Trasformerless (class A) DOA mic preamp - design discussion

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Aleguitarpro

Well-known member
Joined
Dec 28, 2011
Messages
179
Location
ITALY
Hi.

I'm mainly a musician and my interest in building/studying designs is to know what fits best my needs.

I tried a lot of preamps recording in studios, thanks to friends, building my own etc.

I'm a classical guitar player and I record for friends and colleagues classical music and acoustic instruments mainly.

I tried following preamps/design:

NEVE 1073 style preamps (a lot of models/clones...)
API 312 /my own clones and originals
THAT based preamps (my build)

CHANDLER (germanium, LTD1, TG2)

Avalon (737, m5, 2022)

Millennia HV35, GML 8304.

The latest (AVALON m5/2022 and Millennia) are the most detailed and I like a lot my THAT1512 based preamp.
But I'd like to try to make my own class A preamp based on a DOA that I like a lot, APP992.
I tested it in my own Sontec clone and then tried in my preamp leave me very pleased.

Design goals are skip transformers and use if it's possible APP992 DOA class A everywhere.

Attached you'll find a preliminary schematic.

Most difficult part to find from standard suppliers are Q1 and Q2.
For them candidates was: 2sc2547 (I used AMEK 9098 and I liked them a lot), MAT02 (difficult to find...mostly from china), 2n4403, 2sk389 (going into FET-land).
I put BC550 in my schematic because they are very easy to find.
I'm wondering if there's something I could consider in DIP format.
I found these in SMD that could be good: 2SC2713, 2SC4117, KSC1845.

MPSA18 was second candidates in this preliminary schematic.

If is there a way to make something good/better without 3 class A DOA I'm very open to other options.
I need very low distortion, fast transients, no colour at all.

Thanks to those who will be kind enough to participate and contribute their wisdom!
I could share in exchange  some of my guitar secrets, but I don't know anyone is interested in this kind of stuff here!  ;D

 

Attachments

  • DOA trasfless.jpg
    619.1 KB · Views: 197
Aleguitarpro said:
If is there a way to make something good/better without 3 class A DOA I'm very open to other options.
I need very low distortion, fast transients, no colour at all.
You can use Szlikai pairs (sometimes improperly called complementary Darlington") at the input.
Check paragraph 7 in https://sound-au.com/articles/cmpd-vs-darl.htm
And you should add CMRR adjustment.
 
abbey road d enfer said:
You can use Szlikai pairs (sometimes improperly called complementary Darlington") at the input.
Check paragraph 7 in https://sound-au.com/articles/cmpd-vs-darl.htm
And you should add CMRR adjustment.

So I need 8 transistors at the input?
 
I expect the THAT preamp is quite good.

The schematic you shared looks like what is called "Cohen" topology. It can be quite good I have made many preamps using  it or variants, over the last few decades. BTW you can flip it around to accommodate PNP or NPN input devices. The Cohen's noise performance at high gain will be dominated by the input devices, at very low gain the op amps used could make some difference in the noise floor. The low noise input devices I used decades ago  (<1 nV/rt Hz) are now obsolete. Most manufacturers these days just use an IC solution like THAT or TI.

I am not a fan of that specific DC servo topology. First I prefer that the input to the servo has the RC configured as a passive LPF, that way it is impossible to slew rate limit the servo amp. Secondly the servo output feeding the preamp output differential at unity gain means the full noise of the servo amp will appear in the output with unity gain. The servo amp has more than enough DC gain to correct the DC levels using a larger value resistor to reduce the noise contribution, of course the other side of the differential needs to be similarly balanced.

I have used the self bias through the feedback resistors (R20 and R21) technique in a fixed gain MC phono preamp back in the 80s. In a mic preamp you may forfeit a little input voltage swing at low gain.

JR
 
Aleguitarpro said:
Hi.

I'm mainly a musician and my interest in building/studying designs is to know what fits best my needs.

I tried a lot of preamps recording in studios, thanks to friends, building my own etc.

