[tangent]

I had the boards made, from your Gerbers.

What it looks like is that someone took my EAGLE files and ran their own hand-rolled CAM processor against it instead of running the included CAM processor, so that all of the DRC rules have potentially changed.

You didn't, by chance, send the EAGLE files to your board house, too? As a rule, you should only send Gerbers to a board house, specifically in order to prevent them from second-guessing DRC rules and such.

Here's a better screenshot of the Gerbers from the bottom side near BUFR, this time oriented the "natural" direction and with the bottom side silk and drill hits in place:

I've been careful to do this on the Gerbers in the ZIP file on the web site, rather than the original versions here in my personal documents folder, just in case there's some important difference.

As you can see, the fenestrations are present, and there are no stray copper fingers going between the buffer pins.

OPA2134, OPA2132, OPA134.

And just to be sure, one of the OPA213x chips are in OPALR at a time, and the OPA134 is in OPAG, right?

I meant to demonstrate the opamp was in fact increasing gain, and therefore was alive.

You mentioned a scope: if you look at a sine wave going through the op-amp with the buffer missing (not shorted) a sine wave should look like a near-square wave unless the input signal is very tiny, since it'll be at gain of 100x or so.

Actually, it might not be very square, because many chips behave badly when you drive their inputs or outputs to the rails like that.

You really want to do this sort of test with the buffer shorted, so the outer feedback loop dominates, being lower in parallel resistance.

They're hard to track down, at a reasonable price

Maybe it matters that you're in the UK, but I see ready stock of about 20000 of them in onesy twosy quantities.

The question, then, is are those in fact LMH6321MRs?

You can partially answer that question by hooking them up to power and seeing whether they pass a signal with unity voltage gain, as they ought to. It's not absolute proof, because sometimes fraudulent designs will be something like a repurposed op-amp, so the main 4-5 pins will work as they're supposed to at low loads, but they won't put out the full power they're supposed to, and the other 3-4 pins might be disconnected or have other functions than the real thing.

Do your chips have the exposed solder pad on the bottom, or are they fully-encapsulated down there?

is there a way to test that the inner layers are in fact present, and correct?

Sure, but it's best done in conjunction with a copy of EAGLE and the *.brd file loaded, so you can get a picture of what's supposed to be connected where.

The file *.ic1 in the dfiles package is EAGLE layer 2, and it contains the V+ plane. Anything connected to V+ on the schematic goes through this plane.

File *.ic2 is the GND plane, and it's EAGLE layer 15.

The bottom copper layer is *mostly* V-, but it's also used for signal traces, as you see in the above screenshot. (I tried to keep most signal traces on the top layer, but sometimes I needed to pop down to the bottom copper layer to make things work.) This is what those fingers intruding between the buffer pins are.

Mostly you can answer this question by heeding the first commandment of troubleshooting: thou shalt check power and ground connections. For that, you don't need EAGLE, you need the schematic and a clear idea of which pads on the PCB connect to which power or ground nets.

I have a "reusable" SOIC8 - DIP8 adapter with more reasonable pin distances for testing. If I pop a buffer in there, is there a way to test the operation of it you can recommend?

Not on this board. On the PIMETA v1, yes, because there were both DIP-8 and SOIC-8 footprints for the buffers, but not on v2, since the LMH6321 isn't available in DIP-8.

if you think the boards are a lost cause then maybe I won't

I think they're questionable, but you can prove them worthy through testing.

I think there's a pretty good chance your boards are more prone to oscillations than the originals due to parasitic capacitive coupling to V- on the high impedance input pin on the buffers, but realize that that isn't a failure guarantee. If true, it just means there are some configurations that won't work on that board that would on a "real" PIMETA v2.

With those slow-ish op-amps, you're going to have more margin to cope with such problems than if you were trying to use something really fast and touchy.

 

[spbkaizo]

I'll reply more fully once digested, tested and evaluated. For now though:

What it looks like is that someone took my EAGLE files and ran their own hand-rolled CAM processor against it instead of running the included CAM processor, so that all of the DRC rules have potentially changed.

You didn't, by chance, send the EAGLE files to your board house, too? As a rule, you should only send Gerbers to a board house, specifically in order to prevent them from second-guessing DRC rules and such.

I put my hands up - I imported them into EasyEDA and then sent the output to the board house. Mea Culpa, entirely.

And just to be sure, one of the OPA213x chips are in OPALR at a time, and the OPA134 is in OPAG, right?

