Mighty designers, do you think input stage devices have greater impact on noise than PSRR and grounding and shielding?
I think that everything is important, but many designs tend to have more power supply rejection than others. This can be important as well.
My initial topology, used in the JC-1 preamp most probably has a noise level of less than 3nV/rt Hz with lowish value feedback resistor. Unfortunately I use a fairly high value feedback resistor 3.83K, and that coupled with the 7.15K resistor in series with the positive input will put me over 12nV/rt Hz.
Rescaling these resistors would reduce the total noise, but would: Lower input impedance, decrease common mode rejection in balanced, and increase static distortion.
It might be worth a try, not that anyone has complained yet, except for Bob Cordell, up to this time.
My initial topology, used in the JC-1 preamp most probably has a noise level of less than 3nV/rt Hz with lowish value feedback resistor. Unfortunately I use a fairly high value feedback resistor 3.83K, and that coupled with the 7.15K resistor in series with the positive input will put me over 12nV/rt Hz.
Rescaling these resistors would reduce the total noise, but would: Lower input impedance, decrease common mode rejection in balanced, and increase static distortion.
It might be worth a try, not that anyone has complained yet, except for Bob Cordell, up to this time.
john curl said:I think that everything is important, but many designs tend to have more power supply rejection than others. This can be important as well.
My initial topology, used in the JC-1 preamp most probably has a noise level of less than 3nV/rt Hz with lowish value feedback resistor. Unfortunately I use a fairly high value feedback resistor 3.83K, and that coupled with the 7.15K resistor in series with the positive input will put me over 12nV/rt Hz.
Rescaling these resistors would reduce the total noise, but would: Lower input impedance, decrease common mode rejection in balanced, and increase static distortion.
It might be worth a try, not that anyone has complained yet, except for Bob Cordell, up to this time.
This post of yours would have been OK had you not put in that last sentence.
Stop whining, John. I did no more than compare my design, which you baseslessly criticized for noise, with yours. Mine won. Big time. By 20 dB. More than can be explained by the resistors you are talking about. Do the math. Do the SPICE. Do your homework before you throw out red herrings.
If you are such a "noise expert", why were you unable to see that my circuit in fact produces a very good noise performance?
Bob
Because, Bob, it has an active load on the input stage. I simply asked how much compromise the active load added to the overall noise. This is important for others, who might want to use your input stage for other situations, like a phono stage. You told me .1dB. I can accept this. However, IF you had to make a real amplifier for the real world, you would see that your design tradeoffs would not necessarily be appropriate. For example, how would you balance drive your amp? I might recommend using a transformer, like Steve Eddy. It would keep your noise to a minimum, but if you did what I had to do, you would be even worse off than me, I'm afraid.
This long-term peeing contest is filling up threads with utter BS about wich one invented wheel first time and was it 2643 B.C. or 2634 B.C. and if its a wheel or not if its not round.G.Kleinschmidt said:
Now, nowI think the long-standing tension and public posturing between you two guys
🙄
I was hoping to teach some of you on this thread a potential problem with active loads. I made a serious mistake with an active load once, without first understanding it. I thought it might be a good thing to pass this 'wisdom' on to you. However, I stand corrected.
Mr Curl,
Active loads like current mirror as load, could it be made better by your suggestion, making R>100/150ohm, and not less, for RE(emitor degeneration)?
I've been thinking alot about this phrase :
The signals usually enter a transistor via its base. Between base and collector of a transistor, there's Cbc. Between emitor and base there's Cbe. Usually one of the leg, whether is it collector or emitor should be connected to +or- supply rail (common collector or common emitor). This will cause problem if the signal is originally referenced to ground, not to rail. The ripple between rail and ground will enter the signal path.
So, supply ripple will ramp into signal path (base) via Cbc and Cbe. It makes sense the less transistor used, the less this PS garbage enters the signal path (base). The more complicated CCT (more transistors being used), the more injections of PS garbage into signal path. I've experimented this PSgarbage/PSRR, they seems very little on magnitude (on scopes), not too important, but their impact on perceived sound is huge.