I'm a classical guitar player and I record for friends and colleagues classical music and acoustic instruments mainly.

...

I need very low distortion, fast transients, no colour at all.

That circuit is called a Current Feedback Instrumentation Amp (CFIA) and it is described in detail starting on page 44 of this document:

  http://www.thatcorp.com/datashts/AES129_Designing_Mic_Preamps.pdf

It's generally accepted as having the best overall noise / distortion performance of any mic circuit around and as such it's found in all sorts of gear. Pretty much all of the basic mixers you'll find at the music store use it. Mackie for example call it "ONYX".

So for transparent and clean, the whole mic topo debate is pretty much "solved" w/ CFIA. And, once you start to get down to that level of noise and THD, making something discrete just isn't going to matter. It becomes just a fun exercise. Although there is some psychoacoustic merit to using something that you built so it's not completely pointless.

However, what some people seem to be overlooking is that, when you're mic pre noise and THD is very low, everything it's connected to becomes dominant. You have the input coupling, the power supply, the output and you probably need a steep low-cut. You could build a perfect discrete CFIA circuit with fancy transistors and gold wire but if your power supply has 60 Hz hum it will ruin your SNR and all your efforts will be wasted.

The power supply is probably the area for the largest improvement. This is because people are starting to realize that switching mode power supplies (SMPS) are significantly better than linear supplies in just about every way. If implemented properly the noise floor is flat as a pancake in every direction. A good SMPS is an amazing device actually. But there are pitfalls. Probably the most important is loading it properly to keep it out of "hiccup" mode. If you don't load an SMPS, it will pulse the supply on and off at a low, very audible frequency and cause all sorts of little bumps in the noise floor. This is why it's important to get an SMPS that has the right power rating which for a little mic pre is going to be probably the smallest one you can find (maybe 1 square inch!).

But perhaps the most important thing that you need to build a high performance microphone preamplifier is a tool for measuring  it. If you can't see the noise and THD of your device, how will you know if your changes actually improved the performance? You're flying blind. So you need a test rig. There are many opinions about what the best rig is and mine always seems to change each time I do it (which is to say it's not easy). But they're usually all something like this:

  USB Audio Interface Output > Attenuator > DUT > USB Audio Interface Input

So you send a tone or noise or some stimulus out of the USB interface, it gets attenuated 50dB or whatever and into your mic pre or Device Under Test (DUT) which amplifies it back to nominal level and into the USB interface where you can then analyze it with software.

However there are some potential problems here. One is that if you connect the USB interface to your mic pre directly such as through the shield of the cable or pin 3, you will very likely introduce all sorts of nasty ground noise from the computer. Sometimes using a laptop helps (maybe unplugged from the wall) but there is a much better more foolproof solution. The best solution for this that I have found so far is to use a Jensen JT-MB-D transformer as the attenuator. Specifically, I use two resistors, like say 47K, in series with the primary wires and then I parallel the secondaries and load it with 150 ohms. Add switches to disconnect pin 1 from input, output and maybe the shield of the transformer. This rig will give the DUT a really clean, low impedance input signal and should make it possible to make some decent noise and THD measurements.
 
JohnRoberts said:
I expect the THAT preamp is quite good.

The schematic you shared looks like what is called "Cohen" topology. It can be quite good I have made many preamps using  it or variants, over the last few decades. BTW you can flip it around to accommodate PNP or NPN input devices. The Cohen's noise performance at high gain will be dominated by the input devices, at very low gain the op amps used could make some difference in the noise floor. The low noise input devices I used decades ago  (<1 nV/rt Hz) are now obsolete. Most manufacturers these days just use an IC solution like THAT or TI.

I am not a fan of that specific DC servo topology. First I prefer that the input to the servo has the RC configured as a passive LPF, that way it is impossible to slew rate limit the servo amp. Secondly the servo output feeding the preamp output differential at unity gain means the full noise of the servo amp will appear in the output with unity gain. The servo amp has more than enough DC gain to correct the DC levels using a larger value resistor to reduce the noise contribution, of course the other side of the differential needs to be similarly balanced.

