Hi H713,
I don't want to kick dirt on your dream, but 'fully balanced' seems exceedingly ambitious for a first design attempt, especially since the benefits are minuscule and roundly swamped out by any number of other design shortcomings -- plus you're already conceding transformers in and out.
I'd strongly suggest getting an SE design working, with the desired gain, distortion and noise specs you need FIRST. Then when it pleases you, by all means double it to make it balanced. In the mean time you'll have saved HALF your planning, arithmetic, soldering, time AND parts, and the troubleshooting (caps or transistors mounted backwards, resistors in wrong locations) that an experienced designer would guarantee takes a lot more concentration and time (in a balanced design).
A technical bit or two -
- successive preamp stages very rarely operate at the same current density
- equal Collector and Emitter resistors, w/the Base near Gnd, and dual supply rails, severely limits dynamic range, and produces unity gain without RV1
- R13 and R16 are too low in value; instead of global feedback, try using local, per-stage linearizing techniques
- especially troublesome is having RV1 within the global feedback loop -- the negative feedback will fight the gain increase
- the differential 2nd stage will saturate prematurely -- it is unsuited for higher levels
- don't forget to properly load the secondary of T1
There's still plenty of fun to be had -- but maybe a little less after you build 26 of 'em!
Cheers,
Rick
I don't want to kick dirt on your dream, but 'fully balanced' seems exceedingly ambitious for a first design attempt, especially since the benefits are minuscule and roundly swamped out by any number of other design shortcomings -- plus you're already conceding transformers in and out.
I'd strongly suggest getting an SE design working, with the desired gain, distortion and noise specs you need FIRST. Then when it pleases you, by all means double it to make it balanced. In the mean time you'll have saved HALF your planning, arithmetic, soldering, time AND parts, and the troubleshooting (caps or transistors mounted backwards, resistors in wrong locations) that an experienced designer would guarantee takes a lot more concentration and time (in a balanced design).
A technical bit or two -
- successive preamp stages very rarely operate at the same current density
- equal Collector and Emitter resistors, w/the Base near Gnd, and dual supply rails, severely limits dynamic range, and produces unity gain without RV1
- R13 and R16 are too low in value; instead of global feedback, try using local, per-stage linearizing techniques
- especially troublesome is having RV1 within the global feedback loop -- the negative feedback will fight the gain increase
- the differential 2nd stage will saturate prematurely -- it is unsuited for higher levels
- don't forget to properly load the secondary of T1
There's still plenty of fun to be had -- but maybe a little less after you build 26 of 'em!
Cheers,
Rick
Last edited:
Given that I already have a schematic that simulates at least reasonably well, I'm probably going to try bread boarding it this weekend (assuming I have time) to see what happens. This is certainly not the first thing I've designed or built- this is just my first stab at a mic pre. And we shall see- there may be plenty of issues with this design that the simulations did not show. I'm still unsure of how picky the design will be about transistor matching. That alone could render this design impractical. I am, however, at the point where I want to see how it behaves in the real world before I throw too much time into it.
In addition to that, these channel strips are going to be modular. The mic pre and EQ will each be their own boards so as to allow greater flexibility. I don't have to put in 26 of the same mic pres, and I don't intend to, unless this design turns out to be so amazing that I actually want 26 of them in my desk (highly unlikely).
In addition to that, these channel strips are going to be modular. The mic pre and EQ will each be their own boards so as to allow greater flexibility. I don't have to put in 26 of the same mic pres, and I don't intend to, unless this design turns out to be so amazing that I actually want 26 of them in my desk (highly unlikely).
The design is tolerant of mismatched transistors. In general it seems well behaved and supplies can be as low as +/- 9V. Noise simulations are very good. There should be no significant problems with prototyping. I have transformers and 2N4403/2N4401 lying around so I will knock up a prototype as well.
Output transformers seem an extra expense since there is already isolation on the input and caps on the output. There is an advantage in having them if the following stage is not balanced, but I don't know the complete environment. BD139 is a bit big for a 26 channel solution. That is the only issue I see.
Output transformers seem an extra expense since there is already isolation on the input and caps on the output. There is an advantage in having them if the following stage is not balanced, but I don't know the complete environment. BD139 is a bit big for a 26 channel solution. That is the only issue I see.
I spent some time today breadboarding the circuit, and it does have a couple notable issues.
