I've taken your advice regarding higher current, and simulated this. It still simulates very good, and I think I'll go for this for L/R channels.
There's 4 mA to the input diff and 2 mA to each of the cascodes. When using two diodes instead of a LED for biasing the cascode the voltage swings to within 1 V from rails. Now I have to decide if I'm going to breadboard it or order a PCB.
Changing just a couple of resistors and capacitors, I can use bipolars instead of JFETs.
There's 4 mA to the input diff and 2 mA to each of the cascodes. When using two diodes instead of a LED for biasing the cascode the voltage swings to within 1 V from rails. Now I have to decide if I'm going to breadboard it or order a PCB.
Changing just a couple of resistors and capacitors, I can use bipolars instead of JFETs.
Attachments
A nice design. Good work!
I would add some more decoupling caps, though. Say 100n PP caps on the rails on 2 positions on the PCB.
DC offset might need some attention. One way to keep when temp changes, is to glue the two input JFETs together. Or use shrink wrap or a copper tube.
Sigurd
I would add some more decoupling caps, though. Say 100n PP caps on the rails on 2 positions on the PCB.
DC offset might need some attention. One way to keep when temp changes, is to glue the two input JFETs together. Or use shrink wrap or a copper tube.
Sigurd
nelsonvandal said:
I agree with Sigge (our master of speed vehicles design):
A nice design. Good work!
Look a here: Post #61
Post #61 - Latest & Greatest version
Of This Folded Headphone/Line amp
nelsonvandal amplifier, has got design, that has been proven, again and again.
In real hifi Op-Amps by www.analog.com Analog Devices
.... there are so many variants of this way in:
AD797(bipol), AD823 (jfet), AD82X and ADxxxx and AD...
God (if he is there ..) only knows,
how many opamps they have sold, thanks to all variants to this working idea:
Folded second stage + Diamond buffering outQuestion:
What could / should I change in nelsonvandal circuit ?? If I wanted my own High Class Headphone Amp ??Answer: not much.
But I would see (using Simulation, Listening and/or Oscilloscope Live testing)
if these following 2 things, would improve anything.
=====================
1.
Use JFET current sources. Like BF245A or BF245B (To save some money, instead of 'waisting 2SK170' for this.)
As JFET CCS has no resistors to power supply, They have higher PSSR. (psu rejection ratio)
=====================
2.
I would experiment with some 'better' Output transistors.
First Iwould try, if I can use a higher gain variant of BC327-25 + BC337-25.
BC327-40 and BC337-40 ( gain 400 ) are not easy to find.
I could also try to use some Japanese Output transistors.
Hi Speed TO-126 Drivers, like those SC4793 / 2SA1837 and whatever.
=====================
These two try of mine may be an improvement, in some way.
Or it would not improve things worth to mention ... or to do it.
-----------------
Point # 1, use JFET current sources.
It depends very much on the power supply, if we really need higher PSRR.
And we should know, that JFET CCS can have capacitance issues
and so sometimes benefit from cascoding by BC550C.
So to use BC550C current sources, in the first place, is probably a good idea.
=====================
Lineup .. regars to all & the many other swedish in this topic nelsonvandal said:
Sigurd Ruschkow said:
AndrewT said:
Yes, this may do good,
2 JFET in cascoding configuration.
If we need any CCS changes, is dependent on the PSU, that builder wants to use.
For a unregulated supply .. can be a good improvement with higher PSRR.
For a lownoise regulated PSU .. it may not improve Anything, at all.
Another observation in my MultiSim:
Regarding preamps with Voltage Gain ( say +12dB to +20dB )
The input stage, with only <= 1 Volt AC delta, may not benefit hardly any by cascoding of JFET CSS.
But very much different can be in (diamond) Output stage current sources.
Here the delta AC Voltage is like 4-10 times higher ..... (+12 to +20dB)
It is here I have found Cascoding can do a real difference.
At least a difference, that makes it worth while to try Cascoding JFET.
