| Mr Evil |
My turntable has been sitting on a shelf for many years. Finally I have done something about this and designed an RIAA pre-amp. It uses JFETs since I recently bought a load and like them. It has no global feedback partly because it's an integral part of the functioning of the circuit and partly because I never go open-loop and want to do something different for once.
As per the attached schematic it consists of two differential stages each of which implements part of the filter, the first being the pole and zero at 50 and 500Hz, the second being the pole at 2122Hz (with an extra zero at 50kHz as I have seen suggested in a few places). A source-follower buffers the output.
Differential stages use 2SK389/2SJ109, output stage is 2SK170/2SJ74.
It requires some careful pot twiddling to balance the differentials (essential for correct frequency response) and zero offset. The various source degeneration resistors need to be chosen to achieve the desired gain with particular JFETs.
Since I have not listened to vinyl (at home at least) for a few years and have no other pre-amp to compare, I can't really judge how well the circuit performs, but it's certainly not bad. |
|
|
| analog_sa |
| You can actually get stable offset from this circuit?! Looks not bad at all. Have you considered adding current sources in the tails? |
|
|
| choky |
reminds me to Hafler preamp........
I relly dunno what number,but it's on their site anyway |
|
|
| Mr Evil |
| quote: | Originally posted by analog_sa
You can actually get stable offset from this circuit?! Looks not bad at all... |
Apparently yes. I wouldn't trust it with a power amp without an input cap, but the offset does seem to remain constant and low even after an hour or so of running. I suppose that using monolithic dual FETs takes care of thermal matching so offset of each pair should remain low, and since the same current flows in both differentials, any change in their offset should be roughly equal and opposite, leaving nothing at the output.
| quote: | Originally posted by analog_sa
...Have you considered adding current sources in the tails? |
Indeed I have. It's just a prototype at the moment, so still room for change. I will probably change R3 for a CCS (JFET, of course :D) and possibly also R5, which will give a nice boost to PSRR. |
|
|
| ljozsef |
Hallo,
Did anybody think of using it with a balanced pre/amp?
I wonder if it could drive them without the actual complementary stages. Using FET or MOSFET input amps (e.g. Pass or its diy clones) it could succeed.
And with dual FETs only in the circuit the offset will be minimized.
Any thoughts?
Laci |
|
|
| jam |
You might be on to something here...........You might want to consider doubling up the fets in the first differential.
Jam |
|
|
| jam |
......for noise reasons.
Jam |
|
|
| Mr Evil |
| quote: | Originally posted by choky
reminds me to Hafler preamp........
I relly dunno what number,but it's on their site anyway |
I searched around but couldn't seem to find that one. However it is not an unusual design: I have seen plenty of double-differential JFET amps. The only thing about it that I haven't seen before is the way filtering is accomplished.
| quote: | Originally posted by jam
You might be on to something here...........You might want to consider doubling up the fets in the first differential.
Jam |
Do you think that would be worth it? Those JFETs are already low noise and not cheap. Maybe if I were going no-expense-spared, but this is something that will see only occasional use.
That makes me think of something: Is it better to have n-channel or p-channel for the input? The datasheets say noise is the same for both.
I've attached what the sims say the distortion spectrum at 1kHz should look like. Higher order harmonics are very low, which is nice. |
|
|
| Onvinyl |
| quote: | | That makes me think of something: Is it better to have n-channel or p-channel for the input? The datasheets say noise is the same for both. |
I recall borbely opts for npn in his papers (fets-the new frontier in audio), and I think he did for noise reasons (not sure).
| quote: | | The only thing about it that I haven't seen before is the way filtering is accomplished. |
You may want to visit this thread:
http://www.diyaudio.com/forums/show...&threadid=12397
Rüdiger |
|
|
| EUVL |
Nice and Simple Circuit, but enough gain for MC ?
And how about making it fully symmetrical with differential output ? Probably not much more needed than adding another source follower on the opposite arm ??
Patrick |
|
|
| Noobee |
Shouldn't R13 and R14 read around 24k in order to achieve the nominal 47k?
Bernd |
|
|
| Mr Evil |
| quote: | Originally posted by Onvinyl
I recall borbely opts for npn in his papers (fets-the new frontier in audio), and I think he did for noise reasons (not sure)... |
I had a quick glance through that paper and could see no explicit statement of which was quietest, but all the examples of low-noise circuits used n-channel inputs, so I think it can be safely assumed that they are the best choice.
