The ripples are IMO not so crazy as they might first seem for a couple of reasons:
a) When two adjacent bands are boosted or cut by the same amount summation would probably be flatter with the ripples than without.
b) Having a bit of countersteer beside the band would make its effect more audible by contrast.
Today I learned Dick Burwen is a clever fella.
a) When two adjacent bands are boosted or cut by the same amount summation would probably be flatter with the ripples than without.
b) Having a bit of countersteer beside the band would make its effect more audible by contrast.
Today I learned Dick Burwen is a clever fella.
I've got back on this topic a bit recently, looking at the Palette Preamp.
Given that the schematics were posted here way back in this long and winding thread, I inputted the filter chain (U10, 11, 4, 6 7, 9, 1, 2, 5 and 8), numbering from the block diagram.
The downside is that it is noisy. The output noise from that chain is 19uV with an SNR of 94dB in 20kHz BW.
The original design used OPA627 opamps, and in selected areas the opa633 current booster (the filter drive amp uses one). For some reason the OPA627, still available, is eye poppingly expensive.
I used the latest opamp in that generation, the OPA828, in the model. Still expensive, but a fraction of the OPA627. The current booster OPA633 is obsolete, so I used the BUF634 model instead. It is that combination that gives rise to 19uV. The main cause of this is the filter blocks, where the resistors have relatively high values. Each block has around 7uV noise from each. There are 4 summed in parallel, so the effective noise is 2 x 7 = 14uV. Then some additional noise from the summer, and the two series filter blocks that follow and you end up with 19uV.
Now I tried to reduce this by using another latest kid on the block - the OPA1611 - with a voltage noise of a mere 1.1nV/root Hz. It was a noise disaster. Reason is that with the high impedance levels in the filter blocks, current noise is really important. The OPA1611 is bipolar with higher current noise, whereas the OPA828 (and OPA627) are FET input and have very low current noise.
Given that the schematics were posted here way back in this long and winding thread, I inputted the filter chain (U10, 11, 4, 6 7, 9, 1, 2, 5 and 8), numbering from the block diagram.
The downside is that it is noisy. The output noise from that chain is 19uV with an SNR of 94dB in 20kHz BW.
The original design used OPA627 opamps, and in selected areas the opa633 current booster (the filter drive amp uses one). For some reason the OPA627, still available, is eye poppingly expensive.
I used the latest opamp in that generation, the OPA828, in the model. Still expensive, but a fraction of the OPA627. The current booster OPA633 is obsolete, so I used the BUF634 model instead. It is that combination that gives rise to 19uV. The main cause of this is the filter blocks, where the resistors have relatively high values. Each block has around 7uV noise from each. There are 4 summed in parallel, so the effective noise is 2 x 7 = 14uV. Then some additional noise from the summer, and the two series filter blocks that follow and you end up with 19uV.
Now I tried to reduce this by using another latest kid on the block - the OPA1611 - with a voltage noise of a mere 1.1nV/root Hz. It was a noise disaster. Reason is that with the high impedance levels in the filter blocks, current noise is really important. The OPA1611 is bipolar with higher current noise, whereas the OPA828 (and OPA627) are FET input and have very low current noise.
Last edited:
By contrast, I looked in detail at Burwen's original Audio Palette design. That uses four filter bands (15Hz, 500Hz, 2kHz and 25kHz) with the usual bell shape, and followed that with two shelving filters.
Because impedance levels in the filter blocks are low, the OPA1611 really shines. Burwen used what was available in 1984 - the still excellent NE5532A - but much noisier than the much more recent OPA1611 (or the dual OPA1612), introduced in 2010.
The output from the 4-band positive summer is a superb 1.7uV with an SNR of 115dB. And after the shelving filters, 2.3uV and an SNR of 113dB. That is quiet, really quiet.
Sure the bell shaped characteristics are not ideal, with out of band artefacts, but that is a small price to pay for the quietness.
