I used OPA637 in one, and OPA134 in the other, + BD139/140.
The transistors did not get warm, so they can maybe be biased higher? (Or I have wasted a pair beautiful heatsinks? 😉 )
Arne K
The transistors did not get warm, so they can maybe be biased higher? (Or I have wasted a pair beautiful heatsinks? 😉 )
Arne K
You can increase the current but then you must use the Vbe-transistor also. You can skip the LED's.
On the term "Diamond" for this topology
Hi Per-Anders, Thread-Readers,
You've mentioned on your site, that you are unclear about the origin of the term "diamond buffer".
I was under the impression to have read a source for this years ago, but trying to find it now, wasn't that successfull.
The oldest article found by Google would be the Burr Brown OPA660 app note:
http://focus.ti.com/lit/an/sboa071/sboa071.pdf
Note that it's called "diamond transistor" and not "diamond buffer" there. Has some logic, as the circuit clearly has a base an emitter and sort-of collector.
Also it is presented as an "ideal transistor", but I don't think that's the origin of "diamond".
Further down it states the circuit is "called called 'Diamond structure' in laboratory jargon".
My take on this is, that the circuit can be nicely drawn as octagon, see:
http://www.double-diamond.de/scratch/dmnd-pretty.pdf
And the octagon (or mitered square) is often used as symbol for "diamond".
Regards,
Peter Jacobi
Hi Per-Anders, Thread-Readers,
You've mentioned on your site, that you are unclear about the origin of the term "diamond buffer".
I was under the impression to have read a source for this years ago, but trying to find it now, wasn't that successfull.
The oldest article found by Google would be the Burr Brown OPA660 app note:
http://focus.ti.com/lit/an/sboa071/sboa071.pdf
Note that it's called "diamond transistor" and not "diamond buffer" there. Has some logic, as the circuit clearly has a base an emitter and sort-of collector.
Also it is presented as an "ideal transistor", but I don't think that's the origin of "diamond".
Further down it states the circuit is "called called 'Diamond structure' in laboratory jargon".
My take on this is, that the circuit can be nicely drawn as octagon, see:
http://www.double-diamond.de/scratch/dmnd-pretty.pdf
And the octagon (or mitered square) is often used as symbol for "diamond".
Regards,
Peter Jacobi
Arne, from what I see from the pictures you don't need caps for fixing the Dc gain. OPA134 is good and OPA637 is very good in terms of offset voltage. I suggest you just solder in a wire (short the caps) under the pcb. You can leave the caps where they are if you can't desloder them easily.
Re: On the term "Diamond" for this topology
Anyone who really knows, Mr D maybe?
Your idea about the name was newpjacobi said:
Anyone who really knows, Mr D maybe?
Re: On the term "Diamond" for this topology
Pedja
Diamond buffer and diamond transistor are not exactly the same things. The "transistor" includes the "collector" and the buffer doesn't. The article you linked has info about both of them. P-A's output stage is the buffer (and the first stage is the transistor).pjacobi said:
Pedja
Tinkering around with the QRV06
Hi peranders,
Thank you for posting the LTSpice .asc file (in the "True Current Feedback..." thread), so that I was able to easily play with it.
I've switched back to this thread, as it seems more fitting.
As said, as a first test I substituted the current source for the input diamond with a simple 11k resistor, with no noticable effect on distortion. Both version give 0.0068% THD from the".fourier 20k V(vout)" command.
Of course the resistor version is lacking some power supply rejection - but only some. The symmetrical 100Hz ripple in both rails is still rejected fine.
Then I started to re-distribute the "saved" transistors. To make a long story short, here is the pj-fied version:
PDF: http://www.linearaudio.de/scratch/qrv06r0sim-pj-2.pdf
ASC: http://www.linearaudio.de/scratch/qrv06r0sim-pj-2.asc
I've put the output diamond into class A, for which I tripled the output stage and used the BCPs throughout. This will now have an idle power dissipation of 3W, each BCP dissipating 300mW. The maximum current available in class A is 120mA.
