Hi,
Then lets urgently stop talking about these critters, others might get ideas.
Anyway, Scott Wurcer and John Curl have already both declared the LH0032 to be a poorely performing antique, old junk so to speak, so surely it cannot be any good, can it now...
So lets not talk about old rubbish, on with the new N-Jfet 10 Ohm equivalent Noise Design (sorry, cannot show my one, it's a commercial implementation).
Ciao T
It's a NS
Picture came from here:
( ￣?￣)*???: ?????！LH0032
See also:
http://www.diyaudio.com/forums/anal...rch-preamplifier-part-ii-682.html#post2373150
Then lets urgently stop talking about these critters, others might get ideas.
Anyway, Scott Wurcer and John Curl have already both declared the LH0032 to be a poorely performing antique, old junk so to speak, so surely it cannot be any good, can it now...
So lets not talk about old rubbish, on with the new N-Jfet 10 Ohm equivalent Noise Design (sorry, cannot show my one, it's a commercial implementation).
Ciao T
Now trouble begins..or my limited knowledge come into play....
Nice
Not so hard. If you only want the have the 75uS pole in the first stage just calculate a cap to go over the 4.7k and simulate it. Adjust it (R or C) until you get it right. Remember to add the input capacitance of the following stage,or better yet, add the stage to the simulation. A good place to start is 15nF and trim R by parallelling a 5.6k with something until you get it right. With my tube second stage i am right on the money with 5.6k//47k//15n.
Why not place the complete riia network in there and just run the second stage flat and open loop?
BR,
Anders
Last edited:
Hi,
Well, it looks like you are running the 2SK170 at maybe 3mA. Looking at the datasheet it seems you get 0.95nV|/Hz per 2SK170. As you use in a differential configuration this means around 1.35nV|/Hz Ein.
For reference, the same pair of 2SK170 at the same current give 0.67nV|/Hz. So given the same amount of devices balanced is actually 6dB noisier than SE.
With respect to an 0.5mV pickup I would estimate maybe 75dBA SNR, maybe a little less due to the increased noise at LF. So for 0.5mV pickups this noise is okay, much lower outputs and the SNR becomes marginal. Really quiet vinyl gets close to -70dB if you ignore the pop's and clicks.
Now output/gain/EQ. The two 2SK170 at 3mA each have around 22mA/V transconductance. As they are effectively in series WRT to the input signal they deliver this overall 22mA to the folded cascodes.
We loose around 10% due the low value load resistors compared to the folded cascode transistors emitter impedance, so we deliver around 20mA/V into the load. With a 4.7K load we get indeed around 40dB Gain.
If we use a pure transconductance RIAA to get MC gain in a single stage we need to get around 80dB DC Gain, so 10,000/20mA means a 510K Resistor on the output. This in the general vicinity of the collector impedance of the folded cascode, so this resistor will need trimming in reality...
We can now use the RIAA Calculator at KAB USA:
http://www.kabusa.com/riaa.htm
This suggests 75K, 4.3nF & 1.5nF for the EQ, with an extra shunt likely in the region of one to several MOhm to get the lowend right.
I would suggest limiting things to 40dB Gain at 1KHz for the first stage may be a better choice. In this case 44nF (22nF * 2 in parallel) and 15nF would give a close RIAA with a 7.27KOhm resistor (6.8K in series with 470Ohm).
The Shunt resistor that determines the LF gain would be around 51KOhm, DC gain would be 60dB.
A further stage with around 20dB gain would be needed for MC, the circuit has too much input capacitance for MM as is. I guess one could use a linestage fpr 20dB Gain and then pad down other line level sources by 6-12dB to get sensible gain for these.
Ciao T
4.7 Kohm is app 40 db...of gain...easy to ajust..with the resistor to gnd...
still i can't anticipate the noise of the circuit...
Well, it looks like you are running the 2SK170 at maybe 3mA. Looking at the datasheet it seems you get 0.95nV|/Hz per 2SK170. As you use in a differential configuration this means around 1.35nV|/Hz Ein.
For reference, the same pair of 2SK170 at the same current give 0.67nV|/Hz. So given the same amount of devices balanced is actually 6dB noisier than SE.
With respect to an 0.5mV pickup I would estimate maybe 75dBA SNR, maybe a little less due to the increased noise at LF. So for 0.5mV pickups this noise is okay, much lower outputs and the SNR becomes marginal. Really quiet vinyl gets close to -70dB if you ignore the pop's and clicks.
