Hi
I am quite new to electronics and have built few amps before (not my own design). Now I am trying to design my own preamplifier but there are so many ways to create one that I haven't gotten that far.
I am using my cellphone and computer as music source most of the time but I also listen to vinyl records so the preamplifier will need to have quite a lot of voltage gain. I will later build a riaa amplifier so no need to worry about that yet. I want to use discrete transistors because I want to learn precisely how amplifiers work and also be able to be in control of more things.
So far I have gotten best results with 2-stage bjt amplifier where the input stage is common emitter which amplify the voltage and the second stage is common collector aka buffer amplifier which lowers the output impedance. Then I also added a feedback which I adjust to set the output voltage. I got enough voltage gain and THD values of about 0.1% on line level 750mV.
I have also tried cascade in the first stage and common collector on the second stage. Cascade made good frequency response but it had too much voltage gain and I didn't manage to make working feedback to it. The THD was also around >1% on line level which is way worse than the 2 stage bjt amp that I tried. The cascade was made with fet transistors.
So my question is which gain configurations/stages should I use to get lower THD values and also as flat frequency response as possible. You can also just name some different gain stages / transistor configurations which could work as an amplifier. Thank you for your responses!
I am quite new to electronics and have built few amps before (not my own design). Now I am trying to design my own preamplifier but there are so many ways to create one that I haven't gotten that far.
I am using my cellphone and computer as music source most of the time but I also listen to vinyl records so the preamplifier will need to have quite a lot of voltage gain. I will later build a riaa amplifier so no need to worry about that yet. I want to use discrete transistors because I want to learn precisely how amplifiers work and also be able to be in control of more things.
So far I have gotten best results with 2-stage bjt amplifier where the input stage is common emitter which amplify the voltage and the second stage is common collector aka buffer amplifier which lowers the output impedance. Then I also added a feedback which I adjust to set the output voltage. I got enough voltage gain and THD values of about 0.1% on line level 750mV.
I have also tried cascade in the first stage and common collector on the second stage. Cascade made good frequency response but it had too much voltage gain and I didn't manage to make working feedback to it. The THD was also around >1% on line level which is way worse than the 2 stage bjt amp that I tried. The cascade was made with fet transistors.
So my question is which gain configurations/stages should I use to get lower THD values and also as flat frequency response as possible. You can also just name some different gain stages / transistor configurations which could work as an amplifier. Thank you for your responses!
Yes, these are the kind of circuits I was talking about.
Here is the cascode preamp.
And this is the bjt 2-stage ce,cc amplifier which made better results than the cascade one.
Do you have any suggestions of this kind of gain configurations?
Thank you for your reply. I quickly looked through the page and it looked very helpful.
Last edited by a moderator:
Even a common modern op-amp will outperform a simple discrete circuit but if you want to use transistors, I suggest you look at legacy circuits like the Sinclair preamp.
http://diy.torrens.org/Sinclair/P60/issue_1/p06.jpg
Bare in mind that older circuits used NPN transistors exclusively because that was the best performance and cost-effective available devices at that time.
DIYA has posted some ~JFET RIAA preamps but they are also a bit crude. If you need a phono preamp, you will require RIAA equalization, which means a RC network and lots of gain.
You seem to be referencing very simple textbook "common emitter" etc circuits. Real circuits are rarely that simple. A real discrete pre-amp circuit would be something like the LTP (long-tailed-pair) + VAS (voltage-amplifier-stage) stages of an op-amp or VFA (voltage-feedback-amplifier) power amplifier. Simple circuits like the Sinclair RIAA preamp suffer performance limitations because the RIAA equalization network is a significant load at high frequencies for the relatively high impedance output.
There are some simple passive tone control circuits like those in guitar amplifiers, but they require (noisy) gain to compensate for losses, so you are better off using the standard Baxandall tone control topology. Some purists choose to not use any tone controls.
https://www.wiringview.co/4558-ic-audio-equalizer-circuit-diagrams/
The Sinclair preamp is also a Baxandall tone circuit using a single transistor instead of an op-amp.
Aside from RIAA preamp and tone controls, a preamp is just a selector switch or mixer that combines several inputs. In the past, a "tape monitor In-out" switch and jacks allowed the tape recorder to record other inputs and monitor the tape output.
http://diy.torrens.org/Sinclair/P60/issue_1/p06.jpg
Bare in mind that older circuits used NPN transistors exclusively because that was the best performance and cost-effective available devices at that time.
