audio-talk

Thanks but it more likely a semi retirement build for 16 months time when I get state pension but keep some days of work to make feasible paying the bills in a big house until we downsize but have some days off. Maybe 3/5ths contract. Right now it’s 5 arduous days with travel over an hour each way, then I fit in the gym .one or two weekdays after work buy food cook for all. So all projects are stalled.

Oh dear - so having tested out my reverse RIAA on the Whest to prove its accuracy... here's what the response of my build of the Millett stage looks like:



That's rising from -14dB at 18Hz to -6dB at around 500Hz. Then we drop off again until we reach -16dB at 20kHz.

No wonder I thought it sounded a bit down on treble. That's a bloody huge amount down on treble. And low bass.

That is with the reverse-RIAA circuit feeding the MC step-ups as they are hard-wired in - could there be some interaction going on there? Or is that likely to be the actual frequency response?

I have to say, on that Sunday morning at Owston, going from that other stage to the Millett it did sound like there was a good 3dB+ drop in the high treble to my ears - what do you think Ali as you commented on it too?

I built it with all the parts exactly as specced - how on earth can it be so far out???

Jeepers.

You should be able to see that response error without the inverse riaa.

I haven't looked but don't those inverse EQ thingies have to be loaded or should I say not loaded down? Maybe a tx input presents too low a Z for it?

Cheers,
Stephen

Thermionic Idler wrote: ↑Tue Jan 31, 2023 7:31 pm Oh dear - so having tested out my reverse RIAA on the Whest to prove its accuracy... here's what the response of my build of the Millett stage looks like:



That's rising from -14dB at 18Hz to -6dB at around 500Hz. Then we drop off again until we reach -16dB at 20kHz.

No wonder I thought it sounded a bit down on treble. That's a bloody huge amount down on treble. And low bass.

That is with the reverse-RIAA circuit feeding the MC step-ups as they are hard-wired in - could there be some interaction going on there? Or is that likely to be the actual frequency response?

I have to say, on that Sunday morning at Owston, going from that other stage to the Millett it did sound like there was a good 3dB+ drop in the high treble to my ears - what do you think Ali as you commented on it too?

I built it with all the parts exactly as specced - how on earth can it be so far out???

Jeepers.

Yes I did think it was a bit down, but then I’ve always been more sensitive to the higher frequencies. Less so now I’m getting older..

izzy wizzy wrote: ↑Wed Feb 01, 2023 4:24 am I haven't looked but don't those inverse EQ thingies have to be loaded or should I say not loaded down? Maybe a tx input presents too low a Z for it?

Cheers,
Stephen

I did wonder about that myself, mine is currently on 150 ohms.

The issue here is at the frequency extremes. We know the RIAA curve should be +20dB at 20Hz, and -20dB at 20kHz. As Nick says, it ought to be possible to see that scale of deviation without the reverse RIAA in circuit, so I think I'll clip a couple of test leads to the secondaries straight from the AD2, bypassing the SUT and the reverse RIAA. If that gives me a similar result then I know the test method is not giving me misleading results.

I'm still mystified as to what this could be - component tolerances shouldn't throw it that far out. I remember triple checking the orientation of those inductors, all the components are exactly as Pete Millett specified. Curve is about the same on both channels so unlikely to be a faulty component or soldering error. The sole deviation is the insertion of a 100 ohm resistor between the PCB ground plane and Pin 3 of the output XLR sockets to pseudo-balance the output and make it (sort of) AES48 compliant in line with what I'm doing with the rest of the system.

I can only think those inductors are not quite accurate and may be throwing things off. I've got the circuit modelled in Spice so next stage is to play with different values and see if I can replicate the result. Then see what I need to change out to correct it.

Ali Tait wrote: ↑Wed Feb 01, 2023 5:23 am Yes I did think it was a bit down, but then I’ve always been more sensitive to the higher frequencies. Less so now I’m getting older..

I hear up to about 14Khz or so, at that point it's -6dB down from the middle, which seems about in line with what I heard.

Still not been able to replicate it in Spice - measuring the response on the SUT secondaries suggests the reverse-RIAA is OK with the MC loading of 150R (plus I think the Whest is set to 470R and the response was bang on the money). Actually now I've typed that, I should adjust the load down to 50R on the Whest and rerun the measurement to make sure.

I'll take response measurements at various points throughout the circuit which might help pin things down. Not sure when that'll happen, I need to dedicate this weekend to moving the linestage project forward - the big hold-up is waiting for a 2x15V transformer to show up from Toroidy, but I want to get everything else done as far as I can.

Been reading up about the Salas Valve Itch phono stage as that seems to be extremely well regarded and simple to make...

Thermionic Idler wrote: ↑Fri Feb 03, 2023 6:51 pm Still not been able to replicate it in Spice - measuring the response on the SUT secondaries suggests the reverse-RIAA is OK with the MC loading of 150R (plus I think the Whest is set to 470R and the response was bang on the money). Actually now I've typed that, I should adjust the load down to 50R on the Whest and rerun the measurement to make sure.

