Jocko's Circuit
Found it!
Its a Sziklai input folded cascode design. Or two times common base. It is drawn somewhat awkward due to that stupid schematic capture program or my computer blindness.
I did. Even posted kind of merger of Jocko's and Jan Diddens and Rudolf Broertjes circuit though I never built it. Will see if I can find that back and post again saving you to use that lousy searchengine.Charles Hansen said:
Found it!
Its a Sziklai input folded cascode design. Or two times common base. It is drawn somewhat awkward due to that stupid schematic capture program or my computer blindness.
Attachments
Re: Jocko's Circuit
My estimate is just that -- an estimate. There is nothing that substitutes for a real-world measurement, and even more important, real-world listening tests. However in this case, I would recommend using a square wave rather than a sine wave. This will more closely approximate the transients that may be present with music. Also, it is important to make sure that your measuring device have sufficient bandwidth to see what is going on. Please let me know what you find out.
I'm a little unclear from your post if this is the circuit you listened to. If so, I would expect it to sound quite a bit better than any op-amp I-V stage. I'm not so fond of the Sziklai pair, as in my experience, this local feedback loop will color the sound. However, Jocko has found that the lowered input impedance here will increase bass "slam" and authority, presumably because of the lowered input impedance presented to the DAC output.
The rest of this circuit is zero feedback, which is something of a double-edged sword. This makes things much more transparent, but also much more revealing of the actual implementation. In other words, if you use (for example) a bad sounding capacitor with a feedback circuit, the coloration produced by the feedback tends to mask the problems of the capacitor. But in a zero feedback circuit, you will much more clearly what the capacitor is doing to the signal.
Elso Kwak said:
My estimate is just that -- an estimate. There is nothing that substitutes for a real-world measurement, and even more important, real-world listening tests. However in this case, I would recommend using a square wave rather than a sine wave. This will more closely approximate the transients that may be present with music. Also, it is important to make sure that your measuring device have sufficient bandwidth to see what is going on. Please let me know what you find out.
Elso Kwak said:
I'm a little unclear from your post if this is the circuit you listened to. If so, I would expect it to sound quite a bit better than any op-amp I-V stage. I'm not so fond of the Sziklai pair, as in my experience, this local feedback loop will color the sound. However, Jocko has found that the lowered input impedance here will increase bass "slam" and authority, presumably because of the lowered input impedance presented to the DAC output.
The rest of this circuit is zero feedback, which is something of a double-edged sword. This makes things much more transparent, but also much more revealing of the actual implementation. In other words, if you use (for example) a bad sounding capacitor with a feedback circuit, the coloration produced by the feedback tends to mask the problems of the capacitor. But in a zero feedback circuit, you will much more clearly what the capacitor is doing to the signal.
Elso:
Your implementation of the CFP is not quite right......
You have to be very careful to keep it from oscillating. Lower the gain some.......for starters.
As for transparency:
Yes, it must have a clean supply. And guess what......Charlie and I tend to use open loop supplies.
Or something similar....that has constant Z vs frequency. I am firmly convinced that a large part of my sucess is a result of the total implementaion, not just coming up with a common-base, or common-source circuit.
Jocko
Your implementation of the CFP is not quite right......
You have to be very careful to keep it from oscillating. Lower the gain some.......for starters.
As for transparency:
Yes, it must have a clean supply. And guess what......Charlie and I tend to use open loop supplies.
Or something similar....that has constant Z vs frequency. I am firmly convinced that a large part of my sucess is a result of the total implementaion, not just coming up with a common-base, or common-source circuit.
Jocko
Charles Hansen said:
This is why you have to include a capacitor in parallel with the feedback resistor to roll of the frequency response of the I/V converter. If you choose a sensible -3dB point of 50kHz then there will be 100 times attenuation at 5MHz and the voltage at the inverting input drops to 4mV. Now if you choose the AD825 (as recomended by Walt Jung in the article you pointed to) it has a gain of 15 at 5MHz and so this voltage is reduced further to 1.3mV.
I ran it through LTSpice and it confirms this, here is the netlist:
I1 in 0 SINE(0 1.2m 5Meg) AC 1.2m
X§U1 0 in N002 N001 out 1pole Avol=3k GBW=40Meg Slew=125Meg ilimit=25m rail=0 Vos=0 en=0 enk=0 in=0 ink=0
V2 N002 0 15
V3 0 N001 15
R2 out in 1.66k
R3 0 out 10k
C1 out in 2000p
.tran 2m
;ac oct 10 100 100Meg
.lib 1pole.sub
.backanno
.end
Can you tell it is a slow day at work
I couldn't resist some plots...