About active loading, if it uses transistor then it will have Cbe or Cbc.
If one day, there's a transistor with 0 Cbe or Cbc, will this make a better active load?
Active loads like current mirror as load, could it be made better by your suggestion, making R>100/150ohm, and not less, for RE(emitor degeneration)?
I've been thinking alot about this phrase :
Besides the non-linearity of the transistor's magnification factor itself, I think there's another important factor to consider.All else being equal, the simpler the cct, the better the sound
The signals usually enter a transistor via its base. Between base and collector of a transistor, there's Cbc. Between emitor and base there's Cbe. Usually one of the leg, whether is it collector or emitor should be connected to +or- supply rail (common collector or common emitor). This will cause problem if the signal is originally referenced to ground, not to rail. The ripple between rail and ground will enter the signal path.
So, supply ripple will ramp into signal path (base) via Cbc and Cbe. It makes sense the less transistor used, the less this PS garbage enters the signal path (base). The more complicated CCT (more transistors being used), the more injections of PS garbage into signal path. I've experimented this PSgarbage/PSRR, they seems very little on magnitude (on scopes), not too important, but their impact on perceived sound is huge.
About active loading, if it uses transistor then it will have Cbe or Cbc.
If one day, there's a transistor with 0 Cbe or Cbc, will this make a better active load?
john curl said:I was hoping to teach some of you on this thread a potential problem with active loads. I made a serious mistake with an active load once, without first understanding it. I thought it might be a good thing to pass this 'wisdom' on to you. However, I stand corrected.
John, there's nothing wrong with trying to teach people about a potential problem with active loads. After all, teaching is a big part of what this thread is all about.
If you and I did not have a disagreement about the relative merits of the full complementary differential input architecture, this whole useful and educational thread would not exist. There is value in disagreement and discussion to sort it out.
However, just throwing out a red herring without explanation is not teaching. Just asserting, without explanation, that "it can be shown that my toplogy is quieter than yours" is not teaching. Whining about me showing real data that involves my amplifier and your amplifier is not teaching.
Teaching takes work, like the work I did yesterday to actually go and simulate three different versions of my topology. I welcome your questions because they force us to think and come up with answers, and sometimes in the process we end up teaching ourselves as well as others.
Heck, I learned something out of this because I had not previously looked hard at the noise of my architecture. I built it, it measured great on noise, and that was the end of it.
If you want to send me a schematic of the JC-1 input/VAS circuit, I'll be glad to simulate it for noise, with and without the extra circuitry to make it have a balanced input (caveat - I'm not sure how good the MOSFET models are in regard to noise behavior).
Similarly, if you want to send me a partial schematic of your balanced input section, I'll try to bolt it onto my architecture and simulate that for noise. That is how we can all learn something more.
Bob
Hi, Mr. Curl,
Putting cap multiplier for front stage (input differential + VAS), is this helping alot regarding noise performance of the whole power amp?
Putting cap multiplier for front stage (input differential + VAS), is this helping alot regarding noise performance of the whole power amp?
Bob, thank you for your offer. Actually, Dimitri and I have discussed this in some detail and we know how to improve the JC-1's signal to noise by adding an extra fet buffer. It has already been worked out, even without SPICE since the problem is so obvious.
I am glad that you have analyzed your input stage with SPICE analysis because this is almost the only way to be sure about the noise contribution of this kind of active load. At least that is what I found in the analysis of the 741 done by UC Berkeley back in 1970. I still don't know if individual noise contributions are brought out in personal computer versions of SPICE, of if you can only get the total noise of the circuit.
Luman, the power supply buffer removes any hum and transient based sag in the power supply. It is not a very low noise source on a steady state basis.
