Can you quantify the degree of abrasion? The clipping might be preferred. The Telarc 1812 canons at +8dB throws every cart I have ever owned out of the groove sliding to the end of the LP.
EDIT - Telarc Omnidisk that is it's meant to do that. If you look at the groove with a microscope it looks like a piece of fingernail prying the stylus out of the groove.
Last edited:
I could take issue with some of your technical points but that would be non-productive. This is a DIY forum so extra bits, complexity & cost for slightly worse performance is OK if that's what floats your boat. I'm sure your version is good enough for you.
Of course if World Peace etc are important, one should adopt methods that have been validated by DBTs ... like Duraglit
________________________
I did say I was generalisingScott Wurcer said:
The technology & know-how to track the Telarc 1812 has been available since about V15 Type2 Improved .. though the low compliance MC craze of the early 80s chose to ignore it. I was thinking more of HF & mid trackability where the audible effects are more subtle (?) .. and the cartridge design requirements much more difficult.
Your 1812 example is certainly more abrasive if the cartridge jumps out of the groove
I don't have numbers for 'abrasion'. Just a suggestion that we LISTEN to the problem. About all we can really say is that it is different.Clipping has been a non-issue since +/-15V rails & 5532s have been available (see eg Tomlinson) ... It's only da passive RIAA pundits who might run foul of it.
Last edited:
Maybe your not familiar with the Omnidisk, Telarc extracted the cannons and recorded it four times in 2dB steps over re: 5cm/sec. Never had a problem with the 0dB track (the way it was released) but the last one was problematic.
Why, please do so. But not in terms of World Peace and sex life improvements.
If you prefer to tune the Duraglit resistors for each pair of transistors beta, and spend time + money matching transistors for beta and Vbe, then I'm sure the sex life will be affected. Simple and simplistic, designing and cobbling, do not overlap.
You are worrying too much about matching requirements for the Duraglit.
In fact, no matching at all is required. I simmed this by giving the transistors Hfe of 100 and 400 for the other one. In a second sim I changed one base resitor from 12 k to 6k. Still works fine, even distortion is not changing significantly.
What changes is the following:
Vce per transistor is shifting by 100 to 200 mV max from 750/750mV (everything matched) to 550/950 mV worst case.
Why so? The Duraglit PS is floating and there is only one single path for DC current through the transistors. The input (emitters of both BJTs) stays at zero volt. The PS is automatically leveling to achieve balance. In case of the 1.5 V supply, the voltage across the base resistors is approx 180 - 200mV and this is also the max. voltage unbalance for the collector emitter voltages. Worst case, even a drastic mismatch only reduces the max output voltage from 500mV to 300mV - still some headroom left.
As soon as the power supply is splitted into a bipolar ground referenced one, all such auto leveling is gone. As there is now a DC pathway through the cartridge the balancing still works for the transistors but the input drifts away from ground and a substantial current can flow - worst case the possible short circuit current of the power supply.
Not sure if the loss in dynamic range due to the beta/Vbe mismatch is important, but I know sure as hell that a design with beta dependent currents is not my cup of tea. One carefully determines the optimum collector current for a certain MC cartridge, only to realize he has to tweak the base bias resistors, to start with, or each time the transistors are replaced (which, for the ZTX devices, it is to be expected; I've already mentioned I observed a rather strong noise performance variation (in particular 1/f noise), strongly correlated (rather expected) with the beta variation). From a bag of ten, I found only two pairs of ZTX devices that match the noise expectation of <2ohm RB and noise corner frequency under 100Hz. If you think this kind of tweaking is the right way to reach World Peace and your sex life improvement, then who am I to debate?
Now if cost is really a concern in #375, replace Q101/Q103 with two 1N914 diodes, find once and forever a value for R103 that defines your desired current through the gain transistors, then you can go and exercise your (improved or not) sex life, you are otherwise done.
I'm suspecting that other than some possible ego issues, Richard's displeasure for anything that is less simplistic than the original Duraglit is related to a higher difficulty to cobble it in dead bug position, and then stuffing it in a Duraglit can. Other than that, I can't imagine why one would prefer something that doesn't offers beta immunity, for the price of two diodes and three resistors. Ah, and a 0.5ohm 80pV/rtHz noise penalty, someone must be joking here
.But the set up of schematic in # 375 is floating as well ?
It can at least be easily rearranged by splitting R 103 in
two and grounding the midpoint, like in "Hawking/Duraglit
Special" schematic.
Just when I was in the kitchen JG called me to tell that 08 wrote
"now that would be a bad idea" - and he was laughing loud about
this reply. You can not even calmly prepare food "without being hit
on the head by old men" he said (we are semi-old ourselves).
So what makes you think this is "a bad idea" ? Both variants work
the same way ac wise or nearly so.
The only difference between your circuit in post 375 and the other
one is the addition of two trans-diodes Q101, Q103 and some emitter
resistors in yours.
"now that would be a bad idea" - and he was laughing loud about
this reply. You can not even calmly prepare food "without being hit
on the head by old men" he said (we are semi-old ourselves).
So what makes you think this is "a bad idea" ? Both variants work
the same way ac wise or nearly so.
The only difference between your circuit in post 375 and the other
one is the addition of two trans-diodes Q101, Q103 and some emitter
resistors in yours.