Top:
OPA627: Avol=10Meg GBW=16Meg Slew=55Meg
Bottom:
AD825: Avol=4k GBW=40Meg Slew=125Meg
V(in) is inverting input, V(out) is output of opamp. Zf is 1.66kohm || 2nF.
Notice how, in the audio-band, the inverting input is 180 degrees out of phase with the current when using AD825 while it is 90 degrees out of phase when using OPA627. My gut feeling tells me this difference affects the DAC performance in some way. Inverting input voltage @5MHz is ~-50dBV = 3mV.
Top:
OPA627: Avol=10Meg GBW=16Meg Slew=55Meg
Bottom:
AD825: Avol=4k GBW=40Meg Slew=125Meg
V(in) is inverting input, V(out) is output of opamp. Zf is 1.66kohm || 2nF.
Notice how, in the audio-band, the inverting input is 180 degrees out of phase with the current when using AD825 while it is 90 degrees out of phase when using OPA627. My gut feeling tells me this difference affects the DAC performance in some way. Inverting input voltage @5MHz is ~-50dBV = 3mV.
Attachments
Ojg,
Just checked this with MicroCap. Model provided by ADI says at 5MHz the inverting input (closed loop; output impedance of the DAC set to 2k and I/V resistor is 1k - doubled value of the I/V resistor will double the inverting input’s impedance) has impedance of about 150 Ohm. Both AC and transient analysis show this. Since the open loop response of the model looks a bit shorter (11MHz) than shown in the datasheet (25MHz), and the phase shifts earlier, I would expect lower value in reality, maybe about half this value, but not less than that.
Btw, you are misquoting Socrates.
Pedja
Just checked this with MicroCap. Model provided by ADI says at 5MHz the inverting input (closed loop; output impedance of the DAC set to 2k and I/V resistor is 1k - doubled value of the I/V resistor will double the inverting input’s impedance) has impedance of about 150 Ohm. Both AC and transient analysis show this. Since the open loop response of the model looks a bit shorter (11MHz) than shown in the datasheet (25MHz), and the phase shifts earlier, I would expect lower value in reality, maybe about half this value, but not less than that.
Btw, you are misquoting Socrates.
Pedja
Marlowe said:
Maybe I'm tired, but wouldn't you have to swap the inputs?
That is, making the inverting input high impedance and the
non-inverting one low-impedance, which of course you cannot
do with a chip unless you design it.
Edit: Hm, maybe not if you use it in invertin config?? Guess it
is to late to think about it now.
Charles Hansen said:
Interesting article, thanks for the link. In other news, has anyone looked at the AD9631? Open loop bandwidth is about 100kHz, so it's not an integrator at the frequencies of interest as mentioned by Jung. The thing does appear to be a bit finicky with respect to supplies, however, as PSRR is only about 65 to 75 dB. It also seems to be happiest with impedances below 1kohm.
Cheers,
Francois.
1. Operational amplifiers can be used as transimpedance amps.
2. Any amplifier circuit ultimately is an "integrator". Unless, of
course, you have some amp with infinite bandwidth. It
just depends on DC open loop gain and response corner.
3. Why make broad sweeping generalizations? Op-amp are
not inherently better or worse than discrete circuits. It depends
on the op-amp, just like it depends on the discrete circuit.
2. Any amplifier circuit ultimately is an "integrator". Unless, of
course, you have some amp with infinite bandwidth. It
just depends on DC open loop gain and response corner.
3. Why make broad sweeping generalizations? Op-amp are
not inherently better or worse than discrete circuits. It depends
on the op-amp, just like it depends on the discrete circuit.
The '846 was the genesis for the transimpedance amp that I came up with. The dealers did not like the sound of current feedback designs. Not just mine, but other manufacturers as well.
Yes, interesting part, but it somehow never lived up to its expectations.
Can't remember the company name.....out of Fairfield where all the TM goof-balls are.......wrote a "white paper" suggesting that CF was not good for I/V, as the settling times are too slow. Which resulted in their ability to only resolve 12-14 bits. Something like that. (They liked the '5534!)
Their claim, not mine.
Jocko
Yes, interesting part, but it somehow never lived up to its expectations.
Can't remember the company name.....out of Fairfield where all the TM goof-balls are.......wrote a "white paper" suggesting that CF was not good for I/V, as the settling times are too slow. Which resulted in their ability to only resolve 12-14 bits. Something like that. (They liked the '5534!)
Their claim, not mine.