I have used the self bias through the feedback resistors (R20 and R21) technique in a fixed gain MC phono preamp back in the 80s. In a mic preamp you may forfeit a little input voltage swing at low gain.

JR

Thanks JR.
I know that this is Cohen topology.
I'm going in this direction to avoid input transformer (is there another way to make a transformerless preamp with good performances?).

About DC servo I admit I cannot hear the presence in a circuit of a capacitor, so I think I could live with it and avoid completely it.

About R20/21 you suggest to reduce them?

About THAT corp. ICs I like the result with a lot of mics but I feel I'm loosing something in high frequency response/velocity in transient/detail.
This is the main reason I'm searching for something different.

abbey road d enfer said:
Why? Four is enough. (2 Szlikai pair)

Thanks. I see the article but ignoring the benefit using Szlikai I only took a look at the article you linked in which is shown a 4pair dealing with only 1in/out.

squarewave said:
That circuit is called a Current Feedback Instrumentation Amp (CFIA) and it is described in detail starting on page 44 of this document:

  http://www.thatcorp.com/datashts/AES129_Designing_Mic_Preamps.pdf

It's generally accepted as having the best overall noise / distortion performance of any mic circuit around and as such it's found in all sorts of gear. Pretty much all of the basic mixers you'll find at the music store use it. Mackie for example call it "ONYX".

So for transparent and clean, the whole mic topo debate is pretty much "solved" w/ CFIA. And, once you start to get down to that level of noise and THD, making something discrete just isn't going to matter. It becomes just a fun exercise. Although there is some psychoacoustic merit to using something that you built so it's not completely pointless.

However, what some people seem to be overlooking is that, when you're mic pre noise and THD is very low, everything it's connected to becomes dominant. You have the input coupling, the power supply, the output and you probably need a steep low-cut. You could build a perfect discrete CFIA circuit with fancy transistors and gold wire but if your power supply has 60 Hz hum it will ruin your SNR and all your efforts will be wasted.

The power supply is probably the area for the largest improvement. This is because people are starting to realize that switching mode power supplies (SMPS) are significantly better than linear supplies in just about every way. If implemented properly the noise floor is flat as a pancake in every direction. A good SMPS is an amazing device actually. But there are pitfalls. Probably the most important is loading it properly to keep it out of "hiccup" mode. If you don't load an SMPS, it will pulse the supply on and off at a low, very audible frequency and cause all sorts of little bumps in the noise floor. This is why it's important to get an SMPS that has the right power rating which for a little mic pre is going to be probably the smallest one you can find (maybe 1 square inch!).

But perhaps the most important thing that you need to build a high performance microphone preamplifier is a tool for measuring  it. If you can't see the noise and THD of your device, how will you know if your changes actually improved the performance? You're flying blind. So you need a test rig. There are many opinions about what the best rig is and mine always seems to change each time I do it (which is to say it's not easy). But they're usually all something like this:

  USB Audio Interface Output > Attenuator > DUT > USB Audio Interface Input

So you send a tone or noise or some stimulus out of the USB interface, it gets attenuated 50dB or whatever and into your mic pre or Device Under Test (DUT) which amplifies it back to nominal level and into the USB interface where you can then analyze it with software.

However there are some potential problems here. One is that if you connect the USB interface to your mic pre directly such as through the shield of the cable or pin 3, you will very likely introduce all sorts of nasty ground noise from the computer. Sometimes using a laptop helps (maybe unplugged from the wall) but there is a much better more foolproof solution. The best solution for this that I have found so far is to use a Jensen JT-MB-D transformer as the attenuator. Specifically, I use two resistors, like say 47K, in series with the primary wires and then I parallel the secondaries and load it with 150 ohms. Add switches to disconnect pin 1 from input, output and maybe the shield of the transformer. This rig will give the DUT a really clean, low impedance input signal and should make it possible to make some decent noise and THD measurements.

