EHX 2 Second Digital Delay (80's) - Schematic

[Scruffie]

Trace of the analogue board, might have an error or two as a few traces I couldn't follow under chips/sliders & a couple of values were obscured in my photos and I'd given the unit back to its owner after repair some time back but 99% correct.

Would actually be buildable with the digital section as there's no ROM involved, although probably prohibitively expensive to do so.

[rustypinto]

Is there a page 2?

[Scruffie]

Is there a page 2?

Not as of yet... didn't seem like there was that much interest to get to tracing it.

I did post a few photos over on the other website.

[rustypinto]

Ah ok, I was just curious.

Smart move not tracing it, I couldn't imagine anyone would be interested in building this either.

Thanks for all the work you put into it anyway. What an incredible design for the time!

[Scruffie]

Ah ok, I was just curious.

Smart move not tracing it, I couldn't imagine anyone would be interested in building this either.

Thanks for all the work you put into it anyway. What an incredible design for the time!

I would build one (it's an excellent sounding delay) but creating a layout is far beyond my capabilities and it would certainly have to be a labour of love. But hopefully it'll be handy for anyone else that needs to repair one.

The work of David %^&*erell, if you think this design looks impressive, you should see inside a Super Replay.

[Slowpoke101]

When Scruffie first posted his analog schematic I went one a search for the digital board's diagram. I only found a hand drawn diagram of the 16 second delay version.



Unfortunately I do not recall where I found it but I will find out and thank them.

The diagram is a bit hard to read and is rather complicated. It is complicated because the delay uses Dynamic RAM (4164) which has some ridiculous timing requirements for its refresh demands. Also the multiplexing requirements for row and column addressing is a pleasant way to go quietly insane. If you were to make a more modern version I would recommend using CMOS static RAM which is far more simpler to use and would only need about a third of the shown circuitry.

[Scruffie]

When Scruffie first posted his analog schematic I went one a search for the digital board's diagram. I only found a hand drawn diagram of the 16 second delay version.



Unfortunately I do not recall where I found it but I will find out and thank them.

The diagram is a bit hard to read and is rather complicated. It is complicated because the delay uses Dynamic RAM (4164) which has some ridiculous timing requirements for its refresh demands. Also the multiplexing requirements for row and column addressing is a pleasant way to go quietly insane. If you were to make a more modern version I would recommend using CMOS static RAM which is far more simpler to use and would only need about a third of the shown circuitry.

Unfortunately the 2 second isn't just a 16 second with less DRAM, the design is quite different.

There are no schematics out there for the 2 second, instant replay or super replay to the best of my knowledge, hence the trace of this one.

Tell me more about this CMOS ram though... be great if it could retro-fit to get work-a-likes of some of these old digital delays.

[Slowpoke101]

There are no schematics out there for the 2 second, instant replay or super replay to the best of my knowledge, hence the trace of this one.

Tell me more about this CMOS ram though... be great if it could retro-fit to get work-a-likes of some of these old digital delays.

I'll see what I can do.
I found the website again that I got the digital schematic from and got the analog diagram as well. Very similar to your diagram but far less readable. I will see if I can work out the timing diagrams for the 16 second board which will help to interface static RAM to it instead of dynamics (which throws away about a third of the logic ICs and reduces power demands). This could be fun.
In the meantime download some datasheets on 4164 Dynamic RAM and 62256 Static RAM. Have a look and see how easy it is to use the static gear.

[DIY Bass]

I am not sure about modern parts, but from dim memory mucking around with RAM in the 80s, the disadvantage with SRAM was that it was a great deal slower.  If access speed is an issue that you may have to go with DRAM still

[rustypinto]

I can understand why you would want to trace/ reverse-engineer these designs for repair purposes or getting a better understanding of SRAM or DRAM. However, if you want to clone these designs, you are probably better served by emulating the storage media on a microcontroller or PLD.

I once built a TR-909 clone, and I was able to emulate the cymbal ROMs using a PIC with enough equivalent flash, enough GPIO pins to form a parallel interface, and a fast enough instruction clock to mimic the ROM's timing requirements. I then connected up to the DAC like usual, and the results were identical.

I imagine you could do something similar here using a microcontorller's internal SRAM or a SPI based SRAM, then mimic the interface to the converters. You would save a ton of space too.

[PRR]

...the disadvantage with SRAM was ...

It was a LOT more expensive per bit. Dynamic basically used stray capacitance for storage, static needed several actual transistors.

And speed but no audio delay gets anywhere near the speed of 1980s RAM. 

