The Sescom input transformer is definitely suspect. I have yet to measure one, but that's next on my list. I don't have a datasheet for them (I was told they came out of a sescom mic splitter). Given that I have no idea how they were loaded, it is something worth looking into. Good suggestion.
Here's an update: The frequency response was tracked down to the Sescom transformer, which when measured alone has a rise of about 8dB from 10kHz to 20kHz- not ideal. I will have to take some time to figure out the proper way to load it, since I was pleasantly surprised to see that its distortion specs weren't too far off the Jensen.
I was able to address the stability issues (mostly) by adding a 1000pF cap from the base of Q1 to the base of Q2. Frequency response is still good well past 100kHz even with this cap in place. The RC bridging between the collectors of the input transistors did not solve the problem, and in some circumstances made it worse. With the 1000pF cap on the input, it was much more stable. It still doesn't really like having my oscilloscope ground clip attached to one side of the output, but I found that with the 1000pF cap in place, I can short either side of the output transformer to signal ground without issues.
I'm presently working on a PCB layout. I should have that finished in the next day or two. I added spots in the PCB for caps in parallel with the feedback resistor, and a space for a 100nF cap between the +/- 18V rails right next to the output stage. These parts may or may not be needed, but it's easy enough to leave those places unpopulated.
I was able to address the stability issues (mostly) by adding a 1000pF cap from the base of Q1 to the base of Q2. Frequency response is still good well past 100kHz even with this cap in place. The RC bridging between the collectors of the input transistors did not solve the problem, and in some circumstances made it worse. With the 1000pF cap on the input, it was much more stable. It still doesn't really like having my oscilloscope ground clip attached to one side of the output, but I found that with the 1000pF cap in place, I can short either side of the output transformer to signal ground without issues.
I'm presently working on a PCB layout. I should have that finished in the next day or two. I added spots in the PCB for caps in parallel with the feedback resistor, and a space for a 100nF cap between the +/- 18V rails right next to the output stage. These parts may or may not be needed, but it's easy enough to leave those places unpopulated.
That's the way I've been measuring it, but the fact that it can oscillate when a scope ground is clipped to one of the outputs suggests it might not be 100% stable. That said, I also was testing with meter-long cables connecting to the power supply without the 100nF bypass cap. I suspect that the output stage will be more stable with that cap in place.
This is definitely asking for trouble.
0.1uF is definitely a must cap. I would add a 10uF, any place or close to the output stage.
Last edited:
Very interesting transistors.
_As mike preamp front end. Do they have low Rb, low noise ?
_Ouput stage drivers in power amps.
The layout I'm working on is using traditional through-hole parts. BC550 and BC560, then BD139 for the output stage. I estimate PCB size to be about 100mm by 60mm. A good portion of that space is occupied by the input pad and phantom power section. Switches are ALPS SPUJ series or equivalent, and the footprint of the pot is for that of a Clarostat 388 series. I should have a first draft of the layout done tonight.
If I didn't have so many through hole parts that I'd like to use up, I'd probably design the whole thing with SMD.
If a PCB is drawn for the version with the FET input, it would probably be wise to use all SMD since as far as I know there aren't a whole lot of low-noise through-hole JFETs left.
Okay, here's where things are at. I think the values of the feedback resistors needs to be updated, and I believe that R15 is 2.5K instead of 5K. Aside from that, this schematic should be representative of the actual prototype.
This is the first semi-complicated PCB I've designed, so there may well be issues with it. Attached are the Gerber files I came up with.
If I didn't have so many through hole parts that I'd like to use up, I'd probably design the whole thing with SMD.
If a PCB is drawn for the version with the FET input, it would probably be wise to use all SMD since as far as I know there aren't a whole lot of low-noise through-hole JFETs left.
Okay, here's where things are at. I think the values of the feedback resistors needs to be updated, and I believe that R15 is 2.5K instead of 5K. Aside from that, this schematic should be representative of the actual prototype.
This is the first semi-complicated PCB I've designed, so there may well be issues with it. Attached are the Gerber files I came up with.