Thanks. I already read AES129.
I don't know if I earn something with a discrete design instead of opamps but if I have to go into direction of cheap mixers I'll skip this project.
I recently tried amek 9098 and I liked it
So I took a look at the schematic and it's pretty common implementation of Cohen balanced topology.

I already measured THD in other projects using my interface (motu avb series) and REW.
So I'll build an attenuator to use it with a mic preamp.
But I'd like to be sure about what I'm going to prototyping.
I'd like something better than my that 1510/12 based preamp, more detailed in high frequency.

About smps I recently tried a small module by Mean Well (a medical applications one) and I admit that performs super in those particular circuit.
In this case I've to find one with +48vdc.
It is about +125mA at +/-15vdc if I remember well.

Thanks for all your contribution!

 
I think the Forssell mic pres are excellent for classical guitar. Fred has posted here in the past,  there might be some good info archived  to draw ideas from.  There are some schematics on the Forssell website.

Also have a look at the INA103 datasheet.  It's pretty good in its own right,  which you might prefer it to the 1510. But you can also see how the internals are and replace the opamps with 3 DOAs.  If using good low noise opamps I would omit the additional input transistors.
 
Aleguitarpro said:
...I tried a lot of preamps recording in studios, thanks to friends, building my own etc.
I'm a classical guitar player and I record for friends and colleagues classical music and acoustic instruments mainly.
..Most difficult part to find from standard suppliers are Q1 and Q2.

May I ask you what is the average distance of the microphone from the preamp during recording and from which manufacturers did you use the cables when testing the mic preamps? As far as I can see GML is not among the best. Can you say something more about your impressions of that preamp?

You may check out 2SA1084 transistor also (it' PNP).
 
squarewave said:
....The power supply is probably the area for the largest improvement. This is because people are starting to realize that switching mode power supplies (SMPS) are significantly better than linear supplies in just about every way...
Can you, please, tell in more detail what these significant benefits are?

 
Aleguitarpro said:
Thanks JR.
I know that this is Cohen topology.
I used it decades before ever hearing it called "Cohen".
I'm going in this direction to avoid input transformer (is there another way to make a transformerless preamp with good performances?).
IMO no. But if you look inside the popular mic preamp ICs you will find the Cohen topology or variants of.
About DC servo I admit I cannot hear the presence in a circuit of a capacitor, so I think I could live with it and avoid completely it.
Properly done you should not hear capacitors.  About the only capacitor in that circuit that is doing heavy lifting is the one in series with the gain resistor. At max gain that capacitor is feeding low single digit ohms. I have seen servo circuits that support DC coupling that gain resistor, but I make no claims about audibility, or not.
About R20/21 you suggest to reduce them?
umm no... Perhaps I wasn't clear... those resistors are providing the operating current for the input devices. The typical alternative is to use current sources to provide those mA of operating current. Providing the bias current through those resistors is elegant (I have done it before in a MC phono preamp) but there is a trade off in you lose a couple volts of input swing (not an issue for a phono preamp).
About THAT corp. ICs I like the result with a lot of mics but I feel I'm loosing something in high frequency response/velocity in transient/detail.
Measure twice, cut once.  For HF response we used to jangle keys in front of a mic, this will excite an octave plus above the audio pass band. I can't hear up there, but i can hear IMD when the circuitry can't keep up.

If you think you hear a difference between two preamps, try a null test... that will confirm the difference, just not which one is different.

JR
This is the main reason I'm searching for something different.

Thanks. I see the article but ignoring the benefit using Szlikai I only took a look at the article you linked in which is shown a 4pair dealing with only 1in/out.

Thanks. I already read AES129.
I don't know if I earn something with a discrete design instead of opamps but if I have to go into direction of cheap mixers I'll skip this project.
I recently tried amek 9098 and I liked it
So I took a look at the schematic and it's pretty common implementation of Cohen balanced topology.