Lineup
For a Headphone amplifier, however, we often do not need much Voltage gain (16/32 Ohms phones).
And so we can skip cascoding (keeping a simple & good circuit).
And do just as good Hi-Fi Quality output.
Thank you for your kind words
I've used JFETs as CCS before (mainly 2N5484 and 5486), and I agree that they are brilliant as you don't have to connect them to ground, but you do need a resistor because of the large spread of IDSS. I find bipolars more reliable when simulating, but I'll probably use JFETs in real life. The problem here in Sweden, is that they don't come cheap, and are hard to find. 2SK170 is really the cheapest alternative here. They are 4.85 SEK each when bying 25 units and BF245 is 5.52 SEK.
I have lot's of BC327/337, and I have used them as output devices in several diamond buffers. I find the BD's warmer with a weightier bass and smoother treble. But the TO92 package is more attractive in a portable amp, and I was thinking of BC639/640, but I've never heard them or seen anyone else use them. I guess they're a lot like the BD's.
Maybe I should give MJE15030/31 a try. There aren't so many alternatives here. I buy most of my stuff from ELFA, because it's only a couple of km's from where I live, and it's expensive to order from abroad.
The amp is intended for battery use and with an active ground channel, so I don't know if chasing even lower PSRR is worth it.
I've used JFETs as CCS before (mainly 2N5484 and 5486), and I agree that they are brilliant as you don't have to connect them to ground, but you do need a resistor because of the large spread of IDSS. I find bipolars more reliable when simulating, but I'll probably use JFETs in real life. The problem here in Sweden, is that they don't come cheap, and are hard to find. 2SK170 is really the cheapest alternative here. They are 4.85 SEK each when bying 25 units and BF245 is 5.52 SEK.
I have lot's of BC327/337, and I have used them as output devices in several diamond buffers. I find the BD's warmer with a weightier bass and smoother treble. But the TO92 package is more attractive in a portable amp, and I was thinking of BC639/640, but I've never heard them or seen anyone else use them. I guess they're a lot like the BD's.
Maybe I should give MJE15030/31 a try. There aren't so many alternatives here. I buy most of my stuff from ELFA, because it's only a couple of km's from where I live, and it's expensive to order from abroad.
The amp is intended for battery use and with an active ground channel, so I don't know if chasing even lower PSRR is worth it.
nelsonvandal said:Thank you for your kind words
I have lot's of BC327/337, and I have used them as output devices in several diamond buffers. I find the BD's warmer with a weightier bass and smoother treble.
But the TO92 package is more attractive in a portable amp, and I was thinking of BC639/640, but I've never heard them or seen anyone else use them.
I guess they're a lot like the BD's.
a lot like the BD's (BD139/140)
You are exactly right! nelsonvandal
BC639-16, BC639-25 are simply TO-92, 1 Watt versions of BD139-16 & BD139-25.
As a matter of fact you also have 'new' SMD versions, using same die (core of transistor).
All these use basically the same Spice Model.
main difference is the package .. and how much power these packages can take.
See www.elfa.se
Search for BCP56-16 and BCP53-16
https://www1.elfa.se/elfa~eu_en/
They come in SOT-54(TO-92), SOT-223, SOT-89, SOT-23 and as BD139/BD140 in TO-126
Philips: you do not change a Winning Team .. not in football and not in transistors
So Philips want to sell some more of these good PRICE/performance audio transistors
--------------------------
Here are the Philips product links, with all different packages.
You will also find Philips own spice models, for download:
NPN - BC639; BCP56; BCX56
PNP - BC640; BCP53; BCX53
Spice Models:
http://www.nxp.com/models/spicespar/SST.html
I use BCP56-16 and BCP53-16 spice models
for both BC639, BC640, BD139 and BD140
Because they are basically the same transistor, with same model.
Lineup, regards
to nelsonvandal & his High fidelity (hi-fi) amplifier
I took your advice and added trimpots for DC-offset. The amp will look something like this. I wish I was handy with Eagle or a PCB-layout program that can produce gerber files. Do you have any recommendations? Preferably easier to handle than Eagle.