Makes sense to use stock. The BC550/BC560 don't have especially low rb. They work for medium impedance circuits at 200uA. For lower impedance circuits higher current transistors tend to have lower rb. The 2N4401/2N4403 are around 40R rb. BC550/BC560 are above 100R. The matched transistors are higher current and I would expect noise to be good (for low impedance). BD139 is a bit slow. May be contributing to instability. Gaining stability by reducing bandwidth is OK, but HF distortion will tend to rise.
There are options like using constant current loads for the output. This will reduce current for the same distortion performance. Some like color in mic amps. Unbalancing the circuit slightly with a control can give you options.
Bandwidth is fine with the cap. Will help keep RF out.
Check headroom and clipping performance if R15 is lowered. With 2.5k the DC balance is closer to mid rail but it may not be optimal.
There are options like using constant current loads for the output. This will reduce current for the same distortion performance. Some like color in mic amps. Unbalancing the circuit slightly with a control can give you options.
Bandwidth is fine with the cap. Will help keep RF out.
Check headroom and clipping performance if R15 is lowered. With 2.5k the DC balance is closer to mid rail but it may not be optimal.
The layout should not be critical as it is low impedance and not fast. The choice of routing is not what an experienced PCB designer would choose. This is a simple design and 2 layers would be used for routing (outer layers so you can do mods). One layer is horizontal, the other vertical. Inner layers would be power and GND with a plane for GND. The current version replicates routing on multiple layers so mods would be problematic but it should work. If you have time, look at the FET circuit and visualize routing based on that scheme. I.e. +V on one side -V on the other. GND can be a layer so ground connected components can be anywhere. Place components relative to the rails and fill ground last. Hope this makes sense.
This is a first iteration. Don't be too fussy. You can be smarter with a follow up version that can accept SMD as well as TH.
This is a first iteration. Don't be too fussy. You can be smarter with a follow up version that can accept SMD as well as TH.
Last edited:
Time for an update. PC boards arrived, and I finally found time to assemble one and do some real performance testing. I made a couple of bonehead mistakes, most notably routing it such that pushing in the pad switch turns the pad off rather than on and then purchasing momentary switches. A PCB revision will be made to correct this and to add the mounting holes I forgot to add. I will also try to make the silkscreening easier to understand.
Performance was excellent. I did a comparison with my Focusrite ISA One (allegedly based on the original Focusrite ISA 110, a preamp renowned for being very clean and low-distortion). Distortion was in the same league as the Focusrite (ranging from .005 to .1 % depending on gain and output level), and the noise was almost exactly the same. Ultimately I did try replacing the BC550/BC560s with 2N4401/2N4403s, and ultimately I found that the noise floor was improved by ~1.0 dB. Distortion didn't really change. IMO, either transistors are perfectly acceptable. I found that performance is best between +/- 15 and +/- 20V. Maximum output on a +/- 18V supply was about +24dB. When it does clip, it appears (at least on the scope) to do so quite softly, which I consider to be a good characteristic.
I did notice that the last ~1 degree of potentiometer rotation corresponds to a sudden 3dB jump in gain. I tried a couple different pots (including a linear one), and they all exhibited the same behavior. I have the conclude that this is either 1) A fucntion of how pots are made, or 2) Due to the pot being in the NFB loop, though I would think that the resistor in series with the pot would alleviate that. Thoughts?
Lastly, since I'm doing some PCB revisions, any suggestions (without going beyond two layers)?
One thought I did have was that because it works quite well on +/- 16V, if there is interest it would be quite easy to modify the PCB layout and turn it into a 500 series module.
An updated schematic with all the current part values will be posted within the next couple days.