DIYA has posted some ~JFET RIAA preamps but they are also a bit crude. If you need a phono preamp, you will require RIAA equalization, which means a RC network and lots of gain.
You seem to be referencing very simple textbook "common emitter" etc circuits. Real circuits are rarely that simple. A real discrete pre-amp circuit would be something like the LTP (long-tailed-pair) + VAS (voltage-amplifier-stage) stages of an op-amp or VFA (voltage-feedback-amplifier) power amplifier. Simple circuits like the Sinclair RIAA preamp suffer performance limitations because the RIAA equalization network is a significant load at high frequencies for the relatively high impedance output.
There are some simple passive tone control circuits like those in guitar amplifiers, but they require (noisy) gain to compensate for losses, so you are better off using the standard Baxandall tone control topology. Some purists choose to not use any tone controls.
https://www.wiringview.co/4558-ic-audio-equalizer-circuit-diagrams/
The Sinclair preamp is also a Baxandall tone circuit using a single transistor instead of an op-amp.
Aside from RIAA preamp and tone controls, a preamp is just a selector switch or mixer that combines several inputs. In the past, a "tape monitor In-out" switch and jacks allowed the tape recorder to record other inputs and monitor the tape output.
Thanks for the very specific answer.
I have to take op-amps in serious consideration. And I also need to do some more research on fet-transistors since they have many advantages compared to the old NPN-transistors.
I have been experimenting and doing math calculations on those "textbook circuits" because because I think it's the best way to get started. I havent't studied electronics in university yet so I have studied everything by myself from internet.
I will look more into LTP and VAS stages and also analyze the sinclair preamp you included. Then I try to come up with something my own.
Equalization is also something I have been thinking about, whether I should use it or not. I'm using old Bower and Wilkins bookself speakers which has 6.5inch woofer so I might need to emphasize the low frequencies. If I decide to use eq then I'm going to use the baxandall topology.
I have to take op-amps in serious consideration. And I also need to do some more research on fet-transistors since they have many advantages compared to the old NPN-transistors.
I have been experimenting and doing math calculations on those "textbook circuits" because because I think it's the best way to get started. I havent't studied electronics in university yet so I have studied everything by myself from internet.
I will look more into LTP and VAS stages and also analyze the sinclair preamp you included. Then I try to come up with something my own.
Equalization is also something I have been thinking about, whether I should use it or not. I'm using old Bower and Wilkins bookself speakers which has 6.5inch woofer so I might need to emphasize the low frequencies. If I decide to use eq then I'm going to use the baxandall topology.
At the Delft university, they used to have a course called Structured Electronic Design. The title was broader than the contents, as it was actually about structured design of negative-feedback amplifiers. It only covered amplifiers that work in class A, like preamplifiers normally do. It was basically an extension of the Ph.D. thesis of Ernst Nordholt, Design of high-performance negative-feedback amplifiers.
They split up the design process in a couple of steps. If I remember well, it was something like this:
1. Analyse whether the input and output impedances should ideally be very low, very high or some well-defined value that's neither very low nor very high. This depends on the properties of the signal source and the load: with what impedance does it work best? For example, for a line level audio preamplifier, you would normally choose a high input and low output impedance, for a moving-magnet RIAA preamplifier a 47 kohm input impedance and low output impedance. Also decide what gain you want.
2. Choose a feedback topology that matches with the requirements from step 1.
3. Design the input stage. If noise matters, that boils down to choosing a device and bias point that gives minimum noise for the given source impedance. Choose a bias point, but don't worry about how to implement the biasing yet.
4. Design the output stage. It has to be able to handle the voltage and current swings that occur at maximum drive and it is nice if it has large voltage, current, transimpedance and transadmittance gains, so the previous stages don't have to handle large signals. That usually means picking a common emitter or common source stage, or sometimes a long-tailed pair if you need series feedback at the output (which you probably don't in an audio preamplifier). Again, pick the device and the bias point, but don't design the biasing circuit yet.
5. Add intermediate stages to increase loop gain, if needed. Again, common emitter stages and long-tailed pairs are preferred because of their high gain.
6. Analyse the feedback loop stability and add frequency compensation. Add common gate stages (cascodes) between the stages if needed to get rid of Miller effect. For frequency compensation, phantom zeros are usually preferred because they don't reduce loop gain. Examples of phantom zero compensation are lead compensation by reducing the attenuation of the feedback network above a certain frequency, or connecting a resistor in series with a capacitive load when you have an output with shunt feedback, or connecting a capacitor in parallel with a resistive source when you have an input with series feedback. The next preferred methods are those that above a certain frequency, reduce the overall loop gain by local feedback. For example, Miller compensation or pole-zero compensation by local feedback.