As aboos has shown, even a wildly different hfe pair is more than acceptable for THD bla bla .. You'll note in my original document, I don't even mention hfe/Vbe matching in my table of base resistors cos it's unnecessary.Not sure if the loss in dynamic range due to the beta/Vbe mismatch is important, but I know sure as hell that a design with beta dependent currents is not my cup of tea. One carefully determines the optimum collector current for a certain MC cartridge, only to realize he has to tweak the base bias resistors, to start with, or each time the transistors are replaced (which, for the ZTX devices, it is to be expected; I've already mentioned I observed a rather strong noise performance variation (in particular 1/f noise), strongly correlated (rather expected) with the beta variation). From a bag of ten, I found only two pairs of ZTX devices that match the noise expectation of <2ohm RB and noise corner frequency under 100Hz. If you think this kind of tweaking is the right way to reach World Peace and your sex life improvement, then who am I to debate?
....
I'm suspecting that other than some possible ego issues, Richard's displeasure for anything that is less simplistic than the original Duraglit is related to a higher difficulty to cobble it in dead bug position, and then stuffing it in a Duraglit can. Other than that, I can't imagine why one would prefer something that doesn't offers beta immunity, for the price of two diodes and three resistors. Ah, and a 0.5ohm 80pV/rtHz noise penalty, someone must be joking here
One real reason for this immunity was revealed by Guru Hans who showed Duraglit was equivalent to a wonky virtual earth. ie its noise performance is CE rather than CB.
I don't have an aversion to extra complexity/cost .. but usually require this gives better, not worse performance.
But thanks syn08, for posting your experience with ZTX851/951 which is so different from Wayne and H&H. Perhaps your circuit is more sensitive. Please send Bill your prototypes as I'm sure he'll find time to compare your masterpiece to a genuine Duraglit.
Because then the most important feature of Syn08`s circuit is gone. There is then a pathway for DC from the emitter of one of the input BJTs through the cartridge to the emitter of the opposite current mirror BJT and this will spoil the rock solid Ic stabilisation.
A few posts above I clearly stated the purpose and scope of this configuration. It has nothing to do with AC performance, I though that was clear. Obviously it was not.
The purpose is to make the bias as much as possible beta independent and balance as much as possible the DC.
Splitting R103 would do nothing, other than adding another resistor (and potentially damaging the DC balance). Why would one do that, only JG would know.
I don't have an aversion to extra complexity/cost .. but usually require this gives better, not worse performance.
Understood. So in your view having a reproducible and stable bias point is not a performance criteria, or at least it is less important than adding 0.08nV/rtHz of noise.
What can I say?
Syn08,
your circuit has excellent bias stability and is almost immune to beta changes and differences between the 4 BJTs. Noisewise the 0.5R penalty is purely academic - differences in noise between individual transistors even from the same batch will be bigger. And cost is no argument at all as we are speaking here about a few cents.
The only drawback of your circuit in my view is gain. It is limited due to the low 150R base resistors. With 0R source resistance one gets about 20 x gain. With a 10R cart however gain drops to 5...6 fold which is a little bit too low. For curiosity I simmed your proposal to use different transistors or even simple diodes for the mirrors. For some common standard transistors (2N2222 like) one can increase the base resistors into the kOhm range with no noise penalty. Bias points stay stable for different ZTX betas. Gain now is determined by the output resistor and can be set in a wide range again.
Maybe you can check with your boards if this is real.
your circuit has excellent bias stability and is almost immune to beta changes and differences between the 4 BJTs. Noisewise the 0.5R penalty is purely academic - differences in noise between individual transistors even from the same batch will be bigger. And cost is no argument at all as we are speaking here about a few cents.
The only drawback of your circuit in my view is gain. It is limited due to the low 150R base resistors. With 0R source resistance one gets about 20 x gain. With a 10R cart however gain drops to 5...6 fold which is a little bit too low. For curiosity I simmed your proposal to use different transistors or even simple diodes for the mirrors. For some common standard transistors (2N2222 like) one can increase the base resistors into the kOhm range with no noise penalty. Bias points stay stable for different ZTX betas. Gain now is determined by the output resistor and can be set in a wide range again.
Maybe you can check with your boards if this is real.
True.
What I found really, really interesting in this circuit topology is that in a first approximation the gain at Rs=0 doesn't NOT depend on anything but supply voltage. That is because R100/R106 resistors define the collector current (hence the transconductance) and are also the gain stage load. The gain is Go=(Vcc-2Vbe)/(2*Vt) where Vt=kT/q. For Vcc=2.7V, gain @ Rs=0 is about 28dB. I mentioned this in a previous post in this thread.
However, at Rs>0, Rs acts as a degeneration, and for a good range of Rs values (the cartridge resistance) the gain can be approximated (R100/2)/Rs, this time independent of Ic and the supply voltage.
From where I sit, by feeding the circuit using a solar cell, and optimizing for my Benz Micro Gold (0.4mV, Rs=12ohm) I get R100=R106=270ohm and a resulting gain of about 7 (the solar cell has tens of ohms internal resistance which lowers the gain). This is exactly what one would need to feed the output in an MM input (with the RIAA correction. Which is what I did for listening to the result.
P.S. For what is worth, the gain for RS>0 is analitically G=Go/(1+2*Go*Rs/R100) where Go is the above gain at Rs=0. Everything can easily be verified by simulation. Obviously, the larger Rs, the better the degeneration gain approximation.
Last edited:
This is possible the most perturbing post I have seen in a while
Purely personally I think this thread stands out like a refreshing craft beer in a sea of Coors light (other cr*p tinned plastic beers are available). Give or take the odd pages of wibble that always happen a number of solutions are presented, analysed and a few built. Someone finding this thread in the future can read through and have informed information to make a decision on if they want to go hair shirt or civilised.
@Richard: You have to bear in mind I've cut my project backlog from 22 years to 4.5 but progress is still slow
I do not completely agree with your sense of humour.This is possible the most perturbing post I have seen in a while
This is possible the most perturbing post I have seen in a while
Fortunately I recalled the stupid debate starting in #660 and decided it's not worth another one of the same kind.