutual conductance in mA/V)
So none of the gain is from the transistor... this is present only as a support function, that allows the triode to maximise the use of its generated signal.
Its a good topic, but thanks for starting a new thread as I think it is best handled separate.
Back to the SET headamp. I am a bit of a sucker for headphone reviews. This is a good run-down for anyone interested in high end headphones, he touches on some DHT amplification:
Battle Of The Flagships (57 Headphones Compared) UPDATE: HifiMan HE-400 added 12/24/12
Now with a professional masterings I found that and taking the http://www.diyaudio.com/forums/multi-way/204857-test-how-much-voltage-power-do-your-speakers-need.html I need just 6 Vp-p swing with the HE-500's.
So 45 was correct. I was way overestimating my power requirements based on just short listening.
Now this takes me back to Iko's proposal of just a simple 4p1l-4p1l CCS-ac coupled. This will power all the dynamic headphones on the "57 greatest" list except the HE6. As long as the mastering is good there should be enough sensitivity. Where I was running into trouble with the test was with HDCD recording decoded to -6dB and some raw sbd's that were very low level, but I guess that is why we have replay gain and other technology.
So I'm saying is the most efficient lowest distortion design is going to be a 4p1l to 4p1l. If we CCS or Gyrator ac couple then it really makes sense to parafeed. By parafeed the advantages as outlined by dsavitsk, we get a high performance permalloy core transformer with most important a low dcr low inductance secondary.
I can tell you that most singlefeed OPT 32 ohm secondaries have a DCR around 8 ohms. I think this forms a voltage divider when you sketch it out. So if your attach 32 ohm phones you get a (8/32)^2 power loss. Doesn't sound like much till you consider how much harder your amp is working (more distortion.) Also factor in the K factor of the transformer and you find that your single feed transformer is causing your amp to distort more than it should for low level detail.
So I'm saying we want a 4p1l-4p1l parafeed Sowter Type 8665 ($274/pair).
For those on a budget there is a similiar transformer from edcor PCW10K-7K/300-32 for $20 a pair.
I think we may be able to get away without needing a shunt regulation since with parafeed the gyrator or cs would have good psrr, but we still need separate chassis for the power supply.
The toughest part of course is the coupling caps, I leave that to the discretion of the builder but the parafeed cap is going to need to be big, I've had good luck with the russian K75 parallel with the teflon russian. For the coupling cap back to the brits with their ampOhm copper oil.
I can sketch this out for anyone interested, but its simple to the point and does everything we need without breaking the bank. Most important it preserves the low level linearity we want in a DHT headphone amp.