Ed, I can't find any mention of Is parameter in the datasheet, perhaps you mean Ic ?
How does one go about measuring this, say set 1V on base, no base resistor, emitter to ground and measurure collector current accross a load resistor?
Finished drawing schematic in LTspice and done full simulation with a Laplace function voltage source simulating a MM cart as input. Here's a Bode plot of the output at TP11, just before the last series output resistor. That point is where negative feedback is taken for the active riaa eq. Does that makes sense? What worries me is that there's more than 180° phase shift and I learned that can lead to oscillation - But my knowledge is insufficient to know if this is good or bad in this specific case...
Last edited:
Marcel, thank you for this excellent trick! Tested a few transistors, these weren't the actual parts (they're in the mail), only cheap BC170C that were lying around. The voltage drop between samples was less that 10mV but I guess that's sufficient for "good enough" matching?
This (seems 120 Ohms) resistor is shunted @ MC mode by S23+R256=10 Ohms. (and in MM mode it is much less than cart wiring 1k). - not good, but not a big problem. The problem is a input on bjt (even with lower collector current as MarcelvdG wrote) @ MM mode, - jfets here will have much lower noise.
Last edited:
Thanks Nick for the graphs, seems like load capacitance is in an inverse relation to load resistance, so for MC one would need high-ish capacitance with respect to the low input R. Makes sense I guess that explains the high values on top-end preamps and hints that the 100p here is not optimal when in MC mode with 100R load.
For reference the series resistor R246 is a large 1K, the scans are really bad. But in MC mode that resistor is shunted by the 10R yes. However in my design I will drop the 1K and hard-wire the 10R resistor only. The freed DIP switch position will accomodate a 2nd load capacitor.
For reference the series resistor R246 is a large 1K, the scans are really bad. But in MC mode that resistor is shunted by the 10R yes. However in my design I will drop the 1K and hard-wire the 10R resistor only. The freed DIP switch position will accomodate a 2nd load capacitor.
10 mV of difference corresponds to about 40 % of difference between the collector currents at a given base-emitter voltage. Not great, but good enough.
Nick Sukhov likes to use moving magnet cartridges with a non-standard load: far higher resistance and far lower capacitance than usual. A consequence is that the moving magnet amplifier has to be built into the turntable, otherwise the cable capacitance messes up everything. Moving coil (normal or high output) cartridges are much less sensitive to the load capacitance than moving magnet.
That was a maximum range, within that I can probably find exact matches out of the 100 pieces I have ordered. Good thing silicon is so cheap
It seems to me that for critical circuits matching for Is instead of hfe is the way to go? Or is it better to match both? When is hfe matching relevant?
So many questions...
Are you still writing? Just curious
In total, I wrote two audio-related articles for Electronics World, one about a main amplifier in 1996 and the 2003 article about noise-optimizing moving-magnet phono preamplifiers. Years later, I wrote several articles for Linear Audio, the last was an article about a DAC that was published in the very last Linear Audio in 2017. The last published article I wrote was a book review in audioXpress earlier this year, https://audioxpress.com/article/boo...spice-models-to-simulate-vintage-op-amp-noise
The hunderds of microampere currents of post #11 are a compromise between MC and MM. If you just want to optimize it for a 700 mH, 1300 ohm MM cartridge, don't care about MC, and you would use input transistors with hFE = 600, low base resistance and not too much 1/f noise, then the theoretical optimum total collector current would be:
RIAA- and A-weighting: 37.05 uA
RIAA- and ITU-R 468 weighting: 27.58 uA
As you have to divide this between the NPN and the PNP side, the collector currents per transistor get quite small even with only one NPN and only one PNP. Many discrete transistors will have difficulties keeping their hFEs high at such low currents.
The advantage of connecting transistors in parallel is that you reduce the effect of the base spreading resistance. However, there are many transistors with a base spreading resistance far below 1300 ohm, so for this specific case, there is no need to parallel them. All in all, single transistors make more sense in this case.
