No, as long as there is no voltage at the cartridge connection, there is no current through the cartridge.
The current flows "down" through the NPN trs and into and through the PNP trs. The cartridge output current adds or subtracts from this bias current, and the resulting current change creates an AC output voltage across the Trs load Rs, and hence to the output via the isolation caps.
An elegant solution - my only thought is that it loads the cartridge rather heavily, and in my experience MC carts sound better with very light (47k) loads.
Regards, Allen
The current flows "down" through the NPN trs and into and through the PNP trs. The cartridge output current adds or subtracts from this bias current, and the resulting current change creates an AC output voltage across the Trs load Rs, and hence to the output via the isolation caps.
An elegant solution - my only thought is that it loads the cartridge rather heavily, and in my experience MC carts sound better with very light (47k) loads.
Regards, Allen
From a MC cartridge perspective, this approach is a disaster waiting to happen. Every imbalance (temperature, bias, etc...) or failures (e.g. power supply) will trigger a high current through the MC cartridge (the source impedance is low). Even worse, at each power up the MC will take a serious hit (because it's not possible, without very special approaches, to timely synchronize the +/- power supplies voltages.
This is indeed what I got when I experimented with some rabeyrolles-like circuits. What ever I tried, even using those THAT-transistor array, I couldn't get under 2mV input offset, which could rise up to +/-18mV over time. I abandoned this approach because of this, despite the real nice sound it gives.
What offset voltages do you get there, Joachim?
Rüdiger
So the questions that remain have to do with cartridge safety and if the DC component is audible. Another issue is if loading the cartridge down is bad for the sound.
I have to repaet myself here : in all my experiments with dc coupled bipolar transimpedance input stages i never destroyed a cartridge. Anyway, i take your commends seriously so i redesigned the circuit to have a floating supply. That should avoid current through the cartridge even on turnon. Failure of active elements can not be avoided but it is unlikely that this happens. the powersupplies for left and right channel can not be shared in that design and have to be seperated. What i did not tell you is that i experimented with different amounts of DC offset on the cartridge. i used the asymmetric Hiraga Optime for this using my 5.5 Ohm Titan i. As long as the offset was under 0.05mV i could hear no difference. At more then 1mV the difference was obvious: a loss of dynamics and slam in the bass made the sound quite thin.
To Allen : the reason i am so persistent with this circuit: at least with my low impedance cartridge the sound i get is very good, tremendoes dynamics, resolution and extention. I hear no loss of treble. To build the circuit costs virtually nothing and it is a good learning experience if somebody wants to explore transimpedance techniques.
On the first day in Vegas a old genlemen came to my room and gave me some low noise N and P channel Toshiba Fets he had saved from the late 70th. He ecouraged me to continnue so my next adventure is to try a High Z input. I will tell you how i like that sound when i have a circuit that works for me.
One other thing: John Curl was in my room and he seems to have a new Parasound phonostage ready. Actually i have seen a prototype in the Parasound room. I hope he will explain a bit what he did. The input stages are in seperate metal containers and the powersupply looked like a switch mode solution.
To Rüdiger : i get the offset down to under 0.1mV and it is stable after switch on on my DC coupled transimpedance designs. Actually it is the sound that made me so persistent but i move now to High Z to not bore you to much with my obsession.
I have to repaet myself here : in all my experiments with dc coupled bipolar transimpedance input stages i never destroyed a cartridge. Anyway, i take your commends seriously so i redesigned the circuit to have a floating supply. That should avoid current through the cartridge even on turnon. Failure of active elements can not be avoided but it is unlikely that this happens. the powersupplies for left and right channel can not be shared in that design and have to be seperated. What i did not tell you is that i experimented with different amounts of DC offset on the cartridge. i used the asymmetric Hiraga Optime for this using my 5.5 Ohm Titan i. As long as the offset was under 0.05mV i could hear no difference. At more then 1mV the difference was obvious: a loss of dynamics and slam in the bass made the sound quite thin.
To Allen : the reason i am so persistent with this circuit: at least with my low impedance cartridge the sound i get is very good, tremendoes dynamics, resolution and extention. I hear no loss of treble. To build the circuit costs virtually nothing and it is a good learning experience if somebody wants to explore transimpedance techniques.
On the first day in Vegas a old genlemen came to my room and gave me some low noise N and P channel Toshiba Fets he had saved from the late 70th. He ecouraged me to continnue so my next adventure is to try a High Z input. I will tell you how i like that sound when i have a circuit that works for me.
One other thing: John Curl was in my room and he seems to have a new Parasound phonostage ready. Actually i have seen a prototype in the Parasound room. I hope he will explain a bit what he did. The input stages are in seperate metal containers and the powersupply looked like a switch mode solution.
To Rüdiger : i get the offset down to under 0.1mV and it is stable after switch on on my DC coupled transimpedance designs. Actually it is the sound that made me so persistent but i move now to High Z to not bore you to much with my obsession.