The first, and most serious, is stability. The circuit breaks into sustained oscillation when the gain is turned up significantly, and RV1 needs to be limited to a minimum of 150-ish ohms in order to avoid major oscillation. This may not prove to be an issue, as I could put a 180 ohm resistor in series with RV1 and tweak the feedback resistors to allow the 60dB+ of gain that I'd like this circuit to be capable of. I do not think the emitter followers are the problem, as I tried eliminating them from the circuit entirely (fine as long as I don't want to drive any load) and it had minimal impact.
The potentially more difficult issue is related to the gain control, and I believe it is something that Rick hinted at. This was in the back of my mind while I simulated the design. Using a linear taper pot does not correspond to a linear increase in gain the way it should, and I suspect this has to do with how RV1 impacts the feedback loop. The brute-force solution to this would be to use a rotary switch with resistor values that are calculated and tested to provide the proper amount of gain. This really isn't a great solution, since rotary switches aren't cheap and requires a lot more time to wire up. I'm going to take a look at how I can reconfigure the feedback loop to deal with this.
The first, and most serious, is stability. The circuit breaks into sustained oscillation when the gain is turned up significantly, and RV1 needs to be limited to a minimum of 150-ish ohms in order to avoid major oscillation. This may not prove to be an issue, as I could put a 180 ohm resistor in series with RV1 and tweak the feedback resistors to allow the 60dB+ of gain that I'd like this circuit to be capable of. I do not think the emitter followers are the problem, as I tried eliminating them from the circuit entirely (fine as long as I don't want to drive any load) and it had minimal impact.
The potentially more difficult issue is related to the gain control, and I believe it is something that Rick hinted at. This was in the back of my mind while I simulated the design. Using a linear taper pot does not correspond to a linear increase in gain the way it should, and I suspect this has to do with how RV1 impacts the feedback loop. The brute-force solution to this would be to use a rotary switch with resistor values that are calculated and tested to provide the proper amount of gain. This really isn't a great solution, since rotary switches aren't cheap and requires a lot more time to wire up. I'm going to take a look at how I can reconfigure the feedback loop to deal with this.
You could try base stopper resistors or (less noisy) ferrite beads on the first stage to see if that helps for stability. RC series networks to ground might also help. By the way, how did you model the transformers in your simulation?
When the gain control range is large, as it is, RV1 should ideally be a reverse logarithmic potmeter. I thought you had found one because there is R-Log annotated on the schematic.
When the gain control range is large, as it is, RV1 should ideally be a reverse logarithmic potmeter. I thought you had found one because there is R-Log annotated on the schematic.
Yes, a reverse-log is specified in the schematic. I changed that to a linear pot simply for testing purposes, and it didn't behave the way I was expecting it to with a linear taper pot, so I'm not sure that a rev-log taper is going to solve the problem, but we'll see. I'll play around with some of these suggestions to see if they help the stability.
Last edited:
Here's an interesting development: It seems that it will only break into full-power oscillation if I ground one side of the output transformer. This applies to either signal ground, or even to Earth by clipping on an oscilloscope ground lead. If I connect my scope ground lead to signal ground, all is well. I wish I could explain why it behaves this way, and it is indeed a pretty significant issue. Incidentally, the gain control doesn't seem to be an issue anymore either. Output transformer in question is an older Jensen J-T11-FL.
It's still not super stable, and it can suffer from parasitic oscillation into certain loads. That's another issue, and when I have time to mess with this again (probably next weekend), I'm going to try a trick I've had success with on tube amps, which is a zobel network across the secondary of the output transformer.
If I'm understanding you correctly, a resistor and a cap in series from the collector to the base on each of the input transistors?
Beyond the stability issues, I'm satisfied with performance. I'm reading about .04% THD at 40dB of gain (max gain is about 67dB) and that's with completely unmatched transistors and an input transformer out of an old Sescom mic splitter that I have reason to believe is average at best. Before the coupling capacitors, I was measuring about a 3.5 volt (!) difference between the emitters of the two output transistors. Probably my fault for using some random NTE2347s that I had laying around. I swore I had BD139s in stock- guess I have to order some. Noise performance was decent enough too, at maximum gain I read about a -45dB noise floor, at minimum gain it was about -87dB. These figures will only improve when this is on a real PCB rather than a crusty breadboard.
It's still not super stable, and it can suffer from parasitic oscillation into certain loads. That's another issue, and when I have time to mess with this again (probably next weekend), I'm going to try a trick I've had success with on tube amps, which is a zobel network across the secondary of the output transformer.
If I'm understanding you correctly, a resistor and a cap in series from the collector to the base on each of the input transistors?