Indeed

You mentioned a scope: if you look at a sine wave going through the op-amp with the buffer missing (not shorted) a sine wave should look like a near-square wave unless the input signal is very tiny, since it'll be at gain of 100x or so.

Actually, it might not be very square, because many chips behave badly when you drive their inputs or outputs to the rails like that.

Can confirm this happened - I turned to pot down to maybe a 1/4 turn.

You really want to do this sort of test with the buffer shorted, so the outer feedback loop dominates, being lower in parallel resistance.

Interesting - this makes sense.

The question, then, is are those in fact LMH6321MRs?

Do your chips have the exposed solder pad on the bottom, or are they fully-encapsulated down there?

No, they do not. Guessing this is a red flag?

Will investigate the other stuff - apologies for splitting the post.

 

[tangent]

As far as I'm aware, you can't buy an actual SOIC-8 version of the LMH6321. You have a choice of the PowerPad/PSOP version or the DDPAK version, that's it.

I think there's an excellent chance those aren't actually LMH6321MRs. It'd explain everything.

 

[spbkaizo]

As far as I'm aware, you can't buy an actual SOIC-8 version of the LMH6321. You have a choice of the PowerPad/PSOP version or the DDPAK version, that's it.

I think there's an excellent chance those aren't actually LMH6321MRs. It'd explain everything.

Let's assume so. The chip from a reputable supplier arrived today, I only purchased one so I expect the tests will not quite be meaningless, but they did provide a strong level of reassurance.

DC Offset on the lucky recipient (on OPALR buffer) is now down to 23mv, so managed to lose what I assume must be a short from the 'fake' chips. I ran a 10khz signal through it, and the results were the square wave from the tests without the buffer, as with the new buffer installed. The signal is now making it to the buffer.

Time to buy the remaining 5, that I initially need. Confidence remains high, budget remains low

Image of the two chips is side by side. The marking on the original (05AC) is actually worse than the counterfeit (54RD). Shame they couldn't get the insides right...https://photos.app.goo.gl/wbtBD5hnCMaPxvAn8

 

[spbkaizo]

Update: It works

Installed the new chips, and just doing some testing, but it's all good so far. DC Offset is 9.4mv on one channel and 16.2mv on the other channel. Signal is clear, and VPP from 1v input is 5.2v output.

Now I can finally breathe again - thanks for the support. You've got me over this being a doomed project.

I'm going to go through more and more tests, before moving on. But, at least the project has moved back into 'fun' status.

 

[tangent]

I'm glad to hear it.

Care to name-and-shame the retailer who sold you those bogus buffers, so others can avoid them?

 

[spbkaizo]

Absolutely, I sourced them from AliExpress, seller was XYWElectronic Store.

I've known, and found it fascinating, that there are counterfeit chips around, and I'm certain I must have had some in the past.

Another lesson learned, to only buy from reputable suppliers - in my case, RS Components in the UK, who offered free delivery and they arrived in less than 24 hours on both occasions!

Thanks again. I've now got the following back in flight:

1. Desktop/Portable Headphone Amp
2. Purely Portable Headphone Amp
3. Pre-Amp with Dual Rails supply (waiting on some power supply components)
   
4. Mainly SMD based Headphone amp

This leaves me another 6 boards to study and learn with, so should keep me quiet for some time...

 

[tangent]

I sourced them from AliExpress, seller was XYWElectronic Store.

I think I found the listing you mean. It gives the brand name as PSTQE, not NatSemi or TI. A search for that brand on AliExpress suggests they're a maker of knock-off electronics.

Another red flag is that there are only two votes on the product, neither with commentary.

You might give this product the single star it deserves, and a reason why it deserves that. I think the single star is greatly overused, but in a case like this, where the product doesn't even try to do what its label claims, I think it's justified. If it even had a datasheet showing that it provides some other function and that they just HAPPENED to have picked the same part number as something else (yeah, right) it'd be worth more than a single star, but if all they're giving is a misleading part number...yeah, one star it is.

I don't consider the lack of a photo or the 3D rendering that doesn't match the actual item — missing PowerPad — to be a red flag. Legitimate resellers do this, too. EDIT 1: Digi-Key's 3D rendering shows the PowerPad, though; good on 'em!

EDIT 2: The price should also have been a clue. Digi-Key wants US $5.57 per buffer in qty. 10, whereas this AliExpress listing I found says $16.15 for 10. You can't explain a 3.45x difference between resellers any way other than "too good to be legitimate."