Performance was excellent. I did a comparison with my Focusrite ISA One (allegedly based on the original Focusrite ISA 110, a preamp renowned for being very clean and low-distortion). Distortion was in the same league as the Focusrite (ranging from .005 to .1 % depending on gain and output level), and the noise was almost exactly the same. Ultimately I did try replacing the BC550/BC560s with 2N4401/2N4403s, and ultimately I found that the noise floor was improved by ~1.0 dB. Distortion didn't really change. IMO, either transistors are perfectly acceptable. I found that performance is best between +/- 15 and +/- 20V. Maximum output on a +/- 18V supply was about +24dB. When it does clip, it appears (at least on the scope) to do so quite softly, which I consider to be a good characteristic.
I did notice that the last ~1 degree of potentiometer rotation corresponds to a sudden 3dB jump in gain. I tried a couple different pots (including a linear one), and they all exhibited the same behavior. I have the conclude that this is either 1) A fucntion of how pots are made, or 2) Due to the pot being in the NFB loop, though I would think that the resistor in series with the pot would alleviate that. Thoughts?
Lastly, since I'm doing some PCB revisions, any suggestions (without going beyond two layers)?
One thought I did have was that because it works quite well on +/- 16V, if there is interest it would be quite easy to modify the PCB layout and turn it into a 500 series module.
An updated schematic with all the current part values will be posted within the next couple days.
Good move.
Quest for very low noise:
Try ZTX851. A known very low noise transistor.
Another one, but a bit expensive:
AS394, it is a dual matched pair, a remake of LM394 that has been a best.
How is it about mains hum at full gain.
Do you have power supply transformer close to the preamp ?
This is to know about hum pick up by the circuit.
Quest for very low noise:
Try ZTX851. A known very low noise transistor.
Another one, but a bit expensive:
AS394, it is a dual matched pair, a remake of LM394 that has been a best.
How is it about mains hum at full gain.
Do you have power supply transformer close to the preamp ?
This is to know about hum pick up by the circuit.
Last edited:
I have not done much testing near transformers, so I do not know how sensitive the circuit is to hum. The power supplies I've used for testing are all very quiet and are a good 0.5 meters away from the PCB under test. That said, during testing, the most sensitive part on the board was the electrolytic capacitor in series with the gain control pot. Unsurprisingly, touching it resulted in about a 30dB increase in noise. My original plan for this preamp is to put it into a console, which has its power supply sitting in a separate rack. That said, for stand-alone usage in a rack mount enclosure, it would be worthwhile to do some testing. As it stands the boards are relatively compact and it would not be hard to fit five or six of these in a 1u rackmount enclosure, or alternatively it could be made larger into a 500 series module.
For my own purposes, I'm going to create a version compatible with a dbx 900 series rack. It would be trivial to redraw to a 500 series module. I may cross post this on GroupDIY to gauge if there is any interest in such a board.
My interest in designing a version that uses SMD parts is relatively low. If I draw up a version with a FET input then I will have to allow for SMD parts, and likely go SMD all throughout. Now that I've seen the real boards, there are some modifications I'm going to do that should make it slightly more compact (boy would it be nice to get it down to 50mm wide) and to clean up the layout.
One thing I notice in Rod Elliot's design is the 100k resistor in parallel with the pot.
I should note that I tested this design without the input transformer, and it performed quite well. I still prefer to have an input transformer, but it is a perfectly acceptable "transformerless" preamp. That said, the output transformer is, IMO, very important in a design like this. It isn't nearly as nicely behaved without it. Fortunately, it's not too hard to come by decent line output transformers for a reasonable price, at least in the US.
For my own purposes, I'm going to create a version compatible with a dbx 900 series rack. It would be trivial to redraw to a 500 series module. I may cross post this on GroupDIY to gauge if there is any interest in such a board.
My interest in designing a version that uses SMD parts is relatively low. If I draw up a version with a FET input then I will have to allow for SMD parts, and likely go SMD all throughout. Now that I've seen the real boards, there are some modifications I'm going to do that should make it slightly more compact (boy would it be nice to get it down to 50mm wide) and to clean up the layout.
One thing I notice in Rod Elliot's design is the 100k resistor in parallel with the pot.