Nothing new under the sun:) I think it's an elegant way of shaping frequency response.
| quote: | Originally posted by EUVL
Nice and Simple Circuit, but enough gain for MC ?.. |
How much gain do you need for a MC? My prototype has ~40dB of gain at 1kHZ. If the LTP source resistors were made very small then gain could be increased to 50 or 60dB. Alternatively a third LTP could be added.
| quote: | Originally posted by EUVL
...And how about making it fully symmetrical with differential output ? Probably not much more needed than adding another source follower on the opposite arm ??. |
It would indeed be just the addition of an extra pair of FETs to give it a differential output, which would greatly reduce any possible worries about DC offset stability, enhance PSRR and send even harmonic distortion through the floor. No good for me though, since the rest of my kit is all unbalanced.
| quote: | Originally posted by Noobee
Shouldn't R13 and R14 read around 24k in order to achieve the nominal 47k?
Bernd |
You're right. The input resistors appear in series, not in parallel. |
|
|
| jam |
Mr.Evil,
I would opt for n channels myself...........you could use the front end of the Pearl phono stage to get the extra gain for a low output MC.
Great concept, though.
Regards,
Jam |
|
|
| Onvinyl |
Did you measure the riaa's accuracy? If so, how is it?
Rüdiger |
|
|
| Noobee |
Talking about Pearl -- Did anybody try to use a capacitance multiplier like there is in the pearl (and in ono, i suppose) in another surrounding , and if so, did it improve things sonically?
Benrnd |
|
|
| Mr Evil |
| quote: | Originally posted by Onvinyl
Did you measure the riaa's accuracy? If so, how is it?
Rüdiger |
It's about +/-1dB from 10Hz-20kHz. I didn't spend any great effort making it accurate, so I'm sure it would be possible to choose component values that can get closer. Note that it's important that the voltages across the LTP load resistors be balanced as much as possible or effective filter component values will change. |
|
|
| Noobee |
Hi Mr. Evil,
I've found another neat idea of you here: http://www.diyaudio.com/forums/show...3548&highlight=
That leads to the first thing that went through my head loooking at your schematics:
What are the reasons you don't use ccs here. Sonics? Lazyness (I guess not?)
Bernd |
|
|
| Jocko Homo |
I have built several open-loop JFET preamps. Running the input "balanced" makes all the difference in the world. Everyone needs to try that at least once in their DIY career. I used current mirrors in the second stage to convert back to SE, and pick up some dBs in the process.
The uber-nerd that I am..........I trimmed the RIAA to +/- 0.04 dB. Or better.
Jocko |
|
|
| jam |
Jocko,
How about the schematic....................:D |
|
|
| Mr Evil |
| quote: | Originally posted by Noobee
Hi Mr. Evil,
I've found another neat idea of you here: http://www.diyaudio.com/forums/show...3548&highlight=
That leads to the first thing that went through my head loooking at your schematics:
What are the reasons you don't use ccs here. Sonics? Lazyness (I guess not?)
Bernd |
But there is a CCS in there ;) The four input JFETs, being depletion mode devices, maintain a constant voltage across R10/RV1, giving a constant current.
EIT: Or did you mean in the preamp? I haven't used a CCS here for simplicity, since it's still just a prototype. I may use one or two CCSes for the final version.
| quote: | Originally posted by Jocko Homo
I have built several open-loop JFET preamps. Running the input "balanced" makes all the difference in the world. Everyone needs to try that at least once in their DIY career. I used current mirrors in the second stage to convert back to SE, and pick up some dBs in the process.
The uber-nerd that I am..........I trimmed the RIAA to +/- 0.04 dB. Or better.
Jocko |
I considered using a current mirror in the second stage, but how do you control the DC offset? |
|
|
| Noobee |
Mr. Evil,
of course I meant the preamp.
But, can we really speak of open loop here? do R4, C1 and R8, C2 not apply some feed back, since any output in a transistor is an input as well?
@Jocko Homo:
well, that sounds interesting. A while ago, you posted a circuit from the MAT02 datasheet, is that it? Come on, some infos please...
Bernd |
|
|
| Noobee |
Sorry, in case I bore you,
A little Idea: since the folded cascodes are said to be unreached in reproducing trebles, shouldn't it make a perfect first stage for a phono amp where the trebles are at +20dB in respect to 1kHz?
To date, I never tried on, but I may next time I find time to play around a little.
Bernd |
|
|
| Mr Evil |
| quote: | Originally posted by Noobee
...But, can we really speak of open loop here? do R4, C1 and R8, C2 not apply some feed back, since any output in a transistor is an input as well?.. |
It's local feedback. Open-loop implies no global feedback, and there is none of that. |
|
|
| EUVL |
If I understand your circuit correctly, the output source follower, driven by the second diff pair, sees power supply noise of the negative rail 1:1 ??