Dynamic needs every bit to be accessed many times a second before the data bleeds off. This is a distraction if you need to hold, say tax rate, on the side for a moment until the subtotal is added-up. But a constant in/out stream like audio delay "can" be a natural dynamic refresh.

I can't believe we have memory a thousand times cheaper and a thousand times larger than the 1980s but we can't throw-together a large audio delay in minutes.

[Scruffie]

Some photos of the digital board to help give an idea of the circuit to those more knowledgeable on this digital business.

[Slowpoke101]

Thanks for uploading those photos Scruffie, they are very interesting. I notice that unlike the 16 second unit that there is no prom used. It used a AM27S19DC. But I do see that the memory used in the 2 second board are marked as D2117-4. These are actually uPD2117-4 2K STATIC RAM ICs. Twelve of them. So the timing requirements of this board are a lot simpler but the power supply requirements are greater than the 16 second board.

Paul does have a point about memory now being so cheap and so large. So why can't we throw together huge delays easily? Well for me it's because of nostalgia - yes I know it's all in the past. I enjoy reviving and restoring the old gear and also improving it where possible. But the idea of a delay unit as touched on by Paul is also interesting.

Now DIY Bass brought up the old static RAM's problem of access speed. Some of the old stuff such as the 2114 and MC6810 (all 4 bit by the way) has access times as slow as 400nS. Which was a concern as it made even running a processor at 1MHz a risky idea. Modern static RAM access times is measured usually at under 100nS. Dynamic started at about 200nS and got down to about 65nS. Now it's all gotten rather silly. With audio delays you don't need anything like these speeds.

Rustypinto - Well done on the TR-909 clone. Not an easy one to do at all. Your use of the PIC was quite ingenious. I hope that you wrote up what you did and shared it somewhere so others could learn from your efforts. Now using micro-controllers is a good idea and does make for great savings in board space.

[Scruffie]

Edit: Huh, maybe I'm wrong... I was certain when I'd looked before they were listed as 16kB 150nS DRAM.

Both this and the 16 second use the same power supplies though +12, +5 & -15V

[Slowpoke101]

My parts database may have misled me with this. I do apologise. 4116 devices makes more sense when I look at the solder side of the board. I will go and make sure.
Thanks for letting me know.

Edit: But we both may be wrong. I've got some really old Japanese data books lurking somewhere here. This could get worrying. The power rails will give it away. Dynamics are usually +5V devices only where older statics required the full  range of supplies.

Edit Edit: Not to worry. We will get to the bottom of this little mystery. It is an unusual memory chip. Interesting that both the 2 second and the 16 second had the same power supply parts. But that's EHX for you. Never throw out anything if you can get away with re-using it again.

[Scruffie]

My parts database may have misled me with this. I do apologise. 4116 devices makes more sense when I look at the solder side of the board. I will go and make sure.
Thanks for letting me know.

https://www.ebay.com/itm/393631352305
This listing has them as SRAM too, it was a long time back now, not sure where I got the DRAM info from so maybe I got bad info.

But as edited in my previous post in case you missed it, the power supplies between this & the 16 were the same.

[Slowpoke101]

Found it! 

It IS a dynamic RAM. It's by Intel (I should have recognised the symbol on the chips).
It is a 16K by 1 dynamic. So that little mystery is solved. Now to figure out where to go to from here.

[rustypinto]

I should probably also mention some other important aspects of the PIC used in my 909 upgrade:

1. You can get 5V variants, which easily interface to the existing logic.
2. Some have internal SRAM that exceeds the amount used in these delay designs. In the case of the 909, there was enough internal flash to store the cymbal samples.
3. Some variants are available in a DIP package.

These attributes, along with sufficient GPIO and extensive timing analysis led me to selecting a PIC instead of spending too much money on a difficult to find EPROM and a suitable programmer. I also did a lot of previous work with PICs and had a programmer, etc, which helped a lot. But the most important aspect of the design was the continued use of the low-resolution DAC circuit (6-bits!), which gave the cymbals their unique sound. I was also considering eliminating other logic (eg. oscillator and counter) in the design, but I ended up re-using most of the original circuit.

U60 is the "PICROM" shown here, which is PIC18F26K22-I/SP drawn as a EPROM:


I suspect you could do the same with these delay designs if your intention is to clone them and do your own layout. A lot of work would be needed to understand the clocking/ timing and matching it to a suitable microcontroller. If done properly, you can retain the audio characteristics while modernizing the RAM aspect of the design.