I knew what you meant
So you approve? When I have an idea that is different from common practice I wonder who is out of step - me or everyone else!
Yes. I plan to use the power of the 3rd dimension too!
A screened power module underneath the board and power leads fed up perpendicular. Then the power to the output transistors is feed sideways.
Thanks for the confirmation. I know that sometimes a poor practice will become common when it is simply copied without much consideration, but it is a bit of a worry to think "how come no one else does it my way?"
so I appreciate your reply.
Best wishes
David
Yes, I have seen this frontend before.
Your question was, have I ever seen this front end before. Yes I have.
Examples of where I have seen it are:
David Hafler used it in both their preamps and FET output power amps.(DH-100)
Audio Research used it in their Analog Modules in the D100 series and SP-5 preamp.
The year was about 1979 thru 1985.
Harmon Kardon used it in their power amps. Nikko made some preamps with this topology, only using J-Fets. Nakamichi used it in their power amps.
I have seen this used in opamps, the old Harris Semiconductor HA1-5195-5.
Burr Brown uses half on one in their OPA-660, current feedback opamps.
SAE used them in their equipment, back in the 1970s and early 80s.
There is an old driver board in my junk box from an SAE amp that has a 2SA798 and a 2SC1583 dual transistor pacs on it. Repaired this amp in 1981.
A comment about this is that it looks good on paper, and functions good electrically. The sound is very exact, very definitive,
but equipment with this front end sounds like cardboard. No coloration of the
music at all. My thoughs are that coloration is necessary to make an amp different(better than) the rest. It is the good distortion that sets equipment sound apart. If this complimentary symetry front end is used, it is my opinion that it has to be modified to produce the coloration that sounds, well, like tube equipment does.
Mike
Your question was, have I ever seen this front end before. Yes I have.
Examples of where I have seen it are:
David Hafler used it in both their preamps and FET output power amps.(DH-100)
Audio Research used it in their Analog Modules in the D100 series and SP-5 preamp.
The year was about 1979 thru 1985.
Harmon Kardon used it in their power amps. Nikko made some preamps with this topology, only using J-Fets. Nakamichi used it in their power amps.
I have seen this used in opamps, the old Harris Semiconductor HA1-5195-5.
Burr Brown uses half on one in their OPA-660, current feedback opamps.
SAE used them in their equipment, back in the 1970s and early 80s.
There is an old driver board in my junk box from an SAE amp that has a 2SA798 and a 2SC1583 dual transistor pacs on it. Repaired this amp in 1981.
A comment about this is that it looks good on paper, and functions good electrically. The sound is very exact, very definitive,
but equipment with this front end sounds like cardboard. No coloration of the
music at all. My thoughs are that coloration is necessary to make an amp different(better than) the rest. It is the good distortion that sets equipment sound apart. If this complimentary symetry front end is used, it is my opinion that it has to be modified to produce the coloration that sounds, well, like tube equipment does.
Mike
are you sure - I can't find a DH-100 power amp - the preamp uses 5534
the DH-500 does have complementary bjt diff pair with resistor collector load and complementary "VAS" - no cascodes, no current mirrors, is driven from only one side of diff pairs
are you sure you understand all of the features that together constitute what Edmond has presented as his possibly novel circuit?
I don't think he claims all, or any of the features individually are novel – it is the whole, how they work together that make the innovation
the DH-500 does have complementary bjt diff pair with resistor collector load and complementary "VAS" - no cascodes, no current mirrors, is driven from only one side of diff pairs
are you sure you understand all of the features that together constitute what Edmond has presented as his possibly novel circuit?
I don't think he claims all, or any of the features individually are novel – it is the whole, how they work together that make the innovation
Last edited:
Is it safe to assume that the 5532 used vertical pnp's in their circuitry, which did not sound the best from my memory. How do you compare the use of the IC op amps to the discrete circuits here?
I have measured 0.003% at 90 W, 20 kHz test signal 80 kHz bandwidth. This was the measurable limit of my equipment. Fairly standard three-stage topology with emitter-follower output and two-pole compensation. To get near 1 ppm you really need something more "exotic", even in simulation before any of the layout headaches hit.
Last edited:
Oh, Really?
hard to stand that a few active elements amp souds better than ~30 transistors?
Hi Harry,
I fully agree with that! BTW, Halcro does it the same way (though with a piece of solid coax instead of a twisted pair). Also the PGP amp uses a piece of coax to route the output signal back to the inverting input (after I urged syn08 to do so).
@ padamiecki, indeed, hard to stand.
Cheers,
E.