I already measured THD in other projects using my interface (motu avb series) and REW.
So I'll build an attenuator to use it with a mic preamp.
But I'd like to be sure about what I'm going to prototyping.
I'd like something better than my that 1510/12 based preamp, more detailed in high frequency.

About smps I recently tried a small module by Mean Well (a medical applications one) and I admit that performs super in those particular circuit.
In this case I've to find one with +48vdc.
It is about +125mA at +/-15vdc if I remember well.

Thanks for all your contribution!
 
moamps said:
Can you, please, tell in more detail what these significant benefits are?
Less noise. No mains hum. More efficient. Less heat. Smaller. Fewer parts. Cheaper. Safer.

I can only think of two disadvantages of an SMPS vs a linear supply. One is handling large changes in load like that of a power amplifier driving an 8 ohm load. I have never tried to use an SMPS with a power amplifier but I think it could be made to work with the right filtering. The other disadvantage is EMI radiated within the enclosure. The only solution for that is to make sure the SMPS is physically located far enough away. But that's almost never a problem. Even 10 cm or so is good enough.
 
john12ax7 said:
I think the Forssell mic pres are excellent for classical guitar. Fred has posted here in the past,  there might be some good info archived  to draw ideas from.  There are some schematics on the Forssell website.

Also have a look at the INA103 datasheet.  It's pretty good in its own right,  which you might prefer it to the 1510. But you can also see how the internals are and replace the opamps with 3 DOAs.  If using good low noise opamps I would omit the additional input transistors.

Thanks for the tip.
Very interesting reading (the INA103 datasheet).

I took a look at Forssell schematics. On website there's a transformer coupled mic preamp (an elegant way to use only 1DOA in class A, without so much work... but I'm trying to stay transformerless), a transformerless one that use JMP-1 module and the JFETMP1

https://www.forsselltech.com/media/attachments/JFETMP1.PDF.

JMP-1 module catch my interest: probably is NOT so different from the schematic I posted.
Seeing pcb arrangement is about 1 pair of (impossible to find) 2sk389 and 3 DOAs and 1 servo.

JFETMP1 use 2sk389 and a total of 4 opamps.
OPA604 are very good on paper and probably in operation, too.
But I'm trying to make it in class A.

I know it's more than and exercise but I'm trying to see all option on my table before deciding to build and test one of them.

moamps said:
May I ask you what is the average distance of the microphone from the preamp during recording and from which manufacturers did you use the cables when testing the mic preamps? As far as I can see GML is not among the best. Can you say something more about your impressions of that preamp?

You may check out 2SA1084 transistor also (it' PNP).

About cables I use generally Mogami, Sommer or Reference ("ultimo cavo" is my latest addition and in italian means literally "latest cable"... ).
I don't remember series but I can't hear any loss of differences switching between them.
Are generally accepted as good among others.
I haven't mentioned GML between my favourite because I haven't a direct reference on guitar.
I tried it with a violin and I found GML highs are a little too "piercing", if I can use this term.
But I haven't any direct experience on guitar, maybe it helps details and sounds great.

One preamp that from what I hear from demo is great (but here in Italy I can't demo one directly) is buzz audio ma2.2.







 
JohnRoberts said:
Properly done you should not hear capacitors.  About the only capacitor in that circuit that is doing heavy lifting is the one in series with the gain resistor. At max gain that capacitor is feeding low single digit ohms. I have seen servo circuits that support DC coupling that gain resistor, but I make no claims about audibility, or not.

Thanks. I don't want to make something more difficult that can't justify benefit of adding parts if it hasn't a direct effect on quality.

umm no... Perhaps I wasn't clear... those resistors are providing the operating current for the input devices. The typical alternative is to use current sources to provide those mA of operating current. Providing the bias current through those resistors is elegant (I have done it before in a MC phono preamp) but there is a trade off in you lose a couple volts of input swing (not an issue for a phono preamp).