7. Design the biasing. Connect two ideal current sources and two ideal voltage sources to each transistor to bias them, and then start shifting the sources so you can combine them into a limited number of sources at convenient locations. Come up with circuit implementations of the bias sources. Add DC bias loops or common-mode bias loops if needed to make the circuit's biasing less sensitive to small deviations of the bias currents or voltages.
I think the main thing I learned from it is that there are more than two or three ways to hook up a bunch of transistors to make an amplifier.
They split up the design process in a couple of steps. If I remember well, it was something like this:
1. Analyse whether the input and output impedances should ideally be very low, very high or some well-defined value that's neither very low nor very high. This depends on the properties of the signal source and the load: with what impedance does it work best? For example, for a line level audio preamplifier, you would normally choose a high input and low output impedance, for a moving-magnet RIAA preamplifier a 47 kohm input impedance and low output impedance. Also decide what gain you want.
2. Choose a feedback topology that matches with the requirements from step 1.
3. Design the input stage. If noise matters, that boils down to choosing a device and bias point that gives minimum noise for the given source impedance. Choose a bias point, but don't worry about how to implement the biasing yet.
4. Design the output stage. It has to be able to handle the voltage and current swings that occur at maximum drive and it is nice if it has large voltage, current, transimpedance and transadmittance gains, so the previous stages don't have to handle large signals. That usually means picking a common emitter or common source stage, or sometimes a long-tailed pair if you need series feedback at the output (which you probably don't in an audio preamplifier). Again, pick the device and the bias point, but don't design the biasing circuit yet.
5. Add intermediate stages to increase loop gain, if needed. Again, common emitter stages and long-tailed pairs are preferred because of their high gain.
6. Analyse the feedback loop stability and add frequency compensation. Add common gate stages (cascodes) between the stages if needed to get rid of Miller effect. For frequency compensation, phantom zeros are usually preferred because they don't reduce loop gain. Examples of phantom zero compensation are lead compensation by reducing the attenuation of the feedback network above a certain frequency, or connecting a resistor in series with a capacitive load when you have an output with shunt feedback, or connecting a capacitor in parallel with a resistive source when you have an input with series feedback. The next preferred methods are those that above a certain frequency, reduce the overall loop gain by local feedback. For example, Miller compensation or pole-zero compensation by local feedback.
7. Design the biasing. Connect two ideal current sources and two ideal voltage sources to each transistor to bias them, and then start shifting the sources so you can combine them into a limited number of sources at convenient locations. Come up with circuit implementations of the bias sources. Add DC bias loops or common-mode bias loops if needed to make the circuit's biasing less sensitive to small deviations of the bias currents or voltages.
I think the main thing I learned from it is that there are more than two or three ways to hook up a bunch of transistors to make an amplifier.
Thank you for the very comprehensive reply!
Cool that they had a course for structural design of audio preamplifiers in Delft university. I googled the Ernst Nordholt's thesis
The Design of High-Performance Negative-Feedback Amplifiers", it's a shame that it is sold only a paperback book otherwise I would have read it.
I have thought to set the input impedance somewhere in 10-30k, because going higher will rise the noise floor and going lower will load the source and result in poor bass. I have mm-cartridge on my Kenwood record player, so I have to consider increasing the input impedance to 47k as you said. Output impedance will be designed as low as possible.
I have learned now that I need to take a deep look into LTP and feedback methods immediately.
This step by step designing method was very helpful and I think this will get me started quite well. There are indeed enormous amounts of ways to make those few transistors work as amplifier. Thanks for sharing this method and information.
Cool that they had a course for structural design of audio preamplifiers in Delft university. I googled the Ernst Nordholt's thesis
The Design of High-Performance Negative-Feedback Amplifiers", it's a shame that it is sold only a paperback book otherwise I would have read it.
I have thought to set the input impedance somewhere in 10-30k, because going higher will rise the noise floor and going lower will load the source and result in poor bass. I have mm-cartridge on my Kenwood record player, so I have to consider increasing the input impedance to 47k as you said. Output impedance will be designed as low as possible.
I have learned now that I need to take a deep look into LTP and feedback methods immediately.
This step by step designing method was very helpful and I think this will get me started quite well. There are indeed enormous amounts of ways to make those few transistors work as amplifier. Thanks for sharing this method and information.