It's not debugged yet.
It's not debugged yet.
Attachments
After some fiddling with SprintLayout, this is the best I can do. It's not easy to place 3 channels inside a square area. As you understand, it has an active ground channel. Not shown in the schematics is the TLE2426 used as virtual ground. I forgot to place C8 in the ground channel, so it has to be stuffed in as a "bug".
It's intended for throughhole transistors and tombstoned or SMD (805) components.
I know it looks like a rats nest, but do you see any obvious errors? Did I get the JFETs right, or should they be turned 180 degrees?
Do I benefit from using a ground plane and connect it to "shield"? When using star ground, is there any reason to use a ground breaking resistor?
I think it's a bit weird to have 3 rail-to-rail caps in parallel at a distance from each other, should I use small resistors to "isolate" them/the channels from each other?
Top copper
Bottom copper
Left/right channels
Ground channel
It's intended for throughhole transistors and tombstoned or SMD (805) components.
I know it looks like a rats nest, but do you see any obvious errors? Did I get the JFETs right, or should they be turned 180 degrees?
Do I benefit from using a ground plane and connect it to "shield"? When using star ground, is there any reason to use a ground breaking resistor?
I think it's a bit weird to have 3 rail-to-rail caps in parallel at a distance from each other, should I use small resistors to "isolate" them/the channels from each other?
An externally hosted image should be here but it was not working when we last tested it.
Top copper
An externally hosted image should be here but it was not working when we last tested it.
Bottom copper
An externally hosted image should be here but it was not working when we last tested it.
Left/right channels
An externally hosted image should be here but it was not working when we last tested it.
Ground channel
An externally hosted image should be here but it was not working when we last tested it.
Good to see that you are progressing! Is that your first layout? Can't be. Good work!
I think you have forgotten an input filter that will remove any
high freq stuff. I see you have a 3k in series with the input SK170 - do you use that resistor and the input capacitance of the LTP as an LP-filter?
If not, I would add a cap after the 3k resistor to get an RF filter.
The -3dB frreq for the HP filter looks high - what is the -3dB freq for the HP filter?
Not sure what headphones you will be using, but did you consider a Zoebel network at the output?
Good that you put in a pot for adjusting DC-offset!
I would make that a very, very good pot.
I would add the volume pot into the simulation file, and check for any issues with the pot. I always use a 10k pot.
What value do you plan to use?
Using a GND plane is a good idea. You can even have it on both sides of the PCB.
I would use a small resistor to isolate each ch from eachother as you say.
BTW,
do you have a complete schematic which includes the PS?
Sigurd
I think you have forgotten an input filter that will remove any
high freq stuff. I see you have a 3k in series with the input SK170 - do you use that resistor and the input capacitance of the LTP as an LP-filter?
If not, I would add a cap after the 3k resistor to get an RF filter.
The -3dB frreq for the HP filter looks high - what is the -3dB freq for the HP filter?
Not sure what headphones you will be using, but did you consider a Zoebel network at the output?
Good that you put in a pot for adjusting DC-offset!
I would make that a very, very good pot.
I would add the volume pot into the simulation file, and check for any issues with the pot. I always use a 10k pot.
What value do you plan to use?
Using a GND plane is a good idea. You can even have it on both sides of the PCB.
I would use a small resistor to isolate each ch from eachother as you say.
BTW,
do you have a complete schematic which includes the PS?
Sigurd
nelsonvandal said:
Thank you very much for your constructive suggestions. I don't have any in-depth knowledge as you probably already know. I started to DIY because I wasn't happy with the amps I bought. There are absolutely no portable discrete headphone amps out there (and if there was they would be darn expensive), so I have to make it myself.Sigurd Ruschkow said:
Yes, this is actually my first layout. The other amps I've built have been breadboarded or built on PCB's made by others. I don't like Eagle (and it's too expenesive), so I've never bothered to learn it and lay out PCB's, but now I've found SprintLayout.