Because impedance levels in the filter blocks are low, the OPA1611 really shines. Burwen used what was available in 1984 - the still excellent NE5532A - but much noisier than the much more recent OPA1611 (or the dual OPA1612), introduced in 2010.
The output from the 4-band positive summer is a superb 1.7uV with an SNR of 115dB. And after the shelving filters, 2.3uV and an SNR of 113dB. That is quiet, really quiet.
Sure the bell shaped characteristics are not ideal, with out of band artefacts, but that is a small price to pay for the quietness.
Yes. Pick your poinson!
I wonder, with the newer opamps, do you really need the current booster?
Jan
I wonder, with the newer opamps, do you really need the current booster?
Jan
No you don't Jan. Burwen got away with NE5532A's throughout. He kept impedance levels at 1k ohms and above to prevent rising distortion with lower load resistors (although the data sheet says it will drive 600 ohms). Even modern opamps, like the OPA1611 exhibit this effect, particularly if driven close to their rails, which is why distortion graphs in datasheets are at modest levels like 3V.
Guys,
Thanks for your inputs.
Have lived with and used the (opamp) Palette pre (on a 10m umbilical so I can "play" from the listening chair - including in/out switching and gain trim) for years and have NEVER HEARD any problems related to noise, including substituting LME49710's for the 627's on the original boards. Customers who got the other two builds never noticed/said anything either. For potential lower noise, thought about but never tried to implement a parallel (vs serial) filter scheme (a la National Semi) but not experiencing any problems made me leave it alone. Only "theoretical" problem was the accumulation of output offset voltages - being direct-coupled. But again....
Now have a stash or 40 627's. Need any??
FWIW and jollies, look up the Charter Oak PEQ-1
Thanks for your inputs.
Have lived with and used the (opamp) Palette pre (on a 10m umbilical so I can "play" from the listening chair - including in/out switching and gain trim) for years and have NEVER HEARD any problems related to noise, including substituting LME49710's for the 627's on the original boards. Customers who got the other two builds never noticed/said anything either. For potential lower noise, thought about but never tried to implement a parallel (vs serial) filter scheme (a la National Semi) but not experiencing any problems made me leave it alone. Only "theoretical" problem was the accumulation of output offset voltages - being direct-coupled. But again....
Now have a stash or 40 627's. Need any??
FWIW and jollies, look up the Charter Oak PEQ-1
Last edited:
Charles, I read the user manual at the Thomann website.
They (Charter Oak) made some very sensible decisions for that design.
Were you involved?
Jan
They (Charter Oak) made some very sensible decisions for that design.
Were you involved?
Jan
The switchable frequencies on the PEQ-1 are interesting, as is the balanced I/O transformers. The filter topology must be interesting, because usually you can't change band frequency easily without changing the filter Q.
Unless they are using LC filters like Thorsten has championed. Then you could switch frequencies just by switching in a different L/C/R module without messing with the Q.
Anyone notice that the illustration of the band shapes in the manual is identical in all regards to the Stereophile review of the Palette Preamplifier?
Craig
Unless they are using LC filters like Thorsten has championed. Then you could switch frequencies just by switching in a different L/C/R module without messing with the Q.
Anyone notice that the illustration of the band shapes in the manual is identical in all regards to the Stereophile review of the Palette Preamplifier?
Craig
Last edited:
I had considered switchable frequencies on an upcoming Palette variant so it's nice to know other people have the same brain damage 🤪
The filters are most likely either gyrators or state-variable bandpass since a 4pdt switch with centre off can switch two capacitors per channel, which is all you need to change the frequency of either filter topology. Centre positions choosing the highest frequencies is a big ol' hint in that direction.
The filters are most likely either gyrators or state-variable bandpass since a 4pdt switch with centre off can switch two capacitors per channel, which is all you need to change the frequency of either filter topology. Centre positions choosing the highest frequencies is a big ol' hint in that direction.