Last edited by a moderator:
By the way, there is another book about the same subject, Structured Electronic Design: Negative-feedback amplifiers by Chris J.M. Verhoeven, Arie van Staveren, G.L.E. Monna, M.H.L. Kouwenhoven, E. Yildiz. I never read it because I find it much too expensive, but knowing the first four authors, it must be pretty good.
Yep I found that book very expensive too but looking the preview it looked very professional.
One good free book is Douglas Self Small signal audio design which you can find a pdf from the internet. The book is very broad and profound, it is almost 600 pages and pdf so I takes a while to read it through.
One good free book is Douglas Self Small signal audio design which you can find a pdf from the internet. The book is very broad and profound, it is almost 600 pages and pdf so I takes a while to read it through.
Anton Montagne is also continuing the work by Northold on Structured Electronic Design. He's got his own site and he wrote book expanding on what CJM Verhoeven wrote back in 2004. You can find the book (for free) here:
https://analog-electronics.tudelft.nl/
I feel like Anton's book is more complete, but it's lacking an example design like Chris does it in the last chapter of his book.
https://analog-electronics.tudelft.nl/
I feel like Anton's book is more complete, but it's lacking an example design like Chris does it in the last chapter of his book.
Thanks for sharing this book to me. I looked through it and it had a lots of mathematics which is great since self Douglas's book doesn't contain that much of it.
Anton's books approach is bit more theoretical than Douglas's which includes more practical examples. However it's also important to learn theory which makes Anton's book a great pair with Douglas's.
Anton's books approach is bit more theoretical than Douglas's which includes more practical examples. However it's also important to learn theory which makes Anton's book a great pair with Douglas's.
You're welcome.
I'd argue that there's a something really important provided by Anton's book (which is also provided in Ernst's old thesis, but perhaps better presented in Anton's) but not so much in Chris's book. I have also not seen it in any other book on electronics, be it "theoretical" or more cookbook-styled (and mind you, I know lots).
It is the fact that if you understand the 4 basic configurations of feedback with a nullor, you'll be able to also "synthesize" your own configuration, in case, you want, for instance, an inversion on the signal.
If you concentrate on that part, for instance, you'll be able to derive a non-inverting mixer with virtual ground. Once you get those configuration down, "exotic" configurations like that one will follow logically.
I even made a quick implementation with 2 ip-amps of it in ltspice. I have never seen it anywhere, but you'll see that coming up with such configurations will be easy, no need to look it up.
You'll see that there are ways to do non-inverting versions of transconductance or trasimpedance amplifiers, as well as inverted versions of the voltage and current amplifiers.
You'll also see that it is possible to design a double loop amplifier with a well defined input or output impedance (can't be done with one single loop, unless you do brute force, which is pretty much what everyone on the internet shows).
Anyway, enjoy!
I'd argue that there's a something really important provided by Anton's book (which is also provided in Ernst's old thesis, but perhaps better presented in Anton's) but not so much in Chris's book. I have also not seen it in any other book on electronics, be it "theoretical" or more cookbook-styled (and mind you, I know lots).
It is the fact that if you understand the 4 basic configurations of feedback with a nullor, you'll be able to also "synthesize" your own configuration, in case, you want, for instance, an inversion on the signal.
If you concentrate on that part, for instance, you'll be able to derive a non-inverting mixer with virtual ground. Once you get those configuration down, "exotic" configurations like that one will follow logically.
I even made a quick implementation with 2 ip-amps of it in ltspice. I have never seen it anywhere, but you'll see that coming up with such configurations will be easy, no need to look it up.
You'll see that there are ways to do non-inverting versions of transconductance or trasimpedance amplifiers, as well as inverted versions of the voltage and current amplifiers.
You'll also see that it is possible to design a double loop amplifier with a well defined input or output impedance (can't be done with one single loop, unless you do brute force, which is pretty much what everyone on the internet shows).
Anyway, enjoy!
Please do NOT quote the entire post just above yours. This is made clear in the RULES. Help keep the forum clean.Posts have been cleaned up.
@aprendiz007
Now I have a lot of educational material to study, which gives me a great headstart to university. I have looked through the 4 different feedback topologies but deeper understanding demands more time. Lately I have been studying Long tail pairs which I will use as the input stage of my preamplifier because LTP's offer common mode rejection and is somewhat easy to include a feedback network to.
Now I have a lot of educational material to study, which gives me a great headstart to university. I have looked through the 4 different feedback topologies but deeper understanding demands more time. Lately I have been studying Long tail pairs which I will use as the input stage of my preamplifier because LTP's offer common mode rejection and is somewhat easy to include a feedback network to.