Depends what you mean by "fancy". The front end (input stage and VAS/TIS) was mostly surface-mount, apart from VAS/TIS current load transistor, VAS/TIS cascode, Vbe multiplier transistor, and large capacitors. This meant the total footprint of the front end could be minimised, with particular attention paid to try and keep high-impedance areas as small as possible.
Power was brought in on two twisted pairs, one with + and ground, the other with - and ground; I no longer think this is the best way to do it.
If only cardboard did this! Would make designing an amplifier a lot easier!
Surely the best thing to do is have your amplifier as colourless as possible; you then add pre-distortion (linear and/or non-linear) with DSP on the front end. You can then pick colouration suitable to your mood/album rather than have it permanently hard-wired into the amplifier circuitry. Want your amp to sound like your favourite "tube" amp? No problem! Want it to sound like a Bryston/Naim/Audiolab/Roksan/Rotel/Primare/Plinius etc. etc. etc.? Also, no problem! Just select the appropriate DSP algorithm.
perpendicular onto the PCB.
Hi Andrew,
Due to a design flaw *(my fault
) the wiring was extremely critical. Normally, I don't think such drastic measures are necessary.
* R32 and R33 (see: http://home.tiscali.nl/audio/FrontEndSch.html) are tied to the power ground, which appeared to be wrong. They should connected to the signal ground instead.
Cheers,
E.
PS: I'm still busy with DiAna.
Hi Andrew,
Due to a design flaw *(my fault
) the wiring was extremely critical. Normally, I don't think such drastic measures are necessary.* R32 and R33 (see: http://home.tiscali.nl/audio/FrontEndSch.html) are tied to the power ground, which appeared to be wrong. They should connected to the signal ground instead.
Cheers,
E.
PS: I'm still busy with DiAna.
Hi Andrew,
No, you can't let them float, as they determine the voltage gain of the second (cfb) stage.
Please see: http://home.tiscali.nl/audio/front-end.html, 3rd figure from top, R5 and R6. (this is an oversimplified schematic of course). I hope this web page will sufficiently clarify how the PGP front end works.
Cheers,
E.
No, you can't let them float, as they determine the voltage gain of the second (cfb) stage.
Please see: http://home.tiscali.nl/audio/front-end.html, 3rd figure from top, R5 and R6. (this is an oversimplified schematic of course). I hope this web page will sufficiently clarify how the PGP front end works.
Cheers,
E.
DSP sound
LOL
I'm not sure if above scheme will work. If one knows the tube sound is generated by a DSP, I bet those tube freaks will start complaining about a DSP sound.
Cheers,
E.
LOL
I'm not sure if above scheme will work. If one knows the tube sound is generated by a DSP, I bet those tube freaks will start complaining about a DSP sound.
Cheers,
E.
Thoughts on layout
As I mentioned earlier, I'm not convinced by the layout of the e-amp. I hope folk find the critique below useful in general terms rather than just a criticism of this otherwise excellent amplifier. I would like to take this opportunity to once again re-iterate my gratitude to Andrew for publishing details of the e-amp in such depth; if you haven't read his write-up already I encourage you to do so; it gave me plenty of food for thought.
As I understand it, the e-amp layout is as below:
If the wiring is indeed like this, this creates some issues as I've alluded to earlier. In the images below, current pathways for +ve current are highlighted red. Current pathways for -ve currents are similar and not shown. For example, consider the current path for low-frequency currents that will be supplied by the reservoir caps on the PSU board:
Due to the large area enclosed by this loop, inductance is high; but more importantly out and return paths being far apart means the loop will radiate easily.
Now, consider the current pathway for the +ve decoupling capacitor:
Again, the inductance of this loop is large. If you compare this current path with that for the pathway from the main PSU reservoir cap, you will see that enclosed loop area is quite similar. High-frequency impedance of the two loops will therefore be quite similar, defeating the purpose of decoupling caps - to provide a localised, low-impedance path for higher frequency currents.
Next, observe the current path for the Zobel:
Again, large loop area implies high inductance so at high-frequencies this pathway looks inductive rather than resistive as it should.
Finally, with regards to the order in which ground connections are teed in the PSU consider the section highlighted in the image below:
This segment carries current from the decoupling capacitors. This current multiplied by the resistance of this connection generates an error voltage for the so-called "clean ground" (that used for the front-end of the amplifier and presumably as the feedback reference) relative to the reference terminal (ground connection) of the loudspeaker output. The length of this segment is very short, so the error voltage will be very small but may be significant in amplifiers aiming at 1 ppm distortion and below.