I am glad that you have analyzed your input stage with SPICE analysis because this is almost the only way to be sure about the noise contribution of this kind of active load. At least that is what I found in the analysis of the 741 done by UC Berkeley back in 1970. I still don't know if individual noise contributions are brought out in personal computer versions of SPICE, of if you can only get the total noise of the circuit.
Luman, the power supply buffer removes any hum and transient based sag in the power supply. It is not a very low noise source on a steady state basis.
john curl said:Bob, thank you for your offer. Actually, Dimitri and I have discussed this in some detail and we know how to improve the JC-1's signal to noise by adding an extra fet buffer. It has already been worked out, even without SPICE since the problem is so obvious.
I am glad that you have analyzed your input stage with SPICE analysis because this is almost the only way to be sure about the noise contribution of this kind of active load. At least that is what I found in the analysis of the 741 done by UC Berkeley back in 1970. I still don't know if individual noise contributions are brought out in personal computer versions of SPICE, of if you can only get the total noise of the circuit.
Luman, the power supply buffer removes any hum and transient based sag in the power supply. It is not a very low noise source on a steady state basis.
Thanks, John.
Actually, I was wondering the same thing about LTspice - namely whether it will spit out the noise contributions of individual components. I have to admit, I have not looked real hard at the manual for that. It is, indeed important to know the individual contributors in order to get the noise down in a design.
Cheers,
Bob
Bob Cordell said:
Actually, I was wondering the same thing about LTspice - namely whether it will spit out the noise contributions of individual components.
After running a noise simulation, select the schematic pane and point at the component of interest, a message appears at the bottom "left click to plot <noise quantity(ref des)>, the total output referred spectral noise density from <ref des>.", left click and the quantity will be plotted. The same is available from the pull-down menu when the schematic pane is selected, View/Visible Traces, along with other plots such as gain.
The problem with LTSpice noise simulation seems to be the unreliable active device modeling, noise model parameters were not set to match datasheet curves or do not model noise at all.
nuvistor said:
After running a noise simulation, select the schematic pane and point at the component of interest, a message appears at the bottom "left click to plot <noise quantity(ref des)>, the total output referred spectral noise density from <ref des>.", left click and the quantity will be plotted. The same is available from the pull-down menu when the schematic pane is selected, View/Visible Traces, along with other plots such as gain.
The problem with LTSpice noise simulation seems to be the unreliable active device modeling, noise model parameters were not set to match datasheet curves or do not model noise at all.
Thanks for the tip!
Best,
Bob
Balanced inputs
Hi John,
Speaking of balanced inputs for power amplifiers, have you tried the InGenious 1200-series balanced input line receiver IC from That Corporation? I haven't tried it, and I realize you are not a big fan of IC's, but it appears to have very good distortion specifications.
Cheers,
Bob
Hi John,
Speaking of balanced inputs for power amplifiers, have you tried the InGenious 1200-series balanced input line receiver IC from That Corporation? I haven't tried it, and I realize you are not a big fan of IC's, but it appears to have very good distortion specifications.
Cheers,
Bob
Re: Balanced inputs
Ironically, the InGenius, whose primary purpose is to mimic the behavior of input transformers by providing high common mode rejection even in the presence of slight source impedance imbalances and is intended as an alternative to input transformers, was designed by Bill Whitlock, President of... Jensen Transformers. 😀
And I could see where some would prefer a small, $3 chip to a large $70 transformer. But then the InGenius can't provide ground isolation so it ain't THAT smart. 😀
se
Bob Cordell said:Speaking of balanced inputs for power amplifiers, have you tried the InGenious 1200-series balanced input line receiver IC from That Corporation? I haven't tried it, and I realize you are not a big fan of IC's, but it appears to have very good distortion specifications.