Beyond the stability issues, I'm satisfied with performance. I'm reading about .04% THD at 40dB of gain (max gain is about 67dB) and that's with completely unmatched transistors and an input transformer out of an old Sescom mic splitter that I have reason to believe is average at best. Before the coupling capacitors, I was measuring about a 3.5 volt (!) difference between the emitters of the two output transistors. Probably my fault for using some random NTE2347s that I had laying around. I swore I had BD139s in stock- guess I have to order some. Noise performance was decent enough too, at maximum gain I read about a -45dB noise floor, at minimum gain it was about -87dB. These figures will only improve when this is on a real PCB rather than a crusty breadboard.
Last edited:
I mean a resistor and cap in series from collector to collector of first stage.
I have made a first prototype. However this is a different flavor. The first stage is FET 2SK3557. Since this is N, the other stages have swapped to P 2N4403 and resistors re-balanced. The circuit works. No squealing but only a partial test so far since transformers are not with me. Gain is only 21dB at the moment but it demonstrates that the concept is flexible. BW 1.5MHz.
Balance of the system was about 1 volt without feed back, 30mV with. I did match the FETs. Running at +/- 18V.
I have made a first prototype. However this is a different flavor. The first stage is FET 2SK3557. Since this is N, the other stages have swapped to P 2N4403 and resistors re-balanced. The circuit works. No squealing but only a partial test so far since transformers are not with me. Gain is only 21dB at the moment but it demonstrates that the concept is flexible. BW 1.5MHz.
Balance of the system was about 1 volt without feed back, 30mV with. I did match the FETs. Running at +/- 18V.
Last edited:
Alright, that makes sense. I'll try adding that to the first stage next weekend- I suspect will probably take care of the parasitic oscillation.
How's the noise floor with the FET input? I was under the impression that JFETs tend to be noisier with low-impedance sources. Aside from that, I do like the idea of using JFETs in a mic preamp.
I am not yet ready to do noise tests. The source resistance does affect JFET noise performance. However, the 2SK3557 is supposed to be better on this front. We shall see. I chose the FET because it is time something replaced the standard Schoeps affair. It is a very clever scheme but crappy at the same time. A FET front end makes it more general purpose and I can see an in-mic version.
Transformers went missing. Now relocated. A better evaluation in the offing. With +/- 18V, 9V pk-pk before clip, single ended. So the concept is compatible with directly driving a differential A/D.
The master stroke was in swapping out the opamps. The circuit path is reduced from 70+ to 6 transistors. It seems to be completely general purpose and can utilize any technology, BJT, JFET, Enhancement FET, valves and any mix of them. I have simulated all N and P versions. Noise simulation varies from 2nV/rtHz down to 0.4nV/rtHz.
Great concept H713.
The master stroke was in swapping out the opamps. The circuit path is reduced from 70+ to 6 transistors. It seems to be completely general purpose and can utilize any technology, BJT, JFET, Enhancement FET, valves and any mix of them. I have simulated all N and P versions. Noise simulation varies from 2nV/rtHz down to 0.4nV/rtHz.
Great concept H713.
With the input transformer hooked up, gain is now up. Previously I was driving it single ended. I did not fit the cap as I am aiming for a capless circuit. The gain I can measure is 51dB. There is more gain, but the pot is very sensitive near the end stop. Everything is unshielded so hum is breaking through. It is also picking up the florescent light.
The FET gate resistors are 22M so that won't help. Noise will have to wait for a better environment, but I am not too concerned about that. At least there are now two versions with pretty good gain.
The FET gate resistors are 22M so that won't help. Noise will have to wait for a better environment, but I am not too concerned about that. At least there are now two versions with pretty good gain.
Do you have a schematic for this version? I'm a little interested to see how you implemented a FET front end on this. It's an interesting idea and aside from the buffers in condenser microphones you don't see a lot of JFETs in the front-end.
When I have time I may play around with simulations for a tube front end for this and see how it behaves.
Another interesting way that tubes could be used here is for the line drivers, and then use something like a 10K:600 line transformer (depending on tubes in question) for the output.
This weekend I'm probably going to play around with soldering it up onto a piece of perfboard with semi-matched transistors and see if I can't get some more meaningful numbers, and hopefully get the stability issues worked out.
Couple things worth noting.
1) I suspect that the input pad can probably go, since I was able to feed a 0dB signal in pretty comfortably without noticeable distortion.
2) At around 10kHz the frequency response starts climbing. I need to spend some time debugging to track down the cause. As of right now I see nothing that should cause that sort of behavior, so I need to take a careful look at what could be causing this, but I may wait until I get the stability worked out, as that may take care of this issue at the same time.