I expect better from you (please take that in the right spirit I consider you a respected friend). The circuit I am talking about operates always in small signal mode and has no slew rate, it is -90dB or so THD at line level and does not get to 1% distortion until several 100's of mV at the input. All performance aspects are far beyond the capabilities of an LP and stylus.
Even if an op amp with bipolar 15V supplies is used for the first stage, the overload point is not an issue. Say the first op amp can output 13V peak on a 15V supply. From past research by Shure etc. a mm cartridge will output less than 100mV peak on a hot record. Then the maximum gain for the first linear stage before clipping is 13V/0.1V = 130 times or 42dB.
We only need 60dB for both stages combined to get 40dB midband RIAA gain, since there is a 20dB midband loss in the passive RIAA network. If we choose the first stage to have a gain of 31.6 times or 30dB, there is a factor of 130/31.6 = 4.11, or over 12dB, overload margin on even the hottest records. And my thousands of LPs don't have loud pops.
Last edited:
Let's look for a moment at low noise design for phono. First, overall, low noise jfets can cover all the bases, but they are expensive, and difficult to buy. Bipolar transistors can work but never 100% effectively for both MC and MM cartridges. You have to choose, and you have to be careful what bipolar transistors that you use.
Most of the common bipolar transistors have significant Rbb' (a defined term) which is effective base resistivity. Usually from 40 ohms-300ohms. Back in the early 70's, most 'low noise' transistors had an Rbb' of 200 ohms or so, but they did have fairly high beta. Even 50 years ago, some parts were available with 40 ohms or so (2N4403) Rbb' but they had medium beta, and worked poorly for MM cartridges, because the beta would fall off at low operating currents. BUT for MC, they were pretty good.
As the years progressed, manufacturers, (especially the Japanese) offered really low Rbb' and hi beta transistors. In many ways, the best of both worlds, but they could have other problems. Today, we might use a low noise bipolar input based IC, (marginal for hi inductance MM cartridges or tape heads) or a very low noise jfet (LS170), often in parallel for lowest noise, (expensive). There is no 'free lunch'.
Most of the common bipolar transistors have significant Rbb' (a defined term) which is effective base resistivity. Usually from 40 ohms-300ohms. Back in the early 70's, most 'low noise' transistors had an Rbb' of 200 ohms or so, but they did have fairly high beta. Even 50 years ago, some parts were available with 40 ohms or so (2N4403) Rbb' but they had medium beta, and worked poorly for MM cartridges, because the beta would fall off at low operating currents. BUT for MC, they were pretty good.
As the years progressed, manufacturers, (especially the Japanese) offered really low Rbb' and hi beta transistors. In many ways, the best of both worlds, but they could have other problems. Today, we might use a low noise bipolar input based IC, (marginal for hi inductance MM cartridges or tape heads) or a very low noise jfet (LS170), often in parallel for lowest noise, (expensive). There is no 'free lunch'.
You forget to consider the overload recovery after pops. Pops output can be much higher than any hot signal from the grooves. 12dB is IMO not enough, 20dB is a minimum.
Has anyone measured this recently or are we just quoting Shure papers? Scott has found that clipping the ADC input off a flat gain stage causes no problems.
Do I need to post plots? I have found no problem at all with pops, in any ordinary recording set up. First of all they are not pre-emphasized second of all an open-loop gain stage has essentially zero overload recovery time.
Myths and more myths, time to move on.
NPN only, though. And mediocre beta, making the input current noise way too large for any MM application.
The beauty of the now extinct low noise bipolars from Hitachi (2SC2545/2SA1085) was the unique combination of low Rbb (some 3 ohm) and ultra high beta (could even reach 1000, 700-800 was common) meaning an ultra thin base. This was achieved by an unique polysilicon emitter process, more complex than the process for a standard (at the time) IC.
If you don't care about input current noise, almost every medium/high power transistor will heve Rbb of 10ohm or less.
Yes. I found pops outputs as high as 10mV from a Benz Micro Gold MC cartridge (low output, nominal 0.5mV/5cm/s, measured with a vinyl test record to 0.4mV). That's 28dB over nominal. The Benz Micro Gold is a high compliance cartridge and used with a matching SME 3009 Series III tonearm, which doesn't need in this case any extra damping.
Why bother, FETs are good for both.
Which PNP? PNPs are usually lower noise than NPN, all other things being equal (emitter area, etc...).
And syn08 will appreciate, the reason Zetex's devices have such low resistance is: big area (Ic_max = 4 amperes!) plus polysilicon base. Not polysilicon emitter, polysilicon base. They have a writeup on their website, which dedicated and curious people can discover, talking about this design and showing a nice 3D drawing.
That's over 50pA/rtHz, that's monstrous. Measured myself, 2SD786 has about 15pA/rtHz @ 10mA. For the Zetex device, a 4ohm source resistance will match the 3ohm Rbb voltage noise. Put 4 in parallel and that source resistance needs to be under 1 ohm (halve the voltage noise, double the current noise).
Link please, never heard of a "polysilicon base" transistor.
P.S. Polysilicon base electrodes, to lower the extrinsic Rbb, are a different kettle of fish. That's not "polysilicon base".
For MC, low noise bipolars are usually fine, as long as you don't care about the cartridge DC current, or accept a DC blocking capacitor, or implement one of the floating configuration with it's extra complexity and limitations (needs a very low noise floating power supply).
DC current can be cancelled in a symmetrical design, although it is difficult in a discrete implementation, due to temperature coefficient issues. But uncorrelated noise sources cannot cancel, unless you use some chopping cross-correlation topology, with its own issues. Otherwise, DC current cancellation comes in a package with a 3dB current noise penalty.
For MM, I would not touch bipolars with a ten foot pole.
DC current can be cancelled in a symmetrical design, although it is difficult in a discrete implementation, due to temperature coefficient issues. But uncorrelated noise sources cannot cancel, unless you use some chopping cross-correlation topology, with its own issues. Otherwise, DC current cancellation comes in a package with a 3dB current noise penalty.
For MM, I would not touch bipolars with a ten foot pole.
- Status
- Not open for further replies.