I was hoping the input capacitance of the JFET's would do the job as input filter, but maybe I should add some small capacitors there. I think there's room left. Otherwise I could easily parallel the RB resistors. When using Hammond 1455J1201, there's plenty of room on the bottomside of the board.
- 3dB is at 5MHz. Do you think I should use a larger cap? What - 3dB frequency would be appropriate?
I use Sennheiser HD650, 300 Ohm. I've had some trouble before when connecting phones (or even a straight cord), and I've cured it with ferrite beads. I've experimented with Zoebel networks, and it also did the job. I think I can add a Zoebel. Do you think it's for the better, and what values should I use?
I'm planning on using a 10 k Alps RK09 pot. I'll add it to LTSpice. The reason there's a power switch is that ELFA doesn't stock this pot with built in switch.
The batteries are the PSU. It's intended to be portable. I'm using a TLE2426 as railsplitter and an active ground channel instead of large caps sinking current from the phones. I'm just going to use a 24 V AC-DC wallwart. I have one with an added LM317T adjustable from 16 - 20 V.
Hi,
I would add a small filter directly at the inputs. 50 Ohm in series and 100pF to GND will be a nice filter against RF and that directly on the
phono jacks.
If there is space, I would add a Zoebel filter on the outputs. The exact values I cannot give you as that depends on the load, but you can start with 50 Ohms and 10n.
With a 10k pot and your 3k series resistance, I think you will note that maybe 3k is rather high. I use 50 or 100 Ohms but then I also use a cap to create the RF filter.
Your HP filter is OK - I misread the values. That cap will affect sonics, IMO, and you might want to use the best cap you can fit.
2.2u or even 0,22n will be fine, and you can then choose a better cap compared to using a huge 22u cap. Polystyrene or polypropylene are the types I would use.
You might want to consider matching the JFETs for IDss. It is easy to do but you do need a bunch of them to find matches.
It looks like the VR resistors do the same job as the pot, ie adjusting DC offset. Do you really need the VR resistors?
BTW,
ferrite beads is not something that I would use in the signal path.
Do post some photos of your assembled PCB.
Sigurd
I would add a small filter directly at the inputs. 50 Ohm in series and 100pF to GND will be a nice filter against RF and that directly on the
phono jacks.
If there is space, I would add a Zoebel filter on the outputs. The exact values I cannot give you as that depends on the load, but you can start with 50 Ohms and 10n.
With a 10k pot and your 3k series resistance, I think you will note that maybe 3k is rather high. I use 50 or 100 Ohms but then I also use a cap to create the RF filter.
Your HP filter is OK - I misread the values. That cap will affect sonics, IMO, and you might want to use the best cap you can fit.
2.2u or even 0,22n will be fine, and you can then choose a better cap compared to using a huge 22u cap. Polystyrene or polypropylene are the types I would use.
You might want to consider matching the JFETs for IDss. It is easy to do but you do need a bunch of them to find matches.
It looks like the VR resistors do the same job as the pot, ie adjusting DC offset. Do you really need the VR resistors?
BTW,
ferrite beads is not something that I would use in the signal path.
Do post some photos of your assembled PCB.
Sigurd
nelsonvandal said:
I've never thought of how the pot affects the circuit before. Thanks for pointing it out. I've noticed how amplifiers sound different at different volume positions, in another way than just loudness. Now I know why.
I've added Zoebels to the board, 805 SMD parts.
So you're saying that I should use one RF-filter just after the jack, before the pot, and one just before the transistor input? Did I get you right? There's room for it, I think.
I misunderstood your question about HP-filter. I was thinking LP. You mean the 22 uF cap on the input. I have it there to simulate my DAP. I use 22 uF BlackGate HiQ's instead of the original tantalums. I'm not using any input caps in this amp.
VR1/VR2 are the offset pot. Threre's no model for a pot in LTSpice.