I should note that I tested this design without the input transformer, and it performed quite well. I still prefer to have an input transformer, but it is a perfectly acceptable "transformerless" preamp. That said, the output transformer is, IMO, very important in a design like this. It isn't nearly as nicely behaved without it. Fortunately, it's not too hard to come by decent line output transformers for a reasonable price, at least in the US.
From Ron Elliot, about preamp pot.
I think, reverse here does not concern rotation only, it is actually "anti logarithmic".
Log law / Exp law.
The max gain depends of the resistor in serie with the pot + 2Re, the emitter resistance of the transistor. This Re depends of Ie, the emitter current. Re is what is limiting the max gain you can get.
Log pot / Reverse log pot.
I think, reverse here does not concern rotation only, it is actually "anti logarithmic".
Log law / Exp law.
The max gain depends of the resistor in serie with the pot + 2Re, the emitter resistance of the transistor. This Re depends of Ie, the emitter current. Re is what is limiting the max gain you can get.
The prototype seems to performs well.
A first suggestion would be to make a composite FET SMD/BJT pad. This should not have much impact on space. This would allow the option to use FETs for the input pair, populated as desired. The reality is that the world needs another mic amp like a hole in the head. Given the choice, most people would choose to make a 1073 clone. There is great personal satisfaction in producing a design, but to garner more interest there needs to be a wider view.
This design configuration is more flexible than the 1073. So there is a route from a product perspective to make it interesting with a unique blend of features that can make it stand apart from the rest. This is especially true for a rack unit which has avenues to create a channel strip for example. I mentioned previously the option to add color control with progressive distortion. The FET version boadens application to acoustic guitar piezo and wired pickups. A following stage could implement eq based on the same circuit. This creates the channel strip. The gain control can be implemented with LDR and LED so it can be remote capable.
Following on from that, the LDR could be used to create a compressor, with no additional electronics in the signal path. So it all depends where you want to end up. A fully featured unit would be a commercially viable in a boutique way.
See modified gain arrangement. Might do something but haven't tried it.
A first suggestion would be to make a composite FET SMD/BJT pad. This should not have much impact on space. This would allow the option to use FETs for the input pair, populated as desired. The reality is that the world needs another mic amp like a hole in the head. Given the choice, most people would choose to make a 1073 clone. There is great personal satisfaction in producing a design, but to garner more interest there needs to be a wider view.
This design configuration is more flexible than the 1073. So there is a route from a product perspective to make it interesting with a unique blend of features that can make it stand apart from the rest. This is especially true for a rack unit which has avenues to create a channel strip for example. I mentioned previously the option to add color control with progressive distortion. The FET version boadens application to acoustic guitar piezo and wired pickups. A following stage could implement eq based on the same circuit. This creates the channel strip. The gain control can be implemented with LDR and LED so it can be remote capable.
Following on from that, the LDR could be used to create a compressor, with no additional electronics in the signal path. So it all depends where you want to end up. A fully featured unit would be a commercially viable in a boutique way.
See modified gain arrangement. Might do something but haven't tried it.
Last edited:
Agreed, there is certainly no shortage of good preamp designs on the internet. Part of the reason I designed this the way I did is that the world really does not need another transformerless preamp using a BJT front end with an op-amp for gain (see MCI JH600 preamp). I do, however, think the world could benefit from another FET input mic preamp (assuming it can be made quiet), since there really aren't a ton of FET input mic preamps. I know Bob Cordell has one.
The unfortunate truth is that the FET input version is different enough that I think it will likely require its own PCB. I'll take some time to look at pinouts, it might be possible. I have already finished the revisions to the BJT input PCB, and I'm probably going to hold off on ordering boards until I finish the FET input PCB as well as the PCB for my high voltage bench power supply in order to save on shipping costs.
The unfortunate truth is that the FET input version is different enough that I think it will likely require its own PCB. I'll take some time to look at pinouts, it might be possible. I have already finished the revisions to the BJT input PCB, and I'm probably going to hold off on ordering boards until I finish the FET input PCB as well as the PCB for my high voltage bench power supply in order to save on shipping costs.