The current mirror is replaced with one of the designs discussed in the thread "A special current mirror". The Cdom is replaced with a much larger 1nF for stability, putting f(-3dB) to 1.6MHz. This could be pushed higher again.
The voltage source for the current mirror cascode isn't specified - a blue LED would fit the voltage and give a nice look.
The 20kHz THD is now 0.00064%, a nice factor ten improvement.
Another reduction to 0.00037% was possible, by substituting all BCPs with high VAF devices. I simulated with KSC3503/KSA1381 (models available from Fairchild). I don't know what possible SMD types there are for this task.
I assume reaching this low level in simulation only means that real life results are dominated by all sorts of practical problems, not accounted for in simulation. But I nevertheless like the tinkering with the simulator.
Regards,
Peter Jacobi
Hi peranders,
Thank you for posting the LTSpice .asc file (in the "True Current Feedback..." thread), so that I was able to easily play with it.
I've switched back to this thread, as it seems more fitting.
As said, as a first test I substituted the current source for the input diamond with a simple 11k resistor, with no noticable effect on distortion. Both version give 0.0068% THD from the".fourier 20k V(vout)" command.
Of course the resistor version is lacking some power supply rejection - but only some. The symmetrical 100Hz ripple in both rails is still rejected fine.
Then I started to re-distribute the "saved" transistors. To make a long story short, here is the pj-fied version:
PDF: http://www.linearaudio.de/scratch/qrv06r0sim-pj-2.pdf
ASC: http://www.linearaudio.de/scratch/qrv06r0sim-pj-2.asc
I've put the output diamond into class A, for which I tripled the output stage and used the BCPs throughout. This will now have an idle power dissipation of 3W, each BCP dissipating 300mW. The maximum current available in class A is 120mA.
The current mirror is replaced with one of the designs discussed in the thread "A special current mirror". The Cdom is replaced with a much larger 1nF for stability, putting f(-3dB) to 1.6MHz. This could be pushed higher again.
The voltage source for the current mirror cascode isn't specified - a blue LED would fit the voltage and give a nice look.
The 20kHz THD is now 0.00064%, a nice factor ten improvement.
Another reduction to 0.00037% was possible, by substituting all BCPs with high VAF devices. I simulated with KSC3503/KSA1381 (models available from Fairchild). I don't know what possible SMD types there are for this task.
I assume reaching this low level in simulation only means that real life results are dominated by all sorts of practical problems, not accounted for in simulation. But I nevertheless like the tinkering with the simulator.
Regards,
Peter Jacobi
Re: Tinkering around with the QRV06
I haven't simulated P-As amplifier, but I have earlier simulated
a stand alone Diamond buffer (essentially identical to the one
presented by Jung except for some swap of transistors, and
maybe bias current). I just tried replacing the CCSs with
resistors, taking care to get the same currents as in the CCS
case. I get a 20dB difference in PSRR for the negative supply
and 26dB for the positive supply. I would say that is more
than "some" difference. One difference, however, was that
the resistors gave a flat PSRR response up to around 1MHz,
while the BJT CCSs started to lose rejection at or slightly
above 100kHz. Of course it will vary from case to case, depend
on the choice of transistors etc. etc., and it is after all just
simulations. I justed wanted to show that the mileage can
vary here. I haven't tried to redo any distorsion measurements
with resistors, though. I might try that at some occasion.
It should be noted , perhaps, that I measured the AC response
at the buffer output when applying AC stimuli to the rails. I
did not measure the current variations in the CCSs/resistors,
since that is of less interest.
pjacobi said:
I haven't simulated P-As amplifier, but I have earlier simulated
a stand alone Diamond buffer (essentially identical to the one
presented by Jung except for some swap of transistors, and
maybe bias current). I just tried replacing the CCSs with
resistors, taking care to get the same currents as in the CCS
case. I get a 20dB difference in PSRR for the negative supply
and 26dB for the positive supply. I would say that is more
than "some" difference. One difference, however, was that
the resistors gave a flat PSRR response up to around 1MHz,
while the BJT CCSs started to lose rejection at or slightly
above 100kHz. Of course it will vary from case to case, depend
on the choice of transistors etc. etc., and it is after all just
simulations. I justed wanted to show that the mileage can
vary here. I haven't tried to redo any distorsion measurements
with resistors, though. I might try that at some occasion.