Thanks, clear.

Measure twice, cut once.  For HF response we used to jangle keys in front of a mic, this will excite an octave plus above the audio pass band. I can't hear up there, but i can hear IMD when the circuitry can't keep up.

If you think you hear a difference between two preamps, try a null test... that will confirm the difference, just not which one is different.

JR

Great and easy method (jangle keys in front of a mic).
Faster than switch preamp and play a guitar in front of it.
 
I tried here to follow some of your advices.

Let me know what do you think about it.

I'm still searching for good candidates for Q1-Q4.
Requisite is easy availability/good performance in this position.

2n4403/2n4401?

Thanks
 

Attachments

  • Trless class A variant.jpg
    Trless class A variant.jpg
    352.1 KB · Views: 72
squarewave said:
Less noise. No mains hum.
I disagree.
More efficient. Less heat. Smaller.Cheaper.
Agree, but not important in high class preamps.
Fewer parts.  Safer.
I disagree.
One is handling large changes in load like that of a power amplifier driving an 8 ohm load. I have never tried to use an SMPS with a power amplifier but I think it could be made to work with the right filtering.
SMPS are used in power amps with great results (especially in active speakers for PA or studios).
The other disadvantage is EMI radiated within the enclosure.
And back to mains.

Thank you for your reply.
 
You may want to consider using ZTX851/853 transistors, according to my calculations your collector current is around 2mA, with that current each transistor ZTX851/853 exhibits roughly a 0.37 nV/rtHz voltage noise density and 1.9 pA/rtHz current noise density, but for a differential input pair the noise voltage is actually a bit higher  0.52 nV/rtHz, the ideal source impedance ought to be around 200 ohms.  That is not counting the resistor noise, but you get the idea, plug it into a simulator.
 
moamps said:
I disagree.
Having a debate these things is good. But it doesn't work if you don't backup your position with any explanation.

moamps said:
And back to mains.
The EMI is not mains. It's at switching frequency.

A vaguely good SMPS should have absolutely ZERO mains hum in the output. You can very easily inject mains hum from a device you're connected to (like a computer audio interface that isn't properly isolated). But the source will not be the SMPS. SMPS do have switching frequency ripple. But that is easily filtered out with a CLC filter. You can follow that with a capacitance multiplier or regulator if you really want to bring the supply noise way down but for most things it really isn't even necessary.
 
user 37518 said:
You may want to consider using ZTX851/853 transistors, according to my calculations your collector current is around 2mA, with that current each transistor ZTX851/853 exhibits roughly a 0.37 nV/rtHz voltage noise density and 1.9 pA/rtHz current noise density, but for a differential input pair the noise voltage is actually a bit higher  0.52 nV/rtHz, the ideal source impedance ought to be around 200 ohms.  That is not counting the resistor noise, but you get the idea, plug it into a simulator.
+1,,, I am not rigorously following modern low noise  devices but those look better than my old (obsolete) favorites.

The 2n4403 was considered relatively low noise back in the 70s (the 2n4403 was written up in the classic low noise design text by Motchenbacher and Fitchen. )

Using even perfect input devices and op amps will not get your noise lower than the Johnson (thermal) noise of the mic's source impedance (150-200  Ohm). There is maybe one dB or less of total noise improvement available versus canned IC solutions. Then we have room noise and and any mic electronics self noise.

JR   
 
user 37518 said:
You may want to consider using ZTX851/853 transistors, according to my calculations your collector current is around 2mA, with that current each transistor ZTX851/853 exhibits roughly a 0.37 nV/rtHz voltage noise density and 1.9 pA/rtHz current noise density, but for a differential input pair the noise voltage is actually a bit higher  0.52 nV/rtHz, the ideal source impedance ought to be around 200 ohms.  That is not counting the resistor noise, but you get the idea, plug it into a simulator.
Do you know the value of rbb' ? I don't see any noise specs for this.
 

Latest posts

Back
Top