This is a perfect case for the inverse RIAA equalized TIM test or what I call TIM (30,30). If it can pass that with less than .01%, then I won't complain.
Now that we all now know how to parallel FET's to get .4nV of noise, what is the best partitioning of gain and RIAA time constants to maximize headroom? I would consider this far more important.
Count with me, syn08. My feedback eq. cap value is about 100nF. This makes 79.6 ohm at 20kHz. For 5Vpeak, I need 62.8mA from opamp output. Too much for AD797, too much for any opamp. The buffer is absolutely needed.
PMA said:
Still, you don't get it. My input stage is already (partly) RIAA corrected. The gain dros like a stone at HF, therefore the output swing at HF is very small. Under 50cm/s overload conditions, I have measured AD797 25mA peak current in 20Hz-100KHz, remarkably constant (in fact it slightly drops with frequency). Simulation says the same story.
I was not strictly exact, regarding gain division between stages.
At 20kHz, we of course need about 36dB less RIAA gain than for 40Hz.
So - what is your output current into FB at 2OkHz/5Vp output? You can simulate it easily.
At 20kHz, we of course need about 36dB less RIAA gain than for 40Hz.
So - what is your output current into FB at 2OkHz/5Vp output? You can simulate it easily.
PMA said:
Please, tell me, or at least post a link to what he was saying. That might be more helpful than a terse comment.
BTW, what I said about that circuit was not what I think about it, rather what my simulator reported.
PMA said:
5V output where? I hope you remeber there are two more gain stages following the input stage, the full schematic is in the GPP thread.
- At 20KHz, the input stage gain is about 24dB and the following gain stages account for 20dB.
- Also, 5V output @20KHz means 31mV input from the MC.
- Which leads to a (to) high output current at the AD797 output of 60mA.
Read this carefully:
The assumption of 31mV input from a MC cartridge with 0.5mV nominal at 5cm/sec (resulting in a 5V output at 20KHz) is, to me, absurd. Actually, my preamp was designed for a 28dB headroom. 31mV/0.5mV = 36dB, you are already far and away from any reasonable spec or design target.
I have never seen or measured a 36dB headroom in a commercial MC design. If you can show me one (solid state, of course), I promise to adjust my design to be compliant. Is not even a big deal, as you said, a OPA634 with 250mA output current will do easily.
Meantime, I'm happy with 28dB headroom, as designed. With this headroom, the AD797 output current never exceeds 25-30mA which is reasonable.
If this doesn't make it clear, I don't know what I could further explain...
Edit: If not clear enough, my whole MC preamp has a total of 64dB gain @ 1KHz.
syn08 said:
I know my understanding of SNR is below all standards. That perfectly explains why the PGP amp has a better SNR than YAP.
Anyhow, I do care about the noise around the 330ohm resistors. They contribute much more noise than those 1 Ohm source resistors. You better opt for somewhat higher values for the FB network (and unburden the AD797!!!) and lower resistors in the I-mirrors.
syn08 said:
I'm not sure whether 28 or 36dB is sufficient. A phone pre-amp should easily handle ticks as well. I once read a very interesting article about this subject. Was it in Wireless World? I can't remember anymore. But what I do remember is that one needs an unexpectedly large amount of headroom to handle these ticks properly. It had something to do with clipping and recovery.
Edmond Stuart said:
I know my understanding of SNR is below all standards. That perfectly explains why the PGP amp has a better SNR than YAP.
Anyhow, I do care about the noise around the 330ohm resistors. They contribute much more noise than those 1 Ohm source resistors. You better opt for somewhat higher values for the FB network (and unburden the AD797!!!) and lower resistors in the I-mirrors.
I'll answer this one for you:
- The YAP amp has higher noise only and simply because it's current feedback. The PGP has indeed very low noise because I used very low noise devices, not by any front end design features.