 

[spbkaizo]

Yes, I was hoping that mass economies of scale explained away the price. Remarkable how easy it is to delude yourself.

Finished one of my builds now, sounds good and happy so far - 8.4v (2 * 4.2v Li-ion '9v battery') powered portable, running with the buffers and an OPA2227/OPA134 combo. Battery life seems OK down to around 3.8v per cell, and they have a 600mah capacity. Consumption runs at around 60mah, I'm getting about 5 hours of use before things are noticeably `wobbly`. They have a charging circuit built in, and charge from that level in about an hour (at around 600mah, assuming split so 0.5C) which helps via micro-usb, so one less cable to carry....

The biggest surprise was that I seem to have managed to get around 6.5mv DC offset, so pleased with that!

https://photos.app.goo.gl/BhaBHTscjCzYdLSL9
https://photos.app.goo.gl/fUKC9fJV2ecjZmQg6

Going to focus on the pre-amp next, more free space to place components and the PSU.

Thanks again for the help.

 

[tangent]

Consumption runs at around 60mah

mA, not mAh. Divide milliamps into milliamp-hours to get hours.

That tells you in this case that you'd expect 10 hours of runtime, which is enlightening because it means your amp isn't draining the battery fully, since you're only getting half that runtime.

You probably can't fix it with a DC-DC converter. If you double the voltage to the PIMETA v2, you double the current draw on the battery, which nets you the same 5 hour runtime.

If it were me, I'd ignore it, since you don't want to fully drain a lithium chemistry battery anyway.

I don't think I've seen that case before. Has Serpac come out with a new color since I last looked, or has someone cloned the case closely enough that the PIMETA v2 fits into it the same way?

the pre-amp next

You'll want to build that as a 2-channel amp. Preamps normally short IG to OG, since the input and output grounds are normally shorted together, which at best makes the ground channel on the PIMETA v2 useless, and can actually cause it to misbehave.

 

[spbkaizo]

mA, not mAh. Divide milliamps into milliamp-hours to get hours.

That tells you in this case that you'd expect 10 hours of runtime, which is enlightening because it means your amp isn't draining the battery fully, since you're only getting half that runtime.

You probably can't fix it with a DC-DC converter. If you double the voltage to the PIMETA v2, you double the current draw on the battery, which nets you the same 5 hour runtime.

If it were me, I'd ignore it, since you don't want to fully drain a lithium chemistry battery anyway.

Yep, this is my 'travelling amp' for planes, trains, etc. 5 hours is plenty for these scenarios I estimate.

I don't think I've seen that case before. Has Serpac come out with a new color since I last looked, or has someone cloned the case closely enough that the PIMETA v2 fits into it the same way?

It's close, but not the same - there's no mounting posts inside for example, and I've ignored the fixing points as it holds together nicely with just an elastic band. I liked the colour...

You'll want to build that as a 2-channel amp. Preamps normally short IG to OG, since the input and output grounds are normally shorted together, which at best makes the ground channel on the PIMETA v2 useless, and can actually cause it to misbehave.

[/quote]
Indeed - I've read and re-read the docs and believe I know what to do (ditch the ground channel entirely, run real gnd into the voltage divider vgnd location) amongst others.

So, I've now finished the second one, and I'm blown away by the difference between the two sets of op-amps. This one is a semi-permanent (i.e. I take it home) installation on my work desk, running at 18-20v. I get 3 hours on battery power from this one. I'm intending to run it down each day for an hour or so, then plug it into charge throughout the day. I'm using a Lenovo laptop power supply for this, they're pretty quiet from what I can hear. Any tips for measuring actual noise?

I've gone with the AD8065/AD8066 combo in this one.

https://photos.app.goo.gl/QyGUFS5FFdxDb9zn6

I've got quite a few op-amps lying around, so I'll start rolling them around the handheld one to see if it improves. I'm so happy with the desktop one I'm not going to touch it.

 

[spbkaizo]

I've started on another Pimetav2 portable amp, and wanted to just document and seek others opinions into my approach.

POWER SUPPLY

I'm intending to power this one from a Li-On battery, with the battery connected to a DC boost converter with dual voltage output, negating the ground channel and components, using a module I've found. I'm still waiting on delivery of the module, but it is a buck boost converter which outputs to an arbitrary dual voltage power supply (+/-5v in my case). The module is one of these https://www.aliexpress.com/item/32817163270.html

I'm intending to connect the power outputs into the respective rails, with V+ being connected to B+ on the PCB and the power switch being jumpered, with power being controlled via another switch on the batteries + line.
QUESTION: The module claims to have 'soft start' but I'm curious about the supply to the pimeta, give the soft start on the amp is designed to operate with pre-charged capacitors. Insights as to potential issues would be appreciated.
QUESTION: The module has a switching frequency of 400khz. Should I consider attempting to add in a LPF into the circuit, and if so where would be preferable?