Jocko
I just wanted to tell that (else than the fact he has not wrote anything, so in that regard it is impossible to literally quote him), the guy was not agnostic. Actually, he was at the side of achievability of the truth. The point of his lore (related to the sentence in your signature) is that the knowledge is not given by itself nor contained in the common wisdom, but yet has to be achieved; this real knowledge starts by the discovery of the fact the truth is not given directly (plainly) but the effort is needed to come across it – as long as you do not start to (re)investigate, you know nothing.ojg said:Didn't mean to disrespect Socrates, maybe another of his qoutes are more fitting:
I know nothing except the fact of my ignorance - Socrates
Written as it was, it looks like the motto of the agnostic which Socrates certainly has not been. Even worse, it can be understood like the play with the words and from that same ancient time there are sophists that are known both for being prone to such things and for being on Socrates/Plato’s top lists of the bad guys.
Pedja
Re: Re: Jocko's Circuit
This is the circuit I listened to with the NON-OS TDA1543.
I changed it a bit for the TDA1543: replacing D1 by two red LED's in series and R3 =15k. Also a 10µF bypass cap to ground at the base of the transistor, BC550C. R8 is omitted.
I got sound, though pretty low in volume and some hum. I do not like this sound. The timbre of the musical instruments is not right. And the sound gets on my nerves. I know this circuit is very crude. What shall I do, adding a constant current source or build my first circuit without the Sziklai?
BTW with a squarewave and the OPA604 I got the same amplitude (10mV) band of signals on my scope. Could not trigger it well.
Charles,Charles Hansen said:
This is the circuit I listened to with the NON-OS TDA1543.
I changed it a bit for the TDA1543: replacing D1 by two red LED's in series and R3 =15k. Also a 10µF bypass cap to ground at the base of the transistor, BC550C. R8 is omitted.
I got sound, though pretty low in volume and some hum. I do not like this sound. The timbre of the musical instruments is not right. And the sound gets on my nerves. I know this circuit is very crude. What shall I do, adding a constant current source or build my first circuit without the Sziklai?
BTW with a squarewave and the OPA604 I got the same amplitude (10mV) band of signals on my scope. Could not trigger it well.
Attachments
Re: Re: Re: Jocko's Circuit
It sounds to me like that you are probably not providing the DAC with the proper conditions for proper operation. The data sheet for the TDA1543 holds the clues to your problems.
First of all when you are using the op-amp I-V, it will hold the audio outputs at Vref that is also supplied by the DAC. In this case it is 2.2 VDC +/-0.1 VDC. Your circuit (as described) will have the audio outputs held at around 3.2 VDC (depending on the characteristics of the LEDs used). The DC compliance is stated as between 1.8 and 3.8 VDC (assuming a 5 V supply), so this may be all right but maybe not.
The AC compliance is restricted to 25 mV, which means that the input impedance of the I-V circuit must be less than 10 ohms for proper operation. For the simple circuit you have drawn the input impedance should be around 2.5 ohms (assuming ~10 mA bias), which should be adequate. However I would double-check the actual bias current, and also measure the actual signal level at the DAC pin.
A Sziklai pair will help in this regard, but I would probably keep the single transistor and bump the bias current up to 25 mA to get the input impedance down to 1 ohm or so. (You may or may not need a different transistor to do this safely -- I'm not familiar with the European devices.)
Next, the PSRR of your circuit is dismal -- zero! Any hum or noise on the rail is directly coupled into the audio signal.
Finally, all the little details will matter. What brand of capacitors, what kind of PS regulators, et cetera, et cetera. In the end however, I think you will find the journey well worth it.
Good luck!
Elso Kwak said:
It sounds to me like that you are probably not providing the DAC with the proper conditions for proper operation. The data sheet for the TDA1543 holds the clues to your problems.
First of all when you are using the op-amp I-V, it will hold the audio outputs at Vref that is also supplied by the DAC. In this case it is 2.2 VDC +/-0.1 VDC. Your circuit (as described) will have the audio outputs held at around 3.2 VDC (depending on the characteristics of the LEDs used). The DC compliance is stated as between 1.8 and 3.8 VDC (assuming a 5 V supply), so this may be all right but maybe not.
The AC compliance is restricted to 25 mV, which means that the input impedance of the I-V circuit must be less than 10 ohms for proper operation. For the simple circuit you have drawn the input impedance should be around 2.5 ohms (assuming ~10 mA bias), which should be adequate. However I would double-check the actual bias current, and also measure the actual signal level at the DAC pin.
A Sziklai pair will help in this regard, but I would probably keep the single transistor and bump the bias current up to 25 mA to get the input impedance down to 1 ohm or so. (You may or may not need a different transistor to do this safely -- I'm not familiar with the European devices.)
Next, the PSRR of your circuit is dismal -- zero! Any hum or noise on the rail is directly coupled into the audio signal.
Finally, all the little details will matter. What brand of capacitors, what kind of PS regulators, et cetera, et cetera. In the end however, I think you will find the journey well worth it.
Good luck!