Now output/gain/EQ. The two 2SK170 at 3mA each have around 22mA/V transconductance. As they are effectively in series WRT to the input signal they deliver this overall 22mA to the folded cascodes.
We loose around 10% due the low value load resistors compared to the folded cascode transistors emitter impedance, so we deliver around 20mA/V into the load. With a 4.7K load we get indeed around 40dB Gain.
If we use a pure transconductance RIAA to get MC gain in a single stage we need to get around 80dB DC Gain, so 10,000/20mA means a 510K Resistor on the output. This in the general vicinity of the collector impedance of the folded cascode, so this resistor will need trimming in reality...
We can now use the RIAA Calculator at KAB USA:
http://www.kabusa.com/riaa.htm
This suggests 75K, 4.3nF & 1.5nF for the EQ, with an extra shunt likely in the region of one to several MOhm to get the lowend right.
I would suggest limiting things to 40dB Gain at 1KHz for the first stage may be a better choice. In this case 44nF (22nF * 2 in parallel) and 15nF would give a close RIAA with a 7.27KOhm resistor (6.8K in series with 470Ohm).
The Shunt resistor that determines the LF gain would be around 51KOhm, DC gain would be 60dB.
A further stage with around 20dB gain would be needed for MC, the circuit has too much input capacitance for MM as is. I guess one could use a linestage fpr 20dB Gain and then pad down other line level sources by 6-12dB to get sensible gain for these.
Ciao T
Torsten..
To lower noise you can run 2 sets of fets..(here they run at app 5.5 mA).. would require a different CCS though..
I plan to make a design a riaa around a similar second stage...but I need a good output buffer to that stage to make it into voltage source...have a design with a single ended BJT which seems to work Ok...But not quite confident it's the best solution..
pls take a look....and feel free to comment
To lower noise you can run 2 sets of fets..(here they run at app 5.5 mA).. would require a different CCS though..
I plan to make a design a riaa around a similar second stage...but I need a good output buffer to that stage to make it into voltage source...have a design with a single ended BJT which seems to work Ok...But not quite confident it's the best solution..
pls take a look....and feel free to comment
Last edited:
Front end ... modified to be non feedback transconductance
Would like to have the high frequency (75 us) filter passively in front of the second gain block where I'll the insert the base shelf (318-3180us) actively in the feedback loop..
Now trouble begins..or my limited knowledge come into play....
As the Amplification drops with the reduced impedance..The two things are pulling the same direction.. and..difficulty is with the 75 us pole....so so difficult to get right on the money
Calculations anyone..??
Performance still good.. lowish distortion and nice distribution...no harmonics higher an 3.th order and that is already below -80 dB
Hi MiiB,
You're moving in the right direction. No NFB in the first stage is fine since the signal levels are so low. You may have to trim gain of the two channels, however, due to slight differences in the open-loop gain of the first stages.
Consider doing the 75us with a shunt capacitance across the 4.7k output resistor; this way the fast-edge stuff from ticks never makes it into the voltage domain with high amplitude.
Have you simulated this circuit for noise and PSRR?
Watch those low-value emitter resistors and the possible noise contributions of the biasing LEDs. Give up a little voltage swing headroom if you need to. Watch out for PSRR degradation due to the rail-to-rail current mirror output transistor. Check carefully on how much DC offset you get with expected worst-case matching between the two input JFETs.
Cheers,
Bob
The H suffix was for National's hybrid product line. They were never monolithic and at least 30yr. old (pre-SMT). At CMRR, PSRR and AOL of as low as 300-500 and thermal tails up the wazoo (settling only speced to .1%/10bit) of course this amp will compromise the measured performance of any modern audio DAC. But it sounds soooo good, that's the point right? Why do the ESS guys bother?
I take it that means you are trying to discourage others as you are working in secret to develop a modern version!
If anyone has measurements, I suspect it has nice even order distortion.
Hi,
You would need 4pcs per polarity to get even with just 2pcs in SE. And 8pcs in SE...
I can see the benefit in terms of thermal drift over SE circuits of course, though BJT's are easy to compensate and we can bias the K170 a little positive to get the Id towards the Zero Tempco point.
I would keep it all passive or all active, it is easier to get a really coherent sound without mucking about and fine tuning endlessly.
Doing an all-in-one shunt/transconductance RIAA is probably a good idea in the context.
The second stage could probably remain open-loop using source degeneration for the input pair to lower gain to the needed level.
Use a Mosfet Follower (single ended) with a low Crss Mosfet. The gate impedance will be sky high, Ciss will be bootstrapped and largely disappears, so Crss will create the dominant pole in your circuit (unless your cascode transistors have a higher Ccob than the Mosfet's Crss).