Would this not be an argument to go balanced or at least use a current mirror or as some of you might fancy an output transformer ?
Also the DC offset would drift with the negative rail voltage, e.g. when one were to use batteries ??
Patrick |
|
|
| Mr Evil |
| quote: | Originally posted by EUVL
If I understand your circuit correctly, the output source follower, driven by the second diff pair, sees power supply noise of the negative rail 1:1 ??
Would this not be an argument to go balanced or at least use a current mirror or as some of you might fancy an output transformer ?
Also the DC offset would drift with the negative rail voltage, e.g. when one were to use batteries ??
Patrick |
Negative PSRR is around about zero, yes. It's not a problem if the power supply is well regulated.
A balanced output would be great, but is only useful if the following stage has a balanced input.
Change in negative rail voltage would appear directly at the output. AC coupling to the next stage is definitely a good idea. |
|
|
| EUVL |
No matter how much energy one invests in designing the power supply, there are hardly any supply that is both totally noise free (especially at high frequencies) AND voltage stable.
I could not help thinking that a low impedance (meaning therefore high capacity) battery supply, coupled with a servo circuit (using for example another dual JFET) which also serves as a (variable) current source to drive the first diff pair, AND a form of balanced receiver right before the input of the following stage (if singled ended) would improve the performance of this simple and elegant circuit significantly.
And I do not like coupling caps, neither at the input nor at the output.
Patrick |
|
|
| EUVL |
For those who might want exact RIAA values, and if my maths is correct :
C1 = 212n (e.g. 150n // 47n // 15n)
R8 = 1.272k (e.g. 1.24k + 33.2R)
C2 = 2.5n (e.g. 2.2n // 330p)
Philips (or BC Components) does a series of 1% axial PP caps at very reasonable price.
I though find the bias at both diff pairs a bit low, so that one is not really operating at the most linear regime of the JFETs. Probably a fact of 5 to 10 would be better, thermal dissipation allowing. Rs and Cs values need changing of course.
Patrick |
|
|
| Mr Evil |
| quote: | Originally posted by EUVL
No matter how much energy one invests in designing the power supply, there are hardly any supply that is both totally noise free (especially at high frequencies) AND voltage stable.
I could not help thinking that a low impedance (meaning therefore high capacity) battery supply, coupled with a servo circuit (using for example another dual JFET) which also serves as a (variable) current source to drive the first diff pair, AND a form of balanced receiver right before the input of the following stage (if singled ended) would improve the performance of this simple and elegant circuit significantly.
And I do not like coupling caps, neither at the input nor at the output.
Patrick |
Well I suppose it depends on how far you're willing to go. As I mentioned earlier, this is something that will see infrequent use, so although I do want it to be as good as possible, there are limits of complexity and cost.
The noise floor of the circuit as it stands is ~80dB down. That's with a cobbled together LM317/337 regulator with little bypassing, very long wires and no shielding. |
|
|
| Onvinyl |
| quote: | | Well I suppose it depends on how far you're willing to go. As I mentioned earlier, this is something that will see infrequent use, so although I do want it to be as good as possible, there are limits of complexity and cost. |
That is always true, even for the so-called cost no object designs...
But inserting 1-fet currentsources at the two tailes here costs 2 fets and 2 resistors (not that it cannot get much fancier here as you know...)
@Patrick: do you mind to draw a quick sketch of your servo idea?
In AoE there is a conceptual drawing of a LTP with a current mirror (wilson style, if I recall right) as a load (on page 'have-to-look-at-home'), having then a single output, do you think of such a receiver?
I alway wondered why nobody seems to use a dual-diff-pair as MC-Input stage, but that may be too OT here.
Rüdiger |
|
|
| Mr Evil |
I did say I was thinking of using current sources, but all this discussion of PSRR has made me change my mind: A CCS in the first LTP would significantly improve its negative PSRR, but when the preamp as a whole is considered, it actually makes negative PSRR worse!
Say the negative rail increased in magnitude. This would increase the current through the first LTP, increasing the output voltage and thus increasing the current through the second LTP. This would increase the output voltage, partially cancelling the change in rail voltage. In fact a quick calculation says that this effect gives ~28dB of negative PSRR. With a CCS, the negative PSRR would only be as good as that of the second LTP alone, or 0.
A CCS in the second LTP would have very little effect, and so I don't think it's worth it if I'm not also going to use one in the first LTP. |
|
|
|