Ironically, the InGenius, whose primary purpose is to mimic the behavior of input transformers by providing high common mode rejection even in the presence of slight source impedance imbalances and is intended as an alternative to input transformers, was designed by Bill Whitlock, President of... Jensen Transformers. 😀
And I could see where some would prefer a small, $3 chip to a large $70 transformer. But then the InGenius can't provide ground isolation so it ain't THAT smart. 😀
se
Steve,
1) what is your experience with HF and VHF signal throughput in signal transformers - have you measured?
2) what is your experience with voltage induced from stronger LF magnetic fields into signal transformers?
1) what is your experience with HF and VHF signal throughput in signal transformers - have you measured?
2) what is your experience with voltage induced from stronger LF magnetic fields into signal transformers?
PMA said:1) what is your experience with HF and VHF signal throughput in signal transformers - have you measured?
The input and microphone transformers I've used (Jensen JT-11P-1, CineMag CMLI-15/15B, CineMag CMMI-10C) have all had high frequency extension (-3dB) out to 80-100kHz. This is not only shown in their respective datasheets, but I've also verified it myself.
2) what is your experience with voltage induced from stronger LF magnetic fields into signal transformers?
Haven't had any problems with it as long as I wasn't plunking them down on top of a power transformer. They all come in MuMetal cans, but even the "nude" 15/15Bs I've used haven't been problematic.
se
Steve,
ad 1) in fact I meant frequencies in MHz range and higher, as our environment is full of them. I found these frequency components to be increased by signal transformers (antenna effect, interwinding cap, (HF resonance?))
ad 2) this probably depends on core design, I found problems with tansformer isolation box placed on top of power amplifier with toroidal like 300VA or more.
ad 1) in fact I meant frequencies in MHz range and higher, as our environment is full of them. I found these frequency components to be increased by signal transformers (antenna effect, interwinding cap, (HF resonance?))
ad 2) this probably depends on core design, I found problems with tansformer isolation box placed on top of power amplifier with toroidal like 300VA or more.
PMA said:ad 1) in fact I meant frequencies in MHz range and higher, as our environment is full of them.
Ah. Well you said "signal" throughput which I took to mean well... signal, which is what we want to get through, as opposed to something other than signal, such as noise and/or interference.
I found these frequency components to be increased by signal transformers (antenna effect, interwinding cap, (HF resonance?))
Good quality input transformers have a Faraday shield between the primary and secondary windings to prevent such high frequencies from coupling between primary and secondary via interwinding capacitance.
Now, some input transformers are bifilar wound in which case most of the high frequency bandwidth (which can extend into the MHz region) is due to the capacitive coupling between primary and secondary.
ad 2) this probably depends on core design, I found problems with tansformer isolation box placed on top of power amplifier with toroidal like 300VA or more.
Then don't put your isolation box on top of the power amplifier. 😀
se
Bob, why don't you modify your amplifier to do balanced, as well as single ended, depending on the choice and need of the amp's owner? Then we will have a common point of discussion.
Actually, Steve Eddy is right, and I would prefer a transformer to an IC chip, since I once designed out the best IC chip that I could find about 15 years ago, when it failed critical listening tests.
(Sorry, Scott, but the 712 just didn't cut it! It was not your fault, and AD even sent Walt Jung after me because of it. I am now using the AD797 in a new phono design, but the noise at 10KHz worries me a little.)
Thank you, Bob for reminding me of THAT Corp. They make interesting and quality products. We always say hello at conferences.
Actually, Steve Eddy is right, and I would prefer a transformer to an IC chip, since I once designed out the best IC chip that I could find about 15 years ago, when it failed critical listening tests.
(Sorry, Scott, but the 712 just didn't cut it! It was not your fault, and AD even sent Walt Jung after me because of it. I am now using the AD797 in a new phono design, but the noise at 10KHz worries me a little.)
Thank you, Bob for reminding me of THAT Corp. They make interesting and quality products. We always say hello at conferences.
- Status
- Not open for further replies.
- Home
- Amplifiers
- Solid State
- Unipolar vs complementary input stage