The boost converter supply has standard PCB pin spacing, in a row. As a result I'm going to effectively mount this on the scratch pad area, as it will keep things neat and tidy. For the most part, this will be another SMD build so the only topside items will be the C2 caps, and the C1 caps.

BATTERY CHARGING

For this, I'm going to mount a TP4056 onto a SOIC->DIP adapter as again, this give reasonable pins which will sit alongside the above DC boost converter. The pins on the TP4056 are very low component requirements count, and I'm confident I can fit the required pin 2 resistor to ground by bridging it to either Pin1 of Pin2 on the same PCB (both GND), leaving 4,5,6,7,8 with appropriate connections. I think it'll fit alongside the above module neatly.

CLASS A BIASING

I've decided against this part of the circuit to save space, and allow integration of the power module.

BASS BOOST

For the first time, I've decided to incorporate a Bass Boost via the addition of a cap to R4. This unit will be driving some open back headphones, which suffer from weak bass response. R4 will be 1.8k, and the cap will be a Panasonic 1uF SMD PPS film cap. I've gone this route as I don't want too much gain, but still want increased bass response. I'm not adding a switch as I want it on by default.

TARGET HEADPHONES

Headphones have high impedance drivers, 500ohm and are openback.

OPALR

I've decided to give the new(ish) ADA4625-2 a spin in the first instance.

BUFFERS

I'm leaving the BUFL/BUFR out this time, as the ADA4625-2 datasheet indicates it can drive 600ohm loads, and I'd like to see how it performs.

COMPONENT VALUES

For completeness, here's the component values for the rest that I'm considering.
R1 = 1k or 2k, or 4.7k? (QUESTION: looking for aadvice here!)
R2 = 100k
R3 = 1k
R4 = 1.8k (+1uF PPS cap) = 2.8x Gain
R5 = N/A
R6 = N/A
R7 = N/A
POT = 10k
C1 = 0.22uF (==7hz cut off)
C2 = 330uF 16v
C3, C4, C5 as schematic.

Comments solicited, education welcomed

 

[alphaman]

Tangent/anyone:

I built Pimeta 2 from one of your PCB's many years ago. I see you still have the pages up on your site. Very cool!

I've decided to run it dual, omitting the TLE2426 (I removed the device) and the Pimeta2 still functions with or w/o the IG mod you noted here:

https://tangentsoft.com/audio/pimeta2/ps.html

What do you (Tangent) mean connect IG (on scratchpad) to GROUND?

Here is how I have the pwr connected:

Two 9v batts in series (18 v total)

POS NEG-----POS NEG

POS (connected to BATT + on on Pimeta2 PCB) and NEG (connected to BATT- on Pimeta2 PCB)

I assume "NEG-----POS" (where batts are connected in series, the "----") is Ground.

Or did you mean another "Ground"???

Maybe my Pimeta2 was revised before you wrote:
"If you leave IC2 out, you can instead use a genuine dual voltage power supply, with 3 output terminals. You run + and - to the board just as you would for a single-voltage supply, and you run ground to one of the IG pads in the scratchpad area."

I realize it has been a long time since you wrote that or though about the project. Any clarification is hugely appreciate.

BTW: Continue to use PIMETA2 and PPA2 on a daily basis --- over 13 years! Very reliable units.

 

[Voodoochile]

That refers to Input Ground, which typically would float. I seem to recall isolated jacks were a ting for this reason.

(only responding as I got a notification from the wayback machine. I'm not actually back)

 

[tomb]

That refers to Input Ground, which typically would float. I seem to recall isolated jacks were a ting for this reason.

(only responding as I got a notification from the wayback machine. I'm not actually back)

And you ARE a blast from the past!

Yes, you are absolutely right about the PIMETA or any 3-channel amp. If you do not isolate the output jacks from the ground, it will short out the active ground channel to ground, rendering it useless (and sounding not so great, although it will perform OK). Ensuring that the active ground channel is isolated will ensure that the active ground channel is operating correctly.

I believe that is different than a floating ground. With a floating ground, you still have the same reference everywhere in the circuit and can "ground" the output jacks to that common "floating" ground (as in a CMoy).