I'll take response measurements at various points throughout the circuit which might help pin things down. Not sure when that'll happen, I need to dedicate this weekend to moving the linestage project forward - the big hold-up is waiting for a 2x15V transformer to show up from Toroidy, but I want to get everything else done as far as I can.

Been reading up about the Salas Valve Itch phono stage as that seems to be extremely well regarded and simple to make...

I would encourage you to keep checking everything over. Maybe leave it for a bit as looking at the same thing for ages blind you to deviations.

If you can bypass the EQ; lift L1 and wire direct past the EQ and bypass C5, then the unit should measure super flat and have the right gain. It it does, then you know the issue is in the EQ config. If not, seperate the stages and measure each one.

IMO, something like the Itch can be bettered. I've played a lot with mu follower circuits and ultimately, they're not the wonder circuit or free lunch as claimed by many. There's a reason many settle on a hi gm valve in the front with a fairly hi gm valve following. Nothing fancy at all. Maybe a CCS load some place if you want. A 20k nominal EQ in between. That's pretty much Mark's RIAA, mine is similar. There's many builds of similar on HiFiHaven.

Anyhoo, I hope you get the Millet sorted.

Cheer,s
Stephen

izzy wizzy wrote: ↑Sat Feb 04, 2023 4:47 am

Thermionic Idler wrote: ↑Fri Feb 03, 2023 6:51 pm Still not been able to replicate it in Spice - measuring the response on the SUT secondaries suggests the reverse-RIAA is OK with the MC loading of 150R (plus I think the Whest is set to 470R and the response was bang on the money). Actually now I've typed that, I should adjust the load down to 50R on the Whest and rerun the measurement to make sure.

I'll take response measurements at various points throughout the circuit which might help pin things down. Not sure when that'll happen, I need to dedicate this weekend to moving the linestage project forward - the big hold-up is waiting for a 2x15V transformer to show up from Toroidy, but I want to get everything else done as far as I can.

Been reading up about the Salas Valve Itch phono stage as that seems to be extremely well regarded and simple to make...

I would encourage you to keep checking everything over. Maybe leave it for a bit as looking at the same thing for ages blind you to deviations.

If you can bypass the EQ; lift L1 and wire direct past the EQ and bypass C5, then the unit should measure super flat and have the right gain. It it does, then you know the issue is in the EQ config. If not, seperate the stages and measure each one.

IMO, something like the Itch can be bettered. I've played a lot with mu follower circuits and ultimately, they're not the wonder circuit or free lunch as claimed by many. There's a reason many settle on a hi gm valve in the front with a fairly hi gm valve following. Nothing fancy at all. Maybe a CCS load some place if you want. A 20k nominal EQ in between. That's pretty much Mark's RIAA, mine is similar. There's many builds of similar on HiFiHaven.

Anyhoo, I hope you get the Millet sorted.

Cheer,s
Stephen

Thanks Stephen - I was studying your circuit this morning, I rather like the look of it.

Well, I've managed to get the response a lot flatter than before. I got my Spice sim of the circuit up and started to experiment to see if dropping any of the resistor values would act to bring up the missing top end. Practically, this would mean I'd only need to solder an additional resistor in parallel without having to desolder anything.

Lowering the value of the resistor in parallel with the second, smaller, inductor seemed to do the trick according to the sim (R16/R20 on the circuit diagram here). So I got the phonostage onto the bench, hooked up the Analog Discovery and started doing a series of sweep tests with various values out of my resistor kit wound around R20. I started with 15k but it was all the way down at 5.1k before I got the same response at 10kHz as at 200Hz.

I forgot to take a screenshot of the response but it rolls up to 0dB up to about 1khz, then there's a gentle -1dB dip in the presence region up to about 8kHz before it curves up to 0dB at 10Khz, falling to -1dB at 20Khz and then further off, with a slight hump at 40Khz or so. I checked it from 5Hz to 10Mhz and ran a square wave to ensure there were no nasties created by my mucking about with R20, then soldered on the parallel resistors.

I could probably get it a bit flatter if I increased the value of R17/19 slightly and dropped where I have R16/20 now, but that would require more invasive surgery. The bass still rolls off a little early but I'm less concerned about that, the -3dB point is at 35Hz or so.

I'm listening to it now and it's absolutely wonderful, that 'rightness' of sound that comes with inductive RIAA, so the change doesn't seem to have introduced any collateral problems. What mystifies me is why I had to change that resistor to less than a quarter of the specified value to get the response to where it should be. The sim suggests I should have a hugely overbright sound by now.

I had identical behaviour on both channels, I'd have expected differing behaviour if I'd made an error in construction. Maybe Cinemag put the smallest 'tap' in the wrong place and the inductance is way out on L5 and L6, I had no way to check that.