It should be noted , perhaps, that I measured the AC response
at the buffer output when applying AC stimuli to the rails. I
did not measure the current variations in the CCSs/resistors,
since that is of less interest.
Re: Tinkering around with the QRV06
I have never used or tested the "compound" transistor . Maybe I should test it?
So far I can make the conclusion that my amp has lower dist than 0.008% which is quite good.
Nice improvements you have done but what do the mean in real life? I don't know. The dist figures is a little bit too fantastic I think.pjacobi said:
I have never used or tested the "compound" transistor . Maybe I should test it?
So far I can make the conclusion that my amp has lower dist than 0.008% which is quite good.
PSRR of the diamond buffer, CCS vs. resistor simulations
The difference I see is not so huge, 10dB (58 vs. 68dB) for the positive and 6-7dB (57 vs. 63.5dB) for the negative rail. And yes, I also saw that the CCS can lose its steam somewhat earlier but the absolute values I see are even after that better in the case of the CCS. Also, some current sources will keep the maximum PSRR as long as the resistors will do. It is interesting the fact the change of the type of the used current source does not make almost any difference in the term of the nominal (maximum) PSRR.
If someone is interested I uploaded some pics and schematics of two diamond buffers I tried in the GC. (The comment above applies to the complementary diamond buffer.)
http://www.users.verat.net/~pedjarogic/audio/gainclone/stuff.htm
Pedja
The difference I see is not so huge, 10dB (58 vs. 68dB) for the positive and 6-7dB (57 vs. 63.5dB) for the negative rail. And yes, I also saw that the CCS can lose its steam somewhat earlier but the absolute values I see are even after that better in the case of the CCS. Also, some current sources will keep the maximum PSRR as long as the resistors will do. It is interesting the fact the change of the type of the used current source does not make almost any difference in the term of the nominal (maximum) PSRR.
If someone is interested I uploaded some pics and schematics of two diamond buffers I tried in the GC. (The comment above applies to the complementary diamond buffer.)
http://www.users.verat.net/~pedjarogic/audio/gainclone/stuff.htm
Pedja
Christer,
CFP's used instead of Q2, Q4 will make the parameters even better. THD values under 0.001% are quite easily achievable in the real world. And class A operation of the output stage is a great benefit 😉 .
Pavel
CFP's used instead of Q2, Q4 will make the parameters even better. THD values under 0.001% are quite easily achievable in the real world. And class A operation of the output stage is a great benefit 😉 .
Pavel
JFETs are not good idea. The circuit is based on Vbe cancellation between input and output transistors (Q1/Q2, Q3/Q4).
Yes, but you would not have to worry about offset problems due to the base current of the input pair.
Regards,
Jam
Regards,
Jam
The base current is quite low (Ic1/h21). The resulting output DC offset is determined by the difference in Vbe voltages of Q1-Q4. The circuit is pretty stable. I know what I speak about, as I use the circuit based on this topology as an output stage of my class A power amplifiers.
Well, how about posting the schematic of your amplifier?
Regards,
Jam
P.S. I have built a similar circuit and found the offset drift too high for direct coupled preamp applications. I suppose a higher beta device or darlingtons for the input pair might help. I had to use a servo (nasty) to direct couple it. In a power amp application the offset should be small enough not to worry about.
Regards,
Jam
P.S. I have built a similar circuit and found the offset drift too high for direct coupled preamp applications. I suppose a higher beta device or darlingtons for the input pair might help. I had to use a servo (nasty) to direct couple it. In a power amp application the offset should be small enough not to worry about.
- Status
- Not open for further replies.