Very interesting, this is the first time I've seen this mentioned. It confirms my intuition that as long as the MC cartridge bias is well under the AC output level everything is fine.
I have read a few pages back and I finally I clicked (was a disconnect between brain and keyboard). You are currently using BC550 for your pre. These are fine with 1-3mA or even lower. Reason is, Rbb (about 40-60ohm) clearly dominates the voltage noise. Therefore, the ~1/SQRT(Ic) component is anyway negligible, and it makes sense to keep Ic small so that the input current noise ~SQRT(Ic) is small. Otherwise said, as long as Rbb' dominates the voltage noise, it makes sense to keep Ic small.
This is not true for true low noise devices with Rbb'=1-5ohm.
BTW, those Toshiba low noise devices are pretty equivalent (noise wise) to the ROHM devices you mentioned you already have. The only reason to use the Toshiba's would be if they are in a high beta class.
Last edited:
Hi John ! Nice that you came by my room. Sorry that i did not identify the powersuply corectly. When i see a choke, i think there is something switching. A switch mode supply from you i did not expect anyway so i was sratching my head a bit.
Syn08, thanks for clarifying the issue with Bias on the BC550,560. I know from experience that high bias on those is no advantage.
I am now working on high Z solutions because i got the impression that people here like:
Low DC bias on the cartridge
Low noise
Adjustable imput impedance even to 47kOhm
Low distortion
So here come my first circuits. Again with bipolars. I will use the Renaisas ultra low noise ones at the input because they have a good combination of low noise and high Hfe.
The circuit starts with a modified Hawksford cascode that has provision for DC cancelation at the cartridge and adjustable symmetry. The base current of the cascode transistor is supplied by a battery.
There is a feedback version too. The output transistors are strong enough to support a low impedance feedback loop. I do not know yet where to place the miller compensation.
The circuit looks a little chaotic at first but should work just fine.
Values are not final so do not build it yet.
Syn08, thanks for clarifying the issue with Bias on the BC550,560. I know from experience that high bias on those is no advantage.
I am now working on high Z solutions because i got the impression that people here like:
Low DC bias on the cartridge
Low noise
Adjustable imput impedance even to 47kOhm
Low distortion
So here come my first circuits. Again with bipolars. I will use the Renaisas ultra low noise ones at the input because they have a good combination of low noise and high Hfe.
The circuit starts with a modified Hawksford cascode that has provision for DC cancelation at the cartridge and adjustable symmetry. The base current of the cascode transistor is supplied by a battery.
There is a feedback version too. The output transistors are strong enough to support a low impedance feedback loop. I do not know yet where to place the miller compensation.
The circuit looks a little chaotic at first but should work just fine.
Values are not final so do not build it yet.
Joachim, when do you do such boring things like sleeping and eating? 
What a circuit! With your low-z circuits, you had that nice moving from simple to complex. Now you start at the complex side of things...
A few days ago, I build Werners proposal, the single fet input with folded cascode transistor. Very simple, and remarkable (transparent) sound.
I happen to listen through it at the moment, some cool polish jazz.
I ordered parts for syn's HPS 3.1 as well. Your thread made me digging into the grooves again...
Rüdiger
What a circuit! With your low-z circuits, you had that nice moving from simple to complex. Now you start at the complex side of things...
A few days ago, I build Werners proposal, the single fet input with folded cascode transistor. Very simple, and remarkable (transparent) sound.
I happen to listen through it at the moment, some cool polish jazz.
I ordered parts for syn's HPS 3.1 as well. Your thread made me digging into the grooves again...
Rüdiger
As mentioned in a few posts back, I said I would post some stuff I am looking at for this application.
First the circuit, which is a balanced direct coupled input design. I have not included and DC offset nulling, but this would be pretty easy to add. One option I am thinking about is to null the MC headamp at the output in order to avoid injecting execessive noise into the front end stages. Since the output signal is so small, this is a viable option in my view.
This circuit does not avoid the use of capacitors since the bias transistors (Q3 and Q4) have to be decoupled, but at least there are no caps in the direct signal path.
The loop gain can be adjusted by changing the value of R17. Those who like it low can put in a 2.2k resistor or even 1k, but distortion will go up of course.
The distortion performance as you can see from the analysis is fairly good and under 1ppm (20KHz) for 500uV input and very benign 2nd and 3rd Harmonic
I've checked the frequency repsonse, and its flat from circa 1Hz out to 1MHz (with R17 = 100k).
The noise plot is also shown and using the BC847/857's is around 0.5nV per root Hz. This is not bad and could be bettered with the use of better transistors (Toshiba's or Rohm).