- The noise around the 330ohm noise reflects at the input divided by the JFET stage gain, about 2*2*4*Idss*330ohm (2 for two branches, 2 for the gm and 4 for 4 JFETs per branch, Idss=11mA). Which makes a gain of about 58, therefore the equivalent input noise of a 330ohm is a little over 0.1nV/rtHz. Good enough. Don't forget, the P and N branches are actually in parallel!
- Now if this is not good enough for you, you would need to increase the 330ohm resistor. Gain increases linearly with the resistor, while noise increases with SQRT. The higher the resistor, the higher the SNR.
In all seriousness, watching this exercise has convinced me more than ever that tubes are the right technology to use for phono. Very educational.
syn08 said:
One of the front-end design features was NOT using CFB.
That's correct, I got about the same figure.
That's even more correct. The old geezer has overlooked these basics. I always feared the noise contribution of I-mirrors, simply because most 'expert' warned about it.
Now it appears that the noise contribution is not that disastrous.
Lesson: Don't take it for granted what 'experts' say.
Thx for the eye opener.
SY said:
Which raises a very good question: how much headroom is realistically required? I've never seen anything over a 50cm/sec input equivalent even mentioned, while the RIAA standard says max 25cm/sec.
BTW, I'm to lazy now to do calculations, but it seems that even the famous Manley Steelhead (tubes) preamp has less than 36dB headroom. Max output is +27dB and MC gain is 60dB @1KHz.
@syn08
Get an MC with a nice undamped ultrasonic resonance (most of 'em), do some mistracking or a few bits of dust, stuff that happens with real records, figure in the velocity characteristic of the MC; the notion that you can count on the RIAA spec to set your overload is tenuous.
Get an MC with a nice undamped ultrasonic resonance (most of 'em), do some mistracking or a few bits of dust, stuff that happens with real records, figure in the velocity characteristic of the MC; the notion that you can count on the RIAA spec to set your overload is tenuous.
SY said:
Ok, I assume you have an example, how much headroom you think it's required (and based on what)?
I don't know the Manley topology, so can't comment. But I'll go through mine, which is gain block, passive RIAA, gain block, buffer. The first stage has 58dB of gain and a max swing of 70V. Working backward, the overload of this stage at any frequency is 78mV. Referred to a 0.5mV/5cm/s cartridge, that's 44db of headroom. I can feel pretty safe with that. The limitation happens at 1kHz with the second stage (which has an input overload of 1.7V peak); approximating the RIAA cut from the passive network as 20dB at 1kHz, when the first stage clips, the second stage is hit with 7V peak. That degrades the overload by 12dB or so. So I'll only claim 32dB of overload at 1k. But...
1k is not where overload is likely to happen. Mistracking and tick/pop excitation of the cartridge resonance is likely to have the most energy at 40-60kHz, where the RIAA has rolled off that 7V another 25-30dB. So at frequencies where we really need overload performance, the tube stage will be back to 44dB of headroom, which I'd argue is quite enough. Subjectively, with this topology, I find my poor records easier to listen to.
Yes, I sacrifice something in theoretical noise, but it's still so low as to be inaudible. The ENR of the input stage is something like 7 ohms.
1k is not where overload is likely to happen. Mistracking and tick/pop excitation of the cartridge resonance is likely to have the most energy at 40-60kHz, where the RIAA has rolled off that 7V another 25-30dB. So at frequencies where we really need overload performance, the tube stage will be back to 44dB of headroom, which I'd argue is quite enough. Subjectively, with this topology, I find my poor records easier to listen to.
Yes, I sacrifice something in theoretical noise, but it's still so low as to be inaudible. The ENR of the input stage is something like 7 ohms.
I haven't done envelope capture nor do I have a digital scope to do it with (maybe next year). But published data I have seen showed a 22dB peak over the 5cm/s rating in an MC caused by dust, the maximum energy being at the tip-vinyl resonance. This particular cartridge was a fairly well damped one- I would want at least 6-10dB more at ultrasonic frequencies to ensure that with less-well-damped cartridges, the preamp wouldn't overload.
- Status
- Not open for further replies.