An assumption I make and also made in post #11 is that you can actually change the collector currents as you like without substantially having to worsen some other noise contribution. That's not necessarily true for the circuit of post #1. It depends on whether you are prepared to replace R231 and R232 (resistors between the emitters of the input transistors and the feedback network) with series connections of two resistors with a big electrolytic capacitor across one of them.
The issue is that the collector currents of the input transistors are each determined by a subtraction of two larger currents. The current sources with the red boxes around them produce currents of which the average must be about 26 mA (depending on the S.O.T., select on test, resistor). Resistors R231 and R232 conduct currents equal to the VBE of the input transistors divided by their resistances, that is, about 23 mA. The remaining 3 mA goes through the input transistors; that is, 3 mA through the NPNs and 3 mA through the PNPs, so together 6 mA.
This means that when the current sources are off by 5 %, the currents through the input transistors are off by about 43 %. If you would scale down the input transistor collector currents by increasing R223 and R218 (if that is what they are called; the 680 ohm collector resistors) and reducing the current through the current sources in the red boxes until there is again 2 V dropping across R223 and R218, the circuit would become even more sensitive to inaccuracies of the current sources (which have a different temperature dependence than the currents through R231 and R232).
If you would scale down the currents through the current sources in the red boxes, and scale up R231, R232, R223 and R218 by the same factor, then the sensitivity to current source inaccuracies would not get any worse than it is now, but the thermal noise voltage of R231 and R232 would increase. If the noise increase would be big enough to be a problem, you could split R231 and R232 each into two resistors in series with a decoupling capacitor across one of them.
For example, to reduce the sum of the input transistor collector currents from 6 mA to 36 uA, R231 and R232 would then scale up from 24 ohm to 4 kohm each. That would be a substantial noise contribution if you don't decouple a part of them. If you don't want to decouple them, it is probably possible to calculate some new optimum with the constraint that the collector current times R231 has to remain equal.
By the way, I'm neglecting the base currents of T111 and its NPN colleague. They should be around 6 uA, so not entirely negligible compared to 36 uA/2. You can correct for it by increasing R223 and R218 a bit further. I also haven't considered frequency compensation or distortion.
You are right, the 3852 Hz rule or 5179 Hz rule does not apply to the second stage. I only have an obvious advice: try to give the first stage a high enough gain to make the second stage non-dominant at all audio frequencies, but not so high that it clips on loud records (and if it needs a long time to recover from clipping, also not on scratches).
I was about to start a new thread on this, but since we are talking about inaccuracies of the current sources, I was wondering about the effects, if any, of the aging of the LED references. These are your garden variety 5mm cheapo red LEDs. They are not run very hard (3.6mA).
I live in an imaginary world where licorns thrive and silicon junctions are eternal, but I guess they must have a life expectancy. The amp is currently open on the bench and I note that the LEDs are quite dim - much dimmer than a red LED passing 3 mA. Furthermore they are not all equally lit between halves of the bipolar supply on each current source section. Note that the transistors including the power finals and all other diodes don't seem to suffer aging...
Information on these old amps is scarce but once I stumbled upon an article on upgrading a member of the same series and vintage of amps, which has basically the same circuit but with scaled-down components and power (I have an exemplar of this model as well). The author, which apparently had a good deal of experience in restorations and upgrades, wrote that he systematically replaced current source LED references.
So I guess I should too, or was he over-zealous?
I live in an imaginary world where licorns thrive and silicon junctions are eternal, but I guess they must have a life expectancy. The amp is currently open on the bench and I note that the LEDs are quite dim - much dimmer than a red LED passing 3 mA. Furthermore they are not all equally lit between halves of the bipolar supply on each current source section. Note that the transistors including the power finals and all other diodes don't seem to suffer aging...
Information on these old amps is scarce but once I stumbled upon an article on upgrading a member of the same series and vintage of amps, which has basically the same circuit but with scaled-down components and power (I have an exemplar of this model as well). The author, which apparently had a good deal of experience in restorations and upgrades, wrote that he systematically replaced current source LED references.
So I guess I should too, or was he over-zealous?