I used the same gain setting resistors values as Syn08 (1 Ohm and 39 Ohm) for a system gain of 40. This gives an input overload of around 70mV or 43db rev 500uV input
For the power supply (this head amp draws about 17mA pk from each rail), I would just use an LM4562 dual op-amp (one half for +rail and the other half for negative rail. To keep dissipation low, the 4562 would have to be run off +-10V rails or similar.
Circuit in next post.
First the circuit, which is a balanced direct coupled input design. I have not included and DC offset nulling, but this would be pretty easy to add. One option I am thinking about is to null the MC headamp at the output in order to avoid injecting execessive noise into the front end stages. Since the output signal is so small, this is a viable option in my view.
This circuit does not avoid the use of capacitors since the bias transistors (Q3 and Q4) have to be decoupled, but at least there are no caps in the direct signal path.
The loop gain can be adjusted by changing the value of R17. Those who like it low can put in a 2.2k resistor or even 1k, but distortion will go up of course.
The distortion performance as you can see from the analysis is fairly good and under 1ppm (20KHz) for 500uV input and very benign 2nd and 3rd Harmonic
I've checked the frequency repsonse, and its flat from circa 1Hz out to 1MHz (with R17 = 100k).
The noise plot is also shown and using the BC847/857's is around 0.5nV per root Hz. This is not bad and could be bettered with the use of better transistors (Toshiba's or Rohm).
I used the same gain setting resistors values as Syn08 (1 Ohm and 39 Ohm) for a system gain of 40. This gives an input overload of around 70mV or 43db rev 500uV input
For the power supply (this head amp draws about 17mA pk from each rail), I would just use an LM4562 dual op-amp (one half for +rail and the other half for negative rail. To keep dissipation low, the 4562 would have to be run off +-10V rails or similar.
Circuit in next post.
I just took another look at the overload - if powered off +-15V, you can get 200mV input overload. Added benefit is that the midband input equivalent noise decreasess to .388nV/root Hz. (by midband I mean well away from the 1/f corner - so 100Hz and up)
However, overload margins at this level are largely academic, since the power amplifer would be cliping long before this.
However, overload margins at this level are largely academic, since the power amplifer would be cliping long before this.
I have cascoded the inputs on this design. compared to MCHA_1, this:-
1. Reduces the input bias currents to under 40nA (from 600nA)
2. Reduces the equivalent input noise slightly to 0.366nV/root Hz
3. Reduces the distortion to 0.5ppm 20KHz on +-8V rails ref 500uV input
4. Overload is at 120mV input with the 8V rails.
1. Reduces the input bias currents to under 40nA (from 600nA)
2. Reduces the equivalent input noise slightly to 0.366nV/root Hz
3. Reduces the distortion to 0.5ppm 20KHz on +-8V rails ref 500uV input
4. Overload is at 120mV input with the 8V rails.
Last edited:
I'm afraid you are a little optimistic on the noise side. I have checked my BC847, etc... models and they do not include Rbb' (in fact they assume Rbb'=0). Add RB=40 to those models and redo your simulations. You are going to get noise figures of about 0.6-0.7nV/rtHz. You certainly need to massively parallel devices to bring those number down to the current levels.
I would also challenge your statements regarding the overload margins. While I certainly agree with you on the overall MC pre headroom, this is not necessary true regarding the head amp. Remember, due to the RIAA correction, at 20KHz the overall MC gain goes down 20dB, while the MC output goes 20dB up. You certainly don't want clipping to occur in the head amp, so you have to provide a very large headroom here. There is a write-up about, on my web site, under HPS4.1.
As a rule, we would like to always let the last stage in the audio chain to clip, that is the power amp. The power amp is usually the only component in the audio chain that is designed to implement soft clipping. That's why both the MC headamp, the MC pre, and the pre have to implement very large dynamic ranges/headrooms.
Syn08,
I agree with your statement on noise. As I stated above, better transistors are available. Thanks for the point on Rbb - I guess if I add the c. 40Ohm in series with the input transistors I should get a more realistic noise spec. However, if I end up using the Toshiba or Rohm devices, then the noise performance will be closer to the figure I got in the sim.
Regarding overload margin, increasing it is not a problem - just raise the supply rails. At 15V the overlaod is 200mV which is about the same as your design. The cascoded front end devices still keep the bias currents to very low levels (sim'd of course). I suspect there will be a lot of 2nd order effects that would need to be ironed out in a practical implementation.
I agree with your statement on noise. As I stated above, better transistors are available. Thanks for the point on Rbb - I guess if I add the c. 40Ohm in series with the input transistors I should get a more realistic noise spec. However, if I end up using the Toshiba or Rohm devices, then the noise performance will be closer to the figure I got in the sim.
Regarding overload margin, increasing it is not a problem - just raise the supply rails. At 15V the overlaod is 200mV which is about the same as your design. The cascoded front end devices still keep the bias currents to very low levels (sim'd of course). I suspect there will be a lot of 2nd order effects that would need to be ironed out in a practical implementation.