If you had a fully balanced circuit you could probably neutralise Crss (and Ccob) to a large degree as well, then bandwidth (open loop) starts tending towards "DC2Light".
You could also improve the folded cascode transistors by applying a Hawkesford cascode to them, then the output impedance of the cascode goes way up and so does you OLG, control the follower mosfets capacitance and you will have ton's of OLG with tons of Bandwidth...
Probably enough to do single OPA active RIAA EQ for MC in "single stage".
This is now sliding towards Connoisseur / Jonathan Carr like design, instead of something that derives John Curl.
Ciao T
You would need 4pcs per polarity to get even with just 2pcs in SE. And 8pcs in SE...
I can see the benefit in terms of thermal drift over SE circuits of course, though BJT's are easy to compensate and we can bias the K170 a little positive to get the Id towards the Zero Tempco point.
I would keep it all passive or all active, it is easier to get a really coherent sound without mucking about and fine tuning endlessly.
Doing an all-in-one shunt/transconductance RIAA is probably a good idea in the context.
The second stage could probably remain open-loop using source degeneration for the input pair to lower gain to the needed level.
Use a Mosfet Follower (single ended) with a low Crss Mosfet. The gate impedance will be sky high, Ciss will be bootstrapped and largely disappears, so Crss will create the dominant pole in your circuit (unless your cascode transistors have a higher Ccob than the Mosfet's Crss).
If you had a fully balanced circuit you could probably neutralise Crss (and Ccob) to a large degree as well, then bandwidth (open loop) starts tending towards "DC2Light".
You could also improve the folded cascode transistors by applying a Hawkesford cascode to them, then the output impedance of the cascode goes way up and so does you OLG, control the follower mosfets capacitance and you will have ton's of OLG with tons of Bandwidth...
Probably enough to do single OPA active RIAA EQ for MC in "single stage".
This is now sliding towards Connoisseur / Jonathan Carr like design, instead of something that derives John Curl.
Ciao T
I take it that means you are trying to discourage others as you are working in secret to develop a modern version!
If anyone has measurements, I suspect it has nice even order distortion.
We have requests for 1A out modern op-amps. The packaging
guys don't want to support the heatsink packages. Some of our 100mA out early CB amps were the magic sauce for at least one CD maker. Even came and asked for private brands.
So be careful which products you slag, you might be unknowingly praising them elsewhere.
Last edited:
Let us be fair about the LH0032. It is quite possible that I COPIED the topology, adding complementary output fets, instead of bipolars, back in 1973. There are a limited number of optimum approaches and this is a pretty good one, ESPECIALLY when you need significant gain and have only N channel jfets to work with. However, MY version of this design was about 25 times quieter, guaranteed class A, and reasonably high peak current.
The LH0032 was NOT designed for audio, but for speed and other factors. It is NOISY! It 'might' have Xover distortion. Has anybody measured it carefully?
However, it has a VERY HIGH open loop bandwidth, and this helps it work properly as a video amplifier, and this will make it sound good too, in my opinion. Low differential phase distortion.
The LH0032 was NOT designed for audio, but for speed and other factors. It is NOISY! It 'might' have Xover distortion. Has anybody measured it carefully?
However, it has a VERY HIGH open loop bandwidth, and this helps it work properly as a video amplifier, and this will make it sound good too, in my opinion. Low differential phase distortion.
Have you ever seen slag? How about slag being dumped? (Very pretty sight!)
I was not slagging the product, I think you know my stand Picasso vs Rembrandt. Both are great, depends on your taste.
CB? Citizen Band? Common Base? Cricket Bat? Common Bull...?
ES
Yes slag from the drop forges was our "clean fill". Beautiful sound too during third shift in the winter carried miles. CB = complimentary bipolar (not FETs).
I have worked the circuit a bit...put in hawksford cascodes that improved distortion quite a bit...made the full Monty Riaa compensation in oneshot...and added a single ended fet follower..Would loveee to use a better FET like the zetex zvn3310 or similar...but can't get the model working in LT-spice.
Gain is 42 db a 1KHz
Distortion when run full gain is low -85dB 2.order and -100 db 3. and no higher order....
not enough gain for low MC-s but it could be possible to add a buffer with gain..
Gain is 42 db a 1KHz
Distortion when run full gain is low -85dB 2.order and -100 db 3. and no higher order....
not enough gain for low MC-s but it could be possible to add a buffer with gain..
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.