I am going to match the JFET's. Since I'm planning on using them as current sources, I'm going to need many. As current sources, I adjust the current with resistors.
I'll post pictures of the amp later on. I'm not used to order boards, so I hope I get it right. I've only done it once before, but that board were layed out by by Nico Ras, and with gerbers and drill files all ready to roll.
I've learnt a lot today
I've added Zoebels to the board, 805 SMD parts.
So you're saying that I should use one RF-filter just after the jack, before the pot, and one just before the transistor input? Did I get you right? There's room for it, I think.
I misunderstood your question about HP-filter. I was thinking LP. You mean the 22 uF cap on the input. I have it there to simulate my DAP. I use 22 uF BlackGate HiQ's instead of the original tantalums. I'm not using any input caps in this amp.
VR1/VR2 are the offset pot. Threre's no model for a pot in LTSpice.
I am going to match the JFET's. Since I'm planning on using them as current sources, I'm going to need many. As current sources, I adjust the current with resistors.
I'll post pictures of the amp later on. I'm not used to order boards, so I hope I get it right. I've only done it once before, but that board were layed out by by Nico Ras, and with gerbers and drill files all ready to roll.
I've learnt a lot today
nelsonvandal said:
There is a model for pot in LTSwCAD, but first you have to register to http://groups.yahoo.com/ (LTSpice group) then look here for pot symbol and subcircuit model.
http://tech.groups.yahoo.com/group/LTspice/files/ Lib/Potentiometer/
Could you show as a open loop response/phase margine?
Hi,
a larger pot value will affect the circuit even more, so I am glad to read that you use a 10k pot and not a 100k pot.
In my oppinion, a small filter at the input is enough if you have a well-shielded box and no long wires from the connectors to the PCB.
Ah, good to read that you skipped the huge input cap. Do check how much DC comes from the source as you wil amplify the DC offset from the source with your amp's gain, and even though you will adjust your amp with a short circuited input (standard way of doing it) to be zero V, you might get DC at the output of your amp when the real source is concted to your input.
If you have a VR-pot which is to null DC offset, may I ask why there is a pot on the input stage, too?
Sigurd
PS I think you will learn a lot getting a PCB in your hand from the PCB manufacturer with the input files from the layout program you use. It will be fun!
a larger pot value will affect the circuit even more, so I am glad to read that you use a 10k pot and not a 100k pot.
In my oppinion, a small filter at the input is enough if you have a well-shielded box and no long wires from the connectors to the PCB.
Ah, good to read that you skipped the huge input cap. Do check how much DC comes from the source as you wil amplify the DC offset from the source with your amp's gain, and even though you will adjust your amp with a short circuited input (standard way of doing it) to be zero V, you might get DC at the output of your amp when the real source is concted to your input.
If you have a VR-pot which is to null DC offset, may I ask why there is a pot on the input stage, too?
Sigurd
PS I think you will learn a lot getting a PCB in your hand from the PCB manufacturer with the input files from the layout program you use. It will be fun!
nelsonvandal said:I've never thought of how the pot affects the circuit before. Thanks for pointing it out. I've noticed how amplifiers sound different at different volume positions, in another way than just loudness. Now I know why.
I've added Zoebels to the board, 805 SMD parts.
So you're saying that I should use one RF-filter just after the jack, before the pot, and one just before the transistor input? Did I get you right? There's room for it, I think.
I misunderstood your question about HP-filter. I was thinking LP. You mean the 22 uF cap on the input. I have it there to simulate my DAP. I use 22 uF BlackGate HiQ's instead of the original tantalums. I'm not using any input caps in this amp.
VR1/VR2 are the offset pot. Threre's no model for a pot in LTSpice.
I am going to match the JFET's. Since I'm planning on using them as current sources, I'm going to need many. As current sources, I adjust the current with resistors.
I'll post pictures of the amp later on. I'm not used to order boards, so I hope I get it right. I've only done it once before, but that board were layed out by by Nico Ras, and with gerbers and drill files all ready to roll.
I've learnt a lot today
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.