This segment does carry the +ve/-ve decoupling currents once they have "mixed" so with a linear load, the current in the segment will be purely sinusoidal for a purely sinusoidal output voltage - i.e. there will be no "non-linear" voltage generated in the segment resistance. However, loudspeakers are non-linear loads that draw non-sinusoidal currents when a sinusoidal voltage is applied to their terminals*. So in real-world usage, a non-linear distortion voltage will be generated in this segment of ground wiring. Non-linear current drawn by loudspeakers is also relevant in all the talk about trying to make supply-wiring radiation "linear" and is a consideration that I don't recall reading anywhere before; for real-world usage it is therefore important to arrange the supply wiring to minimise radiation from it rather than simply attempting to make the radiation linear.
An improved layout is shown below:
*This may precipitate the thought: what's the best way to drive a loudspeaker to minimise acoustic output non-linear distortion? Should voltage or current be the controlled variable, i.e. should a speaker be driven by a voltage source or current source? Of course this is a whole other kettle of monkeys but let's just say here: For minimum non-linear distortion it should be driven by a current source, but this creates very significant linear distortion (i.e. non-flat frequency response) as the impedance varies across the audio spectrum. When driving a speaker with a current source, motional feedback from the cone is therefore required to deliver a flat frequency response.
As I mentioned earlier, I'm not convinced by the layout of the e-amp. I hope folk find the critique below useful in general terms rather than just a criticism of this otherwise excellent amplifier. I would like to take this opportunity to once again re-iterate my gratitude to Andrew for publishing details of the e-amp in such depth; if you haven't read his write-up already I encourage you to do so; it gave me plenty of food for thought.
As I understand it, the e-amp layout is as below:
An externally hosted image should be here but it was not working when we last tested it.
If the wiring is indeed like this, this creates some issues as I've alluded to earlier. In the images below, current pathways for +ve current are highlighted red. Current pathways for -ve currents are similar and not shown. For example, consider the current path for low-frequency currents that will be supplied by the reservoir caps on the PSU board:
An externally hosted image should be here but it was not working when we last tested it.
Due to the large area enclosed by this loop, inductance is high; but more importantly out and return paths being far apart means the loop will radiate easily.
Now, consider the current pathway for the +ve decoupling capacitor:
An externally hosted image should be here but it was not working when we last tested it.
Again, the inductance of this loop is large. If you compare this current path with that for the pathway from the main PSU reservoir cap, you will see that enclosed loop area is quite similar. High-frequency impedance of the two loops will therefore be quite similar, defeating the purpose of decoupling caps - to provide a localised, low-impedance path for higher frequency currents.
Next, observe the current path for the Zobel:
An externally hosted image should be here but it was not working when we last tested it.
Again, large loop area implies high inductance so at high-frequencies this pathway looks inductive rather than resistive as it should.
Finally, with regards to the order in which ground connections are teed in the PSU consider the section highlighted in the image below:
An externally hosted image should be here but it was not working when we last tested it.
This segment carries current from the decoupling capacitors. This current multiplied by the resistance of this connection generates an error voltage for the so-called "clean ground" (that used for the front-end of the amplifier and presumably as the feedback reference) relative to the reference terminal (ground connection) of the loudspeaker output. The length of this segment is very short, so the error voltage will be very small but may be significant in amplifiers aiming at 1 ppm distortion and below.
This segment does carry the +ve/-ve decoupling currents once they have "mixed" so with a linear load, the current in the segment will be purely sinusoidal for a purely sinusoidal output voltage - i.e. there will be no "non-linear" voltage generated in the segment resistance. However, loudspeakers are non-linear loads that draw non-sinusoidal currents when a sinusoidal voltage is applied to their terminals*. So in real-world usage, a non-linear distortion voltage will be generated in this segment of ground wiring. Non-linear current drawn by loudspeakers is also relevant in all the talk about trying to make supply-wiring radiation "linear" and is a consideration that I don't recall reading anywhere before; for real-world usage it is therefore important to arrange the supply wiring to minimise radiation from it rather than simply attempting to make the radiation linear.
An improved layout is shown below:
An externally hosted image should be here but it was not working when we last tested it.
*This may precipitate the thought: what's the best way to drive a loudspeaker to minimise acoustic output non-linear distortion? Should voltage or current be the controlled variable, i.e. should a speaker be driven by a voltage source or current source? Of course this is a whole other kettle of monkeys but let's just say here: For minimum non-linear distortion it should be driven by a current source, but this creates very significant linear distortion (i.e. non-flat frequency response) as the impedance varies across the audio spectrum. When driving a speaker with a current source, motional feedback from the cone is therefore required to deliver a flat frequency response.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.